! ! ' i SOLAR-TERRESTRIAL PREDICTIONS PROCEEDINGS i VOLUHE i PREDICTION GROUPREPORTS Edited by Richard F. Donnelly Space Environment Laboratory , Boulder, Colorado 80303, U.S.A. August 1979 , The International Solar-Terrestrial Predictions Proceedings : and Workshop Progra_n was hosted by the NOAA Space Environment Laboratery. The workshop was held April 23-27, 1979, at the College Inn in Boulder, Colorado. Science Sponsors of the Program: AAS: American Astronomical Society AGU: American Geophysical Union AMS: An_rican Meteorological Society COSPAR: Committee on Space Research IAGA: International Association of Geomagnetism and Aeronomy IAU: International Astronomical Union lUWDS: International URSIGRAM and World Days Service SCOSTEP: Scientific Committee on Solar-Terrestrial Physics URSI: Union Radio Scientifique Internationale; Commissions E and G Science and Financial Sponsors of the Program: Air Force Geophysics Laboratory :. Air Force Office of Scientific Research Department of Energy National Aeronautics and Space Administration National Science Foundation ! NOAA Environmental Research Laboratories UNITEDSTATES NATIONALOCEANICAND EnvironmentalRe.arch DEPARTMENTOF COMMERCE ATMOSPHERICADMINISTRATIONLaboratums JuanitaM. Kr_ps,Se_retar/ Rh:hardA.Frank.Administrator WdmotN Hess.Director 1980010188-002 https://ntrs.nasa.gov/search.jsp?R=19800010188 2020-03-21T20:08:29+00:00Z d° f NOrlCE lne Environmental Research Laboratories do not approve, recommend, or endorse any proprietary product or proprietary c ;_ material mentioned in this publication. No reference shall be made to the Environmental Research Laboratories or to this publication furnished by the Environmental Research Labora- tories in any advertising or sales promotion which would in- dicate or imply that the Environmental Research Laboratories approve, recommend, or endorse any proprietary product or proprietary material mentioned herein, ot which has as its purpose an intent to cause directly or indirectly the adver- ; tised product to be used or purchased because of this Envi- . ronme_,tal Research Laboratories publication. " _ For sale by the S_peffntendent of Documents, U.S. Government Printing' Omee, Washington, D.C. 20402, - • I _Orderby SD Stock No. 008-023-00041-9) J, • t 1980010188-003 PROGRAMCO/_tlTTEE F. E. Cook, IPS, Sydney, Australia A.B. Severny, Crlmean 0bserv.,U.S.S.R. A. 0. Oanllov, Moscow, U.S.S.R. H.A. Shea, AFGL, Bedford, Mass. Y. Hakura, Tokyo, Japen P. Simon, Meudon, France G. R. Heckman. NOAASEL, Boulder W.C. Shoddy, NASA-MSFC, Huntsville A. P. R|tra, NPL, New Delhi, India H. Tanaka_ Tokyo Observ., Japan G. A. Paulikas, Aerospace, Los Angeles J.M. Wilcox, Stanford University N. V. Pushkov, IZHIRAN, U.S.S.R. D.J. Williams, NOAASEL, Boulder i! WORKINGGROUPSand LEADERSLong-Term Solar Activity Predictions: Mr. P. S. Hclntosh, NOAAERL SEL, Boulder, Colorado 80303 USA Short-Te-m Solar Activity Predictions: Dr. P. Simon, Observatory, Heudon, France Solar Uind and Magnetosphere Interactions: Dr. C. T. Russell, institute of Geophysics and Planetary Physlcs_ University of California, Los Angeles, California 9002k, USA c Geomagnetic Storms: Or. S. I. Akasofu, Geophysical Institute, University of Alaska, College, Alaska 99701, USA * Energetic Particle Disturbances: Dr. G, A. Paulikas, The Aerospace Corporation, P. O. Box 92957, , Los Angeles, California 90009, USA t, Nagnetosphere-lonosphere Interactions: Dr. Richard R. Vondrak, Radio Physics Lab., SRI International, Nenlo Park, California 94025, USA : High-Latltude E- and F-Region Ionospheric Predictions: Dr Robert D. Hunsucker, Geophysical Institute, University of Alaska, College, Alaska 99701, USA ; Midlatitud,- and Equatorial E- and F-Region Ionospheric Predictions: Dr. Charles M. Rush, Institute of Telecommunications Sciences, National Telecommunications and Information Administration, Boulder, Colorado 80303, USA • D-Region Ionospheric Predictions: Dr. Eivind Thrane, 0ivision for Electronics, Norwegian Defense Research Establishment. P. 9. Boks 25, Kjeller, Norway Solar-Weather Predictions: Or. K. H. Schatten, Department of Astronomy, Boston University, 7_5 Commonwealth Avenue, Boston, Massachusetts 02215, USA Communications Predictions: Dr. A. P. Mitra, Radio Science 9ivislon, National Physical Laboratory, Hi.llside Road, New Delhi - 110012, India Subsection on Ionosphere-Reflected Propagation: Dr. B. M. Reddy, Radio Science Division, National Physic-_i Laboratory, Hillside Road, New Delhi IIOOI2, India Subsectioa on Trans-lonosphere Propagation: Dr. John A. Klobuchar, AFGL-PHP, Hanscom Air Force ,% :: Base, Bedford, Massachusetts 01731, USA Geomagnetic Applications: Or. Wallace H. Campbell USGS Box 2S0_6, 0enver Federal Center MS 9_6, _;. Denver, Colorado 80_25, USA Space-Craft Environment and Harmed Space Fllght Applications: Dr. A. L, Vampola, The Aerospace Corporation, P.O. Bo_ 92957, Los Angeles, California 90009, USA [= " ; iii - :: 1980010188-004 PREFACE The International Solar-Terrestrial Prediction Proceedings and Workshop Pro- gram (ISTP/P-W Program) included the following: (1_ an open call for contrib- uted papers on solar-terrestrial predictions; (2) _nvited review papers about (a) the predictlont warning and mopEtoring services of groups that regularly Issue solar-terrestrial predictions; (b) the current and future needs for predictions by groups that use solar-terrestrial pr_dictions, and (c) current knowledge of selected topics in solar-terrestrial physics and applications; (3) working groups on fourteen areas of interest for solar-terrestrial pre- dictions; (4) a preprint exchange from October, 1978 through March, ]979; (5) a workshop of representatives of the working groups; and (6) the Solar- Terrestrial Predictions Proceedings. These proceedings consist of four vo i umes: Volume I. Prediction Group Reports Volume 11. Working Group Reports and Reviews Volume I:!. Solar Activity Predictions Volume IV. Prediction of Terrestrial Effects of Solar Activity Volume I reviews the current practice in solar-terrestrial predictions. Vol- t ume II presents the recommendations and reports developed by the workinggroups at the workshop. Topical reviews and pape:s on the current and future needs for predictions are also included. Volume II reports the results of discussions of questions such as: What predictions are needed? Where are current predictions inadequate? What recent results from solar-terrestrial i research should be applied to improve solar-terrestrial predictions? The pro- gram did not include a conference where the authors presented their papers orally. Working group participants were asked to read the preprints and cor- ! respond with the authors and other group members before meeting at the work- shop. Participants included forecasters, scientists and prediction users. Volumes III and IV present individual suggestions for particular predictionsche . _" The goals of the program were as follows: (i) to determine and document the current state-of-the-art of solar-terrestrial predictions, the applications of i these predictions, and the future needs for solar-terrestrial predictions, (2) tO encourage research, development and evaluation of solar-terrestrial ! predictions, and (3) to provide indepth interaction of prediction users, fore- I _ casters and scientists involved in the research and development of prediction techniques. To achieve the first goal, we invited forecast groups and user _ _ groups to review their activities. The working groups concentrated on deriv- / ing recommendations for future needs pertinent to solar-terrestrial predic ....i ,_ tions. The early call for contributed papers was made to achieve the second goal, i,e. authors had more than a year to orient their work towards a paper " on predictions. The workshop was aimed at the third goal. The criteria used for selecting working-group representatives to attend the workshop included the goal of having at least one forecaster per group. The working groups on ' i solar-activity, magnetospheric physics, ionospheric physics, etc., were aimed at attracting scientists. On the other hand, the working groups on communic_ _ tions, geomagnetic applications, and space-craft and man-in-space applications wer_ntended to attract prediction users as well as applications scientists. , _.___-_)-_ Richard F. Donneily i HOAA/ERL/SEL August 16, 1979 Boulder, Colorado 80303 USA i 1980010188-005 %BOB DOEKER - 1920-1978 In 1965, Bob Dqeker was given the task of defining and developing the : organization in Boulder that became the Space Environment Services Center - (SESC) and the IUWDS World Warning Agency. His pioneering efforts brought : together separate programs in several government agencies and forged them into a single national center providing a full range of space environment services and predictions. His work laid the foundation for the Solar- Terrestrial Predictions Workshop by identifying a broad range of users who : found they needed these services and an organization for delivering them. : The remaining necessary element for the program he envisaged _an improved ability to measure and predict the variations that affect the user's systems _was addressed by this workshop and these proceedings. Bob Doeker passed away September 24, 1978. He had received the first announcement cf this workshop and was excited. He would have liked to come and participate. His presence would have provided a formidable critic for new forecasting models. He would have questioned whether they improve on the existing methods and whether the predictions that result produce more benefits than the cost to provide them. If the answers were clearly positive, then he would have pushed for their adoption. Bob Dceker was born in Covington, Kentucky, on July 20, 1920. He origin- ally trained as a civil engineer, then became a meteorologist specializing in tropical disturbances. By his own admission he was the best weather fore- caster in the Air Weather Service. He was one of the pioneers In setting up the Air Force space environment program. At an opportune time, he received L the call to come to Boulder. His legacy is a dynamic U.S. space environment program that is left for future development. Gary R. Heckman SESC ":. vi % 1980010188-006 0OVERVIEW One goal of the workshop was to determine current practices in solar-ter- restrial predictions. Invitations were sent to groups that routinely make " some type of _olar-terrestrial prediction asking them to review their current prediction techniques. We suggested the reviews of predictions Include the following: the empirical and/or theoretical basis, customers, applications, examples, input data, and accuracy. This volume presents these prediction- group reviews, some describing particular predictions in detail and others _ briefly reviewing all the types of predictions issued by the group, _. Host prediction groups provide predictions for ionosphere-reflected HF radio propagation, which has always been the dominate appiication of predlc- . tions. A few forecast centers provide services for radiation hazaru predic- tlons for manned space flights and high-altitude and high-latitude aircraft flights, Several forecast centers serving countries at high latitudes include t [ geomagnetic activity predictions for geomagnetic survey organizations ancl long-line electric power networks. _'_ _" Host predictions are based on empirical techniques, Our knowledge of the physical processes involved has guided the form of the empirical relations i used. Recently, models based on physical principles have been developed for science research to describe certain parts of solar-terrestrial disturbances. :_ _ In most cases, these models are not yet regularly used at forecast centers because they have not yet demonstrated a consistent improvement over older, simple empirical techniques. On the other hand the route toward improved _ predictions will undoubtedly follow new understanding of the solar-terrestrial environment. Conversely, the major applications area that needs further knowledge of solar activity and its terrestrial effects is the solar-terres-# _rial predictions service. Other applications, e.g. for space system design, ¢_ither already have or are receiving sufficient solar-terrestrial data, or their needs can be met by direct measurement programs. Solar-terrestrial predictions are disreputable in the eyes of some : _ scientists, as if they were crystal-ball fortune telling. We hope to dispel th-is by presenting reviews of current prediction practice. We also hope that scientists will be inspired to suggest better scientific prediction techniques. The predominate use of the root °°predict'° in the scientific literature is / " distorted in the form that theory predicts something, lacking the sense of ! time, i.e. the sense of foretelling has been abandoned. We challenge scien- _ tists to apply their work more for forecasting and to put the future back into their use of "predict". Economic factors are an important influence in solar-terrestrial monitor- ing and prediction services. We are currently capable of certain improvements in short-term forecasting, e.g. through satellite monitoring of the backside of the sun, but the potential improvements have not yet justified the expense. Furthermore, predictions compete with other approaches for minimizing the i, effects of solar activity on particular systems, Some systems can be redesigned to minimize or even eliminate the undesirable effects induced by solar activ- '_ : vii 1980010188-007 lty by reducing the sensitivity of the system to these effects. In other cases, * the systems can be made to automatically adapt to the new conditions outing ! solar activity. Economic factors influence which approach is used. ! Computers are now used extensively in forecast centers to process, com- municate and display data. Some computer terminals are designed to help the user apply the predictions and real-time data to interpret their effect on • his particular problems (see the review paper by P. Argo), Some predictions ; are made by computer algorithms, but so far most predictions are made by exper- ienced forecasters (see the discussion of computer forecasting in the review paper by Severny et al.). In the coming decade, minicomputers with graphical television displays will probably be used more extensively by forecast centers to make routine predictions, as a research tool For developing and testing new prediction schemes, and to improve the forecast-customer interface. The more intense a solar-terrestrial disturbance, the more important it is to forecast centers; the importance is nonlinearly proportional to inten- sity. The number of solar-terrestrial disturbances greater than a particular • intensity is generally irwersely proportional to that intensity. Combining the distribution of events and the importance weighting factor per event still leads to the group of very major events being the most important. The reviews in this volume tend to discuss major disturbances. In contrast, most science : research studies in recent years are dominated by small events because of their large number and ease of observation. The need for long-term solar activity predictions is great, but the accur- acy of current predictions is rather low, especially for times longer than ten years. The work in this area is empirical. We lack a scientific under- standing of the physical causes of long-term solar activity. Consequently, the time scales considered by forecast centers tend to be less than one solar _ cycle. During solar minimum and during the start of a new cycle, there is an increase of activity on predicting the new cycle. At the other end of the time scale, forecast centers predominantly warn of events in progress, or predict particle radiation and geomagnetic storms hours to one or two days : in advance. Solar-flare-activity is forecast for the next several days. Some forecast centers consider one month in advance for recurrent solar activity related to solar rotation. The enclosed review papers do not discuss certain approaches apparently because they are not used, e.g. planetary periods are apparently not used. The great hope of the past decade has been the search for a reliable universal flare precursor. None has been found, although certain "precursors a' are : sometimes helpful in special cases. Also note that many solar flare forecasts depend on ground-based data with fine spatial resolution that are used to study the detailed evolution of active regions. Satellite data have had great suc- cess In real-time monitoring, event detection, and warning applications, but have not yet had a major direct impact on predictions. Recent satellite exper- iments have illustrated how we can observe coronal holes, which should be ; helpful in predicting the solar wind structure at Earth. However, the time delay between new research results and routine applications is large enough In this area that such satellite measurements are not yet routinely part of the operational forecast services. The possibilities for Improved predictions viii t 1980010188-008 s_ " J: through Improved observations from space and through Improved understanding of solar-terrestrial physics are exciting. I hope they will challenge the reader to contribute to making such Improvements a reality. Richard F. Donnelly July 1, 1979 ISTP/P-W Program Chairman \ ACKNOWLEDGEHENTS r: The editor wishes to thank the many persons and organizations who helped conduct or support the International Solar-Terrestrial Proceedings and Work- shop Program. Without their support, this program would not have been pos- sible. I am grateful for the science sponsors and financial sponsors listed on the title page. I appreciate the help of the Program Committee and the ; extensive work of the Working Group Leaders listed on the page after the title page. Numerous persons helped with mailing preprints, including the follow- ing: Dave Bouwer, Ed Green, Bob Hetzel, Klm Hattison, Dan Watters, Jim i Williams, the entire staff of SESC, and my family, Scot, Lora and Betty Ann. ! I'm grateful to the many typists who participated, including the following: : • Gloria Baros, Hillie Bolme, Kristi Brubaker, and Kim Hattison. I wish to thank NOAA ERL Publications Office and Rush Services for editorial aid. I i appreciate the College Inn staff for their help during the workshop. I am thankful for Jim William's help with the bus tour dur;ng the workshop and Kim i Hattlson's help with the banquet. I appreciate Georglna Picketing and Andy Smith for helping to obtain part-time help and solving travel problems during the workshop. ? { , ix f • g 1980010188-009 SOLAR-TERRESTRIALPREDICTIONSPROCEEDINGS VOLUHEI PREDICTIONGROUPREPORTS TABLEOF CONTENTS Program Committee and Working Group Leaders ............................. iii Preface ................................................................... v Bob Dooker 1920-1978 vie6oeeooeotelealJoeoeeeeoloQieelelJooeege, aoo otl0olo Overview ................................................................ vii Acknowledgements......................................................... ix The Forecasting Center of Keudon, France J P. Simon (France) ............. 1 The French Short Term Radiopropagation Predictions in the Decameter Band m p. Lassudrle-Duchesne, A. M. Bourdila, and H. Sizun (France) ....... 12 Propagation Predictions for the HF-Range by the Research Institute of the Deutsche Bundespost _Th. Damboldt (West Germany).......................... 25 i A Hethod of Predicting Geomagnetic Activity Based on a Coronal Hodel ! of Relations Between Solar and Geomagnetic Activities J J. Halenka(Czechoslovakia) ........................................................... 44 ! KAZIA - A Computer System for Solar-Terrestrial Data Processing and ! ObJective Prediction --K. Stasiewicz, H. Haksymienko and H. Jakimiec i (Poland) ................................................................... 61 i Estimation of Solar I_ "e Influence on Radio Circuit Transmission Loss -- Z. Klos and K. Sta_ ....'.z (Poland) ......................................... 67 _ Soviet Short-Term For_,asts of Active Region Evolution and Flare Activity _A. B. Severny, N. N. Stepanyan, and N. V. Steshenko (USSR).............. 72 i Forecasting of Solar Energetic Radiation at the Forecast Center of the _ Institute of Applied Geophysics S I Avdyushin, N. K. Pereyaslova, " F. L. Dlikman and Yu. H. Kulagin (USSR).................................... 8_ : Forecasting Ionospheric and Geomagnetic Conditions at the lAG Fore- casting Center _S. I. Avdyushin, A. D. Danilov, A. B. Halishev, G. N. (USSR) i04Novikova and P. N. Svidsky ......................................... _ Long Term Solar Activity and Ionospheric Prediction Services Rendered : by th_ National Physical Laboratory, New Delhi m B. H. Reddy, S. Aggarwal, D. R. Lakshmi, S. Shastrl and A. P. Hitra (India) ............... 118 ¢ A Simplified Indexing of F-Region Geophysical Noise at Low Latitudes -- S. Aggarwal, D. R. Lakshmi and B. H. Reddy (India) ........................ 134 Spiral Sunspots and Solar Activity Forecasts _Y. Ding and B. Zhang (China) 140_- IIIIIIIIlllIIIlIIIIIIIIllOIIIIIlIII_OIIIlilttO*IIIlIIIlIIlIIIIIOII I ( Solar Activity Predictions at Peking Observatory _Peking Observatory(China) ................................................................... 154# : Long-Term Forecasting of Solar Activity----Z. Xu, A. Zhao, Y. Mei and Q. Guo (China) 163llll_llllllOlllllJllOIIlllllllillltlellllillttllllllllllllllllt xi t 1980010188-010 Prediction of Large Active Regions at the East Solar Limb-- q. Guo (China) ................................................................... 168 Short-Term Solar Activity Forecasting R X. Chen and A. Zhao (China) ...... 176 Forecasts of Geomagnetic Activities and HF Radio Propagation Conditions : Made at Hiraiso, Japan --K. Marubashi, Y. Miyamoto, T. Kidokoro and T. Ishii (Japan) .......................................................... 182 A Review of Short-Term Flare Forecasting Activities at Toyokawa m S. Enome (Japan) .......................................................... 205 Radio Propagation Prediction Services in Japan --R. Maeda (Japan) ........ Zi2 Radio Disturbance Warning Issuance System--R. Maeda and H. Inuki (Japan) ................................................................... 223 A Review of the Operation of the IUWDS Regional Warning Centre at the Ionospheric Prediction Service, Sydney --F. E. Cook and P. Davies (Australia) ............................................................... 229 A Disturbance Forecaster's View of the September 1977 Events --F. E. Cook (Australia) .......................................................... 239 The Use of Ionospheric Indices to Make Real- and Near-Real-Time Forecasts of foF2 Around Australia _L. F. McNamara (Australia) .................... 249 Prediction Limits For foF2 --P. J. Wilkinson (Australia) ................. 259 A Weekly Ionospheric Index--J. F. Turner and P. J. Wilkinson (Australia) ............................................................... 279 Coronal Holes Inferred From the "Fleurs _' East-West Solar Scans --Brian G. Ferguson (Australia) ................................................... 288 PROPHET: An Application of Propagation Forecasting Principles --P. E. Argo and I. J. Rothmuller (USA) ........................................... 312 Predictions of the Space Environment Services Center --G. Heckman (USA)..322 Geophysical Forecasting at AFGWC--R. L. Thompson (USA) .................. 350 Ionospheric Modeling at Air Force Global Weather Central --T. F. Tascione T. W. Flattery, V. G. Patterson, J. A. Secan and J. W. Taylor, Jr. (USA)..367 MSFC Solar Activity Predictions for Satellite Orbital Lifetime Estimation _H. C. Euler, C. A. Lundquist and W. W. Vaughan (USA)................... 378 Examples of "Problem" Flares or Situations in Past Solar-Terrestrial Observations _H. W. Dodson, E. R. Hedeman and O. C. Mohler (USA)........ 385 Forecasts of Geomagnetic Activity by Ottawa Magnetic Observatory: Their Reliability and Application --J. Hruska (Canada) ......................... 39B PPS76 - A Computerized "Event Mode" Solar Proton Forecasting Technique D. F. Smart and M. A. Shea (USA).......................................... 406 Summary of Forecasters' Meeting on April 25, 1979,During the International Solar-Terrestrial Predictions Workshop _J. B. Smith ..................... 428 Xll 1980010188-011 [' 8'0-18463 THE FORECASTING CENTER OF MEUDON, FRANCE Paul Simon Observatoire de Meudon 92190 Meudon I. HISTORICAL BACKGROUND The first daily broadcast of solar and geophysical data, now so called URSIGRAM, began at the Eiffel Tower Station, at Paris, on December Ist, 1928 "in order to study the geomagnetic storms and their related phenomena'*. Formally the Regional Warning Center was established at Bagneux during the AGY (1957) and was moved to the Observatory of Meudon in 1965. Real solar forecasti,lg began in the summer of 1965, at the request of the SPARMO Secretary, Mr. J. P. Legrand, for coordinated balloon fl_^hts carried out in Finland, Norway and Sweden by several European groups interested in the study of auroral X-ray emission and arrival of solar protons. The second campaign was carried out during the winter 1965-1966 for coordinating simultaneous balloon flights in the same countries and in Antarctica. In 1966, during the Proton Flare Project coordinated by Meudon, they succeeded in forecasting three solar proton events: this performance surprised the solar astronomers contributin9 to the project. The taff is made up of two solar astronomers and three technicians operating the Center six days a week and nine hours a day. They currently issue a daily solar and geophysical forecast, around 1500 local ti_e, according to the GEOALERT format of the IUWDS. in addition, for temporary campaigns and on request, they issue special forecasts related to the geo- magnetic activity and to a few selected categories of solar events. The scientific environment is a solar observatory in which, since its establishment in 1876,several astronomers were interested in studying so:ar terrestrial relationship an_, since 1956, after the building of the radio- astronomical station of Nancay, in uniting both the optical c_d radio data for studying solar activity. 2. SEVERAL MAIN FEATURES OF SOLAR ACTIVITY AND THEIR FORECAST Many categories of solar data show "solar activity! spectroheliograms, radio and X-ray data, white light coronal observations, particles data, photospheric images, photospheric magnetic field, etc...However, roughly speaking, all these categories of data contribute to the identification of tw.__.oobasic phenomena: a. The active centers built up by a magnetic field which is the origin of different phenomena observed in all p_fts of the solar aLmosphere. l 1980010188-012 b. The Flares observed in the chron_sphere but so,_etimes accompanied _, by X-ray and radio bursts which are signatures of phenc,nena occurring in the co rona. Of course, in a more complete description, we would have to add the recently identified coronal holes, the streamers and the coronal transients which play a key role in solar terrestrial relationship. ,+ Consequently, in order to follow, day after day, the evolution of the solar activity, we need to have series of data related to all the layers of the solar atmosphere: a single series of data would supply us with a very incomplete view of solar activity. Among the Flares, only a few are energetic events. As an example, from 1974 to the middle of 1977, out of 4685 flares, only 50 flares, i.e. close to one per cent, were, according to the microwave d&ca, true energetic events: they occurred in a few selected active centers and only during a short interval of their life. Consequently, if we issue a daily forecast for all the spotted groups, which are just 50 per cent of the active centers on the solar disk, and if our forecast is "no energetic event" for all these active centers whatever they look like, we will have 95 per cent of right forecast but, of course, we would not succeed in forecasting a single energetic event. Due to the current distribution of +he energetic events among the active centers, the useful forecast concerns only a small number of active centers and, some- times, during a short part of their disk crossing. There is some evidence for a relationship between the time scale of a phenomena and the accuracy of the relevant forecast: with one forecast a day, one can expect to predict accurately the evolution of an active center (time scale of a few weeks) but not the time of a flare (time scale of a few minutes). A_ present we just succeed in identifying the magnetic configura- tion in which "flares like to occur" and we actually forecast (or identify) more the flare buildup than the flare itself: for large flares, it can be done several days in advance. If we succeed in identifying the "flare trigger," we would have the possibility of issuing an accurate forecast but we wuuld have to do it at a very short notice (probably a few minutes or a few seconds in advance). Presently, our knowledge of solar phenomena is so limited that to issue more than one forecast a day does not really imarove ; the accuracy of the prediction. 3. DATA AVAILABLE Our forecasts are established according to a small set of ,ocal data and a large quantity of data circulating in the IUWDS network. 3.1 Local Data The solar monitoring program carried out at both Meudon and Nan_ay supply us with a set of "home made" data: sunspot drawing, (disk s!ze 30 cm.), spectroheliograms in H_, Klv , K3 (disk size 85 mm.) and, during wonderful 1980010188-013 • | intervals, photospheric magnetograms. On average, the weather permits us to get optical data two days out of three: for preparing the forecast we use the negative plates themselves, sometimes not yet dried. In addition, we have a classical flare patrol which can be used, weather permitting, to identify the flares reported by radio-astronomers or by any other flare effects: SlD, SFA, etc... The Calcium plages help us to identify Active Centers, H_ K1v spectroheliograms give the possibility of having the precise sites of sunspots in plages and of the filaments and of the bright H_ features with respect to the spots. Of course, magnetograms when available give addi- , tional pieces of information about new emerging flux (which does not have an inversion line surrounded by a filament) and about the gradients of ma§netic field intensity. We receive daily reports of radio solar observations carried out with the three radioheliographs of Nan§ay working respectively at 9800 Mhz, 508 Mhz and 169 Mhz and the radiometers working at 408, and 169 Mhz (the 9800 Mhz radiometer has not worked for several years but a new one is planned): they give us pieces of information about the corona above active c_nters. We receive daily repo.'ts from an ionospheric station working at Poitiers from a geomagnetic station working at Chambon la Foret and from the various geophysical instruments operated at Kerguelen Island. In addition, we get daily sunspot reports from Catania which has very beautiful weather during all the year. 3.2 Ursigram Reports We receive daily messages from Boulder, Moscow, Tokyo (including the Sydney report), Darmstadt, Praha, and Varsaw. Several categories of data are included in these interchange messages following a special request from our side: the URARL messages from Toyokawa which give flux and spectrum of microwave solar sources; the USSPI and USSPY messages (first studied with the collaboration of the Observatory of Roma and now used in several Russian observatories) in which are described the sunspot magnetic polarities. Of course, the best source of event reports is Boulder and their plain language SDF report is an important source of information about the evolution of active centers and flare occurrence. All these data are used in the preparation of the forecasts. We have in mind the evolution of the active centers, the place and time of flares, the coronal ho]es when they are reported, the specific consequences of the solar activity and of the so]ar events: solar particle arrival, sudden ionospheric disturbance, magnetic storm and re]ated ionospheric perturba- tion. In preparing the forecast, we have in mind models of active centers and of flares and criteria that help in the selection of categories in the relevant models. 4. SCIENTIFIC BACKGROUND 4.1 Active Centers The simplest active centers are the bipolar ones. The leading part of 3 | I " l 1980010188-014 the plage plays the key role, with the leading spots appearing first and remaining longest, and the following spots appearing later and having a shorter life than the leading ones. Its inversion line is roughly parallel to a solar meridian and their two sets of spots of opposite polarities have a slow, diverging motion until they have reached their maximum size. According to space observation, they are surrounded by EUV and X-ray loops but their contribution, if any, to the microwave emission is very weak. Dur- : ing their evolution, a filament is forming above the inversion line and will surround the remaining calcium plage. Very few, if any, subflares are reported in these simple bipolar active centers and they occur at the time of the birtil. As a matter of fact, the X-ray images show that most of the active cen- ters and of even weak plages separated by distance up to 30 degrees can be connected by magnetic loops. This magnetic connection can play a role in the evolution of active centers close together (at less than 15 degrees) and bring some anomaly in the ideal scheme described above as a simple bipolar active center: most of the flares occur in "anomalous" active centers. When two active centers appear close to each other (distance less than 15 degrees) with a lag of several days, and if the younger one is at the west side of the older one, the new sun spot group will evolve in a Bf magnetic type and the two closest sets of spots of both active centers will have a slow approaching motion. If, in addition, the two active centers are not at the same latitude, differential rotation can add its proper motion to the , motion described ab,)ve. Sometimes, the two last sets of spots to remain are the leading spot of the eastern center and the following one of the western center: they build a reversed polarity group. Sometimes, the last spot to remain is the following spot of the western group which is a reversed polarity spot for the cycle. The same kind of influence can be noted in superimposed groups hav g a more or less large time lag: the detail of their evolution is not fully understood, but this relationship can help to forecast evolution inside a complex group. Another anomaly is the "parasitic polarity", which could be a "satellite spot": a small area of one polarity inside the unipolar part of a plage. Finally, among the "anomalous" spot groups, we must note the so- called "B" configuration. In the "A" configuration, we have in the same penumbra two sets of spots of opposite polarity with nearly equal areas; in the "B" configuration, we have just one large round spot _urrounded on one side by an "eyebrow filament" and, sometimes, by a _haplet of tiny spots of opposite polarity situated very near the large spot. An important item in the forecast is the understanding of the course of : evolution of these active centers (A.C.s): are they growing or decaying? During the growing phase, the sunspots, even the large ones, are not seen through the chromosphere. Roughly speaking, during the decay an A. C. becomes simpler and the distance between the opposite polarity spots does not increase. The process of simplification concerns the pattern of the neutral line and could be accompanied by flaring activity. The microwave brightness of complex actice centers (near 3 cm) is linked to the intensity and to the gradient of the magnetic field; their coronal 4 1980010188-015 contribution in terms of I0 cm flux and X-ray emission could be very high. Another important item of the forecasL is the latitudinal extent of the spot group and its mean latitude. Below I0 °, the A. C.s evolve according to their proper motion; above I0 °, differential rotation plays a role and we must take it into account in the forecast of the spot motion. Of course, the same effect can be noted in spot groups at different latitudes but at nearly the same longitude. In the study of anomalous A. C.s for the forecast, all the data must be used in a "cross-checking" process. A strong gradient of the magnetic field is not always easy to identify according to just a crude real time magneto- gram: the microwave and X-ray emission can confirm this identification. In the first step of the forecast, all of these data are not used as criteria, but as pieces of information contributing to success in a jig-saw puzzle. 4.2 Flares Flares occur on both sides of an inversion line of the longitudinal component of the magnetic field and the chromospheric brightenings are the foot points of coronal loops or arches associated with flares. However, we do not observe flares along all inversion lines and they must be linked to a specific evolution of the magnetic field. As a matter of fact, we could have several categories of flares and of magnetic field evolution driving flares. Let us start with one model which could be mostly relevant to small flares. Series of magnetograms with a high spatial resolution show the temporal evolution of the fine structure of the magnetic field. In a uni- polar field, most of the time we can identify definite peaks of field intensity separated by valleys; the spots, if any, are at the field peaks. The intensity of these peaks is evolving continuously, and after one or a few days, we can notice the disappearance of a few peaks and the growth of new ones inside previous valleys. This evolution seems to be linked to the evolution of the velocity field of the photosphere as it can be presently observed, i.e. with just one component of the field. If two of these "evolving features" of opposite polarities are on both sides of an inversion line and evolve in opposite ways (one growing while the other one is decreas- ing), a flare is currently triggered during this evolution. This model could explain that satellite spots, kinks along inversion lines, complex centers with e ,edded polarities are flare producers: all these spatial distributions of the fields increase the number of "evolving features" of opposite polarities on both sides of an inversion line. We do not know if this model is relevant to all the large flares: some- times it does apply and we can notice either the inverse evolution and flare occurrence or a parallel evolution of the spot area and no flare is reported. It justifies our interest as forecasters in the way in which spots evolve according to the relationship described above. However, it seems true that several other categories of field evolution can contribute to the "destabilization" of the coronal magnetic field in an active center: emerging of magnetic flux, birth of a new active center near 5 1980010188-016 the first one, proper motion of the sunspots wlth a field shearing. We use the evolution of the active centers as described above, in order to forecast the "flare build up" which can be linked to the evolution of several active centers, and we check its actuality by using several categories of dat_: Ha brightness, microwave spectrum, X-ray emission, but we cannot announc_ either the hour or the day of the flare: we can have two large fla es ':le same day or wait several days for the second one. In addition, accord g to the observed or "inferred" gradient of the magnetic field, we can forecast a proton event: of course, the proton flux will not always reach t_e Earth environment, but both the hard X-ray burst and the strong microwave burst will give the possibility of identifying the acceleration process. Of :ourse, we cannot pretend to have only two categories of flares, and manT flares may not correspond to either of the above descriptions. Howev_r_ these two models help to prepare the forecast and to understand the Jfficulty we have in getting a high score in the large flare forecast: theq delay between the first forecast and the flare occurrence could reach several days and sometimes we are afraid to "stay" on our forecast; we remove it and the flare occurs at a time when we do not continue to expect it. Summarizing this forecasting technique, the first step is a forecast of the evclution of all A. C.s taking into account their relationship to the past and actual A. C.s and the "solar cycle" and "latitude" effects. In this work, all the categories of data are used to confirm the evaluation in terms of _ coherent A. C. model. The second step concerns the flare forecast itself in which several categories of data supply us with criteria to select the c_tegories of expected flares. 4.3 Geomagnetic Activity t4e_urrently identify two categories of geomagnetic activity. a. Recurrent activity. Typical recurrent storms have their maximum ir_tensity near the equinox and their minimum at solstice. Most of the time, we have tv_ separate streams of recurrent activity in opposite phase: one around Ma,'chand the other one around September, and they can be followed from one year to tho n_xt one. Several among these storms have a sudden \ co_nencement. They .',e apparently linked to the presence of large coronal h_ _s at low lat:_ude and are well observed during the last three years of a sunspot cycle, however, typical recurrence of weak amplitude can be noted at any time. Our knowledge of the coronal holes is just at a beginning, but it seems that their contribution is questionable during the few years pre- ceding _t_nspot maximum (their polarity is opposite to the one of the relevant polar hole) and during the two or three years following sunspot maximum: duri thi_ interval there are no polar holes at all or they have a very sn._l'size. b. Sporadic storms. Most of them have a sudden commencement and follow a large fl:r¢ However, several of the storms which are not in a recurrent stream hava no parent flare: sometimes they can be associated with a 6 t L 1980010188-017 disparition brusque. Looking at the flares, we would notice that most of the large flares or those accompanied by a large microwave burst are followed by a sudden com- mencement occurring in a delay related to both their order in the flaring activity (the first flare shock wave travels more slowly than the following flare shocks) and the integrated flux of the microwave burst (the larger the integrated flux, the fas*er the shock wave). Most of them are accompanied by a metric type II burst and the directivity effect is weak: flares from 600 East to 60 o West can be followed by an SSC. However, the intensity of the geomagnetic storm is not easy to forecast: sometimes the sudden commencement is followed by no definitive increase of the geomagnetic activity, even for large flares near the center of the disk. 5. THE FORECASTING TECHNIQUE ! i 5.1 Data Used for Checking the Model a. Calcium plage ; a.l. A bright and compact plage of small area means a birth or a large temporal evolution. 3 ? a.2. Between two bright plages, the brighter one corresponds to the : _ more complex group or to the more evolving group a.3. A large scattered plage shows the remnant magnetic field of " 1 • ! an old group. ' , a.4. Many filaments are seen on calcium spectrohe!iograms as filaments or as corridors. b. H_ image i b.l. Flares occur in bright H_ features; however, all the H_i features will not have flares. -i b.2. In a simple group, the bright features begin between the two polarities, and later on, on both sides of the filament, if present. b.3. Any bright persisting point which is at some distance from : the "normal" bright features described above shows a "parasitic" polarity. . b.4. A very bright feature indicates a point of fast evolution in- side the group: increase of one polarity, new birth, young group, parasitic and "huge" polarity. , ! ; b.5. The filaments show the inversion line, but they exist only 7 1980010188-018 following the evolution of the group (never at the beginning). b,6. A spot can be seen in H_ mostly when the group has passed its maximum sunspot area. b.7. An isolated spot surrounded by a filament is a B configuration i (mostly in the decreasing phasc of the solar cycle). c. Radio wave emission The X-ray and the microwave emission supply us with nearly the same kind of information. c.I. Strong microwave emission (three centimeters) can correspond either to a large spot(strong field) or to a complex configura- tion (magnetic field gradlent). c.2. High flux at three centimeters (> 25 f.u.) and a strong ratio (> 1) between the ] cm and the 8,5 cm flux of a group indicate a strong gradient of the magnetic field. c.3. Metric wave emission fo]iows flaring activity: strong temporary emission for a large flare, weak or middle-quasi permanent emission for many subfiares. However, a group could be a subflare producer without any metric storm (coronal structure and directivity effect). d. Radio bursts d.I. A large microwave burst (cm.) corresponds to a strong hard X-ray burst and to strong emission of soft X-rays (SID, etc...). d.2. A large microwave burst of long duration is accompanied by a large type IV burst (which may not be reported due to a directivity effect). This microwave burst can be used to forecast the time of arrival of the relevant geomagnetic storm. d.3. The largest microwave bursts correspond to proton accelera- tions; however, amongst the microwave bursts, many events are not followed by proton arriva] at the earth. (propagation , effect) d.4. Any radio burst indicates a particle acceleration, mostly electrons (type III), or a shock wave (type II). e. e.l. The hard X-rays show strong gradients of magnetic fie]d. e.2. The X-ray bursts show particle acceleration and flares occur- ring in strong gradients (sotnetimessubflares). 8 1980010188-019 5.2 Current Forecast QUIET Simple bipolar group born far (> I0 °) from another group or from a scattered plage. The inversion line is nearly parallel to a meridian. No radio emission. Fore_ast: A few subflares can be expected during the first days. ERUPTIVE Someweak anomaly in comparison with a quiet group: bright H_ points, born inside an old plage, etc .... They could have weak microwave or metric wave emission. Forecast: Many subflares, sometimes with weak microwave bursts. Could evolve after several days into an active or proton center. ACTIVE Large sunspots in white light. B or A configuration (strong magnetic gradient d/D << I). _right Ha and Calcium features between spots. Definite microwave emission (= 25 f.u.). Ratio 3 cm to 8,5 cm weak (_ 1.00). Some hard X-ray flux. Forecast: Flare with large microwave and X-ray emission. However, the evolution can be slow and the flare could be reported only on the second or third day after the forecast. PROTON An A configuration with all the features of an Active center but: The inversion line is more strongly drawn. The gradient is stronger but the spot areas are not necessarily larger. The microwave emission is stronger (> 25 fu). The Ratio 3 to 8,5 cm flux is high (> I.O0). i The Center is between 300 East and the West limb. Forecast: A Proton Event in the next few days. MAGNETICSTORM I. Followin 9 a flare with large microwave burst (> 200 f.u.) and some duration (> 15 min.) occurring between 60o East and 60o West. Forecast: the delay evaluated according to the microwave importance. 2. Recurrence: Two streams by solar rotation, one having its maximumat the first equinox, the other one at the second equinox. Forecast: taking into account the seasonal effect. iEN CM FLUX We try to evaluate the contribution of each center and to usP the solar activity Forecast in order to predict the evolution of the fldx for _he next three days. 6. FORECASTAND USERS We :ompared our series of forecasts, issued from July 1968 to December 1969, to the actual events reported later on. Our four classes of daily spot group activity _re those currently used in the IUWDS: the solar activity is not distributed equally among these four classes since, during this inter- val, we had 17l Quiet groups, 32 Eruptive groups and 3 Active grouh_ per one Proton group. Consequently, the Quiet Activity is easier to forecast than the Eruptive one, the Eruptive Activity than the Active one, and the 9 1980010188-020 : "Active" Activity than the Proton Activity: the study confirmed this expecta- tion with respectively 90 per cent of success in the Quiet Forecast, 50 per cent in the Eruptive one, 18 per cent in the Active one and 30 per cent in the Proton forecast. Starting from the actual activity reported, 92 per cent of the Quiet groups were correctly forecast, and respectively 37, 30 and 64 4 per cent of the Eruptive, Active and Proton groups. These figures can be linked to the duration of each class of activity during the disk crossing of the sunspot groups: a group is currently eruptive during intervals of two • : or three consecutive days. This interval is long in comparison to the dura- , tion of Active and Proton gro_,ps, but very short in comparison to the Quiet : groups which can cross the entire disk with only a few subflares reported. In conclusion, we would like to emphasize that even these poor performances are very useful for users: if they follow the forecast, they increase definitely their chance of observing a specific category of event, especially the unusual events like the large X-ray flares or the Proton flares. We did not have a more recent opportunity of evaluating the current accuracy of our forecasts. We can classify our customers in three categories: ; l. Those using a forecast in order to begin or to continue a program of operation: Coordinated International Scientific Programs such as the Proton Flare Project, the Incoherent Scatter WG, the SPARMOBalloon Flights, Solar Satellites and technical programs such as oit or water researches. In order to succeed in these enterprises, we need to have a close and fast con- _ tact with our customers: we can add to our forecast special advice, taking into account both the actual trend of solar activity and the constraints under which our customer is put. 2. Those using a forecast as a warning: scientists firing rockets into : an aurora, engineers controlling an electric power distribution, operators of radio transmitters, etc .... The current problem is to issue a forecast easy to understand. 3. Those using our forecast in order to issue their own forecast: forecasters of ionospheric propagation, of orbits of low apogee satellites, etc .... The main problem is to understand the accuracy they expect from our forecast. Acknowledgements The Forecasting Center of Meudon could not operate without the technical assistance of the Centre National d'Etudes et Telecommunications, the scientific expertise of our colleagues of the Solar Department, the technical assistance of observers operating at Meudon and Nancay, the data supply of our IUWDS colleagues In charge of Regional Warning Centers, the unexpected requests of our customers and the diffuse scientific support of the International Solar Community. I0 == 1980010188-021 REFERENCES Avignon, Y., M. J. Martres, and M• Pick (1964): Identification de classes d'eruptions chromospheriques associees aux emissions de rayons cosmiques eta I'activite radioelectrique. Ann. Astro•, 27:23• Caroubalos, C. (1964): Contribution a I'etude de l'activite solaire e0 rela- tion avec ses effets geophysiques. Ann. Astro., 27:333. Kundu, M. R. (1959): Structures et proprietes des sources d'activite solaire sur ondes centimetriques. Ann. Astro., 22:1. Martres, M. J. (1968): Origine des regions solalres actives "anormales", n° 35 IAU Symposium, 25. Martres, M. J•, R. Michard, and I. Soru-lscovici (1966): Etude morphologique de la structure magnetique des regions actives en relation avec les pheno-menes chromospheriques et les erupti solaires. Ann. Astro., 29:245. . Martres, M. J•, R. Michard, I. Soru-lscovici, and T. Tsap (1968): Structure L _. evolutive magnetique des regions actives solaires. Solar Phys., 5:187. Martres, M. J., I• Soru-Escaut, and J. Rayrole (1971): An attempt to associate observed photospheric motions with the magnetic field structure and flare occurrence in an active region. IAU Symp. n° 43, 435./ Martres, M. J., I. Soru-Escaut, and J. Rayrole (1973): Relationship between _ome photospheric motions and the evolution of active centers. Solar Phys•, 32:365. Martres, M. J., J. Rayrole, E. Ribes, M. Semel, and I. Soru-Escaut (1974): On the importance of photospheric velocities in relation to flares. Flare- Related Magnetic Field Dynamics, HAO/NCAR, Boulder, Colo., p. 333. Martres, M. J., I. Soru-Escaut, and Y. Nakagawa (1977): H_ off-band pre-flare activities. Astron. Astrophys., 59:255. i Pick, M. (1961): Evolution des emissions radioelectriques solaires de type IV et leur relation avec d'autres phenomenes solaires et geophysiques. : Ann Astro , 24:183. t Pick, M., and P. Simon (1973): Solar activity and its forecast. Proc. IEEE, ,. 61:1303. Ribes, E. (1969): Study of the correlation between flares and the evolution i of magnetic field• Astron. Astrophys., 2:316. i Simon, P. (1979): Polar coronal holes and solar cycles, Solar Phys., in press. ; Svestka, Z. (1976): Solar Flares. Reidel Publ. Co., Dordrecht, Holland• Tanaka, H., and T. Kakinuma (1964): The relation between the spectrum of slowly component of solar radio emission and solar proton event. Rep. Ion. Space Res. Japan, 18:32. I1 1980010188-022 ; N8,0-18464 THE FRENCH SHORT TERH RADIOPROPAGATION PREDICTIONS IN THE DECAHETER BAND P. Lassudrle-Duchesne, A. H. Bourdlla, H. Sizun C. N. E. T. D_partement "Mesures Ionosph6riques et Radlo61ectrlques" Route de Trggastel : 22300 Lannion (France) Ionospheric short term predictions are issued regularly by the . "Centre National d'Etudes des T_16communications" located in Lannlon. They are concerned with both weekly and daily predictions of the ionospheric characteristics relevant to radlopropagation. A description of the techniques involved is given as well as examples showing how the prediction nmssages are prepared. I. INTRODUCTION The "Centre National d'Etudes des T_l_communications" (C.N.E.T.) in Lannion prepares and issues short term radiopropagation predictions for the decameter band on a regular basis. Th,= prediction messages are sent by telex to the users. They are mainly the French Army and Post Office as wel_ as a small number of scientific institutes in Europe. The short term predictions are confined to restricted geographical areas or "zones", namely the "European zone" and the "North European zone". Figure i shows the geographical extent of both zones. The predictions are : relevant to radio circuits whose terminators are both located inside the same zone, As a matter of fact, they can be used with a reasonable approximation for circuits less than 3000 km of length whose reflection point lies within a given zone. With the exceptions of special notices and alerts, the short term pre- dictions are of two kinds--weekly and daily predictions. I.I Weekly Predictions They are issued every week and contain a prediction for the next seven days of the radiopropagation conditions. An example of weekly prediction message Is given in Figure 2. The first part of the message is a review of the main solar and ionospheric events that occurred during the past week and which are of interest to radlopropagatlon (S.I.D., P.C.A., auroral absorption, .... etc.). 12 1980010188-023 r_ lS'W 0 30"E 45" "-'-" _..._... EUROPEAN ZONE NORTH EUROPEAN ZONE Figure I. European and North European zones. The second part of the message of Figure 2 is a prediction of the median radlopropagation conditions (MUF) for the next seven day period. It is con- cerned with the European and North European zones, as well as with radio circuits starting from France. In this part, reference is made to a "perma- nent series of bulletins" which the users have at hand and which are normally used for long term predictions. A more detailed description of this permanent series of bulletins will be given in the following section. For the moment we shall simply mention that these bulletins contain monthly median predictions ut: the HUF and LUF as a function of local time for circuits lying within the zone of interest (Figure 3). The short term prediction message of Figure 2 indicates which of these bulletins is to be selected and the " percent corrections to be brought to the curves of this bulletin to match the median HUF expected for the next week. Reference is also made in the message to radio circuits starting from France. The message ends with a brief description of the expected development of the radioprupagation conditions for the future week. 1.2 Daily Predictions • They are issued every day at 15 LT (except weeks end) and contain a more detailed evaluation of the HUF and LUF for the next 24 hours. Figure 4 shows an example of a daily prediction message. The predictions are presented for the European and North European zones separately. The next 24 hours are 13 1980010188-024 ¢{ i t i I Constatatlon d'anomalies falte au cours d la derni_re p6,;ode heb- i domadaire : ! - Propagation m_diocre le 9 Septembre 1977 entre 16 H 30 et 18 H 30 • et le IO en d_but de matinee. - Faiblesse de l'ionisation durant toute la journ_e du I1. PH3..__8- Pr_visions de propagation ior-sph_rique valable du vendredi 16 12 H OO TU au vendredi 23 Septembre 1977. Stop. Primo : Zone Europe - Utiliser les bulletins de la s_rie permanente Septembre IR5 _gale 40. Dlminuer de 5 Z les valeurs de !a HUF 90 Z. Stop. Secondo : Zone Eurore du Nord - Utiliser les bulletins de la s6rie perma- nente Septembre IR5 6gale 40. Augmenter ie IO Z les valeurs de , la MUF 90 %. Stop. Tertio : Point & point _ partie de la H6tropole. Utiliser es bu11_ _ns de Septembre. Augmenter de 5 Z les valeurs de la MUF 90 Z. Stop. Quarto : Instabilit_ de l'ionisation en d6but et en fin de p_riode hebdo- madaire. Disparition des nombre_ses r6flexions par la couche E sporadique. Figure 2. Example of weekly prediction message. divided into 6-hour intervals beginning at 18 LT and ending at 18 LT on the next day. During each of the 6-hour intervals (say 18 to O0 LT in the European zone, for instance), a descriptive cypher is used to evaluate the expected MUF level ("2" placed between 18 and O0 in our instance). The des- criptive cyphers refer again to the permanent series bulletin mentioned above. In each 6 hour time interval the descriptive cypher can take the following values (Figure 3): (a) I if the MUF is expected to be above the MUF 30% (monthly predicted MUF with a 30% probability of occurrence, (b) 2 if the MUF is expected to be comprised between the MUF 30% and the MUF 90%, and (c) 3 if the MUF is expected to be under the MUF 90%. Roughly speaking, the descriptive cypher is I if the expected MUF is higher than normal (i.e. monthly predictions), 2 if it is normal, 3 if it is lower tha,, normal. In Figure 4 the terms "short distances" ("petites distances") and "median distanc¢_'("moyennes distances") refer to the MUF(O)F2 and the MUF(3000)F2, respectively. A similar presentation is used for the LUF during the day hours (6-12 and 12-18 intervals). J 2. AVAILABLE DATA 2.1 Vertical Sounding Data The vertical ionosonde of LANNION (48.75°N, 3.45nW) is located close to 14 I 1980010188-025 I MHz EUROPE e:40 SEPTEMeRe 250 K. * 8; i ? ___3O% F 4 MUF 3 ._ 90'9O% _ 2 _ 0 3 6 9 12 15 18 21 24 TL Figure 3a. Example of Permanent Series Bulletin. ' MHz _. 30 {. ' EUROPE R:40 SEPTEMBRE 3000 KM 2S y 30% 20 f I 2 , _o MUF 0 3 6 9 12 15 18 21 24 TL . Figure 3b. Example of Permanent Series Bulletin. ; 15 _ f ] 9800 ] 0] 88-026 ....... "..........- ...........' .............lib .................... i-:*'<................ II 1 PJ263 Prlvislons de propagation ionosph_rique Journali_res Europe moyenne lndice 40 Septembre HUF Petites Distances 18 2 O0 3 06 2 12 2 18 HUF Royennes Distances 18 2 O0 3 06 3 12 3 18 LUF 06 ! 12 I 18 Europe du Nord indice 40 Septembre HUF Petites Distances 18 3 O0 3 06 2 12 2 18 HUF Hoyennes Distances 18 3 O0 3 06 3 12 3 18 LUF 06 i 12 I 18 Figure 4. Example of Daily Prediction Message. 4 the prediction center and provides the prediction group with ic,iospheric measurements. The ionosonde of POITIERS (46.58°N, O.33°E) (about 320 km away from LANNION) also sends ionospheric data by way of telex or telephone. F,'om Northern Europe, the data of the ionosondes of UPPSALA (60°N, 18°E) or LYCKSELE (65°N, 19°E) are sent daily to LANNION by telex. 2.2 Hagnetometer A magnetometer is located in LANNION and prov;des the predictors with real time geomagnetic data. 2.3 URSIGRAHS URSIGRAMS relevant to data measured in observatories from all over the world are received by telex. The data used for short term predictions are concerned with: (a) sunspots, solar flares, solar radio flux, description of solar active centers, (b) geomagnetic indices of activity, and (c) ionospheric critical frequencies, indices of propagation, absorption measurements. Figure 5 gives a synoptic view of the data collected by the prediction center in LANNION. 2.4 Permanent Series Bulletins The "Permanent series _j bulletins mentioned in the previous section are available to the users for long term (monthly) predictions. They also serve as a reference to display short term prediction final results, the latter being presented as a correction to long term predictions. It should be noticed that this is the only interference between short term and long term predictions, the two methods being otherwise independent. The "Permanent series _' bulletins or _Permanent Zo_al Predictions" are consistent with the definition of the geographical zones. These are chosen _ 16 t <_- 1980010188-027 ! ,o.o.ph..,ci,o.o.ph..,cl !'.,"'t o"'.. 1I .o..di.g I .°""da"e ILI'.""l°n I L po,J... _,L.C.._n_o.,___.j• / Uppsale ; _ TIME I PREDICTION " I CENTER :: [ Lonnion • I TELEX l prediction messages , alerts .... : for EUROPE and NORTHERN EUROPE _ { URSIGRAMS ] ,: _;!, J p__bserv [ _ f _""_dictions sun leomagnet ism Ionosphere SUn geomagnetism e . sunspots . Indices of / . cr',tlcai . radio flux . Indices of ' activity | frequencies . proton omissions activity _ radio flux / _ X ray-emissions . . __o'oPsgszlo_'" _ M regions . alerts .; indices . sunspots ' Figure 5. Data collection for the Prediction Center in LANNION. : ,_ ;* in such a way that the monthly median foF2 can be regarded as roughly constant ,, (i.e. independent of the geographical conrdinates for a given local time) inside a definite zone. In other words, both latitudinal and longitudinal i effects are assumed to be negligible within the limits of a zone (the earthOs : i rotation being accounted for by the use or local time). This has made it " i possible to construct regression curves between the solar index IR_ and the : : foF2 for given month and local time whe-e IR5 is defined by the fo lowingi relation: 1 + ) (1)i IR5 =_ (Rm_3 + Rm_2 + Rm_I + Rm Rm+l t with R - Wolf number for month m. Any one of these curves is considered to : _ apply ?or the whole zone. The results of this analysis is presented to the users in the form of a ; series of bulletins, each one being relevant to a particular solar index. i 17 .2 1980010188-028 The whole series covers the range of solar indices from IRs - i0 to IR5 - 160 and is applicable to a definite zone. For any given solar-index, the corresponding bulletin contains diagrams similar to those of Figure 3 dis- playing the HUF and LUF as a Function of local time. The parameters for each diagram are the month of the year, the distance between the end points of the radio circuit and the attenuation which depends on the particular equipment to which the prediction is applied. The advantage of such a method for long term predictions is that the user can be given the series of bulletins corresponding to a certain zone once for all. The only information which remains to be regularly sent to him is then the expected solar index for the prediction period (i.e. the bulletin , which has been chosen). On the other hand, the drawbacks are that the ! prediction is available within a restricted area, and that there is a certain loss in accuracy as compared with other methods which calculate the predicted foF2 at the exact point of reflection of the waves. 3. WEEKLY PREDICTIONS 3.1 Foundation of the Method The method for weekly predictions summarized below has been in use since 1971 and was detailed by Juy (1972). It is based on a linear regression analysis over past data between the 7 day median foF2 at local time h and for week S : T h (S) and the solar index IR5 (S) corresponding to this week (i.e. corresponding to the month which includes week S). Here, the week number S goes from I to 52 in the course of a /ear. Juy (1972) has shown, by using the data o _ Freiburg from 1948 to 1968, that there exists a set of linear regression functions L_ such that: S S S (2) _h(S) = Lh (IR 5 (S)) = ah IRs(S) + bh In this analysis the most magnetically disturbed weeks were eliminated so that the linear functions L_ apply to an approximately quiet ionosphere. Moreover, the L_ functions are assumed to be characteristic of the ionosphere not only over Frelburg, but also over the whole European zone. 3.2 Prediction Technique . The prediction techn!que makes use of the linear regression functions L_ as follows: Let S be the number of the last week for which data are avail- a_le and (S + I) the week to be predicted. Since fh (S) is known for every hour h, eq. (I) can be used to determine a value R_ of the solar index for week S: H Rh (T h(s)) = --S ah : 18 l 1980010188-029 _ The value R_ has no physical meaning in that sense that it merely comes from a mathemati_al relation. It is not in general equal to the actual solar ._i index IR5(S) , which is to be measured for the month including week S. S does not change from week S toThe main assumption is that the value Rhweek S + i: _S+i S Kh = Rh (4) Now, eq. (2) is again used at step S + i to determine f--h(S+l) from S .S+I The sam_ proce=s can be repeated for every hour h ofIR5(s+l) = R = Kh the day. In practice, Juy (1972) states that the overall level of ionization during the day can be summarized by the values of only three part{c,.,lar hours: h = i0, h = 12 and h = 14 LT. Hence, for simplicity, the level of ionization fo_ the week S + i is characterized by the index p(S+l) d_fined by: p(s+i) = (Ti0(s+l)+ Tiz(S+l) + Ti4(S+l)) Let pm be the value of the p index computed from the Permanent Series , bulletin values. The percent deviation of p(S+l) from pm is: r(S+l) = I00 p(S+l)-pm (6) : pm A further assumption is that r(S+l) applies also for the percent deviation y of the MUF during the same period. 2" : The correlation between the geomagnetic activity and the ionosphere •L critical frequency is accounted for according to the following criteria: Let AAp = Ap - Ap(S+I) where A--p(S+I)is the predicted average planetary geo- • Ii magnetic index for week S+I and Ap the observed index for week S. Then:If &Ap <--30, then r(S+l) is reduced by a factor 6%. _ i If -30 < AAp =< -20, " " " " " " " 3%. ,, If I Apl< 20, then r(S+l) is unaltered. _ If 20 =< _gp < 30, then r(S+l) is increased by a factor 3%. If AAp -> 30, " " " " " " " 6%.i A similar technique is used for the North European zone. i 3.3 Example of Prediction ; : _, The technique will be more clearly explained with tile following example " , ,_ which describes how the weekly predictions were made for the week September :Ji 16th to September 23rd, 1977 (S+l = 38). Let us take h = I0 LT as an examplef • _, The median foF2 value for the week S = 37 at h = lO is ?-10(37)= 5.7 MHz. " ,; 19 : 5 1980010188-030 ++ Figure 6b, we get that T!0(38)_ - 5,6 HHz_v The same process ca_ be r_peated ,_ for h = 12 and h = 14 LT which gives _19(38) = 5.6 HHz, and _1_(38) = 5.4 MHz. ! / Hence: p(38) - __ ._(I/3)(T10(38) + TiZ(38)';_ = ---f--'4(38)) = 5.5 I_tz. "_et us now • compute the value of pm. Figure 3 yields the following values for the HUF 50_: (h - IO, HUF 50_ = 6.1 HHz), (h - 12, HUF 50_ = 6.4 HHz), and (h - I_,+ NUF 50_ = 6.4 MHz). Hence the value of pm is given by pm = I/3(6.1 + 6.4 + + 6.4) - i/2 fH = 5.7 _z, where fH = gyrofrequency = 1.2 HHz. This gives r(38) = 100(p(38) - pm)/pm = "3.5_. The predicted Ap was A--p(38) = 18 with Ap(37) = 14, hence AAp = -4 and r(38) remains unchanged, i.e. r(38) = -3.5_ rounded to -5_. Figure 2 shows '_ the corresponding weekly prediction message, i 3.4 Discussion of the Method and Future Prospects 7 Figure 7 shows a histogram of the prediction errors for the period + 1973 - 1977 for h = iO, 12, 14. The total number of values is N = 713. The 90_ confidence interval is approximately (-i HHz, I HHz). The main advantage of the _ethod lies in its simplicity. The regression curves, having been established once, for all the weekly predictio,,=, can be done by hand within a few tens of minutes. For criticism it seems that the main causes of errors are the following: a) The method uses mainly a fixed sample of data (through regression curve_) relative to past solar cycles and which ate not renewed automatically as time elapses. The only renewed datum is fh(S), i.e. the last available value of the signal. This only value is, of course, insufficient to take account of the statistical peculiarities of the signal at the time of predic- tion. foF2 EuropeonZone foF2 EuropeonZone 15 MHz h=lO LT !_ MHz h=10 LT + S=37 S=38 -_ 10 J 10 f l I i I IR5 0 IR5 00 _ J t I - 0[u I I _ I :_3_ 50 100 150 50 I00 150 ; RIO R_ (O) (b) '_ Figure 6. Examples of regression lines for weekly predictions. 20 = "Z -':+ 1980010188-031 N60I _0 7 m "_ . _foF2 i_ L _ 20 10 0 I0 20 Figure 7. Histogram of the prediction errors for the weekly predictions. S • b) Although the regression functions Lh, regarded as a serie in S, seem _: to be able to reproduce adequately the mean annual variation of the signal, it should be noticed that the regression analysis has an averaging effect, which tends to rub out any small scale variation. Hence, errors would rise even if the signal to be predicted had still exactly the same statistical _ characteristics as that used to produce the regression curves. z_ ; These considerations have brought us to the conclusion that ionospheric predictions would be better by means of adaptative methods. So, in the _. system modelization and identification theory context, ionospheric representa- tion models, particularly fit to prediction settlement, have been recently _ established. These models have been elaborated by Y. H. Le Roux and will be reported in a contributed paper (2). In fact, the main purpose of this paper will be the weekly predictive case of the foF2 parameter, although itr • briefly deals with the daily prediction problems. It should be noted that such adaptative predictor models may be established for only one particular parameter (ionospheric, solar or geophysical) or several ones together. A test of this new weekly method, over past data, has shown an apprecia- ble improvement in the resulting errors with respect to the present method. : The predictions algorithms are, moreover, recursive, a fact which can be taken advantage of, whenever desired, to reduce the amount of calculation involved in the prediction to a level sufficiently small to be done manually. _ 21 • i 1980010188-032 ,/ @ I ! 4, DAILY PREDICTIONS t !, 4.1 Prediction Technique i i No automatic technique is available. The predictions are heavily depend- ent on the operator's skill and Intuition of the future development of j ionospheric events. The following example will give an idea of the way the ] daily messages are prepared, i 4.2 Example of Prediction i Let us take as an example the prediction for September 21st, 1977. The message shown in Figure 4 was prepared on September 20th at 14 L.T. (12 UT). A magnetic sudden storm commencement (S.S.C.) had been reported on Septem- ber 19th at ilH45 UT. Figure 8 shows the value of the HUF(0)F2 measured in ! LANNION plotted on the Permanent series Bulletin curves corresponding to this • case. Figure 9 shows the magnetogram of LANNION for this period. A positive ! phase is observed on September 19th, a few hours after the S.S.C., followed _ by a decrease in the morning of September 20th. ! The predictor has suspected the HUF values to remain at a rather low level during the night and the following day (21st). More precisely, the ; HUF was expected to remain below the HUF 30%. Hence the descriptive cyphers equal to 2 or 3 in the correspondir.g message (Figure 4). The operator's intuition has been partially confirmed as shown on Figure 9, where the dotted line represents the values observed in the afternoon of September 20th and the following day. t EUROPE _=40 SEPTEMBRE 2SO KM MHz 9 ,0 3 • 9 12 tB 18 21 24 TL Figure 8. MUF(O)F2 values in LANNION for September 19, 20 and 21, 1977. 22 _ eL" 1980010188-ogg J20 )" 19 SEPTEMBER 20 SEPTEMBER Q • • 0 . . 1,_ . . l_e . . 9 Figure 9, Magnetogram of LANNION (E-W horizontal component) for September 19 and 20, 1977. F; 4.3 Discussion of the Technique Telling that the daily ionospheric predictions are an Exact Science would be a questionable assertion. A number of factors are liable to influ- _ ence the operator's judgment whose individual weight in the final decision ! can hardly be appreciated. As a matter of fact, two different operators may i_ well produce a somewhat different prediction in the same conditions. Never- i! theless, the overall results of the daily predictions are fair (see Figure I0). _ The future development in this field tends to make the prediction in a more automatic manner. They are concerned with the modeling of ionospheric disturbances on the one hand and better and more rapid data collection on the , other. _, REFERENCES ,:, Juy, M. (1972): La prSvision ionosph6rique hebdomadaire. (Etude sur une base ! de sept jours de la caract_ristique ionosph_rique foF2 mesurSe en Europe au cours des deux derniers cycles d'activit_ solaire). Th_se de , Doctorat de 30 cycle, Universit_ de PARIS Vl. Le Roux, Y. M. (1979): Forecastin_ methods by means of representation predictive models: Ap.plication to the short term ionospheric case. (P r iva te Communi ca t ion). 23 • I !, 1980010188-034 I1' LANNION 1976 300 N-- 3157 200 100 _foF2 3 2 1 O 1 _ 3 Figure |0. Histogram of the prediction errors for the daily predictions. 24 I 1980010188-035 p., /lIIIm_ __ b ........ _ - -- _-, ]t8,0 -18,465 PROPAGATION PREDICTIONS FOR THE HF-RANGE BY THE RESEARCH INSTITUTE OF THE DEUTSCHE BUNDESPOST Th. Damboldt Forschungslnstitut der Deutschen Bundespost P.O. Box 5000 D-6100 Darmstadt This paper desc_ibes the long-term and short-term HF propagation prediction service of the Deutsche Bundespost. The long-term pre- diction method Is computer-based and was devised on a philosophy independent from the C.C.I.R. - or any other method. The short- term prediction method ls still a manual method which relies mainly on a number of continuous field-strength recordings and on i the ability to Interpret the recordings and to combineforecaster's this information with all available solar-geophysical data. Ex- ! amples of the predictions are given and are compared with actually observed HF propagation conditions. i l. INTRODUCTION Even in our times, where most communications are set up on non-iono- spheric circuits, there is a continuing need for reliable rad;o propagation predictions by a number of users of the HF spectrum. The Research Institute of the Deutsche Bundespost (Post and Telecommunlcatlon Administration of the Federal Republic of Germany) provides these users wiLn both long-term and short-term predictions, e.g. for use by its own frequency planning department and coastal radio station. In the following sections a survey of the pre- diction methods and of the underlying geophyslcal data will be given. 2. LONG-TERM PRF_ICTIONS In this paper the word long-term prediction will be used only for pre- dictions based on published data of the worldwide characteristics of the ionosphere (C.C.I.R., 1974). Long-term predictions are usually computer- based. With the aid of a suitable computer program (e.g.C.C.I.R., 1_74) the cri'tical frequencies or the field strength or other propagation data for any month and path can be estimated, provided the only var;able, the sunspot number, for the relevant month was reliably predicted. 25 t 1980010188-036 2.1 The Field-Strength Prediction Method of the ,_ Deutsche Bundespost Current prediction methods (e.g.C.C.I.R., 1974) calculate effective propagation paths (geometrical optics) and compute either transmission loss or field strength taking account of the main geophysical influences on ionospheric radio-wave propagation. In contrast, the method developed here estimates the field strength empirically based on the expresssons for devia- tive and non-devlative absorption (Beckmenn, 1965). A major disadvantage of the other methods of calculating effective propagation paths is that they are incapable of yielding the field strength at frequencies above the "classical MUF" (maximumusable frequency). The classical HUF denotes the highest frequency at which a _ave propagates by ionospheric refractlon between speclfied terminals. Experience has shown, however, that reception of the transmitted signal is possible also above the classical RUF. The receiving field strength does not abruptly fall to zero when the frequency of the transmitted slgnal exceeds the classical MUF, but it decreases gradually. In most cases, therefore, the "operational MUF" is substantially higher than the classical MUF (Beckmann et al., 1965). Several mechanisms are responslble for the propagation of HF radio waves at frequencies above the classical MUF. At present, these are not taken into account by prediction techniques based on theoretical considerations of effective propagation modes. The main mechanisms are: a) ionospheric scatter propagation by irregularities in the F-layer (e.g. spread-F) and D-layer (ionoscatter), b) meteor-scatter and auroral scatter, c) trans-equatorlal propagation, d) scatter on field-aligned irregularities, e) side-scatter due to ground irregularlties, f) sporadic-E layer refraction, g) off-great circle propagation, caused e.g. by horizontal gradients in the ionization, and h) ducting above the maximum of the F-layer. Apparently several mechanisms are effective at the same time. As a rule, the effectiveness of these modes increases with _he length of the radio link and the width of the radiation patterns of the transmitting and receiving antennas. It has not yet been possible to determine these influences quanti:a- lively by adequate formulae. That is why an empirical factor is applied to the I'standard-HUF" values (which are an approximation to the "Classical-MUF" value) taken from the C.C.I.R. Atias to yield the operational MUF. Z6 1980010188-037 i: The variation of the field strength within the transmission frequency range (i.e. between LUF (lowest usable frequency) and MUF) is computed ac- _ cording to a formula developed by Beckmann, based on the expressions for j devlative and non-deviative absorption (Beckmann, 1965). Together with an empirical factor the prediction method now comprises all factors influencing HF radio-wave propagation (Beckmann and Ochs, 1969; Oamboldt, 1976). ( ! 2.2 Representation of the Predictions A variety of delineations of the long-term predictions is available to the customers. The oldes_ and best-known version is the "basic" frequency- versus-time diagram indica,ing the field strength in the third dimension. Figure I shows an example for a certain radio link which is self-explanatory with its legend. Figure 2 sMows an alternative representation v'here the field strength " for the same radio link a_ in Figure I is given in _abular form. In some cases the user of the HF part of the spectrum has available or|y a limited number of frequencies. ;n Figure 2 the field strength is given for specified frequencies as a function of time. One advantage of this repre-entation is that transmitter power and antenna gain can be chosen independently for each frequency, whereas in Figure i the field strength is normalized to a transmitter power of I kW and antenna gains of 0 dB. Figure 3 shows a printout, where receiver i put voltage (upper table) and signal-to-noise ratio (lower table) for the path Tokyo-Germany are given. The receiver input voltage is calzulated from the predicted field strength taking into account the antenna gain. Here the signal-to-noise ratio is the ratio of the predicted receiver signal and the noise calculated according to data contained in C.C.I.R. Report 322-I (C.C.I.R., 1974). The prediction of Figure 3 is shown here to point out the development from field-strength predictions to more user orientated predictions. It is an intermediate stage of the prediction for the coastal radio station of the Deutsche 3undespost at Norddeich and al; maritime radio stations of the merchant marine of the Federal Republic of Germany. This prediction, shown in Figure 4, gives for several paths that maritiJ,_ frequency band which yields the highest receiver input voltage (taking into account the radiation pattern and gain of vertical rod antennas usually u,,edon board ships). Also given is the signal-to-noise ratio in an accompanying symbol A, B, C or _'_,denoting a s'gnal-to-noise ratio of more than 20 aB, 11-20 dB. 0-I0 dB and less than 0 dB, respectively. The latest version of the predictions is aimed at users with n_bile or portable radio stations. Figure 5 shows a sample of these predictions which are accompanied by the text below up to section 2.3: Explanatory notes on the monthly HF radio propagation predictions for central Europe issued by the Research Institute of the Post and Telecom- munication Administration of the Federal Republic of Germany. Each page nf the prediction refers to a certain time of the d_y of a certain month. The time (UT=Universal Time) and the n_nth as well as the sunspot number R upon which the prediction is based are given in the upper right corner. The Tables indicate for different distances between transmitter and rece!ver that 27 I 1980010188-038 28 1980010188-039 ...... -_*()I,_ TIIE ,,IIGINAL pAGE IS POOR i;.; _ e o,.oo. ! | | B _ N N A @ v | .,,1' I", II O I I E m le, m 0 _ 0 lib N * I I I I N N -,* N I t I _.. ,=- I | l I I Cl "III L I N --'s_ I I I I N,,".iI I l -.*I_I I I _ ,.4 (11 m ,ir 4' lql m O l_ qlD J O O O _ • I I I Iq .-m I I NgD I)- O O • O I I Ir:) 113 O O N ..4 L IL • ,-* O O • m ,4" --* St _ 0' --* IM IR I_ N e | II I N I ,,.e .-e .,.e I I_ .Z .,4"* I O r.. * .4 • I I I t_ I ,,-4 _ Z '_ _ m __ O _ (_ I _ 0 • ._" ID 0" In .1_ I_ _ I_l ,,-e I"- N !$1 I,- m r- o O hl ,,.4 ,, I I I _I I .-., .=, _ lu .Q • _i- _ 4-; _ .II ILU I-- ,,=s . I I Isl =-* I l N tU _' w 2' C" C -- _ I- _m .4 I[:3 I- lu _o N l I I l 3. 4=_ -_ O *T I)- ,4" P- In ...* N C' N 4" I_ I-- ._ _ IIm _$t I_- IN I_- _" O O ,,-m * I I N ,..* I l ,.4 N I I :_ ,,-. I_ 1.5 ),- If', C .,.i C: O IN .,-4 =Jr 0 :IC _" Z v lq ,4' Ill * _ I_t 0 C" 0 _ _' (_ .4" --* ,-_ I-- k. Ih .=* rl .... I_I I:, ,,-* ,, l I_I .-4 I I l .-* I_ I I I[ __' c_ ..4 _ I- u_ I I I I - 0 IlU I_ C O O N If'. C_* 4r _D 4" I" I_ -"- 4-I : - El II@ _ _ iN 4_ I I l " 1:3 I_ I( . .... .... _I[ ,*4 _ • I N @ I I @ N I I I I _I" l I I I I C 4-I ". I)- Ir_ C O -- C) I_ In IN IL C- I I_ w_ I IL _. N_, O O N .'* e I N _ I .-4 .-m I_ I i I I "Jr" "1_ _ .-, _ * _ 4[ I I l m I l.- u. I- .* l _. _ _ _ i_ l l I I l _ I_ II_ Z (3 O O m I • I' I I r-.Z . r- C ¢, _.e _ I-- I- "-_ I_ ,.-* C: C _ "_ O' ao ID _0 N m r=" " I "" m I- * .'_ * I N -'_ -'_ N III l I I I I _-, O C_ I)=- _ ,,11 m r- I I I I I s_ "-- II. i_i _I" Im i,_ ..-4 ..4 ul 4=_ i_ I/_ ii.i i'_ .-4 ,.-,C. _ ,11"/_ i_ (0 : _ -J O T.¢ O C * If" l I I I I *-') I_ E" I-- W If" ,-i _ ,-i I1_ I I | | I l- v =[ --I C C _' m'l _ C IF' I¢I -41' _'._ ! T II I_ O O ,, - l II III N _I I_ I I II II If O LJ q • .__ _ _ (Nf _ I I I I _ *-- llJ I-- _ _. | l I r. e-_ a: u e,j _ ,v ,-. I.-* ILl _ _ _ _ dD .-* _ID 0_ *=, ._ I_ ,dl .I_ -- --. Ik' _ _ IP_ I I I C; VJ Q. &.D r4, _[ ll-- _ _ m-m N NI r_i 1,,4 i,,4 i,.4 N _ 141 1.4 INI N lla :;_ I_, t- I.) _i Z *- 2. _ * Z l-- -r _-, Z O * • t- u_ • I-- u_ tu O O O O O O O O O O O O V r.J "m I-- ,.4 ,-* I_ ,,_ __. _ __. 2. :3 _I, 0 O C. C_ _ O ¢: O O O O _'. Itl _1N II I _ I i _.4 w, X X il,- ,.J II m N 0 _ ql - 29 t ,I I I I I I I I I I I I I I I I I I I I I I I I N _N_ N _eN_ llNI _ ._ li#lll o " III I III I ,_ III II 0 _e_N_ 0 _e_ II II ._ III II 0 _ _I_N_ _ _DI_ I II II -- 0 _ O_I_ _ __N E I I I I .-- Z E !1 I -- I I Z 0 I I I I _ I I _ o-- | -- _ _11#_ _ _ _ I_ / illlll I I m __ _ _0_ 0 III II I II I III il IIII f U .I III II IIIIII 7 Ill II IIIIII III II III!11 - N _O_O_ N __ "-- III II IIIIII I I I i I I I I I i I I i iliIII i IIIi_I 30 t 1980010188-041 =_ R&DIO PR_PJGAT[ON PI_DICT|ON FOR CENTRAL EUROPE MARCH 19?9 S Z ISSUED |V FTZ DARNS?ROT R • [25 TELEPHONY SKY HAVE GROUND NAVE Fi|R GO00 VERY GOOD F&|R GOOD VERY GOOD 10 KH 1.5- 3,9 1,S- |,9 [.5- 3.9 1o5°14,0 1.$-16.0 1,5" S.O 30 KH l.S- 3.9 1.5- i.9 1.5- 3,9 1.3- S.S 1.S- 2.5 " 60 K_ 1,S- 4,0 t.S- 4.0 1.5- 4,0 1.5° 1,8 * * IUO KH I,$- 4,O 1.S- 6,0 1.5- 4.0 - - " 20Q KH 1.S- 6.2 1.3- 4,2 l.S- 4.2 - - " |OO KM ].5- 4.3 1.5- 6.3 _.5- 4.3 - - " 600 KH ].5" _.g 1,S- 4,S 1,7- *,5 " " " SOD KH I.S- 4.7 1.5- 6.7 Z.L- _.7 - " " 600 KH ].5- 5.0 1.8- S.D 2.9- S,O - - " ?OO KR 1.5- 5.3 2.3- 5.3 3.3- 5.3 - - " |00 KH ].8- 5.6 2.9- 5,6 .... 900 KR ?.0- 5.9 3.5- 5.9 6,8- 5.9 - - " 1000 Kfl ?.5- 6,2 4,1- O,Z _,8- 6,2 - - " TELEPRIHT_R SKY MAre GROUND M_VE FAIR GOOD VERY GOOD F^IR GOOD VERY GOOD 10 KH 1.S- 3,9 1,S- 3,q L.b- 3,9 1.5-30,0 t,5-_e,O 1.5- e,O 3_ KH 1.5- 3,9 _.5- 3,9 1,5- 3,9 1,5-_3,0 1,5- 6.Z 1.5 o X,e bO KN 1.5- 4.0 1.5- 4,O 1.S- 4,0 1.5- 2,6 - 100 K_ _.5- 4,0 t.5- 4,0 1,5- 4,5 - - " 200 KR t.g- _._ t.5- 4.2 I,5- 4,Z - ." " 300 Kfl l.§- 4,_ [.5- 4,3 1.5- 6.3 - " " 4O0 K_ 1.5- 4,_ 1.S- k,5 1,5- 4,5 - - " |00 KM t.5- 4.7 1.5- 6,7 1.8- 4,T - - - 600 KR i.5- 5.0 t.S- _.O Z.1- 5,0 - - " 700 KR 1.5- _.] 1.B- S,3 3,0- 5.Y - - " lO0 KH 1.5- $.6 2,1- 5,6 ).O- 5,6 - - - 900 KH 1.8- 5.9 2.6- 5.9 6.3- 5.9 - - LO00 KR 1.9" 6,2 3.2" 602 t,5- 6.2 " " " HORSE SKY WAVE GKO_ NAVE fAIR GODD VERY GOOD WAIR ;OOO VERY GOOD _O KR 1.S- 3.9 I._- 1,9 L,3- 3,9 1.5-30,0 1,5-26.0 1.S*Z6,8 30 K_ 1.5- 3.9 1.5- 3,9 1,5- 3,_ 1.5-18,0 1,5- ?.O 1.5- 2,8 60 KH 1.5- _.0 1.5- 4,0 1o5- _,0 1._- 4,8 1,_- 2.1 " 190 KH I._- -.0 1.5- *,O 1o5- _,0 1._- _,7 - - ZoO K_ 1.5- 4.2 1,5- 6,Z l,S- k,Z - - " 300 KH 1._- 4._ 1.5- 6.3 1.5- _.3 - 4gO KR 1.5- &,_ 1o5- 4+5 1°5- 605 - " " SO0 KR 1.5" _,7 1.5" _.7 1,5- _,? - " " 600 KH 1.5- S,O 1,5" _,O 1,T" S,O - " " TO0 K_ 1.5- _,3 1.3" 5.3 Z,O" _,3 - " " 8_0 KR 1,5- 5,b 1.7- _,b 2,0- P,6 - " " 900 KN 1.5- 5.9 1,9- S.9 3.0- 5,9 - " " 1000 KR 1.6- 6.2 2.3- 6,_ 3.8- 6,Z - " " Figure 5. Prediction for mobile and portable radio, ststi¢o_'l_.l_ Ri+_p_Ox.L _;_ .... _ _ PooP, + + "+"° 1980010188-043 frequency band which has the sigt,al-to-noise ra_io required for various modes of operation (telephony, radioteletype, telegraphy)• The following technical parameters are considered in the calculation of the predictions: Signal-to-noise ratio: As grades for the signal-to-noise ratio (i.e. the ratio of the wanted signal to the atmospheric or man-made noise level at the receiver site) the terms fair, good, and very good were chosen. They are defined as follows: fair good very good minimum S/N ratio tO dB 20 dB 30 dB The man-made noise level was calculated for rural areas. In the vicinity of large cities or factories this noise incneases and, consequently, the signal- ! to-noise ratio deteriorates. Transmitter power: The prediction is calculated for a transmitter power of lO0 W. For |0 W the frequency range with the next less favourable signal- to-noise ratio anplies to all distances (e.g good instead of very good)"f • , _ for I000 W the frequency range with the next better signal-to-noise ratio is i, valid (e.g. very good instead of good). t _ Antenna: The ground wave bridges only short distances. The prediction calculation assunes a vertical rod antenna (length 8 m). The sky wave is : suited for short and long distances. Here the calculations are based on a : half-wave dipole (height 12 m above ground). : Bandwidth: The prediction is calculated for a receiver bandwidth of 3 kHz'-(telephony), !.1 kHz (radioteletype) and 0.2 kHz (radiutelegraphy). Electric ground conductivity: The electric conductivity of the ground is i the determining factor for the propagation of the ground wave. The higher : the conductivity the longer the range of the ground wave (longest over salt ! Y water, shortest over rocky terrain). The prediction calculation assumes poorly conductive ground with a conductivity of 0.003 mho/m and a permit- tivity of 4.0. Location: The prediction is calculated for a path-midpoint approxi- : mately at 50 degrees north and 10 degrees east Ground-wave propagation requires a nearly free line-of-sight between transmitter and receiver. This does not apply to the sky wave. Here, for instance, the distant station can also be reached from a valley. Reliability of prediction: The degree of reliability of the prediction depends on all factors outlined above. Some of them, for instance the noise level and the ground conductivity, are not easily assessible. In the case of the sky wave, the same applies to the unknown actual state of the ionosphere. There is, however, sufficient experience from observations which allows a reliability of about 90% to be expected; i.e. on about 3 days of a month the upper limits of the frequency ranges (i.e. the classical MUF) given for the sky wave will be somewhat lower. Altogether, it can be said that with a careful choice of the technical 33 r 1980010188-044 parameters (antenna, transmitter power, bandwidth, location) the non- |nfluencable quantities (noise level, ground conductivity, ionosphere) have been so taken into account in the calculation, that the predicted frequency bands are useful in at least 90_ of all cases. 2.3 Comparison of the predictions with measured data The basis of the activities carried out by the Prediction Group at the : Research Institute of the Deutsche Bundespost is the continuous measurement of the signal strength of 26 distant HF transmitters. ; For these measurements only radio stations transmitting 24 hours a day are suitable. A number of meteorological broadcast transmitters meets this requirement. Information on these transmitters was taken either from ITU's International Frequency List (power, frequency, site) or from the "Nautischer Funkdienst" (contents of transmissions). Unfortunately, reliable details about transmitter powers and antennas are not always available because most -' stations are operated by the military. At the present time the field strength : of the following stations is recorded: Norfolk (USA), Tokyo (J), Canberra (AUS), Bracknell (GB), and Moscow (SU). These stations transmit simultaneously on several frequencies. The fre- quencies on which the field sLrengths are recorded were so chosen that the HF band (from 3 to 23 MHz) is well represented. The AGC voltages which are a measure for the field strength are recorded on chart recorders, and the charts are evaluated daily. A fully automatic analyzer or a direct digitization of the AGC-voltage \ is not advisable because of unforseeable interference. The correct trace on the chart characterized by certain features is marked by an experienced : operator. This is possible because all transmitters use frequency shift : keying (shift ± 400 Hz). Since the recordings are made with receivers having an IF bandwidth of only I00 Hz, the trace on the chart shows characteristic changes from mark to space. The transmissions received at different fre- quencies from the same station contain the same information. Therefore, the changes from mark to space must be recognizable on all chart recordings of that path. If a trace does not show these characteristics, there is either interference, technical trouble, or the frequency is above the MUF or below the LUF. i The strip charts are cut off every workday in the morning and hourly values of the measured receiver input voltage are read from the revised charts. The conversion from receiver input voltage into field strength is somewhat complicated because the receiving antennas are rhombic antennas of ! high directivity and the direction of the incident wave varies unpredictably (Damboldt,1977). The conversion is made by computer which stores all neces- sary data to estimate the field strength normalized to I kW transmitter power and an isotropic transmitting antenna. These values are then tabulated and depicted in different suitable ways. One way of representing these data is the "Monthly Report", which 34 1980010188-045 presents data of the preceding month and is published within about ten days of the following month. The Tables contained in the Monthly Report can be used to compare measured and predicted field strength. Figure 6 shows an example for four frequencies recorded on the path Canberra (Australia) - Germany for September 1977. Table I shows the corresponding numerical values. The median value for the differences amounts to + 4 dB, - 4 dB, - I dB and + 4 dB for the frequencies 19.6, 13.9, 11.0 and 5.1MHz, respectively. For a11 four frequencies the median value turns out to be O dB. The corresponding rms values for the difference between prediction and measurement are 6.3 dB, 9.6 dB, 6.7 dB and 6.] dB. For a11 four frequencies (a total of 62 values) the rms value is 7.9 dB. This value is an estimate for the reliability of the prediction. It is fairly good compared with another prediction method i (for comparison see: P_r_s, 1978). : 3. SHORT-TERM PREDICTIONS Besides the long-term predictions, the Research Institute of the Deutsche Bundespost also makes short-term predictions for the HF-range. These predic- tions are issued each workday and are distributed as two separate messages at different times of the day. In the following sections the so]ar-terres- trial data underlying the short-term predictions as well as a sample of them will be described. Finally a comparison of the predicted radio propagation conditions with observed propagation cnnditions will b^ given. 3.1 Solar-terrestrial data available for short-term predictions The main source of information on the state of the ionospheric trans- mission channel are the continuous field-strength measurements mentioned in section 2.3. These measurements are available in real time and even a glimpse at the chart recordings reveals details of changes in propagation conditions. Before noon the evaluated measurements of a11 26 frequencies for the past 24 hours are available as hourly median values of both field strength and receiver input voltage. Also available at that time are the readings of a local magnetometer, the readings of the Geomagnetic Observatory at Wingst (near Hamburg), hourly values of the vertical incidence critical frequencies for the past 24 hours from the ionosonde at Lindau and--weather permitting--a chart of sunspots from the observatory on the Wendelstein mountain near Salzburg. Moreover, a variety of data measured in different parts of the world is received through the Ursigram Service. The disadvantage of these data is that they are at least several hours old. Further, very useful information is supplied by the solar daily forecast (SDF) received regu]arly from Boulder. With all this information at hand, two experienced forecasters predict the propagation conditions expected for the next 24 hours. The present form of the short-term predictions gives most of the predicted values in relation to the monthly predictions. So, the monthly predicted value is regarded as "normal" and, therefore, it may happen that the short-term predictions 35 1980010188-046 19690 13920 , /.,. 0 O, i t . tlk .lO -,o ', /;,:_ . a\/l-_, '. -30 _X_. 0 2 4 6 8 10 12 1/, 16 16 20 22 2/.UT 0 2 4 6 8 10 12 1/. 16 16 20 22 2/,UT 11030 , 5100 40, /.0, 30 30. 20 20' 10. 10 ' , I __ /,, b'_', "10' ,,;,,, -20 -20 • , __,/ _, _ ,o _ -/.0 _ -/.0 • , ..................... 0 2 /. § 8 10 12 14 16 18 20 22 2/.UT 2 /. I_ 8 10 12 1/. 16 16 20 22 2/,UT Canberra-LQchowSept.lg77 Measured Predic ted Figure 6. Comparison of measured and predicted field strength. 36 t 1980010188-047 737 R[':PR,)_,,,, • "" " _' "l'i_'OR_GI,, _, , , • "" ".... _ i'OOR 1980010188-048 Indicate some values as above normal each day if the monthly prediction wa_ too pessimistic. The addressees of the short-term predictions each month receive the explanations given In the next section. 3.2 Explanations of short-term predictions Explanations regarding the Preliminary Report and the Daily Report of the Research Group Ionosphere (FI 34) in the Research Institute of the Post and Telecommunication Administration of the Federal Republic of Germany. Harch 1979: Please keep until next month. Each workday two telex messages are issued on HF radio propagation conditions, The Preliminary Re- port is made at about 0730 hrs local time and contains in plain language a review of the conditions for the period after the last Daily Report. The basis for the Preliminary Report is provided by all values being available at this early time of the day. The Preliminary Report is intended to give a : first picture of the propagation conditions ard a forecast until noon. The Daily Report is issued at about noon and comprises a detailed review for the : past 24 hours and a prediction for the next 24 hours (on Fridays for the next 3 days). All times are given in UT. Situation: Short text describing the general propagation conditions. Solar activity: The degree of solar activity is characterized by one expression. The following term_ are used: very low, low, moderate, high, very high. Generally, the critical frequencies of the F2-layer increase with increasing solar activity. Consequently, the propagation conditions improve. However, Lhere is a greater probability of disturbances. 10 cm Ottawa: The value given is a measure for the sol,,r radiation at a wavelength of i0 cm. If one deducts 50 from this value, one obtains an estimate for the sunspot number R which is taken as a bas;s for the monthly propagation predictions. Critical frequencies in central Europe: Data on the critical fre- quencies of the F2-]ayer are given in percent for four 6-hour periods. The predicted monthly median value, as given in the table below, is taken as lOO percent. The percentages apply not only to the vertical incidence critical frequencies, but also approximately to the critical frequencies at oblique incidence for any distance. Propagation conditions day/night: The day is defined here as the time between 06 and 18 UT and the night between 18 and 06 UT. The data are given for 5 different paths. The grades used are defined as follows, where mp stands for monthly prediction: very good: much better than mp, i.e. difference more than + 20 0B. good: slightly better than mp, i.e. difference between + 10 and + 20 dB. normal: corresponds to the mp. 38 -L 1980010188-049 Fair: slightly worse than mp, i.e. difference between I0 and - 20 dB. very poor: considerably worse than mp, i.e. difference more than - 30 dB. Spe£1al e.vents: Under this heading special events are mentioned, which are of Importance For radio-wave propagation, e.g. moderate or strong short wave fadeouts ($WF). An SWF is an increase in absorption at low frequencies. It is therefore necessary to use as high a frequency as possible during an SWF. In the predectlon part the probability of occurrence of an SWF is given in percent. Table of critical frequencies for the current month with an assumed sunspot-number of R = 122 i O0 OI 02 03 04 05 06 07 08 09 I0 II UT4.8 4.8 4.6 4.2 3.8 4.0 5.0 6.5 7.9 8.9 9.7 10.2 HHz • _ 12 13 14 15 16 17 18 19 20 21 22 23 UT 10.4 10.4 10.2 10.1 9.9 9.4 8.6 7.6 6.6 5.8 5.3 5.0 MHz ; Forschungsinstitut der DBP Darmstadt 3.3 Sample of the Daily Report Figure 7 shows a sample of the Daily Report which is issued in German. It is, however, easily understandable if one consults the explanations given ' in the previous section. 3.4 Comparison of short-term predictions with observed propagation conditions A daily quality figure is calculated for each circuit from the 24 hourly values of the received field strength of the transmitters recorded. For each : hour the field strength of the frequency with the highest f;e!d strength is taken. This is done on the assumption that the optimum frequency would be used for communication. Now the quality figure Q is computed from : Q = 6 + 0.2 (F 1 + F2 +...+ F24)/24, where the F[ denote the hourly value3 of the field strength in dB related to I/uV/m. It follows that a 24-hour : average field strength of 0 dB (juV/m) corresponds to a quality figure of 6.0 and that an increase of the qualety figure by 0.1 correspor, ds to a field- strength increase of 0.5 dB. This yields handy values and the quality figures of the paths monitored usually vary between 2.0 (equal to an average field strength of - 20 dB (/uV/m) under poor conditions on the long-distance path Australia - Germany ahd 14.0 (equal to an average field strength of + 40 dB (/uV/m) under good conditions on the short-distance path England - Germany. These quality figures were also forecast in the past and a detailed descrip- tion of the procedure was published (Damboldt, 1979). One disadvantage of the quality figure is apparent, viz the mean of the ]ogarlthmic Field-strength values tends to disregard extreme values and this 39 , 1980010188-050 ' 419209 ion 0 kurzweLLenfunkverblndungen tages0ertcht hr. _7 vom 21.02.79 11.45 mez rueckbLlck lo_ o Q_D_ zeit_aum : 20/0600-21/0600 ut situation : aLLgemetn normaLe Oedtngungen sonnenaktlvitaet: maesslg 10 ¢m ottawa : 2]8 gPenzfrequenzen in mltteLeuropa: 06 - 12 1_ - 1_ 18 - 24 O0 - 06 ut 1_0-150 175-145 150-17C 140-170 prozant ausbreltungs0eOl_gungen tag/nac_t: mttteLauropa : normaL/normaL nordeuroDa : normaL/normaL nordamarlka : noPmaL-maesslg/normaL, max +4 mhz ostasien : normat, max. +4 mhz/normaL; max. mhz australian : 06-12 ut + 20-06 u_ um normaL, 12-20 ut maessig beSOnOeres : magnststoerung seit 21/0300 ut vorhersage zeltrau_ : 21/0600-22/0600 ut situation : maesstge 0eOln_ungen sonnenaktlvltaet: maesslg Ots hoch 10 ¢m ottawa : 23_ g_e,zfrequenzen in mlttaLeuropa: 06 - 12 12 - 1_ 1_ - 24 O0 - 06 ut um 100 100-110 90-110 90o110 prozent ausbreltungsbedt_gunge_ tag/nacnt:mltteteuropa . maesslg/maessig nor0euroDa : maesslg/maesstg noPdam_rika : maesslg/maesslg ostasien : maesstg/maessig austraLter : 06-10 ut normaL, Oanach maesstg wanrschelnLlcnkelt fuer swf: _0 prozent besonderes : an0auern 0er magnetstoerung, in 0eren foLge nledPlge grenzfrequenzen forschungslnstltut 0er Obp darmstadt fi _4-1 gez. suessmann 419209 1on d 11.55 mez nob Figure 7. Sample of daily report. 40 1980010188-051 mean value has no physical meaning. Strictly speaking, the quality figure is connected directly with the field strength only if all 24 hourly values are Identical. It is also questionable whether the users of the predictions are ! satisfied with one value represen_ing 24 hours of a day, if traffic may have been possible during a limited time only (e.g. on long-distance paths). : A comparlsor of predicted and measured quality figures for the montn of September 1977 is shown in Table II. Predicted and measured values agree for 86.7_ of the predictions within ± 5 dB. In Table II the wrong predictions are marked with the number I, correct predictions with a period, ".". As shown Jr0 section 3.3, the Daily Report no longer contains predicted qualitt figures for the reasons mentloned above. The quali_y figures, however, are still calculated from the measurements and are published i_ the Monthly Report and in Solar Geophysical Data. 4. CONCLUSIONS This paper describes the long-term predictions o _ the Deut3che Bundespost which are computed up to several months ahead using as a da_a base the C.C.I.R. Atlas of ionospheric characteristics. These predictions are useful for frequency planning purposes. Also described are the short-term predictions and the solar-geophysical data underlyin£ them. These predictions are useful for the operator to change the frequencies in use according to changing iropaga_ion conditions. There is good agreement between predicted and observed field _trengths and propagation conditions. However, a problem--not yet _olved--is that different users of the predictions have different technical training and different demands for either more detailed or more easily understandable predictions. A closer contact between the forecaster and user is to be aimed at in order to maximize the applicability and the benefit of the predictions. REFERENCES Beckmann, B. (1965): Bemerkungen zur Abh_ngigkeit der Empfangsfeidst_rke von den Grenzen des Ubeltragungsfrequenzbereiches, NTZ 19, S. 64_-653. Beckmann, B., R. Eyfrlg, A. Ochs, W. Plage and H. Rauch (1965): Radio propagation predictions by the FTZ Darmstadt. Submitted to C.C.I.R. ,,? 6/I, 20 Dec. 1965. Beckmann, B. and A. Ochs (1969): Vorhetsage f_r die ionosph_rische Kurzwellen- Ausbreitung. In: Jahrbuch des elektrischen Fernmeldewesens, Bad Wlndsheim, Germany, F. R. C.C.I.R. (1974): Report ]_O-2, Atlas of ionospheric characteristics. C.C.I.R. (1974): Report 322-1, World distribution and characteristics of atmospheric radio noise. 41 t 1980010188-052 i| Table II Predicted and measured quality figures for the month of September 1977 Prod Meas Prod Meas Pred Meas Pred Meas Pred Meas Norfolk Tokyo Canberra _scow Bracknell J Thu 1 5.8 7.1 1 5.0 7. _ 2 4.3 5.8 11.4 11.9 1! .6 12.4 . Fri 2 6.5 7.2 6.6 6.7 . 5.4 _.0 11.8 12.0 12.2 12.6 . ! Sat 3 6.3 6.4 6.3 6.2 . 5.2 6.2 11.6 11.5 12.0 12.0 . Sun 4 6.0 6.& 6.0 7.1 1 5.0 6.& 11.& 11.9 11.8 12.4 Mon 5 6.& 6.8 6.7 6.7 . 6.0 5.2 11.7 12.4 12.2 11.9 • Tue 6 6.8 7.1 6.8 7.4 . 5.3 5.9 12.2 12.1 12.0 12.5 . Wed 7 7.0 7.2 7.2 6.7 . 5.6 &.9 12.2 11.9 12.3 11.8 Thu 8 7.0 6.6 6.6 6.6 5.0 5.2 11.8 12.0 11.8 11.9 Frl 9 6.5 5.0 1 6.5 6.3 . 5.0 5.3 • 12.0 11.7 • 12.0 11.& Sat 10 5.5 4.7 . 5.5 4.9 . 4.0 &.4 . 11.5 11.5 • 11.5 11.5 • Sun 11 4.5 &.O &.5 &.8 . 3.0 3.7 • 11.0 11.8 . 11.0 11.7 . Mon 12 &.& 4.6 5.2 6.5 I 4.6 4.6 . 11.8 12.3 . 11.8 11.7 . Tue 13 4.2 4.3 . 5.5 4.1 4.2 4.8 . 11.8 !1.8 . 11.6 11.2 Wed 14 3.8 5.3 I 3.& &.l &.O 3.2 . 11,4 11.0 . 11.0 11.4 . Ti_u 15 5.5 5.7 5.2 5;5 4.0 4.6 . 11.& 11.5 • 11.6 11.2 . Fri 16 5.8 6.9 1 5.6 4.9 4.5 3.8 . 11.5 12.4 . 11.& 11.7 . ; Sat 17 6.0 ;.0 5.6 /4.2 1 4.5 4.9 . 11.5 11.8 . 11.4 11.4 Sun 18 6.2 6.2 5.6 5.8 4.5 5.2 • 11.5 11.7 • 11.4 11.6 Mon 19 5.5 5.3 5.2 4.0 1 4.6 4.8 11.6 11.4 . 11.4 11.6 i" Tue 20 4.6 3.8 3.6 z.3 1 4.3 z'.7 11.0 10.3 11.2 11.3 ; Wed 21 3.8 3.8 2.5 2.4 4.5 3.3 I 10.5 10.3 11.4 10.7 Thu 22 4.0 3.1 2.4 2.2 3.3 3.0 10.6 11.0 11.0 11.1 Fri 23 3.5 2.8 2.5 2.6 3.2 2.9 11.0 10.2 11.3 10.6 : Sat 2/4 ',.0 3.2 3.0 1.3 1 3.5 3.2 11.0 10.7 11.3 9.4 1 Sun 25 4.5 3.3 1 3.5 3.5 3.8 2.5 1 11.0 11.0 11.3 11.2 Mo_ 26 3.5 &.l 3.8 2.8 3.2 3.8 . 11.0 !1.4 1!.2 11.3 . Tue 2? 4.0 3.3 3.0 3.7 3.6 3.5 . 11.2 I0._ 11.2 10.5 . Wed 28 4.0 &.2 . &.2 &.l 3.8 4.9 1 10.8 11.8 11.0 11.8 . Thu 29 4.5 5.3 . 4.8 3.9 5.0 /4.3 . 1111.8 1i.7 ti.8 11.3 . Fri 30 5.5 5.5 4.3 /4.8 4.5 5.5 i 1.8 12.3 11.6 12.0 42 _- . .., 1980010188-053 ; ! C.C.I.R. (1974): Report 252-2, C.C.I .R. interim method for estimating sky- 1_ wave field strength and transmission loss at frequencies between the . approximate limits of 2 and 30 MHz. Damboldt Th (1976): A comparison between the Deutsche Bundespost ionospheric' • HF radio propagation predictions and measured field strengths. AGARD Conference Proceedings No. 173. Damboldt, Th. (1977): L_chow receiving-site measurements: conversion from r.m.s, receiver input voltage to sky-wave field strength, C.C.I.R., IWP 6/I, Doc. 54. Damboldt, Th. (1978): HF short-term field strength predictions and their agreement with observations. AGARD Conference Proceedings No. 238. : Monthly Report: Published by the Research Institute, FTZ Darmstadt, P.O. Box 5000, D-6100 Darmstadt. P_r_s, H. {i978): Results of comparisons between measured RF field-strength _ data and values estimated by the method of Report 252-2, C._.I.R., IWP 6/I Doc. 76. 43 'r 1980010188-054 fN8,0-18466 " _ A _THOD OF PREDICTING GEOT:_GNETIC ACTIVITY BASED ) ON A CORONAL MODEL OF RELATIONS BET'_EN SOYAR ANDGEOV_GNETIC ACTIVITIES) I Jaroslav Halenka 1 Geophysical Institute, Czechosl. Acad. Sci. ) Bo_nf II., 1401 _ 141 31 Praha, Czechoslovakia For the predicting of both disturbed and quiet periods the method uses solar situations along the CM with the key role of filaments, giving indirect evidence of types of directly unobservable coronal structures above them. The time lag, not to be interpreted in terms of propagation speed, between the CM activity and the commencement of the geomagnetic response is about 1 - 2 days. Solar phenomena serve as indicators within ap- proximately lO_ of the CM and up to the zone of high- -latitude filaments. " INTRODUCTION The method of predicting geomagnetic activity, being used by the Heliogeophysical Group of the Geophysical Institute of the Czechoslovak Academy of Sciences, was developed during a long-term research into the effect of solar activity on geo-- magnetic activity, which was begun in 1952. After partial re- sults had been obtained, the new method of predictions based " on them was elaborated by Bedn_ov_ (1964) The method tries to connect directly observable phenomena in the Sun s photosphere and chromosphere with the variations _- of the geomagnetic field obse_ed at the Earth's surface and ii measured as geomagnetic activity. Predictions can usually be _ made 1-2 days in advance with great reliability, depending on the solar situation and availability of solar data, or 7-8 days in advance with lower reliability from east limb observations. ) This method has been used to issue test forecasts after each observation since 1959. The more than lO00 test forecasts with a reliability of about 90%, issued so far, have mainly been used in verifying and refining our solar-terrestrial re- lationship results, as we are unable, for technical reasons, to maintain a regular warning service_ '/ The use of the predictions is obvious in all areas of geo- physical and related research, as well as in human activities, sensitive directly or indirectly to the condition of the geo- : 44 C" 1980010188-055 magnetic field. The predictions need not be restricted to the _ Earth and its environment, but, without fm_damental limitations, i they can also be issued even for the half-space above the vis- ible solar hemisphere (Halenka, 1967a). BASIS POR THE PREDICTION TECHNIQUE A good tulderstanding of the true character of the relation between solar and geomagnetic activities is obviously the neces- sary condition for producing a workable method of predicting geomaguetic activity. The investigation of the connection bet- ween the solar and the geomagnetic activities, carried out using material from several solar activity cycles, in the dir- ection from solar to geomagnetic activity and vice _ersa, as well as the study of the associations between the individual displays of solar activity and of their relation to _he corona, yielded the following principal result: a geomagnetic disturb- ance will occur only if a coronal formation is pointed at the Earth (Bedn_ov_ and Haienka, 1969). There are three possibilities, basically determined by the local magnetic fields in the neighbourhood of the solar central meridian (CM). If the corona is split above the centre of the solar disk, no coronal formation is pointed at the Earth and geomagnetic calm is to be expected. If there is a coronal stream, or a coronal formation of the minimum type above the centre of the disk, a geomagnetic disturbance will follow in both cases. A sudden commencement (SC) disturbance may be ex- pected especially if a coronal stream is generated directly .; over the centre of the disk, i.e. if the already existing stream is not displaced towards the centre by solar rotation. The character of coronal formations above the centre of the disk, which cannot be observed from the Earth, must be de- rived from the groupings and variability of other displays of solar activity in the lower levels of the solar atmosphere. The coronal stream is generated above regions of enhanced instab- ility of local magnetic fields, the instability being indicated by the generation and decay of spots, by changes in area and by the brightness of plages, by the changes of the chromospheric structure in the neighbourhood of the active region, by the oc- currence of flares, by changes and, in the most expressive cases, also by temporary or permanent vanishing of filaments. On the contrary, if there is a relatively stable local magnetic field, without indications of an instability mentioned above, the coro- na is split. A temporary coronal formation of th_ .miiLimumtype may be formed above the centre of the disk, provided that there are no strong local magnetic fields, especially bipolar ones, along the CM. The coronal model, just outlined, is based mainly on the established significance of filaments and their relations to the coronal structures above them. Filaments, according to 45 t 1980010188-056 , I _ their geomagnetic effects, have been d_¢ided into three types, t signifled by characteristic behaviour, shape, and position with respect to spots, and these were later shown to correspond to 7arJous phases of development of the active centre. Bound fil- aments occur in active centres of little variability, keeping to their basic shape for longer periods and, when observed at the limb, displaying the structure of closed magnetic lines of force. Bound filaments of the first group occur in strong mag- netic fields, limited to zones of occurrence of spots in lower and middle heliographic latitudes. The second group occurs in weaker magnetic fields, often already lacking spots, even in high latituaes. Unstable filaments develop from bound filaments of both groups during sudden changes of local magnetic fields, _ which results in changes of shape and position of the filaments at higher levels of the solar atmosphere and, in extreme cases, in their permanent or temporary disappearance and a change in the fine chromospheric structure, which temporarily loses its magnetic-line-of-force structure, characteristic for the neigh- bourhood of the spots. Free filaments develop from bound filam- ents, which have survived the period of sudden changes and, after the disintegration of spots and the weakening of the lo- cal magnetic field in the appropriate active centre, they shift to higher latitudes, where they remain as bush-like structures for a period of several solar rotations and gradually dissolve. Proceeding to the limb, it can be shown from total solar eclipse observations that prominences of the bound type can be observed in the base of the split corona, whereas prominences of the unstable type occur in the base of the coronal stream. The transition of prominences of the first type into the sec- ond, over an interval of a few minutes or tens of minutes, is quite usual, and also the corona should respond accordingly: a rapid reconstruction of the originally split corona into a cor- onal stream may take place. Frequent coronal transients, often characterized by major local injections of mass and energy into the outer corona from the lower solar atmosphere, and some by apparent rearrangements of material in the corona, which are associated mainly with eruptive prominences, have been observed ; by MacQueen et al. (1976). If the transition in the corona takes place above the centre of the disc, it may explain the occurrence of geomagnetic calm immediately followed by an SC geomagnetic storm. Above prominences bound in weak magnetic fields, helmet- -like or wide radial streams occur, depending on the orient- ation of the long axis of the prominence, directed radially away from the Sun. Free prominences do not disturb the normal structure of the corona, i.e. a minimum-type corona, character- ized by a mighty equatorial streamer. The central meridian pas- sages (CMP) of hel_Let-like streamers, owing to their position and shape, are rarely followed by geomagnetic disturbances. The CMP of a streamer of a minimum-type corona results in a M-re- gion geomagnetic disturbance, which may be recurrent if the so- lar situation is stable enough for more rotations, but is in- 46 t 1980010188-057 .......... io _errupted immediately after the CMP of a strong magnetic field, inactive on its own. When no free prominences are present du- ring the CMP of a minimum corona equatorial streamer, the geo- magnetic activity becomes only slightly unsettled, but no pro- nounced geomagnetic calm occurs. Details on many various investigations, the results of " which have served in elaborating the basis for the method of _ predicting geomagnetic activity, are conveniently summarized in a recent paper by Bedn_ov_ and Halenka (1974), where all re- levant papers are referenced| references on earlier results are • given in a summarizing paper by Bedn_ov_ et al. (1963), as well. PREDICTION TECH_IQUE In elaborating a prediction of geomagnetic activity one tries to estimate the possibility of a coronal formation being pointed at the Earth. This can be done with greater or lesser reliability, depending on the availability of data and com- plexity of the particular situation, by analysing individual indicators with their changes and grouping in the central part of the solar disk and along the CM. Indicators of geomagnetic distu_bance Central. The most reliable and expressive is the sudden disappearance of a filament (SDF). Always followed by a geo- magnetic storm or disturbance, usually with an SC. The CMP of a variable filament, especially oriented in the direction of the " solar meridian. The duration of the disturbance, which follows, may be short. The CMP of a filament oriented mainly in the dir- ection of a solar parallel may not be reliable! apparently, owing to the slightest deflection by nearby magnetic fields, the related streamer may not point at the Earth. The CMP of a sunspot group with considerable activity of surges. The result- ing storm or disturbance need not be an SC one. The generation or decay of sunspots--an SC disturbance or storm is probable. _ Non-central, but on the CM! no other local magnetic field • in between the indicator and the centre of the disk! decreased reliability. The best indicator in this group is the free fil- ament in a very weak magnetic field far outside of sunspots, the CMP of which is followed by a disturbance or storm, usually long-lasting and non-SC. The duration of the disturbed period corresponds to the duration of the CMP. The disturbance vanish- es as soon as a sunspot or bright plage appears in the neigh- bourhood of the CM during the CMP of a free filament. The SDF and other unstable filaments in weaker magnetic fields (plages without spots) also belong to this group, as well as free fil- aments of meridional direction with the low-latitude end not too distant from the centre of the disk. Sometimes, especially when solar activity is distributed asymmetrically with respect 47 1980010188-058 to the solar equator (an active region in one solar hemisphere only), other strong changes mentioned above may be effective,apparently as the result of a widening or deflection of the co- ronal stream, or of the temporary formation of a minimum-type i corona. However, as there is no firm evidence of the actual ex- tent and direction of the resulting coronal stream, the prog- nostic significance of these changes is rather limited. All ab- rupt processes may cause SC storms or disturbances, as is the case with central indicators. Indicators of geomagnetic calm Central. The situation is relatively simple here. The CMP of every sunspot or bright plage, provided ultimately no signs of changes like flares and surges or unstable filaments occur, indicates a period of geomagnetic calm. This is most frequent with early bipolar types of sunspot groups according to the Zu- rich classification, as well as the oldest stable ones (B, C, H, and J! also type A, if not generated at the CM). Even the higher developed types D, E, F, and G may be used, if suf- ficiently stable, which is less probable. Non-central, but on the C_, no other indicator of disturb- ance at lower latitudes. The same applies as to the central in- dicators. A stable local magnetic field safely interrupts a disturbance originating from a high-latitude free filament. The effect may be less pronounced with increasing heliographic la- titude, as the otherwise undisturbed central part of the solar surface manifests itself by slightly unsettled geomagnetic act- ivity. '_en there is a pair of indicators, symmetrically locat- ed with respect to the solar equator, the probability of sudden changes, mentioned above under non-central indicators of dis- turbance, synchronous in both, decreases considerably and the _ resulting geomagnetic calm can be expected more reliably. Aspects of position, time and intensity The efficiency of all indicators decreases very rapidly with increasing distance from the CM. Some dependence on the size of indicators may also exist, larger ones being not so : sensitive to the decrease. _ sudden changes are taken intoconsideration within about of the CZ, whereas for the long- -lasting indicators the CMP of their centre of gravity is dec- isive. Sometimes, even more distant large changes, such as big flares, may influence the behaviour of indicators along the CM, apparently due to the magnetic connection of neighbouring act- ive regions which should, therefore, also be kept under observ- ation. As regards the time interval between the decisive indica- tor and the commencement of a magnetic disturbance or quiet pe- : riod, a good typical value is 1.5 d. In well-defined cases, • such as the SDP or the CMP of variable and free meridional fil- aments, values of between 28 and 53 h have been observed. There _ is a tendency to longer time intervals with decreasing filam- 48 1980010188-059 | ent activity. In the best defined cases, when the coronal stream is being newly formed over the centre of the disk, the observed length of the interval is between 28 and ]8 h, which is used in the case of unstable filaments. In the remaining cases, when the coronal formation has been displaced over the , C_ by solar rotation, and, as a result of a longer lifetime, is not sharply defined, the time interval observed is 1 to 2 days. It is evident, that the value of the time interval may be influenced by several disturbing factors such a£ the width, deflection from the Sun-Earth direction along solar parallels, _ or by the asymmetrical position of the coronal formation with respect to the indicator. The duration of the disturbed period, i associated with high-latitude free filaments, corresponds to _ that of the CMP of the indicator. The presence of unstable filaments in the base of the co- ronal stream and their occurrence prior to geomagnetic storms indicates that the material escaping from the filaments con- tributes, after ionization in the coronal region, to the en- richening of the geomagnetically active stream of the solar plasma. This might form the basis for predicting the intensity of magnetic disturbances. Unfortunately, it is not known which part of the prominence material is really ejected and does not return back to the solar surface, as is often observed. More- over, it is not clear, whether the enrichening is the quality of the coronal stream needed later for generating a magnetic storm during the encounter with the geomagnetic field. At the time, we do not see any objective possibility of predicting the intensity of the expected disturbance on the basis of the in- dicators mentioned in this paper. The predictions of the max- imumKp index value expected during a disturbance, covered by : test forecasts, are, therefore, only rough subjective estim- ates, based on experience, intended more or less for replacing verbal descriptions of geomagnetic activity. The suitable lev- els are Kp below or equal to 1 (geomagnetic calm), Kp about 2 (unsettled activity), Kp equal to 3 or 4 (disturbance), Kp equal to or above 5 (storm). Conditions for issuing predictions In predicting, the most serious factor limiting the re- liability of the forecasts is the availability of the appro- priate data. According to the principles explained above, the forecasts should only be issued, when the situation in the co- rona is derived on the basis of the observation of indicators and decisive changes along the CM. This requires a continuous detailed observation of the Sun, which cannot be ensured at one station. Therefore, our test forecasts have mainly been issued as ' the second kind of reliability forp_asts, i.e. the decisive changes were not observed directly but only the possibility of such changes was estimated from the character of the solar 49 1980010188-060 situation along the CM. : Even less accurate are forecasts of the third kind, based on observations of the solar situation between the eastern limb and the CM. These forecasts may be of greater importance only in periods of lower solar activity, when the initial data for predicting are mainly long-living free filaments. In any case, their accuracy should be improved later, when the indicators arrive at the CM. Another possibility often used for tentative long-term forecasts is the observation of the occurrence of prominences at the eastern limb in about the latitude of the centre of the solar disk. Taking into account their types, one can anticip- ate the possible behaviour of the geomagnetic field 7 or 8 days in advance. Moreover, eastern limb observations provide additional valuable information about the future occurrence of filaments on the disk, since a prominence need not be observed as a filament against the disk due to decreased contrast and Doppler shifts. Iu general, although the final judgement should always be made on the basis of the latest possible observations in t_e vicinity of the CE and, therefore, no period longer than 1 cr 2 days can be covered by a reliable forecast, the knowledge of the development between the eastern limb and the CM of all in- dicators coming into consideration greatly facilitates the task. EXAMPLES OF USING THE NLETHOD TO MAKE A PREDICTION Some contrasting solar-geomagnetic situations will be dis- : cussed here in order to illustrate the method of predicting and the role of several indicators. In Pig. 1 we have an example of sudden changes in the vic- inity of the disk centre. One of the most severe geomagnetic storms followed (Kp maximum = 9o), basically of an impulsive character typical of sporadic storms, starting on Feb. ll, 1958 with an SC at 01.25 UT. Suspicious circumstances: sharp geomagnetic activity increase (Kp = 50) as early as on the 6th, and also in the following three-hour intervals on Feb. lO. Solar situation: Determining factor - clearly a mi6_ty meridional filament, stretching across the centre of the disk and across the equator, connected with a large active region (E 55) displaying flare activity. Considerable chan6es of the filament in the centre region during C_,_Pat about 06 UT on Feb. 9. According to the _eudon Observatory report even an SDF was observed in the interval from 09.40 to 12.55 UT on Feb. 9. Last flare before CMP on Feb. 8, at 17.40 UT, lO°E,o20°S,o i = 2, first flare after ChiPon Peb. 9, 13.37 U_, 5 W_ 18 S, i = 2, another flare on Feb. 9, at 21.08 UT, 14_W, 13_S, i = 2, and many other flares of lower importances around C_,_P. 50 t 1980010188-061 i ,.,.,; es_3 " A4 - h- -! ;, a, I i : %; - ./ r "k i f958 Feb.8 Feb.9 Feb.10 Fig. 1. Unstable filament in the centre of the solar disk prior to the geomagnetic storm of Feb. ll, 1958. Central sections of Preibu_g solar daily^maps (only spots and filaments are shown), l_ange 60_ in hello- graphic longitude at the equator and lO0 _ in helio- grapktc latitude along the CM. The equator and the 0M are marked. N at the top, E to the left. _ Discussion: From observations on a day-to-day basis, on comparing the pictures from Feb. 8 and Feb. 9, we see a clear unstable filament at the centre of the disk. This makes it nearly certain that at least a disturbance would occur in the next 1 or 2 days, measured from the CMP of the filament. Fur- ther, an SC of the disturbance cannot be ruled out owing to the presence of a near sunspot group E 55 where sudden changes of magnetic field, indicated by flares, are probable. Now, in act- : ually observing flares, we can take the moment of the one near- est to the CMP as the decisive change which Hay cause the strong instability of the filament. Ve can then apply the 28 to 38 h inter_'al measured from the flare at 13.37 UT and expect an SC storm (considering the small distance of the filament to the CM, its meridional direction, dimensions, and probable strong change associated with the near flare a higher level of geomag- netic activity may be assumed), commencing sometime between Feb. lO, 18 UT and Peb. ll, 04 UT. If the latter value is de- layed by some ten hours, it will not be surprising as longer time delays also occur. If, however, the SDF is actually obser- ved, which is the highest priority indicator in this situation, it is nearly certaln that a geomGgnetic storm will take place with a very probable SC, which should occur sometime between _J 1980010188-062 kJ Feb. I0, 17 UT and Feb. ii, 03 UT (taking the moment when the , SDP is over as the origin for measuring the delay time). ! In an ex-post analysis one is tempted to associate the I sharp activity increase on Feb. i0, preceeding the main SC-in- crease, with the CMP of the filament in its unstable tate be- fore the SDF. On the contrary, in Pig. 2 we have an example of no sudden changes at the centre of the solar disk, with no filaments, even bound ones, so there is no danger of an unnoticed transit- ion to unstable filaments. Instead, a small stable bipolar B 6 sunspot group is present in a common plage with the following i D 19 grou_, which is flare-inactive before and during the CMP of the B 6 group. Moreover, the area around the centre of the • disk is screened on both sides by plages from the possible in- fluence of h_gher latitude events. Unde_ such circumstances a i_ period of geomagnetic calm should appear 1 or 2 days later. Using our typical value of the time interval (1.5 d) and taking the C_P of the leading edge of the plage, in which B 6 is situ- ated (April 20, around 12 UT), as the origin, we arrive at the almost exact beginning of the actual geomagnetic calm. This oc- curred on April 21 in the last three-hour interval as a sharp decrease in the surrounding, slightly disturbed activity and consisted of five consecutive values of Kp equal to or below O+ (see also Pig. 8). b s_ Fig. 2. Stable active region with no filaments at the centre of the disk prior to the geomagnetic calm of April 22, 1972. _or explanation see Pig. !. 52 1g80010188-063 a b c f .. !i Fig. 3. a) Occurrence of the H ii group in the centre of the solar disk on Oct. 4, 1971, which preceded the SC of Oct. 5, 1971. b) and c) show the situations on 7 and 8 October during the developme_,t of newly ob- served groups A 2 and A 5 close to the CM, the latter being directly in _e centre. The SC followed on Oct. 8, 1971. d) and e) show the changes in the centre of the disk and in its neighbourhood on 17 and 18 March 1971, prior to the SC of March 19, 1971. For explan- ation see Fig. 1. 53 t 1980010188-064 Other characteristic situations prior to SC-disturbances or small SC-storms are given in Fig. 3 - 7 taken from a paper by Bedn_ov_ and Halenka (1972), showing that even smaller geo- magnetic events obey the same laws as the severe ones. The dif- ferent types of situations are characterized by the presence of unstable groups of sunspots (Fig. 3), or only by unstable pla- es without spots (Fig. 4), yet again by unstable filaments Fig. 5), by conditions for a temporary minim,un-type corona after the vanishing of disturbing local magnetic fields (Fig. 6), and by the presence of non-central unstable active resions (Fig. 7). __ Fig. 4. Plages with a filuunent, _" _ v_ithout sunspots, in the centre _ of the solar disk on Aug. 29, " " 1971, which p_eceded the SC _f Aug. 30, 1971. _1orexplanation see _,_ig.1. 2_ _ig. 5. Unstable fii_ent of Jan. 25, 197! just prior to the CMP. This situation w_ i ¢_- followed by the SC of Jan. 27, 1971. _or expl nation see Fig. i. 54 1980010188-065 I a B • a b C i_ , 5 P C 11 • d e _ : F_g. 6. a, b, c) Situation related to _he SC of July 8, 1971. Vanishing of the sunspot group A 2 and other changes on the CM on July 6, 7, and 8, 1971 created conditions for a sudden and temporary forming of a minimum-type corona, d) and _) show the unstable fil- ament between the active regions on Feb. 21 and 22, 1971, when conditions existed for forming a m_nimum- : -type corona above the CM. Followed by two SC s on Feb. 23, 1971. For explanation see Fig. lo 55 1980010188-066 _'_ Pig. 7. Situation on Aug. 28, 1971, _ suitable for the action of an in- _= clined coronal stream, preceding L- the SC of Aug. 29, 1971. Por ex- planation see Pig. i. Nemymore examples, where the solar situation is analysed in individual cases using daily maps with respect to subsequent eomagnetic activity, may be found in papers by Bedn_t_ov_ 1969, 1972) and Halenka (1967b, 1969). A detailed analysis of solar patrol H-alpha films together with solar daily maps re- sulted in a detailed co-ordination of geomagnetic activity with central solar situations (Halenka 1975),- the main result is reproduced as Pig. 8 here. The types of solar situations associated with geomagnetic activity are given in other papers by Bedn_ov_ (1963, _973). The forecasts of some remarkable geomagnetic storms are discussed in section 12 of (Bedn_ov_ and Halenka 1974). z SUN_ARY OF EVALUATIONS OF PREDICTIONS There are some problems in evaluating predictions, de- pending on different kinds of reliability of forecasts, their formulations and criteria used. As mentioned, the reliability is determined by the availability of suitable data, length of period to be covered, and complexity of the particular solar situation. Prom the point of view of checking the accuracy of forecasts the most convenient way of indicating the Kp msximum values expected has no firm objective basis. Sometimes, espe- cially when issuing a forecast without continuous solar observ- ations, an active region may be followed by a storm instead of the expected calm and vice versa, and, sometimes, the expected storm or calm may be displaced to neighbouring days, due to ; tolerances in the time interval, etc. Simple evaluations of the test forecasts have shown that agreement between reality and forecasts tends to occur in 56 \. i I 1980010188-067 i • >-----CI @ _-.-,C I I)uC I zz' :zzs× -+ • :- Pig. 8. Co-ordination of g_omagnetic activity phenom- ena (periods of geomagnetic calm# other decreases, re- vivals and increases of activity), represented by the _< time pattern of the three-hour Kp-indices, with the solar situations in the immediate vicinity of the so- lar disk centre, divided into five fundamental types. The times of the individual indicators of the type of solar situation are represented by the following sym- bols: black circle - group of sunspots (regardless of characteristic but in absence of a variable filament), white circle - plages (in the absence of sunspots and ., variable filaments), pair of semi-circles - undisturb- ed area around the disk centre, thin vertical line - ; undisturbed filament (outside sunspots and plages, un- variable), bold vertical line - variable filament (any changes, arbitrary neighbourhood). The duration of the CMP indicated as required, and combined symbols were used for plages and sunspots in two cases when the sun- spot vanished, or appeared close to the CM. about 90% of the cases. As an example, Table 1 from (Bedn_ov_ : and Halenka 1968) is reproduced here to show the result of the _ evaluation from the period between April 1964 and November 1967, covering 94 days w_th an expected disturbance and 42 days with expected calm inte- vals. 57 q. i_-" )y_ 1980010188-068 Table i. Distribution of the departures of the Kp index from forecast values. u i_ , | i i iii , | Forecast (Actual Kp max.) - (Forecast Kp max.) Totaldisturbance -3 -2 -1 0 +l +2 +3 Number of cases 4 4 25 39 16 6 0 94 ; Per cent 4 4 27 42 17 6 0 lO0 ii , i i , i |i j:j ..... i : Forecast (Actual Kp min.) - (Forecast Kp min.) Total calm -3 -2 -1 0 +l +2 +3 N_mber of cases - - - 37 3 1 0 41 Per cent - - - 90 7 3 0 100 i , , , , w , u , , , , : , The published test forecasts for the years 1959-1963, 1964, 1968 and 1969 and their comparison with reality may be found in (Bednarova 1964, 1965, 1970). The ones issued during the retrospective world interval 20 _arch - 5 _ay 1976 are evaluated in (Bedn_ov_ et al. 1977). It should be stressed, however, that a straightfo_vard evaluation of the fulfilment of predictions according to any criteria is not enough, regardless of the good or poor accu- racy found. One should be able to determine the reasons for discrepancies between predictions and reality which, in our case, can generally be done on the basis of additional analy- sis using more complete data and which shows the inherent soundness of the predicting method used. " CONCLUDING _KS Experience gained so far has shown that, inspire of _e mostly adverse conditions for issuing the test forecast, _e method of predicting geomagnetic activity works relativel_ well. Further improvement and especia_,ly its practical use de- . ponds on more suitable solar observations and this again on _ world-wide co-operation. We have had to be content with neces- sarily seriously limited data, obtained with our spectrohelio- scope and prcminence-ooronagraph. Even if all sunshine periods were used at one well-located station, which is not otl_case, it would not be sufficient for a regular predicting service. As explained earlier, in connection with the possible ex- tension of the predicting method to the half-space above the visible solar hemisphere, the requirements as to "space" pre- dictions do not differ essentially from the requirements for an optimum terrestrial predicting service (Halenka 1967a). A network of several stations, well distributed around the world, 58 1980010188-069 i _ monitoring the solar image at the centre and wings of the H- -alpha line with a time resolution usual for chromospheric pat- rol films and transmitting the image in near real time to a i predicting centre for recording by videorecorder and evaluation ! by repeated accelerated projection, may be considered as suf- _ ficient for the regular issuance of forecasts of the first kind 'i of reliabi]ity. _ A very efficient supplement weald be the observation of _ the distant white corona, similar to that performed during the _ SKYLAB mission. In addition to the fundamental importance for further research into the relations between coronal sbructure and low-level phenomena, otherwise restricted to rare total so- i lar eclipses, such observations would be readily usable for long-term predictions of geomagnetic activity, owing to the stability of some coronal formations. Even more valuable to the regular predicting service would be_the use of a space probe with a c0ronagraph in Earth orbit, 90° away from the Sun-Earth line, monitoring the white-light coronal structure above the CM. REFERENCES Bedn_ov_-Nov_kov_, B. (1963): The active solar region and its relation to geomagnetic activity in the period around the minimum of 1954. Geof_sik_ln_ sboru_k, 11:383. Bedn_ov_-Nov_kov_, B. (1964): A new method of forecasting geo- _ magnetic activity and its practical try-out. Geof_sik_ln_ sborn_k, 12:277. Bedn_ov_-Nov_kov_, B. (1965): An attempt at refining the method of forecasting geomagnetic activity on material from 1964. Geof_sik_ln_ sborn_k, 13:419. Bedn_ov_-Nov_kov_, B. (1969): Sudden commencements of geomag- netic storms at the turn of two solar cycles. Geof_sik_l- n_ sborn_k, 17:203. Bedn_ov_-Nov_kov_, B. (1970): _orecast of geomagnetic activity in 1968 and 1969. GeofFsik_In_ sborn_k, 18: 379. Bedn_ov_-Nov_kov_, B. (1972): Comments on the special inter- vals of January 24 and September l, 1971. In: Data on so- lar-Keophysical activity associated with the ma_or greed level cosmic ray events of 24 Januar_ and 1-September l_71, Boulder, Wbc A _eport UAG-24, Part I, 277-278. .... . Bedn_ov_-Nov°'=ov_, B. (1973): A contribution to the problem of prominen_ _s-filaments as indicators of geomagnetic storms. Geof2sik_ln_ sborn_k, 21:369. 59 I ! 1980010188-070 Bedr_ov_-Nov_kov_, B., V. Bucha, J. Halenka, and M. Kone_r_ (1963): On the problems of the origin of geomagnetic storms. Geof2sik_in_ sborr_k, 11:407. Bedn_ov_-Nov_kov_, B., and J. Halenka (1968): New method of forecasting geomagnetic activity using features of the so- lar corona. Nature, 220:250. Bedng_ov_-Nov_kov_, B., and J. Halenka (1969): A universal in- terpretation of the generation of geomagnetic storms using features of the solar corona. Planet. Space Sci., 17:1039. Bedn_ov_-Nov_kov_, B., and J. Halenka (1972): The properties and causes of geomagnetic disturbances with sudden com- mencements. Geof_sik_lni sbornfk, 20:391. i Bedn_ov_-Nov_kov_, B., and J. Halenka (1974): Results of in- vestigating the relations between solar and geomagnetic activity obtained at the Geophysical Institute, CSAS. Geof_sik_In_ sborn_k, 22:277. Bedn_ov_-Nov_kov_, B., J. Halenka, M. Pavluchov_, and J. P#cha (1977): Forecasts of geomagnetic activity. In: Collected data reports for STIP interval II 20 March - _ Boulder, WDC _ Report UAG-S1," 2722274. ..... " _ Halenka, J. (1967a): Prognoses of geomagnetic activity and cos- mic research. Studia geoph, et geod., ll:lO1. Halenka, J. (1967b): On correlating "cosmic" flares with geo- ! magnetic storms. Geofysik_ln_ sborn_k, 15:367. f Halenka, J. (1969): Geomagnetic activity and the solar situ- ation in the neighbourhood of proton effects. Geofysik_lnl sbornik, 17:181. Halenka, J. (1975): A detailed co-ordination of geomagnetic activity with central solar situations during February, Narch and April 1972. Geof_sik_lnf sborn_k, 23:169. ! _cQueen, R. M., J. T. Gosling, E. Hildner, R. H. _unro, A. I. Poland, and C. L. Ross (1976): Initial results from the High Altitude Observatory white light coronagraph on Sky- lab - a progress report. Phil. Trans. _. Soc. Lond. A, 281:405. 6O 1980010188-071 L1 8'0-18467 ; KAZIA - A COMPUTER SYSTEM FOR SOLAR-TERRESTRIAL : DATA PROCESSING AND OBJECTIVE PREDICTION ; _ K. Stasiewicz, M. Maksymienko Polish Academy of Sciences, Space Research Centre, 0rdona 21, 01-237 Warsaw, Poland ' M. Jakimiec Astronomical Institute, University of Wroclaw Kopernika 11, 51-622 Wroclaw, Poland c The present paper describes briefly a computer system for solar- terrestrial data processing, which is currently developed at the _ Space Research Centre. The data base consists mainly of inforn_- tion contained in Ursigrams exchanged between th_ IUWDS Regional Warning Centres. The system structure and its functions are presented in this paper. Solar flare forecasting is described, as one of the prediction techniques incorporated in the system. i i. INTRODUCTION For short-term ionospheric prediction, a vast amount of solar and geo- physical data in near real-time is required. The data, describing solar activity, interplanetary medium, the magnetosphere and ionosphere, are monitored by various observatories and satellites. These data are exchanged between IUWDS Regional Warning Centres, usually on a daily basis. The : IUWDS - which reads International Ursigram and World Days Service - is an URSI Service established in 1962 for rapid solar-terrestrial data inter- change. The exchanged data cover a wide range of solar optical, magnetic and radio parameters, satellite measurements of solar wind, X-ray emission and many ground based ionospheric and geophysical data. These data are coded in 5-digit numbers, according to the IUWDS booklet on Synoptic Codes for Solar-Geophysical Data, and exchanged through teletype service between ob- servatories and Warning Centres in different countries. Anyone who wants to use these data for quick interpretation of other measurements or for short- term forecasting of ionospheric parameters has to decode the data using the IUWDS booklet. The laboriousness of manual decoding of information contained in Ursigrams prompted us into the development of an automatic technique for processing these data. A group of programs named KAZIA performs the following functions: 61 1980010188-072 (I) decoding of information contained in URSIGRAMS; (2) creation of a data base for further application, i.e. predictions; (3) production of summarized reports on solar, geomagnetic and ionospheric activity; and (4) objective forecasting of solar, geomagnetic and ionospheric activity. The data base for the system is now being expanded by including local measurements of tile geomagnetic field, ionospheric absorption, field strength of LW, SW broad- casting stations and solar radio flux on 2800 MHz. 2. SYSTEM DESCRIPTION KAZIA is wrEtten in FORTRAN IV, implemented and tested on a CDC 6000 computer. Figure I shows a general flow chart of the system. After entering Ursigrams from the teletype into the computer, Ursigrams are processed by a special program which converts all key words to corresponding numbers, and the whole code content is written numerically on a disk file D2. The sort- modify module checks the date of information and updates it in a permanent 4isk file D3. The decoding module containing 12 subroutines is capable to decode information from 42 codes described in the IUWDS booklet. The system can generate up to 26 summarized reports on solar, geomagnetic and ionospheric activity. Figures 2 and 3 show exan_ples of such reports. The data necessary for the display reside on a file F2 in a fixed format. This function, is similar to that of the SELDADS system, developed in the Space Environment Laboratory in Boulder (Williams, 1976). The latter was of great importance for us in constructing the KAZlA system. File 2 was designed to serve both display purposes and as a data base for forecasting programs. File 3 in Figure 1 contains values of the locally • measured geomagnetic field, riometer data and field strength of LW and SW broadcasting stations. They will be used to complement the global data, near real-time n_nitoring of solar-geomagnetic activity and radio wave propagation conditions. These data are also incorporated in the prediction procedures. This part of KAZIA is not yet operating. We plan to implement all available prediction methods on three different levels: (I) solar activity, (2) geo- magnetic and ionospheric effects, and (3) radio circuits performance. In the next section a technique for solar flares prediction is presented as an example of forecasting programs, which are under development. 3. AN OBJECTIVE TECHNIQUE FOR SOLAR FLARE PREDICTION An algorithm for predicting the number of flares from a particular active region is incorporated in KAZlA, using an analysis done by Jaklmiec and Wasiucionek (1979). The prediction itself is made on the basis for several _ . widely used parameters of active regions observed one day earlier. It does not forecast the exact position, time of occurrence and size of a flare but the probable number and class of the flares which might occur during the considered day. Second-order regression equations were derived from a statistical analysis of solar data for the period 1971-1975. The following 62 1980010188-073 'L : ! ....... # _ The flow chart of KAZIA _, PREDICTIONS : local _ SUMMARY measurement s REPORTS DECODING _ SORTMODIFY Ursigram_ _teletype/ RF._ING : Figure i. The flow chart of KAZIA. 63 1980010188-074 \25/ 2 IONOS EVENTS STATION ..... TIME ...... 0B8 EVENT DESCRIPTIUN \ : DIXUN 1.18 6.18 8.00 RIO AUR.A ABSORP 3 DD DIXON 3.15 4.58 5.00 RIO PCA-P ADSORP 1DB KODAIK 6.02 6.36 8.00 RIO PCA-P ADSORP 2 DB PANSKA 11.53 12.02 12.16 --- SPA 5ID 1+ Q1 DIXON 19.15 20.30 23.00 RIO AUR._ ABSORP 1DB KEROUL 17.2 18.0 VIS AURORA TYP-8 COLOR-3 0-4 18.0 19.0 VIS AURORA TYP-9 CULOR-7 Q-4 19.0 20.0 VI8 AURORA TYP-O COLOR-3 0-4 20.0 21.0 VIS AURORA TYP-O COLOR-3 g-4 J Figure 2. Dlsp!ay sh.,wing ionospheric events observed at different stations. _ 14/ 2 SOLAR FLARES .........TIME....... LOCATION IMP INT AREA t 2.45 2.49 3.02 (22:-16) 0 BRIGHT 40 5.25 5.28 5.35 ( 20: ?0) 0 BRIGHT 40 ) 6.00 ***** 6.06 ( 3: 60) 0 NORMAL 80 6.17 6.19 6.24 ( 19: 46J 0 NORMAL 40 6.56 6.58 7.04 ( 20: 76) 0 NORMAL 60 : > 8.05 ***** 8.20 ( 18: 75) ] NORMAL 80 8.22 ***** 8.27 ( 3: 59) I NORMAL 120 Figure 3. Display containing times of solar flares with their location, impor- tance and area, 64 1980010188-075 parameters have been chosen as the most relevant: i Xi - the magnetic classification of sunspot group, • X2 - the area of sunspot group, t X3 - the calcium plage intensity, X4 - parameter defined as MFl: (field strength) x (neutral line run) / _ (distance between the spots), X5 - the observed number of subflares (sn + sb), ! _6 - the observed number of flares of importance I (In + lb), and i X? - the observed number of flares of importance 2 and higher (2n + 2b _ _ 3f + 3n + 3b) ; The predicted number of subflares (Ns), flares of importance i (Ni), and ; flares of importance 2 and higher (N2) are expressed by the second order regression formulae. For groups classified as D (modified Zurich class) the predicted numbers of flares are as follows: : Ns = 1.758 + 0.976"xS"x7- 0.467"xS'x6+ O.154"x3"x5 + 1.019'x2" •x6 - 0.097"x42 + !.458'x4 (t) Ni = 0.027 + O.O98"x4.x5 + 0.021-x3"x4 (2) N2 = -O.0015 - O.014"x_'×6 + 0.014.x4 (3) For _etails see Jakimiec et al. (1979). The above formulae are suitable only for periods of high solar activity. Groups A, B, C have not been considered in this analysis because the probability of a class 2 flare is negligible. It is known chat geophysical effects are correlated rather to X-ray yield than to H= importance. The transition to the X-ray classification class (C, M, X) is done after Jakimiec (1976) using expressions similar to Bayes formula PC = P(sf)'P(C/sf) + P(1)'P(C/1) + P(2)'P(C/2) + P(3)'P(C/3). (4) The same form of equation applies to Ph and PX. where PC, PM, PX are the probability of a class C, H, X flare, respectively, P(sf), P(I), etc., are the probability of a subflare, flare of importance I, etc. P(C/I) is the : conditional probability that a flare of importance I is a class C flare. _ Conditional probabilities are given in Table 1 after Jakimiec (1976). Cal- culations, according to formulae (I - 4) are performed for all sunspot groups and summed up to give the total values Ns, N1, N2. 65 1980010188-076 Table 1. The probability that the X-ray flare is a partic- ular class (C, M, or X) given that an Ha flare of a particular importance has occurred. : X-ray yield H_ importance C M X sf 0.993 0.007 0.O00 I 0.914 0.082 0.004 : 2 O.156 0.674 0.170 3 0.062 0.438 0.500 4. SYSTEM EVALUATION The usefulness of the system, in its decoding and report generation functions, is beyond discussion. Preliminary tests on real Ursigr_ms have shown, however, that improvements of coding forms is necessary to facilitate computer processing. The staff of ARWC Warsaw have sent to the IUWDS a proposal concerning code booklet revision. It is aimed to set more precise and consistent rules for coding in order to avoid ambiguities of present description. Prediction accuracy depends on the algorithms used in the calculations. We do not have sufficient information to evaluate tilewhole system. It is designed to be flexible, to permit easy implementation on a minicomputer, ready incorporation of future advances in prediction techniques, and data base expansion. REFERENCES Jakimiec, M. (1976): unpub:ished manLscript, University of Wroclaw. Jakimiec, M., and J. Wasiucionek (1979): Short-term flare predictions and J their stationarity during the ll-years cycle, Solar-Terrestrial !I Predictions Proceedings, Vol. III. I t Williams, D. J. (1976): SELDADS: An Operational Real-Time Solar-Terrestrial Environment Monitoring System, ;,JAA Technical Report ERL 357 - SEL 37. 66 t r "_ r*. =% 1980010188-077 • _llJilillilq__ _ __ _ ..... . .... 88,0 -18468 ._. ESTIMATION OF SOLAR FLARE INFLUENCE ON RADIO CIRCUIT TRANSMISSION LOSS C _ Z. Klos, K. Stasiewicz Space Research Centre, Polish Academy _, 4- of Sciences :" Ordona 21, Warsaw, PolandI. A provisional method for determining the hours and frequencies . at which solar flares cause major radiopath absorption increase is presented. This technique is incorporated in a computer : program for monthly radio propagation prediction, developed in the Space Research Cenfre in Poland. 1. INTRODUCTION L By increasing the electron production rate in the D and E regions of the ionosphere, solar flare radiation perturbs the propagation of radio waves in : the HF band. This effc_t is manlfested in the increase of ic.ospher_c absorption of HF waves, ,nich in relation to specific radio circuits, results in higher tra_smis"ion ,oss or complete fadeout (SWF). Although the dura- tion of a sola.- flare is relatively short during high solar activity periods, the peuturbations in the HF-transmission may be significant. Naturally, the radio circuit perturbzt[on effect is closely related to the solar flare intensity, as well as the solar zenith angle along the radio path, during th_ flare occurrence. For this reason a method of determining the frequencies and times at which a solar flare of a given class can cause significant perturbations (absorption increase) for a specific radio circuit wou_d ba very usefu]. This paper presents a provisional technique incorporated in the monthly radio pro_..gation prediction computer system developed at the Space Research Centre of the Polish [.ademy of Sciences. 2. BASIS FOR PREDICTION The absorp.lon of HF waves prcpagated at oblique incidence via the )nosphere may be estimated if che height distributions of the ionization and electron collision frequency in the D and E region are known. In practice, such calculations are too complex for routine predictions and the knowledge : of the electron density distribution profiles for all times and places is poor. Th,'s current normal absorption predictions use semi-empirical 67 l 1980010188-078 ,7 _w,i," formulae based on measurements for different geophysical conditions (Schuitz Gallet, 1970; Barghausen et al., 1969, and recently George S Bradley, 1974). " To determine the absorption during a solar flare, model calculations are ;. necessary because of the lack of a vertical-incidence absorption map for flare conditions based on measured data. In the case of vertical incidence propagation at a frequency fv, in the _, frame of magnetoionic theory, the absorption of the ordinary wave is given approximately by the expression for quasi - longitudinal propagation (Ratcliffe, 1962), N L (fv) = 2.33 x 10 -9 f eve dh U (fv + fl)2 (db) (1) where fi is th= Larmor frequency in ffHz, Ne is the electron concentration in electrons/cm _, ve is the electron coll_sion frequency in s -1, p is the refrac- tive index. The lower part of the D-region and the height of reflection at frequency fv' constitute the integration limits. During a solar flare the main increase in ionization, as well as absorp- tion, occurs in the D and E region. Thus, for the frequencies reflected : from the F region, the enhancement of nondeviative absorption (for fv > plasma frequencies; _ _ !) is crucial. In this case, expression (I) takes the form: L (fv) = _,_3-10-'+ fl)2 f Ne ve dh = A(Fi, X)(fv + fl)_ (2) Here, the upper limit of the integration is the top of the Z region, and k is , the f,_nction of the flare class (F1) and solar zenith angle (X). Model calculations of A(FI, X) were made using the neutral atn_sphere model CIRA-1972, effective electron recombination coeff_'ents ouring a solar flare (Mitra, 1974), and the flare spectrum of X-ray solar radiation in the range !-20 _. The shape of the electron concentration p_ofile Ne(h) in the lower ionosphere during flare conditions depends on the intensity and spectral : shape of flare radiation, as different parts of the spectrum contribute at different altitudes. As the X-rays ,.,m be considered as the main source of disturbances in the/lower ionosphere, the classification of flares according to the radiation intensity in the range !-8 _ is introduced. According to this criterion three classes of flares are taken into account: C, M and X. This classifica- tion obviously leaves out the spectral shape of the flare. However, the integration on the altitude range in A(F1) in equation (2) moderates the dependence of HF radio wave absorption on the Ne profile shape, and conse- quently on the spectral shape of the X-ray flare. On the other hand, the mode1 calculations use X-ray intensities measured by satellite detectors. In this case in the deduction of radiation intensity from the telemetric data, the selection of a standard wavelength-averaged transfer function G is used. IC is selected for a given detector to minimize the error resulting from the 68 J 1980010188-079 variation in radiation spectral shape (Oonnel!y et al., 1977). Thus the Intensity classification can be used as an indicator of HF radio wave absorption increase. The spectra of flares which occurred on August 7, 1972 at 14.36 UT and on August II, 1972 at 12.16 UT were taken as typical for the X and M class, respectively. Calculations were performed for the maximum phase of the flare and different solar zenith angles (Dymek, 1976). The result of this calculation is presented in Figure i. 104 X =: 103 "U - N 2 10"20"30" 40' 50"60° 70"80090° SOLAR ZENITH ANGLE 3( Figure I. The dependence of A = 2.33 I0-9 _"N V dh • e e on solar zenith angle and flare class. t The absorption Lob(fob) of the ord;nary wave reflected obliquely from the ionosphere at frequency fob can be related, quite accurately, to the absorption - Lv(fv) of the ord,nary wave at vertical incidence at frequency fv (George and Bradley, 1974) as follows: (fv + fl)2 sec Iio 0 , (3) L°b(f°b) = Lv(fv) (fob + fl_2 where Ilo0 is the angle of the radio wave incidence at the height of I00 km : and f = • TIO0.v fob sec Using expression (2) the absorption for oblique incidence in flare condi- tions can be written as: 69 I 1980010188-080 __ F! A(F X) _ Lob(fob) = (_f'ob1'+ fl )2 sec IlO0 (4) The increase of the absorption at oblique incidence is determined by: A L LFl(foh,Fl,X) LN= - (fob,X,R) (5) where LN is the absorption in normal conditions given by George and Bradley (1974). The increase in the oblique-path aborption is then the sum of the absorption increase on separaze hops. 3. EVALUAFION PROCEDURE The deve_)ped computer procedure, which includes the numerical repre- sentative _f Figure I as well as calculations in normal absor;_tion conditions (following George and Bradley, 1974) estimates the possible increase in flare : absorption on the oblique radio path in the following way: (l) The geo- ' graphic coordinates of all the points at which the radio ray hops cross the height of I00 km as we _I as the incidence angle Iloo are determined" (2) For a given hour UT of a s cific month, the solar zen;th ang!e is determined for each of the above points; "3) From formulae (4) and (5) the increase in the absorption due to flares of different classes at each point is estimated; " (4) Next, the summation for all the separate hops determines the total absorption increase & Lto t along the oblique path; (5) The calculations for a given radio frequency are repeated for all hours of the day, and the hours ,. at which & Lto t > Lp are determined, where Lo is an assumed value which , depends on the technical parameters of the r_dio circuit, and (6) The cal- culations are then repeated for all frequencies in the HF band. The results obtained fcr a given radio circuit appear in a table which indicates at which hours and on which Frequencies the specific type of flare can cause a major (> Lo) absorption increase. Of course, if a flare class M has already caused a major absorption increase, a more intense flare clas: X will cause an even Table I Times and frequencies when an X-ray flare of class C, M or X could produce a major absorption increase. .,. O_ O_ O_ |_ I_ I_ MHz UT 1 - 2 - 4- 5- 6- 7- 8- X X 9- M X I0 - M M X X II - M X X 12 - M M 70 1980010188-081 larger increase. Thus, if in the daily forecast the information about the expected solar flares and their classes is contained, we know cn which fre- i quencies and at which hours for a given radio circuit the trouble r may appear. 4. CONCLUSIONS : We have presented a provisional method for determining for a specific radio circuit, the hours and frequencies at which solar flare_ can increase transmission loss more than the assumed value Lo. We have found this method useful in practice. This information in a form similar to Table I is : included in the monthly prediction of radio circuit transmission parameters issued by the Space Research Centre of the Polish Academy of Sciences. A separate daily forecast of solar flare occurrence is sent to users and thus they can know the hours and frequencies at which they may have troubles, This method has, however, some disadvantages. First, the model calcula- tion for flare conditions takes into account only nondeviative absorption, which is compared to the deviative and nonoeviative absorption for normal conditions. This yields some errors. It is also necessary to analyze many Ne-profiles for different flare classes and for different solar activity periods, to make it possible to _etermine a typical absorption increase. The elimination of these disadvantages in the presented algorithm is under devel- opment. REFERENCES Barghausen, A. F., J. W. Finney, L. L. Proctor, L. D. Schultz (1969): Pre- dicting long-term operational parameters of high-frequency sky-wave telecommunication system, ESSA Tech. Rept. ERL IIO-ITS 78. : Donnelly, R. F., R. N. Grubb, F. C. Cowley (1977): Solar X-ray measurements from SMS-I, SMS-2 and GOES-I information for data users, NOAA Technical Memorandum ERL SEL-48. Dymek, M. (1976): The dependence of ionospheric absorption coefficient on solar flare class and zenith angle. Artificial Satellites, Vol. II, No. 2. George. P. L., P. A. Bradley (1974): A new method of predicting the iono- spheric absorption of high frequency waves at oblique incidence. Telecommunication Journal, _ol. 41, No. 5. Mitra, A. P. (1974): Ionospheric effects of solar flares. D. Reidei Pub- lishing Company. Ratcliffe, J. A. (1962): The magneto-lonic theory and its applications LO the ionosphere. Cambridge University Press. Schultz, L. D., R. M. Gallet (1970): A survey and analysis of normal iono- spheric absorption meas,rements obtained from radio pulse reflections. ESSA Professional Pape, 4. : 71 1980010188-082 gl 8,o- s t 6a,SOVIET SHORT-TERM FORECASTSOF ACTIVE REGION EVOLUTION AND FLARE ACTIVITY A. B. Severny, N. N. Stepanyan, N. V. Steshenko Crimean Astrophysical Observatory p/o Nauchny, Crimea, 334413, USSR !. CRITERIA FOR FORECASTS The birth, development and decay of any active region is determined by the evuiution of the magnetic field of that region. That is the reason why the most essential criteria for a short-term forecast can be obtained from study of the structure and strength of the magnetic fields of the sun's active regions. The intensity forecast can be made using a gradient of the magnetic field between sunspots of opposite polarity. It was found that proton flares occur in the regions with the field gradient ?H > 0.1 gs km-1 (Severny, 1958, 1960). After these flares, the gradient decreased to 0.01 -0.02 gs km"l in the same parts o_ the active region. This result was obtained for many flares. The mean gradients of the longitudinal magnetic field before and after a flare were obtained by Gopasyuk et al. (1963), as shown in Table I. Table I. Mean gradients of longitudinal magnetic field before and after flares. Field 9radient Glkm Flare Number class Flare type of flares Before a flare After a flare 1 Proton (cosmic rays) 13 0.78 0.27 Proton (PCA) 11 0 45 0.27 3 Importance 3 with- 12 0.18 0.13 out PCA 4 Importance 2 and 2+ 15 0.054 0.038 " The regular records of the transverse fields in active regions performed with the Crimean magnetograph showed a highly inhomogeneous transverse 72 1980010188-083 J: _ field He and a much more inhomogeneous He structure than than of H . An examination of the position of flares relative to the H. fields (Severny, _ 1964, 1965; Howard and Severny, 1963) showed the tendency of flares to appear In regions of peculiar behavior of the transversal field such as strong rota- tion of the He vector, or the "inclusion" of transverse fields in regions occupied only by H. or in unipolar regions. Flares also occur in regions _ where the directio_ of H_ appears to cross. The probability of flare appear- ance in the points of He showinn oeculiar behavior is _ 83 percent. ;' The characteristics of the neutral line (H,, = O, H± = O) determined by • ; the magnetic structure of the active region canllbe the criteria for a fore- cast. Morton and Severny (1968) showed that bright flare knots appear mostly (_ 70 percent) near (_ i0") the neutral line of the longitudinal magnetic field (fig. l). The same result was obtained by T. T. Tsap together with French scientists (M.-J. Martres et al., 1968). In figure 2, it is shown that the flare fibrils (low right-hand corner) occur parallel to the neutral line of the longitudinal magnetic field (top left-hand corner) (Zvereva and Severny, 1970). There is a bifurcation of transversal fields in the same regions (lower left-hand corner, fig. 2). Near the Hi= 0 neutral line, as well as in the places of adjoining re- gions opposite1_ directed He and bifurcation of transversal fleld, mustaches appear more frequently (Severny, 1965; Koval, 1965). Koval found that plage brightening and flares appear in the same places as mustaches. Surges are : also more frequent in the period before the flares (Ogir, 1967). The probability that a flare will occur increases in active regions having their neutral line parallel to the equator. Koval (1974) has analyzed the dependence of flare occurrence on neutral line configuration. She has : found out that the occurrence of the neutral line, or even a small part of : it, pointed parallel to the equator, will give birth to a flare of importance > I with a probability of 80 percent (figure 3). North polor,ty South polar,ty 12- L II 21 18 15 12 9 6 3 0 3 6 g 12 15 18 21 D,stonce of flore- knots from neutral line (err sac) Figure I. The distribution of 60 separate flare knots according to distance from neutral line. 73 1980010188-084 • )___.,M.._m o 4m.mRul r 11 4.N-I.MUr 1./I,, ,, ,/ l.l;,l',v'.ll ,,,,.,• , ". -:_-__N __,4 " l-Sn,ol6l l-J., el # Figure 2. Transverse and longitudinal magnetic fields, spots, and flares in the active region, September 2, 1966. _Y ...... ^1 _ /,.7 e= t,. _ 8 • _ • ,, _/ . . O / - _ e/* - e- Ln /.>,¢u 4,. "'" '_" °" • • -o 7_ [ , t , i i , ' | | I_ 0 ._ g 6 6 10 IE n : number of days with neutral line parallel to equator Figure 3. Correlation of neutral l lne and flares. 74 1980010188-085 ,¢ 4 It is known that the emergence of new magnetic fields into (or near) an existing area of strong fields is followed by flares. Th_ evidence of field emergence may be Indicated, for example, arch filament system, bursts, appearance of new pores, _pots, group-satellites, and plage brightenings; but these events are not always connected with flares. They occur frequently during the quiet evolJtion of active regions. A number of Soviet authors have considered the relation between these events and flares. The aim of such studies is to find quantitative parameters for flare forecasting. The relations between the number of flares and the age of the group, on the one hand (fig. 4), and piage brightness and the number of nuclei of sun- spot umbre, on the other hand (fig. 5), have been obtained by Slonim (1957) for the period 1937-1948. The maximum flare occurrence coincides with the seventeenth or eighteenth day of existence of the active region. Similar _ results were obtained earlier by D*Asambuga (1939)with observations over : _ two years. Sionim derives the empirical equations for calculation of the • number of flares Y from the number of umbre nuclei in the group, u, as : Y = O.035u _ and from area of the group, A, as Y = 0.026A. The last formula is obtained for the areas measured in the Greenwich system. For a later period the analysis of observational data was undertaken by • Brailcvskaya et al. (1972). The dependence of the flare activity on the plage characteristics was confirmed. In addition the solar charts at _ cm were compared to the solar image in the CaII K - line. It was assumed that there is an interaction between the plages if they are _urrounded by the same isoline in the 9 cm chart. The interactions of N-N, S-S type (both plages are in the same hemisphere) and N-S type (the plages are in different hemispheres, with the isoline crossing the equator) were ccnsidered (fig. 6). N • 6 ! T 0 1 2 3 4 5 Age of the Sunspot Group solarrotat ions Figure 4. Dependence of the number of flares on age of group. 75 I 1980010188-086 Number of spot Nuclei Figure 5. Dependence of number of Flgure 6. Example of in- flares on number of spot teracting plages. nuclei. if there is a N-S interaction, it is always followed by flare activity brighter than a certain value. In the absence of a N-S interaction, the flare activity was lower than the certain value in I0 of 12 cases. The ratio of the plage brightness in the Ha and K line of CaII can be a criterion of the flare activity as well (Stepanyan, 1974), It was found that in 80 percent of the cases, the intensity ratio of H_ to K permits the determination (with the probability _ 85 percent) of the level of flare activity of the plage during its pass across the solar disc. The study of magnetic field strength at two levels inside the sun3pots before and after the flares was carried out by Koval and Stepanyan (1972), using the CaI A 6!03 line (top level) and FeI ? 6302 line (low level). It was determined for 70 flares that A H (=H6103-H6302) in the nearest sunspots increases befor, the flare and decreases and becomes negative after the flare. Sunspots of the same group but as far from the flare as the sunspots of other groups on the disc do not exhibit this effect (fig. 7). Kuleshova (i972) has found some characteristics of sunspot groups which are important for the prediction of proton flares. The first characteristic i= the diurnal rotation _ (in degrees) of the line connecting two stable sunspots in the group (fl_. 8). The second is & (V H), I.e. the diurnal change of the horizontal gradient of the magnetic field of the same stable _ sunspot pair, where V H_(H n - H )/1 (fig. 9). The mean diurnal changes I&¢_ and !&(VH) I for the active P_gions with proton flares exceed by two or three times the values for the non-proton flare regions, the J&_l : and J& (VH)J values increase greatly ti,e day before a flare. The same v_ues remain nearly constant the day b :re a non-proton flare. 76 0 1980010188-087 AI_ Spots near the flares 10 8 " I I 6 :'" . l'o ..... • ,.-, •.... • ,*'l...I. ... ...._t...._;._-. • Or -'* - ' .L I :_,"-. I GI e qllll el • _ • • v • • . "" " "1""1": .... • • ( r'..". • :.. I I. • _0 Spots for from flcres I I 4 I I 2 .... I.I" • 0• ell _ ooen | • * 0 .... ,._ • -4 ..i • • ,.),.. .... "-_" • ":; l"ii I iii I -4 ,. • • I -I "; I -8 -8 .4 -2 0 2 4 6h /_* (ToE) 5- Tf ) Figure 7. Change of magnetic field difference at two photospheric levels associated with flares. Group-satellites, which appear 15 to 20 degrees from the main grou" can be a criterion for flare prediction. Kusinsky (1972) showed that 65 percent of the cases the moment of group-satellite appearance kre(_;_ the flare. There is good correlation (coefficient of 0.8-0.9) betwee _i._ positional angle of a satellite and a mean positio_Jl angle of f_ares " o- rive to the center of the main group. Ogir and Shaposhnikova (i965) sh_wea that strong flares follow the appearance of new sunspots or thc growing _ sunspots near the flare. Bumba and Obridko (I_69); Vladimirsky, Levitsky and Severny (1971); Vladimirsky and Levitsky (1974); and Ogir (1978) cont, buted to the study of the distribution of flare activity relative to the sector structure boundaries of the interplanetary magnetic field, extrapolated to the sun. Flares, especially proton flares, concentrate near sector boundaries; moreover, the most powerful events appear near the boundaries (-,+) in the northern and southern hemispheres• Summarizing, the authors discussed above have formulated the following criteria for the prediction of flore appearance: i. the longitudinal magnetic field gradient; 77 1980010188-088 a_ (avH ) 20" x 3 X Q 16 " 2 x x....-.-_\ a .i I i l = ' 8 or -3 -2 -1 0 I da,,s°_o -o . T Figure 9. The AVH change4 o before the proton flares. 0 is the day of proton flare appearance i _ J I I j days In Figures 8 and 9, the a curve _ -3 -2 -1 0 1 7" is for active rec_ions with proton flares, and the 6 curve is with- Figure 8. The change in diurnal ro- out proton flares. tation A_ of the group's axis before proton fleres. _. the places of anomalous transverse magnetic fields; 3. neutral line, parallel to the solar equator; 4. the existence of mustaches anO surjes; 5. the plage brightness in Ha; 6. The relat've brightness ef Ha and K CalI plages; 7. the interaction of plages on 9 cm wavelength; 8. the change in difference of _aagnetic field strength in two levels of sunspot depth; 9. the sunspot group rot3tion; 78 I "1 1980010188-089 10. the appearance of satellite groups; and il. =e birth of new sunspot.= in the group. 2. FORECAST METHODS In the USSR, flare forecasts have been carried out in the Crimean astrophysical observatory during the cosmonaut flights. Such forecasts are based on the measurements of magnetic fields and their gradients in active regions. In addition, data of chromosphere activity (the observations in Ha and K rail) and radio emTssion al: cm and m wavelengths were used. These data were also obtained in othe: bservatories of the USSR and were sent to the Crimea in real time. Recently two new forecast centers were ; established in the USSR--In the I_stitute of Applied Geophysics in Moscow, and at the Astronomical Observat_.y of the Kiev University. The forecasts were given three days beforehand in a three-class scale. Forecast I corresponds to the absence of Flares or Co flares of importance < I. Forecast II correspond3 to flares of importance 2. And forecast III corresponds to flares of importance > 3. If forecast III is given, the forecaster indicates at the same t!_e the probability (in percentage) of proton appearance. The description of this method and the estimation of its probability are given by Severny (1965); Severny and Steshenko (]972a,b); Steshenkc (1973); and Gopasyuk et al. (1964). The comnarison of flare patrol data with forecasts issued in Crimea for : the period 1965-1971, which includes the =uI_r activity maximum, shows that 78 percent of all forecasts were correct, 14 percent were overestimated, and S percent were um_erestimated. Comparison of our fnrecasts with those _f other fore(ast centers is difficult because of the difference in forecast criteria and ohjects. Thus in Boulder, following Simon and Mclntosh (1976), the correct forecasts for M, X and proton events occurred in 69, 78, ahd 83 percent of the cases, respectively. In Meudon in 1968 and 1969 there were 82.5 percent correct forecasts; 8.4 percent overestimat J forecasts, and 9.1 percent uncer- estimated ones. In conn{ction with .he beginning of daily forecast in the USSR, mathe- • matica_ prediction methods utilizing _mputers flare been developing. The characteristic peculiarity of these methods is the necessity of utilizing the abu.dant statistical data to formulate the forecast rules. In practice,i -; all investigators have had to use some catalogs of solar data, althou,h it _: diminished the number of initial parameters. The most essential forecast parameters, e.g. magnetic field gradients, Hr_ fine structure, X-ray emissioq of separate regions, and plage brightness, are not yet published with the necessary accuracy and time resolution in catalogs of solar data. Neverthe- less, the problems of computer-predicticn methcds are important to consider. For their decision, one uses some kind of mathelatical method, such as linear-statistical methods, ;ogical methods, and different m(,difications of 79 1980010188-090 {? • pattern recognition methods. Host of these methods give a 70-80 percent probability of correct forecasts. They permit the comparison of predictions ,! based on various parameters, to evaluate the usefulness of each parameter-- ! this is a valuable contribution to the practical application of forecast. But because of the uncertainty or poor quality of the initial data, the most informative parameters in any of the above-mentioned mLthods cannot be i essential in the physical process of a flare. Two forecast problems were solved by different methods by Slutskaya and : Stepanyan (1974, 1976), Burov (1976), Dolgoarchinnyh end Kuklin (1975) on 'he basis r the same solar observation data from 1967 through 1969. The , t rst problem involved using the plage characteristics in the day of its : bir_h on the solar disc to predict whether it would decay on the disc or _ ss to the invisible hemisphere and have spots and flares. The second problem concerned using the observations of a plage born on the solar disc during its pass across _he disc up to its setting behind the west limb to predict whether it would pass from the east limb or not. The first problem, 4 basic parameters were considered, 15 in the second one. Slutskaya and Stepanyan (1974, 1976) established the construction of the extreme algorithm. The coefficients characterizing the meaning of forecast parameters and typicality of the basic objects are found in the process of computer training. The algorithm of Burov (1976) constructed the discriminative hiper-planes. The solution was in part a logical function of the conjunctions of i-parame- ters. Dolgoarchinnyh and Kuklin (1975) constructed the logical rule of a "tree" of sequential dissections to subsequences which consist of objects i of only one class. The results of the above-discussed problems were compared by Burov, Kuklin, Siutskaya, and Stepanyan (1976). In conclusion it was assumed that all methods discussed are effective enough, but the best one is the extremal algorithm method. Along with high forecast probability, it permits the choice of the optimal number and k:nd of parameters. The forecast of the further evolution of an active region from the moment _f a spot appearance in it is discussed by Berlyand, Burov, and Stepanyan (1978) and Stepanyan and Jakimiec (1978). From the same observa- tional data (sunspot groups appeared on the disc from 1967 through 1969), the following problems were solvcd: the forecast flare activity, the existence duration, and the maximum Z_rich class, using the methods of multiple regression and pattern recognition (algorithm "TOPOL"). A number of works on long- and short-term flare forecast were performed in the Astronomical Observatory in Kiev University. Izotov, Izotova, Romanchuk (1976a,b) addressed the long-term forecast. On the basis of the statistical treatment of the sunspot group data and solar flares during 1959-1961 two methods of forecasting the monthly flare number were proposed. The first method was based on the dependence of the number of solar flares on Wolf's number; the second method was based on the dependence of the number of flares in a spot group on the stage of its evolution. ; Two forecast methods proposed by Romanchuk, Izotov, Izotova, and Krivodubsky (1977) for a ]-to-7-day period are based on research about the statistical relationship of flares with Z_rich class and the maximum size of 1980010188-091 the group. The forecast ls accurate for 90 percent of the flares In the : period from July, 1974, to November, 1975. Studies by Efimenko, Efimenko and Telnyuk-Adamchuk (1977, 1976); Izotov, Izotova and Romanchuk (1976); Efimenko (1976); and Vaschenko, Efimenko, and Efimenko (1976) were dedicated to the short-term forecast based on sunspot group characteristics. On the basis of these studies, diurnal flare fore- " casting has been carried out in the Astronomical Observatory of Kiev University. From July, 1974, through December, 1977, 61] forecasts were issued one day in advance of 154 sunspot groups. The occurrence of "no flares", subflares (S) or flares of importance I, and flares of importance • > 2 was forecast, as shown in Table 2. Flares were forecast correctly in 71.6 percent, overestimated in 20.6 percent, and underestimated in 7.8 percent of the cases. Table 2. Forecast justification for every event classification. Forecast observations no flares s;.____l >__2 Total no flares 285 26 1 312 S; 1 117 149 21 287 >2 I 8 5 14 Total 403 183 27 61] During the same period the probability of correct forecasts of the number • of subflares and flares of importance 1 given 1-7 days in advance is 82 per- cent. It is assumed that the forecast is true if the forecast error is not higher than the stardard error of the method (the absolute error is 1.5 flares per day and the root mean square error is 2 flares per day). The forecast and detection of proton flares are of great interest from , the point of view of their geoeffectiveness. The studies of IZMIRAN investi- ; gators Akinyan, Fomichev, and Chertock (1976) are dedicated to the creation of a method of solar cosmic ray intensity investigation from radio burst characteristics. Earlier, radio burst data were apalyzed in connection with Flare diagnosis by Straka (1970), Croom (1971a,b), B6hme (1972), : Castelli et ai. (1973), and Sakurai (1974). But the main advantages of the radio emission were not utilized at,d a number of important properties of solar cosmic ray propagation were not considered. The method proposed by _- Akinyan et al. (1976) included the following new points: Successive considerations of the dependence of the proton flux parameters and the character of their relations with radio bursts in the cm- and m-wavelength ranges on the flare heliographic longitude were made. Information in solar 81 0 1980010188-092 Uradio data at meter wavelengths about the injection of energetic protons into • the interplanetary space was considered. A unified method was used to esti- i _ mate the proton flux parameters in different energy ranges (E > I0, 30 and I 60 Rev). Provision was made for the estimation of the time parameters (the : l time of beginning and the maximum phase of the proton event in relation to 1 the maximum of microwave flux, the value of decreasing of particle stream), ! making possible the calculation of particle flux-time profiles. The char-\ acteristics of microwave radio emission on 3 and 9 GHz were used to obtain ; information on the quantity of protons. • The intensity of the m-component was used to estimate the conditions of protons emerging into interplanetary space and the frequency spectrum of the microwave burst was used to determine the energy spectrum of protons. Proton events from 1965 to 1969 were analyzed to determine quantita- : tively the functions of intensity and longitudinal attenuation and the : basic time functions of the proton flux near Earth, Then a control study was conducted using independent observations of the events from 1970 to 1976, which confirmed the main features of the relation between proton flux ' parameters and radio bursts. Kobrin et al. (1978) have been studying the fluctuations of the solar radiation spectral index in the sm-range. They determined that some days -: , before the proton flare, the amplitude of fluctuations with a period of 30 min increases rapidly (fig. 10). One can make use of this fact for proton flare prediction, At the present time there are some theoretical treatments of the flare forecast. Sirovatsky (1977a,b,c) and $omov and Sirovatsky (1977) have carried out the current sheet theory for a flare mechanism. They make calculations which must permit the determination of the characteristic features of the flare from the observations of pre-flare motions and magnetic fields in the active region. Another method of forecasting the probability of an event is considered by Hogilevsky (1978a,b). The evolution of the active region is considered as a stochastic process, satisfying the qualification of Rarkov chains in that the following states are defined only by the preceding states. This method allows the calculation of the probability of the possible follow- ing states of the active region from the state being observed. To indicate the probability of the given state one must have a parameter that reflects : objectively the energy of the active region. The Fourier spectrum of quasi- periodical oscillations of the magnetic field or radio emission can be used as this parameter. See for example figure lO. i 1980010188-093 " -'f t ',I,,. gl. 111. ilI . q_l"l" _l. ,.' i _ 27 Zg29_0Ol02OJO_050607O?09 June-July , 1974 Figure I0. The change of spectral index fluctuation amplitude before proton flares. The moments of flares are noted as arrows along the abscissa. i REFERENCES : Akinyan, S. T., V. V. Fomichev, and I. H. Chertock (1976): The estimation of solar cosmic rays intensity by the characteristics of the radio bursts. In: The Birth and Evolution of Solar Active Regions, Moscow, "Nauka", 190. _ Berlyand, B. O. (1978): Characteristics of active region producing high energy proton flares. In: Publ. of IX Consultation on Solar Physics, : Wrocl aw, Poland. Berlyand, B. O., and V. A. Burov (1977): The choice of parameters charac- terizing the solar flares with injections of high energy protons. Trudy IPG, 35:28. : Berlyand, B. 0., V. A. Burov, and N. N. Stepanyan (1978): Flare activity of the developing active regions and its prediction. In: Publ. of IX Consultation on Solar Physics, Wroclaw, Poland. f, _ 83 o 1980010188-094 ,_ Bernshtejn, P. B. (1978): On long-term prediction of the proton-flare productivity of active regions. In: Pub]. of IX Consultation on Solar Physics, Warsaw, Poland. Bezruchenkova, T. N., N. K. Pereyaslova, end S. G. Frolov (1978): Diagnos- ; tics of solar flares and predictions of solar cosmic ray characteristics deduced from data on electromagnetic radiation. In: Pubi. of IX Con- sultation on Solar Physics, Wrociaw, Poland. B_hme, A. (1972): Spectral behavior and proton effects of the type IV broadband continua. Sol. Physics, 25:478. Brailovskaya, I. Yu., A. N. Kova], N. B. Ogir, and N. N. Stepanyan (1972): Some parameters of fiocculi characteristic for flare activity. Solnechnye Dannye, 6:88. Bumba, V., t,,id V. N. Obridko (1969): "Bartels" active longitudes, sector . boundaries and flare activity. Solar Physics, 6:104. Burov, V. A. (1976): Forecast of some evolution characteristics of active regions. In: Birth and Evolution of So.]ar Active Regions, Hoscow, "Nauka" 142. Burov, V. A., G. V. Kukiin, T. L. Stutskaya, and N. N. Stepanyan (]976): The comparison of various methods of forecasting of flocculi appearance at the limb. In: Birth and Evolution o! Solar A£tive RF_ions, Moscow, : "Nauka", 150. Casteili, J. P., J. Aarons, D. A. Guidice, and R. M. Stzaka (1973): The solar radio patrol network of USAF and its application. Proc. IEEE, 61:1307. Croom, D. L. (1971a): Solar microwave bursts as indicators of the occurrence of solar proton emission. Solar Physics, 19:152. Croom, D. L. (1971b): Forecasting the intensity of solar proton events from the time characteristics of solar microwave bursts. SoiarPhyslcs, 19:171. D'Azambuga, L. (1939): La cooperation internationale pour i'observation continue du Solei] et ses premiers resultats. L'Astronomie, 97:1. Doigoarshinnich, B. G., and G. V. Kuklin (1975): Forecast of active region evolution by constructing logical recognizing graphs. Res. Geom. Aeron. So1. Phys., 37:24. Eflmenko, V. I., V. H. Efimenko, and V. V. Telnyuk-Adamchuk (]976): Short- term forecast of flare activity of sunspot groups. In: The Birth and Evolution of Solar Active Re_ions, Hoscow, "Nauka", 182. Efimenko, V. I., V. M. Efimenko, and V. V. Telnyuk-Adamchuk (1977): On the : short-term forecast of the chromospheric flares on the basis of sunspot group characteristics. Vestnik Kievskogp Universi.teta_ Ast.ronomia, 19:!9. 84 i 1980010188-095 Efimenko, V. H. (1976): On prehlstory Interval for flare forecast. Solnechnye Oannye, 11:63. Frolov, S. G. (1977): The solar proton flare diagnosis on the following electromagnetic emission. [Judy IPG, 35:37. Gopasyuk, S. I., H. B. Ogir, A. B. Severny, and E. F. Shaposhnlkova (1963): The structure of magnetic fields and its variations in flare regions. Izv. Krim. Astrofiz. Obs., 29:15. Gopasyuk_ S. I., H. B. Ogir, A. B. Severny, E. F. Shaposhnikova, and N. V. Steshenko (1964): On the forecasting of solar flares. In: Proc. Xll International Astlonautical Con_resse Varna 1962, Wien and New York, 63. Howard, R., and A. Severny (1963): Solar magnetic fields and the great flare of July 16, 1959. Astrophys. Journ., 137:i242. Izotov, Yu.l., I.Yu. Izotova, and P. R. Romanchuk (1976a): Flare activity in the process of the group evolution. In: The Birth and Evolution of Solar Active Regions, Hoscow, "Nauka", 162. Izotov, Yu.l., I.Yu. Izotova, and P. R. Romanchuk (1976b): The long-term forecast of the chromosphere flares several months in advance. Solnechnye Dannye, 6:67. Kasinsky, V. V. (1972): On the large scale interrelation between the : chromospheric flare generating sunspots and the satellite groups. • Solar-Terrestrial Physics, USSR, 3:296. Kobrin, H. H., A. I. Korshunov, S. I. Arbuzov, V. V. Pakhomov, V. H. _ ,i Fridman, and Yu.V. Tikhonov (1978): Hanifestation of pulsation in- stability in solar radio emission preceding proton flares. So1. _ Physics, 56:359. Koval, A. N. (1965): The position of moustaches in a spot group relative to • the magnetic field. Izv. Krim. AstrofIz. Obs., 34:278. Koval, A. N. (1974): A dependence between flare activity of the sunspot groups and the orientation of the line dividing the magnetic field polarities. Izv. Krim. Astrofiz. Obs., 51:13. Koval, A. N., and N. N. Stepanyan (1972): Variation of the sunspot magnetic fields at two levels and the development of the active regions. Solnechno-zemnaya F.izlka, 3:210. Kuleshova, V. L. (1972): Some features of development of active regions with proton flares. Geomag. Aeronom., 12:328. Levitsky, L. S. (1979): The flare distribution relative to the sectoral boundaries of the interplanetary field, extrapolated to the sun. Izvestiya Krim. Astrofiz. 0bs., 62. 85 1980010188-096 Martres, M.-J., R. Mlchard, U. Soru-lscovicl, and T. T. Tsap (1968): Etude de ia localisation des 6ruptions dans la structure magn6tique 6volutive des r6gions actlves solaires. Sola ! Physics, 5:187. Mogllevsky, E. I. (1978a): A basis for a stochastic method of quantitative flare prediction. In: Publ. of IX Consultation on Solar Physics, Wroclaw, Poland. Hogilevsky, E. I. (1978b): Solar magnetic fields and a short-term quantita- tive forecast of the solar activity. Sbornlk "Fizika Solnca", Moscow, "Nauka". Horeton, G. E., and A. B. Severny (1968): Magnetic fields and flares in the Region CMP, 20 September 1963. Solar Physics, 3:282. Ogir, M. B. (1967): On some peculiarities of the origin and motion of flares accompanied by type IV bursts. Izv. Krim. Astrofiz. Obs., 37:94. Ogir, M. B. (1978): The relationship between solar flares and the sector boundaries of the interplanetary magnetic field for maximum and decrease phases of zoth solar activity cycle. Solnechnye Dan_ye, 9. Pereyasiova, N. K., L. A. Kolesnikova, M. N. Nazarova, and I. E. Petrenko (1978): Some features of the development of active regions connected with proton events. In: Publ. of IX Consultation on Solar Physics, Wroclaw, Poland. Romanchuk, P. R., Yu.I, Izotov, I.Yu. Izotova, and V. N. Krivodubsky (1977): The forecast of the chromospheric flares from ] to 7 days ahead. Vest- nik Kievskogo Universiteta, Astronomia, 19:29. Sakurai, K. (1974): Energetic particles from tl-e sun. Astro_hys. and Space Sci., 28:375. Severny, A. B. (1958): The appearance of flares in neutral points of the Solar magnetic field and the pinch-effect. Izv. Krim. Astrofiz. 0bs., 20:22. Severny, A. B. (1960): An investigation of magnetic fields connected wlth solar flares. Izv. Krim. Astrofiz. Obs., 22:12. Severny, A. B. (1964): Observations of transversal and longitudinal magnetic fields connected with solar flares. Izv. Krim. Astrofiz. Obs., 31:159. Severny, A. B. (1965): On the changes of magnetic fields connected with flares. IAU Symp. 22, 358. Severny, A. B., and N. V. Steshenko (1972a): Method for forecasting of solar flares. Solar Terrestrial Physics 1970: Part I:173. Severny, A. B., and N. V. Steshenko (1972b): Development of the methods of solar flare forecasting. In: XV Plenary COSPAR, Madrid, May 1972, 158. 86 1980010188-097 Simon, P., and P. S. Mclntosh (1972): Survey of current solar forecast centers. In: Solar Activity Observations and Predictions. IProgress in Astronautics and Aeronautics, V. 30). Mclntosh, P. S., Dryer, M., (Eds_, 33. Slonim, Yu.M. (1957): Active regions of the sun and chromospheric flares, Trudy Tashkentskoy Astronom. Obs., Series 2, 6:61. Slutskaya, T. L., and N. N. Stepanyan (1974): Forecast of active region development by the method of potential functions. Izv. Krim. Astrofiz. Obs., 52:131. Siutskaya, T. L., and N. N. Stepanyan (1976): Estimation of original para- : meters significance for forecasts of floccul: appearance from behind the limb. Solnechnye Dannye, !:70. Somov, B. V., and S. S. Sirovatsky (19773: Quasi-steady magnetic field dissipation in current sheets as the source of heating for solar active region. Sol. Physics, 55:393. Stepanov, V. E., and A. B. Severny (1962): A photoelectric method for measurements of the magnitude and direction of the solar magnetic field. Izv. Krim. Astrofiz. Obs., 28:166. Stepanyan, N. N. (19743: Quantitative measurements of the H and K brightness of the plages as a criterion for forecasting of active region develop- ment. Izv. Krim. Astrofiz. Obs., 51:8. Stepanyan, N. N., and M. Jakimiec (1978): Predictions of the development and flare activity of active regions based on the method of multiple .egres- sion. In: Pubi. of IX Consultation on Solar Physics, Wroclaw, Poland. Steshenko, N. V. (19733: The study of solar flare appearance. Vestnik A. N. ssS___RR,3:42. Straka, R. M. (1970): lIicrowave spectral observations of coronal condensa- tions, Report AFC RL-70-O002, Air Force Cambridge Research Labs., Bed- ford, Mass. : Syrovatsky, S. I. (1977a): On the possibilities to observe preflare current sheets on the sun. Pisma v Astr. Jurnai (USSR), 3:133. Syrovatsky, S. I. (1977b): The problems of solar activity and cosmic system "Prognoz," Moskva, "Nauka," 5. Vaschenko, N. D., V. I. Efimenko, and V. M. Efimenko (1976): The formation of conception algorithms and the forecast of chromospheric flares. The Astronomical Observatory of Kiev University, Preprint 12. Vladimirsky, B. M., and L. S. Levitsky (19743: A solar flare activity dis- tribution around the sector-structure boundaries extrapolated to the sun. Izvestia Krim. Astrofiz. Obs., 51:74. 87 o 1980010188-098 ,+ : Vladlmlrsky, B. M., L. S. Levltsky, and A. B. Severny (1971): In: The Generation of Cosmic Rays on the Sun, Hoscow, 314. Zvereva, A. H., and A. B. Severny (1970): Magnetic fields and proton flares 6 July and 2 September 1966. Izv. Krim. Astrofiz. 0bs., 41-42:97. 88 • t 1980010188-099 ....... _._mN=_ _-=_ i_ ,- . _ :.:'7 i __ J_ , - " , -18470, FORECASTING OF SOLAR ENERGETIC RADIATION AT THE FORECAST CENTER OF THE INSTITUTE OF APPLIED GEOPHYSICS S. I. Avdyushin, N. K. Pereyaslova, F. L. Dl|kman, and Yu.H. Kulagin Institute of Applied Geophysics Goscomhydromet, Hoscow, USSR Charged particle fluxes generated in solar flares are the main source of radiation hazard. However, a variety of conditions for injection and absorption of charged particle fluxes exists in a flare region, and even in large flares flux injection does not necessarily take place. The process of particle propagati_j accompanied by acceleration in interplanetary space is also of Importance. Consequently, solar activity disturbances are not unambiguously related to the radiation environment near Earth. The physical me_hanism of solar flares is not clearly understood and it is impossible toVdevelop a unique and reliable technique for forecasting the radiation environment. Investigation of the relationships between the characteristics of solar cosmic ray events and parameters of heliogeophysical phenomena accompanying a flare has made it possible to find out some regularities which are used as the basis for forecasting techniques. For solving forecast problems, a method of simultaneous investigation of a complex of heliogeophysical phenomena accompanying a solar flare Is used. Required date on solar activity, the interplanetary medium, and penetrating radiation fluxes in interplanetary space and in the magnetosphere are provided by space, strato- spheric and ground-based means, which form the regular observational network of the Forecast Service of the Institute of Applied Geophysics (lAG). Collection, processing and transmission of initial data for analysis are performed with the help of a ground-based automatic system developed on the basis of the °'HINSK-32" computer. The following forecasts are regularly issued by the lAG Service: (I) forecasts of proton flux evolution in solar cosmic-ray (SCR) flares (extrapolation methods), (2) estimations of the integrated SCR flux from a flare based on the accompanying electromagnetic radiation, and (3) qualita- tive forecasts of flare activity and the magnetosphere and radiation environment for a two day period (issued daily) and for a 27 day period (issued weekly). Observational data from natlonal and foreign sources used to determine the state of solar activity and the Interplanetary space include flare data in visible, X-ray and radio bands, and data on sunspot groups and integrated characteristics of solar activity. For penetrating radiation fluxes, the primary data come from measutpments on the AES "Heteor" system and supplementary data from stratospheric soundings, riometer data (USA), and neutron monitor data.oFor the magnetosphere environ- ment, operational_Jnformation are obtained from magnetic observatories and diagnostics from the AES "Heteor" system. 89 1980010188-100 A method of forecasting SCR flux evolution an :ntegrated flux for the whole event has been developed and Is currently used by the lAG Service. This method utilizes a model of lsotropic diffusion and SCR flux data from initial _ccords in polar zones on a "Meteor" satellite (Bezruchenkova et el., , 1976). To forecast proton fluxes, a linear relation is used, i In(I x t )/(2-B)) - -air + b, (I) which was obtained from a diffusion equation, where I is the proton flux density determined according to the special program of the operational control of SCR and galactic cosmic ray ,GCR) fluxes which was developed for the "MINSK-32 _l computer (program "Flux averaging in the polar zone"). ] = tmgas - tot where tmgas is the time of measurement of flux density in the polar zones. The SOlar source of particle injection is identified from heliogeophysical data operationally available in the lAG Service. The moment of proton Injection - t o is determined. InjecLion time can be deter-- mined from data on X-ray and radio wave emission, accompanying a proton flare, and optical observations in H_. The time of a hard X-ray burst is preferable. If such data are not available, the beginning of flare observa- tion in Ha is taken as t o . In case of recording fluxes from beyond-limb flares, the time of radio burst observation (especially type IV radio burst) is taken as to . The diagnostics of the current state of the radiation environment along spacecraft paths in near Earth orbits involves the use of observations of penetrating radiation fluxes in the AES "Meteor" orbit in high-latitude zones of the Earth's magnetosphere to estimate the general structure of particle fluxes, specifically in the orbits of manned space vehicles (MSV) "Saljut"-"Soyuz I'. This estimate should be primarily made for protons of solar cosmic radiation with regard to the real distribution of threshold geomagnetic rigidlties_ which depend on the extent of the magnetosphere disturbance. The observed energy spectrum of SCR particles in the polar caps, characteristics of the spatial distribution of cut-off rigidity and th_ location of the satellite making the measurements are used as the basis for estimating the radiation environment for another prescribed orbit. In view of the considerable time variations of th_ magnetospheric state, the solution of this problem r_quires a dynamic model of threshold rigidities. Such a model is analyti,ally _rltten as (Dorman et el., 1972) 15 F L-I 3 I._ R - O for L _ Lo. (3) In this expression parameters 6 and Lo include the general characteristics of magnetospherlc state. Lo is the polar cap boundary at the nightside of the magnetnsphere and 6 is quantitatively related to the magnitude of the Dst variation which reflects the ring current intensity. For quiet condi- tions Lo - 7 (or = 68o invarlant latitude) and 6 - O. An empirical relation given in Table I has been obtained by comparing the SCR proton observations 9_ $ 1980010188-101 I ° J with the value of Dst-varlation. The determination of the Dst-Variation | Table I Ring current Intensity parameter as an empirical function of D t. s i Dst(Y) = 8 • 16 26 hO 60 I00 160 270 380 6 = 0.0] 0.10 0.15 0.20 0.24 0.30 0.38 0.47 0.60 magnitude is based on the regularity found from the AES "Meteor" observations and reflected by regular displacement of the b_nd of cosmic ray latitudinal v_riatlon with the increase cf ring current intensity (Avdyushin et al., 1976). This relation can be quantitatively described by A L - (L H - l).io'2JDstl2/3 (4) where A L is the displacement of the bend from a quiet level, and LH is the quiet level of the bend in undisturbed magnetosphere. It shc _ be noted that the main parameters, such as the energy spectru_ of particles and char- acteristics of the distribution of threshold geomagnetic rigidities are obtained simultaneously from AES "Meteor" measurements. In the Institute of Applied Geophysics,_s model was used to calculate an operating catalogue of SCR proton inte!gra-] wspectra for a set of turns of MSV "$aljut"-"Soyuz" with different proton initial spectra in the polar caps and the magneto- spheric parameters _ and Lo. While further developing forecasting techniques consideration is given to both physically substantiated premises for forecasting an_ possibilities of operatio=Jal avaiiabillty of heliogeophysical phenomenon characteristics and parameters. Recent investigations have shown that photospheric magnetic fields are the significant factor determining the propagation of flare- generated protons. This was first established by Snyder and Neugebauer (1966) when analyzing the experimental data obtained on "Mariner-2" In subsequent works (Krimigis et al., i971; Roelof and Gold, 1976), comparison of SCR profiles at proton energies greater than one MeV in interplanetary space with equatorial high-coronal structure was made. It was found that low- energy solar protons are distributed over unipolar magnetic regions in the corona, and the interfaces of magnetic polarities are a serious obstacle to their longitudinal propagation. We analyzed the observed characteristics of of the solar protons with energies of 5-40 MeV recorded in high-latitude zon_of the Earth magnetosphere at altitudes 700_1OO0 km with the configura- tion_f large-scale magnetic fields on the Sun. The investigation of SCR events along with pho_,_spheric magnetic fields for the maximum and declining portions of the 20th solar cycle, as well as during the 21st cycle increase, made it possible to establish a relationship between the peculiarities of 5-90 MeV proton propagation and large-scale magnetic structure. The follow- lng regularities have been revealed: 91 1980010188-102 (!) The SCR time profile Is delermined by the temporal location of a flare coupling longitude, i.e. the solar longFtude from which the solar wind flows to the Earth, and neuti'al lines of the longitudinal solar magnetic field, which limit regions filled by flare particles. An equation was devel- oped that relates the decrease constant T to the time of a coupling-longitude displacement to the unipolar region boundary (A t) and unipolar region extent (D). In this case the decrease constant was calculated separately from the intensity maximum Jmax to Jmax/2 according to the following rela- tionship T I = (log D A t) 2"7 (5) and from the intensity maximum to the intensity at the unipolar region reoion boundary according to T2 = 1.6 (log D A t) 2"3. (6) (2) A sharp front of SCR increase up to the maximum was observed for the events in which SCR arrival was recorded after the coupling longitude had crossed the neutral force line. While describing the temporal distribu- tion of proton fluxes by an exponential function, the constant of an increase rate changed from !.2 to 2 hours. Times up to the maximum did not depend on the flare helioiongitude. (3) An increase of SCR angular distribution anisotropy up to 30 per cent was observed in the events in which the coupling longitude, after crossing the neutral force line, returned to the region filled by flare particles. (4) The analysis of SCR spectral characteristics variations showed an effect of spectrum hardening ahead of neutral force lines. (5) The beginning of a record or sharp increase of 5-40 MeV proton fluxes near the Earth was observed after the coupling longitude had crossed the neutral force line (when moving eastward) separating the unipolar region where protons were injected. A relationship between the spectrum index Yinit and Ymax and the decrease time constant T I is obtained for the events in which the coupling longltude and flare were in one unipolar region. The effect of the coronal structure on peculiarities of proton propagation with low energies during the solar cycle increase is established in Krimlgis et al., 1971. Assuming that the large-scale coronal structure is relatively regular during the whole cyc!e, the forecasting technique based o.; the r_gularities found in the 20th sunspot cycle has been applied to the events of the 21st cycle. :_ When analyzing the experimental data, the n_ps of photospheric magnetic fields published in Solar Geophysical Data (1973-1976) were used; their drawing technique is presented in works by Mclntosh (1972). The solar 1 synoptical maps drawn on the basis of the pictures taken in H_, reveal a 92 1980010188-103 iarge-scule structure of weak magnetic fields and feature continuity over large extent (Mclntosh, 1972). The heiloiongitude of the base of the force line connecting the Earth with the equatorial corona was calculated according to the procedure which was called EQRH by the authors Noite and Roelof (1973). This procedure allo_s one to estimate the longitude of coupling or that of the solar wind source by extrapolating instantaneous ideal spirals to the Sun. The possibility of such an approximation of the coupling longitude was theoreti- cally supported in Sakurai (1971) and Matsuda and Sakurai (1972), and it was shown in Nolte and Roelof (1973) that the extrapolation of the trajectory of plasma particle motion according to this procedure actually _st|mates the high-coronal longitude and latitude of a solar wind source. The relationship for calculating the Carrington longitude of the base of the Sun-Earth force line takes the form: I £ r g _o = ¢(t) + _ , (7) where _ - mean angular velocity of the solar rotation, ! r g - heliocentric radius of the detector, ¢(t) - Carrington longitude of the central solar meridian at t moment. The error in determining ¢o will be of the form: % = % " % = T j = (8) , ]= kv ! where Ar = r g - to, r ° is the distance from the Sun, where the solar wind is radial, and v = (r_ - r o)/(t - to), where to is the moment of the solar wind _utflow. Using the known peculiarities of solar proton propagation and taking into consideration the effect of the configuration of large-scale solar magnetic fields (Mikirova and Pereyaslova, 1977; Mikirova and Pereyasiova, 1977; Rikirova and Pereyaslova, 1978; Hikirova, 1977) on the characteristics of SCR proton fluxes, the technique has been developed to forecast the following proton event parameters: (a) decrease constants T for the fluxes of protons with energies E > 5 HeY;P (b) times of the maximum of SCR event intensity with E > 5 MeV;P 93 1980010188-104 (c) integrated flux of protons with Ep > 5-10 ReV and Ep > 40-60 MeV, _ normalized to the maximum intensity; (d) event beginnin9 and duration; and (e) indices of the integral spectrum rigidity at the intensity maximum Nmax and at the beginning of the event at the half-maximum level Nmax/2. Confidence intervals for reliability values _ = 0.9, _ = 0.95, correlation coefficients and the relative error in a forecasted parameter are given in T_ble 2 for all quantitative relationships involved in the technique. The flow chart of forecasting the SCR event parameters is illustrated in Figure 1. Examples of the time dependence of proton fluxes in the SCR events on September 17-20, 1977, January 7-10, 1978, and April 7-12, 1978, are given in Figures 2, 3 and 4. The analysis of the SCR temporal distribution is performed with the use of the maps of photospheric magnetic fields. Table 2. Proton Event Forecast Performance Proton Relative energy Reliability values Correlation error Parameter MeV _ = 0.90 _ = 0.95 coefficient per cent Decrease constant (hour) T I > 5 6 8 0.95 T2 > 5 7 9 0.90 45 Times of the max:mum (hour) • (class I) t > 5-10 7 8 0.80 70 max (class ll) tmax > 5 3 4 0.87 30 ,. Integrated flux F > 5-I0 0.2 0.3 0.89 50 log Nmax3600 > 40-60 0.3 0.4 0.78 Spectrum index > 10-30 1.0 1.1 0.78 90 Yinit > 30-60 0.7 0.8 0.94 > 10-30 0.9 1.0 0.79 80 Ymax > 30-60 0.7 0.8 0.94 94 1980010188-105 J Solar flare j_ data J H_ Synopticali solar maps J Forecasting time of/ Criterion of _j Conditions of SCR arrival at Earthi flare propagation near q and SCR event proton q the Sun durat ion1i ! Quantitative I ! Function of Iongitu-diagnostics of dlnal attenuation ! i proton flare Jmax i FORECAST _]U Decrease constant _ _ J Time of maximum L q intensity r- J Integrated proton LI q flux _J Indices of integral q spect ra Figure I. Flow chart of proton event forecasts. The characteristics of microwave radio bursts and their relation to the proton fluxes in SCR events were investigated as a function of the conditions in the emitting regions and injection source heliolongitude. The following results were obtained (Beiovskii et al., 1977; Belovskii et ai., 1978; Belovskii and Ochelkov, 1977; Beiovskii and Ochelkov, 1977; Belovskii and Ochelkov, 1978; Belovskii and Pereyaslova, 1977) and were used to develop the forecast techniques: (a) The relation between the maximum proton flux -Jp and peak radio burst flux for v = 9100 MHz and E > IO MeV was established. The correlation coefficient is 0.6h. p ; (b) The considerable dependence of the maximum proton intensity on the flare heiioiongitude (@) was determined. The mean attenuation co- efficient for protons with E > 30 and 60 MeV is P : K lO (9)p (c) An east-west asymmetry of the microwave burst intensity with the maximum in heliolongitudes of 30°-60°E was revealed. The attenua- tion coefficients of radio emission fluxes Kr (_) were obtained. 95 f 1980010188-106 \ q) ! I%\ " ! i \,.., ,,JO4 ,,, '_1_. _ _ toe _g ,M tV September, 1977 H_ - Synoptic Map September 1977 (_2.), IO 8 ..6_ . ._) ao 60 , _0 _O d_ JgO _0 _O I i I I I I , , I Heliographic Longitude Fir ....e 2. SCR temporal distribution with Ep > 15 MeV in the event of September 17-20,-1977. (d) The relations between the integral flux for the whole event, the : maximum proton flux and the time up to the event maximum was found. The techniques for forecasting proton event parameters have been devel- oped on the basis of radio observations in the centimeter range. i. The assessment of the probability of 60 MeV proton record based on the observations on frequencies 88OO and 4995MHz. The parameters used: J& - radio emission flux in the maximum; 96 1980010188-107 r, ), 5 HeV ,LO|! .", ,,, \ 'E I_" 'J _t " ""'" I r. \-' !e- _t • ? II 9 _ January, 1978 "* January 1978 _ g - 7 __L ?,_ _ _ _ ,= _.,o ( 9_i I ...1 9_o(,.., 9,__ Hel iographic Longitude Figure 3. Temporal distribution of proton fluxes with I E > 5 HeY in the event of January 7-10, 1978. l ._ P I - integral flux for the whole event determined as I = Jrd, where d is the effective duration. 2. The forecast of the lower limit of the flux of protons with E > 10 MeV in the event maximum on the basis of observations on • v p= 9100 MHz. Besides data on the flux, the flare coordinates are requi red. For the probability of correct forecast being 0.8tt, 0.89, or 0.92, the ,- value of the lower limit of 10 MeV proton flux is determined with regard to 97 1980010188-108 _0_ 0 I 0 • • • t m April 1978 April 1978 .i_ _ 9 s _ a s 4 3 2 ?ON ...... /I 06? " ":" • .. N..._;_.-F\ '""_'2 I :o t0 L .... L_.___ _ _ 0 ¢ . .:_ ._.+ "_ -._._ _ - _.,:.__o / / 4- / _ "::"" 350 3_ttO _0 _ _0 50 _0 ?0 80 90 JO0 _t0 Heliographic Longitude Figure 4. SCR temporal profile with Eo > 5 ReV in the event of April 7-12,ri978. the attenuation of radio burst intensity by introducing the coefficient K_(_) from the respective relationships: 7 1980010188-109 1.279 {log Jr max- 2.26 K_(¢)} - i.B6 log jpmaX = 1.279 {log Jr max - 2.26 Kr(¢)} - !.45 (IO) 1.279 {log j max . 2.26 K_(¢)} - 1.54 The data necessary for forecasting can be obtained from the station Zimenki (USSR) where the patrol is carried out on 91OO HHz with the aid of two identical radio telescopes. ]. Rapid forecast of the integral proton flux for the event. The data on the flux in the event maximum for protons with Ep > 30 and 60 HeV and the time up to the maximum are required. The integral flux of protons ! is calculated according to the equation presented by Belovskii and Pereyaslova (1977): I = Jmto= [TBO(I - T) + 0(_ - I)] (II) where T = t /t , J is the proton flux at the maximum, c - time up to theo m m maximum, O Is HeavisideSs function, and d, 6 and t o are numerical constants. The accuracy of forecasting I is not worse than 20-30 per cent. The data required for forecasting are obtained from the _JHeteor 8_ satellite. X-ray emission is a highly informative parameter for predicting the proton flux. Provided that the data on X-ray emission in the range of 0.5-4 _ and I-8 _ come regularly, it is possible to precaiculate the integral fluxes in SCR events before the protons arrive at the Earth's orbit, i.e. before the moment of the accumulation of the maximum dose. By applying regression analysis, the radiation characteristics of exposure dose and spectrum rigidity are forecast step by step. We shall not dwell here in detail upon the method, since this is clearly presented by Bezruchenkova et al,, 1977 and Pereyasiova et al., 1978. It should be noted that regular and timely availability of the following information is necessary to enable one to use the proposed method operationally: (a) more complete characteris- tics of X-ray emission bursts, (b) the maps of photospheric magnetic fields, (c) the solac wind velocity, and (d) flare heliocoordinates. As stated above, the propagation of solar protons is considerably influenced by the state of the interplanetary space. Therefore questions concerning the diagnostics and forecast of the parameters responsible for the state of the interplanetary space play an important role in forecasting the radiation situation. In a number of papers it was established that SCR characteristics are related to the direction of the radial component of the interplanetary magnetic field (IRF)(Pereyasiova et ai., 1976; Darchieva et al., 1973). The conditions in the interplanetary space are mainly governed by the solar wind density and velocity. The dependence of high-latitude magnetic disturbances on the magnitude and direction of the IHF (Bx, By, Bz) components and quiet solar wind (SW) 99 1980010188-110 velocity forms a physical basis for the forecasti_lg of the SW state. Energy Input to the magnetosphere is controlled by the seasonal Inclination of the dipole to the SW flux and reorlentation of the hello- and geomagnetic equator planes in the Earth orbit. Such geomagnetic disturbances are formed by the superposition of the three current systems: DP2, DP_ and DP4. Each of them depends on Bz < O; B. > 0 and the signs of Bx and B_, respectively. In the opinion of Frils_Christensen et al. (1971), ti:e DP4 currents are to a greater extent governed by ± By and to a lesser extent by ± Bx (IHF compo- nents). The maximum of the DP4 currents is localized at _0 _ 800 ' and the magnitude and direction are determined by the sign and magnitude of ± By (and ± Bx). The delay between the IHF variations and OP4 disturbances is 25 minutes, and this determines the threshold length of the IMF forecasting by the behavior of the geomagnetic field (GMF). By fluctuations measured in the ecliptic plane are known to produce the additional Bz component of IHF in the solar-magnetospheric system. This "addition" of A Bz < 0 is controlled by the dipole inclination to the SW flux. The IHF forecasting techniques based on the well-known Svalgaard-Manssurov effect (Hansurov and Mansurova, 1971; Svalgaard, 1968), should also take into consideration the seasonal dipole variations as a regular effect. The dependence of the high-latitude GHF variations upon the IHF sectoral structure is more pronounced for 6 Z and less pronounced for 6H. This structure manifests itself as the decrease or increase of 6Z by 200-300 x lO "s gauss in the N-circumpolar region with IHF directed from or towards the Sun. At the near-polar boundary of the daytime cusp the decrease or increase nf H is observed for the same IMF sign reversals. Below _0 = ± 750 the effects of the IHF on GMF are insignificant, and this fact determines the selection of stations for observing 6Z and 6H with the object of the IMF forecasting (stations Tule, Vostok, etc.). To realize the forecasts of ± By (or Bx) for the period of I-2 days, it is reasonable to provide the regular income of 3-hour data on 6Z and 6H from the stations located at _0 = 770 to 800 and near the boundaries of the daytime cusp. The preliminary testing of such stations with the use of 6Z and 6H data confirmed the feasibility of the ± By (or Bx) forecasts. The income of data from some stations by seasons is insufficiently operational, and the effects o_ Bz and By on the GHF are superposed. However, having enough information on the morning and evening DP4 current vortices (since By _ 0 with Bz < O) one can attempt to separate Bz and By effects. The technique of the 1t4F forecasting is based on the extrapolation of the gradients of 6Z deviations from the Z trend. The trend takes into account the level of seasonal and recurrence variations of Z (or H) in local warm seasons. The trend averaging takes into consideration the time scales of the Earth's and solar rotations and the maximum duration of IMF sectors _ 15 days. The rate of changes of 6Z deviation gradients from the trend is a measure of the approach of the IHF sign reversal. For long-range forecasting it is possible to use the method based on ;he results of the statistical analysis of the information on the solar p_oton flux characteristics. The peculiarity of the method is in analyzing separately the moments of proton flare occurrence and values of proton fluxes in each flare. This analysis has resulted in the determination of the distribution law for these values, besides the relationships have been found allowing to determine the simultaneous probability of the occurrence of events. 100 1980010188-111 The development and Improvement of efforts on direct diagnosis and forecasting are primarily directed towards the application of automated (recorded on machine processable media) global dynamic models of the radia- tion environment with a limited number of controlling parameters. The estimation and forecast of these parameters is performed on the basis of t_ operational computer-oriented data bank, which is updated with current observational results, actively using the means of information presentation. Models of the radiation situation for each phase of the solar cycle are assumed to be developed on the basis of the experimental investigations with the purpose of using them when forecasting the radiation characteristics. REFERENCES Avdyushln, S. I., Yu. H. Kulagin, E. L. D1ikman, V. A. Kuzmina, G. A. Kirdlna, A. B. Halyshev, V. A. Hokhova, and P. H. Svidskii (1976): Variations of threshold geomagnetic rigidities in the period March-Hay, 1976, and their connection with the ring current. Report UAG-6I. Collected data reports for stip interval II 20 Harch - 5 Hay 1976, p. 183. Beiovskii, H. N., and Yu. P. Ochelkov (1977): Zavisimost nablyudaemoi pikovoi intensivnosti soinechnykh kosmicheskikh luchei ot geliogeografi- cheskoi dolgoty. Soinechnye dannye, 1976, 12:73. Belovskii, H. N., and Yu. P. Ochelkov (1977): 0 dlagramme naprav]ennostl mlkrovoinovykh radiovspleskov Solnza. Doklady AN SSSR, tom 236, N. 6, 51. Belovskii, M. N., and Yu. P. Ochelkov (1978): 0 svyazi mehzclu parametraml protonnykh sobytii. Solnechnye dannye, 2:89. Belovskii, H. N., Yu. P. Ocheikov, and T. S. Podstigach (1977): t4etodolo- gicheskie voprosy prognoza protonnyh yavlenii po radiovspleskam Solnza. Tezisy X Vsesoyusnoi konferenzii po radioastronomicheskim issledovani- yam solnechnoi sistemy (1977, Irkutsk), 51. Belovskli, H. N., Yu. P. Ochelkov, T. S. Podstigach, and M. A. Fasakhova (1978) : 0 prognozirovani i protonnykh yavleni i po radiovspleskam. V sb.: Radioizluchenie Solnza, vyp. 4, izd. LGU, 44. Belovskii, M. N., and N. K. Pereyasiova (1977): Nekotorye parametry radio- vspleskov, svyazannykh so vspyshkami SKI. Trudy IP_G, vyp. 35, Gydrometizdat, H., s. 51. Bezruchenkova, T. H., N. A. Hikryukova, N. K. Pereyaslova, and S. G. Frolov (1977): Diagnostlka _[otonov solnechnykh vspyshek po soprovohzdayush- chemu electromagnitnom izlucheniyu. Geoma_lnetizm i aeronomiya, N. 5. Bezruchenkova, T. H., H. N. Nasarova, and N. K. Pereyaslova (1976): Primenenle model i izotopnoi diffuzii dlya prognozirovanlya razvitiya potokov solnechnykh kosmicheskikh iuchei v vysokoshirotnykh zonakh 101 t 1980010188-112 magnitosfery Zemli. Geoma_netizm I aeronomlya, tom XVI, N. 4, s. 592. Darchieva, L. A., T. A. Ivanova, E. N. Sosnovets, and L. V. Tverskaya (1973): 0 strukturnykh i dynamlcheskykh osobennostyakh pronlknoveniya ° solnechnykh kosmichesklkh iuchei v polyarnye shapki. Izv. AN SSSRr ser. , fiz., tom 37, 1313. Dorman, L. I., V. S. Smirnov, and H. I. Tyasto (1971): Kosmicheskie luchi v magnitnom pole Zemli. "Nauka '_, H. Friis-Christensen, E., K. Lassen, J. Wilcox, etc. (1971): Interplanetary magnetic field, Polarity inferred from polar geomagnetic field observa- tions. Nature Physical Science, Vol. 233, 38:48. Krimigis, S. H., E. C. Roelof, T. P. Armstrong, and J. A. Vcn Allen (1971): Low energy solar particle observations at widely separated points during 1967. J. Geophys. Res., 7.66, 5921. Hansurov, S. H., and L. G. Hansurova (]971): Svyaz mehzdu magnitnymi polyami kosmicheskogo prostranstva i Zemli. Geoma_netizm i aeronomiya, tom II, No. I, s. 115. Hatsuda, T., and T. Sakurai (1972): Dynamics of the azimuthal]y dependente. Solar wind. Cosmic Electr. dyn., 3:97. Hclntosh, P. S. (1972): Inference of Solar Hagnetic Polarities from H-alpha observations, Solar Activity Observations and Predictions, ed. P. S. Mclntosh and H. Dryer, HIT Press, pp. 65-92. Hikirova, N. A. (197_1_. Osobennosti prostran_tvennykh i vremennykh kharak- teristik SKI v sobytiyakh 19.IV 1972 i 5.V11974. Trudy IPG, 35:24. Hikirova, N. A., and N. K. Pereyaslova (1977): Vliyanie fotosfernykh magni- tnykh polei na osobennosti rasprostranenlya protonov solnechnikh vspishek. Dokiadi AN SSSR, 234, N. 4, s. 798. Hikirova, N. A., and N. K. Pereyaslova (1977): Effect of coronal fields on particularities of distribution of protons from the solar bursts. Abstract. Jaga/JAMAP Joint Assembly 1977. Mikirova, N. A., and N. K. Pereyaslova (1978): Osobennosti vremennykh i spektrainykh kharakteri_;ik solnechnykh kosmicheskikh luchei, obuslovlennye vluyaniemVsektornoi st struktury mehz planetnogo magnitnogo polya. Geoma_netizm i aeronomiya, 4:581. Nolte, J. T., and E. C. Roelof (1973): Large-scale structure of the inter- planetary medium. Solar Phys., 3._3:483. Pereyaslova, N. K., S. G. Frolov, and T. M. Bp=zruchenkova (1978): Diagnosti- cheskoe prognozirovanie radiazsionnykh kqllJ_racteristik solnechnykh vspyshek po parametram electromagnitnogo Izlucheniya. Geoma_netizm i aeronomiya, N. 6. 102 := 'T 1980010188-113 Pereyasiova, N. K., M. N. Nasarova, S. M. Mansurov, and L. 6. Hansurova (1976): Nekotorye zakonomernostl svyazi solnechnykh kosmicheskikh luchel v polyarnykh oblastyakh s sektornoi strukturoi mehzpianetnogo magnitnogo polya. Geomagnetizm i aeronomiya, tom XVI, N. 3, 407. Roelof, E. C., and R. E. Gold (1976): Prediction of solar ener_eti£ particle event histories using real-time particle and solar wind measurement. Preprint AFGL-TR-76-O136, Johns Hopkins Univ. Sakurai, T. (1971): Quasl-radial hyper-veiocity approximation of the azi- muthally dependent solar wind. Cosmic Electr. dyn., ]:460. Snyder, C. W., and H. Neugebauer (1966): The relation for Mariner-2 plasma data to solar phenomena. Solar wind, P_l_amon Press, 25. Solar Geophysical Data, 1973-1976.0 Svalgaard, L. (1968): Sector structure of the interplanetary magnetic field and daily varlation' of the _eoma_netic field at high latitudes. Inst. Geophys. Paper No. R-6, Danish Meteorol. Inst. Chartlottenlund. p. II. 0 @ 103 1980010188-114 tlIB,o-1B71 Q FORECASTING IONOSPHERIC AND GEOMAGNETIC CONDITIONS AT THE IAG FORECASTING CENTER S. I._yushin, A. D. Danilov, A. B. Ma3yshev, G. N. Novlkova, and P. M. Svidsky Institute of Applied Geophysics Glebovskaya 20b, Moscow 107258, USSR O A description of the Moscow Regional Warning Center activity in the field cf geomagnetic and ionospheric forecasting. % 1. INTRODUCTION Since 1975, the Institute of Applied Geophysics (IAG) has been serving as a forecasting center of the national ionospheric and geomagnetic service and as the Eurasian Regional Warning Center of IUWDS. To predict geomagn<, ic ac- tivity and ionospheric conditions, the methods developed earlier in IZMIRAN (Zevakina et al., 1975) are being used, although during the last four years some alterations and additions have been made. The _orecaling Center issues three principal types of predictions: i. Short-_rm predictions of ionospheric and magnetic disturbances; 2. Long-term map-like predictions, giving the MUF (maximum usable frequency) distributions all around the world; 3. Long-term predictions of the radlowave propa_tion along fixed paths In this work the main emphasis will be made on t_ description of the first type of forecastin, 3, because on the one hand thi_" type is the mo_t im- portant for customers, and on the other hand, short-term predictions are the most complicated ones, requirin G consideration of the whole set of helio- geophysical events. The methods of long-term forecasting are briefly de- scribed at the end of the paper. Short-term forecasting of ionospheric and magnetic disturbances is prin- cipally based on the predations of different kinds of solar activity. The Forecasting Center collects observational data on the current state uf solar activity and the interplanetary medium to produce its own solar fo_casts, and also receives solar predictions and geoalerts from the other IUW_e- gional centers. In this paper, in speaking about ionospheric and geOo_ neticforecasting, we wi_ assume that both solar predictions and cur/ent .a_ data are available. We will not discuss the problems of solar activity f_- casting. ' Although most ionospheric disturbances are genetically connected to geo- magnetic disturban-.es (storms, substorms, etc.), for our customers the most important are the different kinds of ionospheric disturbance_ because of their drastic influence on such important aspects of everyday life as broad- casting, radionavigation, and so on. Therefore, the main emphasis of this b--- . 1980010188-115 paper is the different types of ionospheric predictions produced by the Fore- casting Center. Collection and distribution of ionospheric data as _Ii as production of ionospheric forecasts comprise the largest part of its opera- tional activity. Any forecasting network should be based on current experimental data. In the case of magnetic predictions, the work of the center is based mainly on real-time data on three-hour K-indexes for Moscow (IAG), as well as on diurnal reports on ak-values an_ three-hour K-indexes from i0 Soviet stations and !0 stations outside the USSR. Ionospheric forecasting has a large bank J of real-time data, collected in the Forecasting Center. The heart of the bank consists of vertical sounding data from the Soviet network of iono- spheric stations. A list of the stations sending current informatien hourly is given in Table 1 (Sukhodolskaya and Shchuka, 1978), and the list of the collected parameters in Table 2. The center receives the vertical sounding data from 14 stations abroad also. Table i. List of the vertical sounding stations used in real-time service. 4, NN Name Station Type of number - telegram I. Alma-Ata 38401 II 2. Arkhangelsk 34601 I 3. Ashkhabad 36401 II 4. Dixon 38701 I 5. Irkutsk 40501 III 6. Kiev 33501 III 7. Volaograd 34505 I 8. Lenlngrad 33603 I 9. Magadan 45601 III i0. Moscow 34502 II Ii. Murmansk 33702 III 12. Novosibirsk 38501 II 13. Podk. Tunguska 39601 I 14. Salekhard 37701 III 15. Sverdlovsk 36602 II 16. Rostov 34506 I i 17. Tashkent 37401 II = 18. Tomsk 38601 III 19. Yakutsk 43601 III 20. Khabarovsk 43501 III 21. Krenkel 36801 I 22. Cape Schmidt 28701 III In addition to the vertical sounding data, the Forecasting Center re- ceives the following real-time ionospheric information: riometer observations in the high-latitude zone (Murmansk, Dixon, Krenkel, Amderma), data on sudden ionospheric disturbances (SID) observed in short wave fade-outs (SWF),and VLF phase anomalies (SPA). The collection of data is arranged on a regional basis. Together with the Forecasting Center in IAG, four more subcenters operate in Murmansk 105 1980010188-116 Table 2. List of ionospheric parameters sent to the Forecasting Center. Type of telegram Parameter I II III foF2--critical frequency of the F2 layer + + + h'F2--minimum virtual height of the F2 layer + + M-3000-F2--Maximumusable frequency factor for the F2 layer + + + fmin--minimum reqistered frequency + + + foEs--critical frequency of Es + + + foFl--critical frequency of the F1 layer + h'F--minimumvirtual height of the F layer + M-3000-F1--maximum usable frequency factor for the F1 layer + hpF2--height of the electron concentration maxlmum + foE--critical frequency of the E layer + h'E--minimum virtual frequency of the E layer fbES--blanketing frequency of the Es layer + + type of Es + + (Polar region), Khabarovsk (East region), Novosibirsk (Siberian region), and Tashkent (Central Asian region). The functions of the subcenters are col- lection an4 transmission to the center of current data in the region and pro- vision ho the local customers of current information and of short-term (less than one day) forecasts. All the data collected in the Forecasting Center are used both for im- mediate transfer to the customers and for producing and correctirg the short- telm predictions described below. 2. VARIATIONS OF THE IONOSPHERIC PARAMETERS Any detailed consideration of ionospheric aeronomy is beyond the scope of this paper; therefore in this section we mention only the main variations of the ionospheric parameters and the probable sources of their appearance. Electron concentration [el in the F2-1ayer maximum deterndnes the most important parameter of ionospheric forecasting, critical frequency (foF2), and undergoes several well-known regular variations. These are (I) diurnal variations, where [el decreases from daytime to nighttime following the in- crease of solar zenith angle X; (2) variations with solar activity, in which noon values of [el (under other equal conditions) are higher with higher solar activity index (radiowave flux at 10.7 cm, FI0, or Wolf number, W); (3) seasonal variations, due to which noon concentrations in winter are high- er than in summer (so-called "winter anomaly" in the F2 layer); and (4) some other less important but known re@ular changes in electron concentration. All such changes are being accounted for in long-term forecasts (these forecasts include all regular variations because they are based on'the analysis of the huge set of experimental data from the previous years) and in medians in short-termpredictions. Short-termpredictions themselves are 106 1980010188-117 needed because of the existence of irregular variations of the ionospheric behavior due to different solar and geomagnetic events. Immediately after a rather strong solar flare the X-ray fluxes with = 1-20A intrude the region of the lower ionosphere, producing a sharp in- crease of [e] at the D-region heights (60-90 km) in the entire sunlit iono- sphere. The increase of electron concentrati_)n leads to a strong increase of short-wave absorption as well as to the appearance of some anomalies in the propagation of low frequency and very low frequency radiowaves. The entire i set of these events is usually called "sudden ionospheric disturbances" (SID). The duration of SID may vary from several minutes up to an hour or more. If during the solar flare energetic protons with the energy of tens of mev have been emitted, several hours after the flare they enter the Earth's atmosphere in the polar cap region and produce strong enhancement of the D-region ionization. This event is called "polar cap absorption" (PCA). In its influence on radiowave propagation it is similar to SID, but has much longer duration, vezy often as iohg as several days. The flux of the kev corpuscular radiation (low energy plasma) emitted during the solar flare, as well as that borne in the flare shock waves, ar- rives in the vicinity of the Earth much later--several tens of hours--after the flare. The low energy plasma and the shock waves influence the Earth's magnetosphere, which causes different events: magnetic storms and substorms, "precipitation" of the particles from the radiation belts, increase of the ring current, and so on. They in their turn produce a whole set of dis- turbances in the ionosphere. As a result of the intrusion of kev-energy electrons, the electron density at 70-80 km in the auroral zone increases, which produces the event of "auroral absorption '' (AA). Simultaneously the heating of the upper atmosphere in the auroral zone takes place because of several factors (Joule heating, absorption of the corpuscular radiation, and so on), which causes a change in the neutral composition of the thermo- spheric gas. The [O]/[N2] ratio, which in the F2 region during daytime conditions governs the electron concentratiQn ([e] _ [O]/[N2]), decreases because of the temperature increase. The effect of the [O]/[N 2] decrease due to dynamic pcocesses, the exact nature of which is still obscure (mer- idional circulation, gravity waves) is transferred with diminishing amplitude to lower latitudes, leading to a corresponding depletion of [e] in the mid- latitude F2 region. In other words, "negative ionospheric disturbance" takes place. Negative disturbances oft_, follow recurrent magnetic disturbances, and are not connected directly with solar flares, but with the passing through the central meridian of the sun of the geoactive regions. The principal mechanism of the negative ionospheric disturbance in this case is probably the same, but the intensity as well as the duration and the velocity of the _i penetration to the lower latitudes might differ very much depending on the type of active region and the corresponding geomagnetic disturbance. Still very poorly understood is the nature of the other type of F2- region events--positive disturbances (enhancements of [e] and correspondingly of foF2). Often positive disturbances appear just prior to or immediately after the so-called "sudden commencement" (SC) of the magnetic storm detected by the ground-based observations. The possible mechanism of this disturbance is associated with the increase of charged particle fluxes from the plasma- sphere and with the changes in atmospheric circulation at the F2-region al- titudes. Fortunately, since positive ionospheric disturbances do not influ- ence the radiowave propagation in so drastic a way as negative ones, the question of their prediction is not so critical as the prediction of the latter. 107 t 1980010188-118 : A very important event in the ionosphere is the appearance of narrow layers of electron concentration at altitudes of 100-120 km, _e so-called sporadic-E layer (Es). Es may influence essentially radiowave propagation, causing an absorption (and sometimes a complete screening) of the short-wave radio signals reflected from the F2 region. The processes of Es formation are still very poorly known (they may be quite different for, say, the mid- latitude narrow layer formed by the wind-shear mechanism than for thick Es in the auroral zone formed probably by corpuscular ionization) and so there is no basis now for Es forecasting. 3. SHORT-TERM FORECASTS The Forecasting Center issues four types of short-term forecasts of ionospheric and magnetic disturbances: 1. Monthly 2. Five-days 3. Two-days 4. Semidiurnal. In short-term forecasting a very important device is a synoptic map of helio-geophysical events that is being currently built in the Center. A schematic example of such a map is shown in Figure i. The following informa- tion is presented on the map: active formations on the sun (spots, flares, etc.); proton fluxes with E = 40-500 Mev and X-ray fluxes with I = 1-8_; solar radiowave flux at 10.7 cm and 1.2 m, 3-hour magnetic K-index for Moscow; critical frequency deviations 6foF2 for Moscow; and intensity of the cosmic rays according to space monitor data. For short-term forecasting, maps of the 27-day recurrence of geo- magnetic and ionospheric events ("Bartels carpets") are also currently being built in the Center. Such maps are being built for the following parameters: the degree of disturbance of the geomagnetic field for Moscow station; critical frequency deviations _foF2 and intensity of Es for Moscow, Alma-Ata, Yakutsk, Murmansk, and Krenkel stations; and the probability of the appear- ance of absorption and diffusivity for Murmansk and Krenkel stations. An example of "carpets" for magnetic disturbances (Moscow) is shown in Figure 2. While preparing monthly forecasts with the help of corresponding Bartels carpets, the periods of probable geomagnetic disturbances are determined and the probabilities of different kinds of ionospheric events (foF2 disturbances, Es appearance, absorption) are estimated. Cazpets, as a rule, show the ten- dency of the development of the recurrent e_unts--decreasing or increasing-- and thus can possibly correct the prediction relative to the prrvious solar rotation (the previous line in the carpet). It is worth mentioning that, as is widely known, recurrent forecasting works reliably enough during the years of decreasing solar activity--at the beginning of a new solar cycle the recurrence is often distorted because of the permanent birth of new act.ve regions. Synoptic mapping allows correction of the above-mentioned recurrent forecast, taking into account the new active formatiols that have appeared on the sun during the pxevious rotation. Monthly forecasts, produced by the Forecasting Center, consist of pre- dictions (for each day of the coming month) of the following parameters: geomagnetic distu_'bances for Moscow; critical frequency deviations 6foF2 for Moscow, Kiev, Alma-Ata, Irkutsk, Khabarovsk, Krenkel, and Murmansk 108 1980010188-119 ***j*** ,', *** 0% ,,, ** "" "" 000 OeO leOe go 0 • • lore 000! 8oe nee oo ° t Ii It l J • * !l. I±! L, r . L ....!, .ILI l / Figure I. A schematic example of the synoptic map. • 1 - Eastern hemisphere of the sun 2 - Western hemisphere of the sun 3 - Solar radiowave flux at 10.7 cm 4 -Sclar radiowave flux at 1.2 cm 5 - Geomagnetic K-index for Moscow station 6 - 6foF2 7 - Cosmic rays 8 - Sign of the interplanetary magnetic field 9 - X-ray fluxes with _,= I-8A and proton fluxes with E = 40-500 Mev. 2 _* stations; intensity and probability of appearance of Es for Murmansk and * Krenkel stations; and probability of absorption and diffusivity for Murmansk t and Krenkel stations. Monthly predictions are issued on the 25th of each * month and are sent to customers by mail and to the regional subcenters by telegraph. ; The five-day forecast is a specification of the monthly prediction. The basis of forecasting remains the same as that of the monthly prediction, but emphasis is changed from recurrence (Bartels carpets) to the real events (synoptic maps). Both the events which took place on the sun during the pre- vious days and the forecast of solar activity are taken into account. 109 /- \. 1980010188-120 : 19?h I1'.I. I. I. [_1.I-I. l. I__._..-:-_.12eV///J.I. I _,/Y,f,_ t6_o _/_ _ m_1,_/1_!; :_" YIINI/_"i_el. I. I- I- I_dx/_ //Z/_. _ _w_ 1l- I. I. !_51.i.I. __i" I" [lli_il_IG_ll/dll_lll,_IlPtlA_1" i" II _ "_ I. I. I+el.I. i. I. Vf_/ll_•_II#IIX_ "7ll_tl_ilAill_._ _Pll#. I '1.1_9 +_ I.I , ._ rf/,__. _I/A.__ V_%I/A_tl. I,l-I_6_o" iXll#1li, i!" _°1I. _ !_I. •_ill_/#ll#/A I _1._/_i#77i_'_//71_//_ _2 <_ Figure 2. "Bartels carpet," 1974, geomagnetic disturbances, Moscow station. 1 - quiet 2 -weak disturbance _-_ _-3 3 - moderate disturbance 4 - strong disturbance _-2 If a geoeffective solar flare is expected, the forecast includes all the corresponding effects mentioned above (magnetic disturbance, PCA, AA, ! negative ionospheric disturbance). It is assumed that PCA events will be_in 1-10 hours after the flare _nd a geomagnetic storm 25-30 hours after the storm. When flares are relatively weak (not followed by the 4th type radio outbursts), PCA events are not predicted and the time delay for the magnetic storm is taken to be about 50 hours. Special attempts have been made (Frolov, 1979) to develop an algorithm , to decide whether during the flare a geoeffective shock wave has been emitted. Such an algorithm considers the characteristics of the X-ray and the radio- emission of the flare (class and brightness, times of increase and decay o_ X-ray emission, spectral characteristics of radioemission) to decide if a shock wave has been emitted and what is the expected time of its arrival to the Earth. The average time delay in such a case is about 50 hours and with the above-mentioned method it may be estimated with an accuracy of about i0 hours. : Magnetic storms may also appear 4-5 days after a geoactive region crosses the central meridian of the sun. The duration of the storm depends on the level of solar activity. During solar minimum it generally is 4-5 days, de- creasing with an increase of solar activity. The five-day forecast includes the following parameters for each day: magnetic disturbances; critical frequency deviations 6foF2 for 22 Soviet sta- 110 L 1980010188-121 i i tions and 3 stations outside the USSR (Wallops, 3uliusru, and Boulder); in- i tensity and probability of Es appearance for Moscow, Murmansk, Alma-Ata,Yakutsk, Krenkel, and Dixon stations; and appearance probability of absorption and diffusivity for Krenkel, Dixon, and Murmansk stations. In the five-day forecast the mean values of some parameters are included. These values are calculated each five days for the previous five days (run- ning median) taking into account only magnetically quiet days. These mean values serve as a forecast of quiet ionospheric conditions. All the devia- i tions (due to the appearance of some disturbances) given by five-day, two- ! day, and semi-diurnal predictions, are scaled from these mean values. The five-day forecast includes medians of foF2 and the M-3000-F2 coefficient for 22 Soviet stations and 3 stations outside the USSR (Wallops, Boulder, and Juliusru). The five-day forecast is issued by the Forecasting Center on days i 5, i0, 15, 20, 25, and 30 (31) of each month and is sent to customers by telegraph and broadcast. In the regional subcenters, at the request of customers, _he short-term ] (five-day) forecast of the radiowave propagation conditions for the fixed path is also issued. The forecast is based on two groups of data: mean values of foF2 and M-3000-F2 in the reflection point (if there is no iono- spheric station at this place, interpolation is made between the nearest sta- tions) and the five-day forecast of the deviation of Doth parameters from the median. As a result, the optimal working frequency (OWF) for this particular path for the coming five days in quiet conditions as well as the expected deviations 6(OWF) for each day are obtained directly on the computer. The next specification of the expected situation in the ionosphere and the geomagnetic field is made in the two-day forecast. This forecast is based on the same current data on solar and geomagnetic activity ds the five- day prediction. The two-day forecast gives a non-technical description of the disturbance degree of the ionospheric F2 region for five principal re- gions of the USSR: Polar cap (Krenkel), Auroral zone (Murmansk), European part (Moscow), East (Yakutsk), and Central Asia (Alma-Ata). The final specification of the ionospheric prediction is given in the semi-diurnal forecast. In producing this forecast the main emphasis is on the current magnetic and ionospheric data. When predicting the negative ionospheric disturbance the velocity o= movement of such a disturbance from the high to the middle latitudes is taken into account. Thus, during the daytime the first minimum on the foF2 curve (maximum of the negative distur- bance) appears in Moscow about an hour later than in Leningrad, and in Kiev an hour later than in Moscow. The time delays of 6foF2 minimum relative to the beginning of the magnetic storm for several stations, situated at differ- ent latitudes, are shown in Figure 3. It is seen from the figure that the average velocity of the negative disturbance expansion southward along the meridian is about i00 m/sec, which is close to the mean velocity of the meri- dional circulation at the F2-region altitudes (Kazimirovsky, 1978). In the southern regions of the Soviet Union the effects of magnetic storms as a rule are not registered in the form of pronounced negative or positive disturbances. The reaction in the low-latitude ionosphere is the appearance of instability of the F2 layer, when the critical frequencies suf- fer irregular variations with amplitude not higher than 15-20 percent. This fact shows that wave-like disturbances (travelling ionospheric disturbances, for example) generated during the storm in the auroral zone have obviously arrived at these low latitudes. 111 0 lg80010188-122 ++ |o'+. t - 5, 1975 io+. 9 - 10,, 19"1"j 4,% \\o+\++ \ \ \ \ +o +s U 0"+ '3 \ +o ..!m +.. C = 'V ,,o\ " '+ao\ _ .++\ oS _ o5 40 40' t, hOUPll t, hou.'ll 1;+_ > ' ;0 10 bel:J.n_tn/¢ bef_inrlir_ of the storm o1"the Jtozst Figure 3. Time delay between the beginning of the magnetic storm and the maximum of negative ionospheric disturbance for different sta- tions (numbers correspond to the station list in Table i). When there is no well-pronounced ionospheric disturbance the ionospheric behavior is predicted based on the comparison of the current observational data and the median given by the five-day forecast. The forecaster compares the predicted diurnal curve of foF2 and the experimental points and, based on the behavior of the experimental data, makes corrections to the mean values to produce the forecast for the next 12 hours. The semidiurnal forecast includes the following parameters: for Polar Cap (Krenkel) and Auroral Zone (Murmansk) stations--6foF2, intensity and ap- pearance probability of Es, absorption and diffusivity; for European area (Moscow), East (Yakutsk), and Central Asia (Alma-Ata) stations--6foF2, in- tensity and appearance probability of Es. The semidiurnal forecast is broad- cast twice a day. The verification of short-termpredictions is rather difficult. First of all, whether a forecast proves to be true depends on the method of its estimation. There is no commonly accepted method and creating one seems to be a rather difficult task both because it is very difficult to determine the degree of coincidence between the forecast and the real situation in the iono- sphere and also because the deviation of the real situation from the predic- tion in different directions (for example towards higher or lower values of foF2) have often quite different meanings for different customers. + Table 3 shows the percent of different types of short-term predictions (described above) proved to be true for two months of 1976 (solar minimum) and / two months of 1978 (increasing solar activity). The principle of verification : 112 # _,_.\ _+ lg80010188-123 t .._._.._.._N,.,La.L.--- ....... J..J' " _ .... Table 3. Verification of the forecasts of different types (monthly average in percent) September 1978 September 1976 July 1978 July 1976 Magnetic forecasts for Moscow Monthly 63 87 68 100 Five-day 87 91 55 100 Diurnal 92 95 87 100 Ionospheric forecasts for Moscow Monthly 82 82 97 100 Five-day 83 82 93 100 Semidiurnal 95 95 93 100 is the one suggested ana zormerly used by IZMIKAN: the forecast for uhis par- ticular day and for this particular parameter is considered proved true if the difference between the forecast and the observation does not exceed one step. As one can see from Table 3, in the years of decreasing and minimum solar activity, the percentage of the forecasts based on recurrence (monthly, five-day) proved true is essentially higher than during the increasing phase of the cycle. This difference confirms the above-mentioned fact that the recurrence "works" well during the second part of each solar cycle and poorly in the beginning of a new one. For the shorter predictions the difference is smaller and evidently reflects the higher probability of error in the forecast when the number of events to be predicted is higher. 4. LONG-TERM FORECASTS AS mentioned previously, the Forecasting Center issues two types of long- term prediction: map-like forecasts of maximum usable frequencies (MUF) all around the globe and forecasts of radiowave propagation along fixed paths. Both types of prediction are based on the work done by IZMIRAN (Chernyshov and Vasilieva, 1973). In that study, data collected during 15 years (up to 1971) from the world-wide network of ionospheric vertical so_._ngs (129 ", stations) was analyzed. This analysis allowed a description of the global distribution of the F2-parameters in space and time in the form of a set of spherics functions. Corresponding series coefficients for these functions (Chernyshov and Vasilieva, 1973) provide the basis of _ong-term predictions. Monthly forecasts of MUF are issued by the Center 4-6 months in ad- vance and contain maps of the distribution all around the globe of two main parameters, the critical frequency of the F2 layer for vertical incidence of radiowaves, foF2, and the critical frequency of the F2 layer for oblique incidence of radiowaves for the distance to the reflection point equal to 4000 km, MUF-4000-F2. The maps of these two parameters allow calculation, using the well- known formulas, of values of MUF for any given distance if the radiowaves are reflected from the F2 layer. The maps of foF2 and MUF-4000-F2 are given for even hours of Moscow time. Figures 4 and 5 show examples of the maps. 113 i" 1980010188-124 N II0 IN m N _ N N N _ _S0|m H N _ U N vo N N IN .0 IN IN I_ ON 0U _Y001NU_f0 ON iN I_ N! W M N M R N M 1O m N N N 1O 0 00 ¢ 00 I0 N N N N 6 u _ u _ N N N N I_ I_ II0 I_ _ _ 70 _0 _ I0 N N O0 Figure 4. An example of the critical frequency (foF2) distribution (in Mhz) around the globe (long-term forecast; February 1977, 1400 Moscow time). When r_diowaves are propagated a small distance, the reflection from lower layers (FI and E) may appear to be essential, so in the monthly long- , term forecast, figures for MUF determination in cases of Fl-reflection (a distance less than 3000 km) and E-reflection (a distance less than 2000 km) are also presented. The key parameter in long-te_ forecasting is the smoothed Wolf number of solar spots, W, which is forecast for the center by the Main Astronomlcal Observatory (Pulkovo) six months in advance. I_ng-termpredictions of radiowave propagation along fixed paths are made on the request of customers on the basis of the same methods. The pre- dictions contain diurnal variations of MUF and lower usable frequency (LUF) for the particular path and the given month. The variation of LUF is cal- culated for the fixed transmitter power of 1 kW. An example of prediction of this kind for the path Moscow-Yakutsk (March 1979) is shown in Figure 6. However, LUF appeared not to be a very convenient parameter for the customers, so now work is being done to substitute LUF by more informative parameters, which characterize the expected weakening of the signal, or its amplitude directly. 114 1980010188-125 Figure 5. An example of MUF-4000-F2 distribution (in Mhz) m_ the globe (long-term forecast; February 1977, 0400 Moscow time_R_Lm" In conclusion, two points should be noted: _Q_'_ i • All the work on long-term forecasting in the center is done on com- ll,lliL. puters. • All the long-term predictions, as was already mentioned, giva a mean picture of the forecast behavior of the parameters and thus reflect only regular variations (with latitude, longitude, season, etc.) of the ionospheric layers. 5. CONCLUSION The above-described activity of the Forecasting Center provides the necessary current information and forecasts for most of the customers. But to increase the amount of distributed information and to improve the reli- ability of the fozecast_ further development of the observational systenl and the forecasting methods is greatly needed. Principally, the new method of forecasting should be based not on the previous year's statistics, but on the physical models of the ionosphere (Vlasov, 1979). Presently in the Forecasting Center, appraisal of the I15 1980010188-126 lloooow t2me, hott._. Figure 6. An example of a long-term forecas_ for the fixed path (Moscow- Yakutsk, March 1979), upper lines--MUFi lower line--LUF. specially made physical models is being made. The models are being checke_ from the point of view of the possibility of using them to produce different kinds of forecasts. Evidently such models may be used to predict regular varia!'t_sns of the ionospheric parameters (long-_erm forecasts) even now. Howeve-, _ 3joon as we get more J formation abouu the aeronomical parameter variations _ring ionospheric disturbances, such models may be involved in the forecasting of disturbed conditions (short-term forecasts). There are very few methods, if any, of D-region state forecasting. At the same time, such forecasts are more and more urgently needed. Presently there is even no real-tAme information about the lower ionosphere. In ver- tical-incidence soundings, the only parameter connected with the lower iono- sphere is fmin (minimum observable frequency), which is not reliable enough because it depends on the characteristics of each particular ionosonde. The Forecasting Center plans in the immediate future to receive real- time information on the electron concentration profiles in the D region near Leningrad, obtained by _he partial reflection method. Having obtained these |]6 O 1980010188-127 O% data, the Center would be able to provide customers with information on the current state of the lower ionosphere and also to start developing methods for its prediction. Forecasting of the lower ionosphere (D region) may appear to be some- } thing quite different from F2-region forecasting because of the strong in- ! fluence on the D-region behavior of meteorological factors. According to modern views this behavior is controlled by two gro_s of parameters: usual geophysical parameters (solar activity, magnetic activity, local time, and so on) and meteorological parameters (stratospheric warmin_s, circulation i at stratospheric and mesospheric levels, thermal regime of the mesosphere, and so on), which makes the problem of D-region forecasting a real burden. Vertical sounding is still the principal method of getting information i about the F2 region. To increase the amount of this information, several additional steps should be made: developing new ionosondes with better po- tentials; t,_ing the network of oblique sounding stations; and launching an ionosonde on-b_ a satellite. In the latter case, the system of trans- ionospheric sounding may be created, where the signals of the ionosonde are being received by the ground-based network of the ionospheric stations, thus obtaining the real-time distribution over a large range of critical fre- quencies for different distances. REFERENCES Chernyshov, T. V., and T. N. Vasilieva (1973): Forecast of m_imum usablefrequencies, Nauka Publishing House, Moscow (in Rus. Jn_ Frolov, S. G. (1978): Doklady Academii Nauk 243, N. 3. Kazim_rovsky, E. S. (1978): The dynamics of the quiet thermosphere. Review paper presented at STP Symposium, Innsbruk, May 1978. Sukhodolskaya, A. N., and T. I. Shchuka (1978): Submitted to Bull. INAG. Vlasov, M. N. (197g): Paper presented at this workshop. Zevakina, R. A. et al., (1975): Methods of short-term forecasting of magnetic activity and ionospheric conditiors. Preprint I_4IRAN N055(02) (in Russian). 117 1980010188-128 _, LONG TERR SOLAR ACTIVITY AND IONOSPHERIC PREDICTION SERVICES RENDEREDBY THE NATIONAL PHYSICAL LABORATORY, NEW DELHI B. H. Reddy, S. Aggarwal, D. R. Lakshmi, S. Shastri and A. P. Hitra Radio Science Division National Physical Laboratory New Delhi - 110012, India ! The Radio Science Division (RSD) of the National Physical Labora- tory is rendering solar and ionospheric prediction services since 1955. RSD is one of the few groups in the world with independent solar prediction activity essentially intended to improve the quality of ionospheric communications. The scope of this paper is to project the data base used in the prediction exercise, to de- scribe the prediction technique and to compare with actual observa- tions. No examples of specific circuit predictions are described, for they are derived routinely from contour outputs with the aid of standard nomograms and conversion tables. Emphasis is laid on Che data base and the technique itself so that the present state of art can be appreciated and critically examined. Future plans include more computer-oriented solar predictions and computerisation of all link predictions including required radiated power, field strength and lowest useable frequency calculations. I. INTRODUCTION One of the basic objectives of the Radio Science Division (RSD) of the National Physical Laboratory, N. Delhi is to predict the radio environment and advise on communication problems. These include ionospheric and tropo- spheric radio communications. This paper is purported to summarize our efforts in predicting the ionospheric communication parameters only. Of the various ionospheric regions, the D-region is relevant only in estimating the attenuation, while the E layer is more systematic and easily predictable. This leaves us essentially with the F-region with its two mos_ important communication parameters, the critical frequency and the heighc factors. Some work has been done at RSD in predicting the sporadic E layer also, but that is not included in this presentation. ¢ The charged particle population in the ionosphere naturally depends upon the neutral atmosphere which presents the ionizable constituents and on the ionizing from the sun (particle effects are also important at high latitudes). While all radiations including the visible and infrared may be responsible ll8 : t 1980010188-129 for the neutral atmospheric profiles, the ionizing radiations are essentially :. the EUV and X-ray radiations which undergo large variations with the solar _' epoch, The ideal solution of course would be to measure these ionizing radia- tions and then predict their levels for the future years. These measurements are impossible on a routine basis. One alternative, therefore, is to monitor the sun using those radiations (like the VHF and SHF radio waves) which reach the ground-based seasors unhindered by the intervening terrestrial ° atmosphere and correlate these measurements with ionosphere producing radia- tions. However, because of the availability of long series of observations, the sunspot number has been found by us as well as by other groups in the world, to be most suitable for statistical prediction techniques. The solar activity study, which was started 200 years ago out of scien- tific curiosity, had assumed great importance by the second quarter of this century because of ionospheric communications. Since recent years, the interest in this study has sky-rocketed because of its potential applications in a wide variety of areas of human interest, such as environmental effects and biological aspects, sun-weather relationships, satellite lifetimes and • manned space flights. It was found necessary to start our own solar predic- tions at RSD because of the special requirements of different user organiza- tions in the country. For example, it is required to predict the activity at different advance intervals like one year, two years or three years to enable long-term planning and frequency clearance from the appropriate governmental authority. Though the ionospheric prediction technique used at RSD is not very different from other techniques, the wealth of data that has gone into the system and the continuous updating as new data pours in, has instilled a certain measure of confidence among the users in the East zone and especially in India where H.F. communications still remain as the backbone of long distance communications. This brings in enormous pressure in conserving the H.F. spectrum and consequently the assignments of frequencies to users is i rather involved and time-consuming. This is one reason why short-term fore-C , casting was not very much in demand until recently. We are presently giving short-term forecasting also, but those details are beyond the scope of this presentation. 2. SOLAR ACTIVITY PREDICTIONS Ever since cyclical variations in solar activity were recognized, there ' have been many attempts to predict them. When they were found to have close - correlation with ionization density in Earth's upper atmosphere, the studies gained impetus from their practical application to radio communications. Sunspots, which are seen as dark patches on the photosphere of the sun, are centers of high magnetic activity that inhibit convection. Many characteris- tics of the sun such as sunspot number, sunspot coordinates, sunspot area, flare br; ntness, radio emission level, etc., are observed directly and depict solar activity changes of long- and short-term durations. However, sunspot numbers have the longest series of observed data available in a standard coded form, and this is the single most important reason to continue sunspot numbers for prediction purposes in spite of its subjective nature. Sunspot number prediction methods to-date have been based on analysis of past data as a function of time, the prediction being an extrapolation of the 119 1980010188-130 statistical analysis to some future time The generally accepted series of observation extends back to over 200 ye_ and when this data is separated into twenty II-year cycles, the variabili=y of these cycles is found to be so much that each cycle seems to be unique in some way or the other. 2.l Data Base and its Salient Features 2 The basic data that supports our long-term solar activity predictions is naturally the Zurich ser;es of sunspot data dating back to 1749. This series has been developed by Rudolf Wolfe in 1848 when he introduced a sunspot num- ber formul_ and applied it to old records of observed data retrospectively from 1749. There have also been some attempts (Shove, 1955) to obtain maxima and minima for sunspot cycles for periods earl ier than 1749 by using other indirect geophysical observations like Aurora. Considerable efforts have gone into keeping the method of handling observations of sunspot data as nearly uniform as possible and just this feature of the Zurich series makes it particularly useful as a forecasting tool. Other geophysical observations have much shorter histories. Even the ionospheric electron density observations cover something less than four solar cycles. Because these other direct observations cover so few cycles, it is almost impossible to identify and evolve from them any statistical relationships among successive cycles. Instead, it has been noted that during their short histories they exhibit a high correlation with the sunspot number data. It is generally assumed that observed correlations will be equally good over the long-term. In order to study the ]ong-term changes in sunspot activity, the data is smoothed to 12-month running average numbers defined as below: 1 [ _----n+5 R Rn L JFRI21 = 1-2 L _'-- Rk + n-6+2 +6° n n-5 for the month n, where Rk is the monthly average sunspot number of the month k. It is this smoothed sunspot number RI2 with which most of the ionospheric and atlnospheric density parameters show a high degree of correla- tion. The main feature observed in these long series of observations is that besides the II-years cycle, there is another cycle of period 78-to-80 years in the peaks of the cycle; but cycle no. 19 has been very unusual in all its\ aspects suggesting that perhaps there is another periodicity of a longer duration. Different methods have been used at different times to sort out : the different periodicities in this series. Cole (1973) has used power spectral analysis for detecting the significant periodicities in relative sunspot numbers. The resultant spectra reveal that the mean cycle length of il.06 years and the 190-year phase modulation are a resultant of two periods of 11.8 and 10.45 years representing a characteristic free running time for the solar cycle and a well-defined periodicity which excites the new solar cycle. Kuki in (1976) has identified several cyclic, secular and super " secular variations in solar activity viz. Ii, 22, 80-90, 1000 and 1700-1900 year periodicities. 120 1980010188-131 2.2 RSD Prediction Procedure A number of mathematical and statistical schemes were evolved in this century to arrive at some empirical relationships to predict the sunspot number (Waldmeier, 1946; Stetson, 1947). All of them basically revolve around the game of guessing based on the general pattern of recurrences dis- cernible in the available series of observations. The sunspot prediction activity was started at RSD during 1955 when the 19th cycle was in progress and the predictions continue ever since. The general principle involved in making a pr_.diction is to predict an average cycle. Instead of mechanically averaging all the previous cycles, the emphasis is laid on the known long-term variations in arriving at the average cycle. Statistical tests are made to identify and evaluate the validity of past cycles and those that belong to an entirely different statistical popu- lation are rejected (see for example Slutz et al., 1971). As the cycle progresses, all the observed values from the start of the current cycle are used to update the pattern of growth. Since these constitute standard _i statistical procedures, the details are not worked out here. However, some empirical relationships were also evolved from the then available eighteen cycles and these relationships were used in the predictions. , Some of the main features observed in the first eighteen cycles which were used as a basis for evolving the empirical relationships are: a) There is an inverse relationship between the peak value of the cycle and the growth period. b) Though the cycles with high peaks rise sharply, the decay is slower, thereby maintaining the total duration of the cycle approximately constant. c) As a corollary to (a) above, the area under the sunspot curve from minimum to maximum will be approximately constant. d) The initial gradient of a cycle has a definite relationship with the peak of the sunspot cycle. It may be seen from Figure I that a fairly significant straight line relationship exists between the peak value of the sunspot number and the corresponding reciprocal of growth period (plotted as ordinate on the left side as the reciprocal of growth period in months). However, cycle nineteen was an exception and was not bound by this simple relationship. Once it is known that we are in the ascending phase of the cycle, a small right angle triangle is formed in the new cycle with the mini_num and the most recently measured value forming the two ends of the diagonal. A triangle is also formed for the previous cycle with the previous cycle with the minimum and : the peak of the cycle constituting the two extreme points of the diagonal. With the aid of these two triangles and employing the empirical relationship that the area under the sunspot curve from minimum to maximum is approxi- mately constant, the growth period and the peak of the new cycle can be predicted. Of course, the periodicities other than II years are also kept in _ 121 1980010188-132 j: .04 I I I 26 @ - .08 o _ • D .0! 40 t0 eO _0 i20 NO tiiO tllo : PEAK OF ANNUAL AVERAGE SUNSPOT NUMBER Figure i. Reciprocal of the growth period (in months) plotted against peak of annual average sunspot numbers for first eighteen cycles. Period of growth, d, (in months) is also shown on the right. A linear relationship can be seen between peak : sunspot numbers and (I/d). mind while making the final estimate of the peak of the cycle. These esti- mates are updated every month using the most recently observed data. The characteristic cumulative growth rate from the start of a cycle in successive steps of 6 months is plotted against the peak sunspot number (Figure 2) for the first eighteen cycles and this empirical relationship is used to construct the growing part of the cycle as new observations are added. Once an identifiable peak is reached in the current cycle, the earlier • cycles are scanned through to pick up cycles of approximately the same peak value as of the current cycle (within ± 20). These cycles are then super- posed on the peaks and an average cycle is evolved which is used to predict the declining phase of the cycle. For example, while predicting the declining phase of the 19th cycle, the average of 2nd and 18th cycles in the series were considered, For these happened to be the two high-peaked cycles in the • Zurich series. A comparison is made of the preceding twelve observed values of the current cycle with those predicted from the average cycle. The ratio of the observed values to the values predicted for the declining phase from the average cycle are found to be approximately constant for about I2 months; and this feature is used to predict and update the declining phase continu- ously. As the trend of the declining phase emerges out more definitely, once again appropriate cycles are chosen from the past cycles to match with the now-emerging trend. This procedure is repeated as often as is found necessary and new adjustments are effected in the average cycle and the prospectivei minimum. Obviously, near the peak and the grough where the gradients almost disappear, prediction is rather speculative and no reliable method exists to improve this situation. Comparisons of the RSD predictions and Zurich predictions are presented in Figures 3(a) and 3(b) for cycles 19 and 20 respectively, along with the observed values. It may be noticed that during the unprecedented high ! 122 F a 1980010188-133 l I ooo B!/I /i • i _J/A_'ai_ -':" '.N_==.,='- ,_.;-O 30 SO 70 tO ilO 130 I_ 170 1110 : PEAK OF ANNUAL AVERAGE SUNSPOT NUMBER _ Figure 2. Gradients in sunspot numbers at different stages (6, 12, : 18, ......... months from the minimum) for growing part of solar cycles are shown against peak sunspct numbers. - o YEANS-'_ Figure 3a. Running average of Zurich observed sunspot numbers for _, nineteenth cycle with RSDand Zurich predictions. Near ] the peak, RSD values Fared better. activity of cycle 19, RSD method fared better. Figure 4 shows the entire predicted twenty-first cycle using just about one year of observations in the cycle. 2.3 Future Plan All the present methods of long-term solar forecasting essentially cen- ter around various forms of statistical analysis of previous cycles. In addition, furtner improvements can be effected by following the growth patterns of the active centers on the solar disk. These centers start lZ3 t 1980010188-134 l IOULlt ¢VC_[ Im i0P _ |_lCd ¢OIMtV|O ),*m_ 000 _$0 IPR_CIXO| i. ! , AliiB4dilNm_ Valtq_?4.MiU 4AN69 _UITO _tltT_ _tll_'| aAN,Y) _NT4 4AN?S .,tNT_ Y[ARS'---'_ Figure 3b. Running average of Zurich observed sunspot numbers for twentieth cvcle with RSD and Zurich predictions. .m..., , .. _) T SOLAmCYCLE_ lu 03 LSO Is "h, T ([W_I_SPOR_OI_I¢_ ._''_ ',_ -- ,U,'C.._ 0iS(RYE0t. '_ i t .... ._((_crto cycLE /S0 YEA_S Figure 4. Complete prediction of the twenty-first solar cycle using RSD technique is shown in this diagram. Error bars show 70_; probability ranges. developing around 40-450 latitude and drift towards the equator as the cycle progresses. Yet another plan is to use the 22-year periodicity more effec- tively. For example, one can start with an odd cycle and treat every even cycle as an anti-cycle and predict it as a continuation of the decaying part of the odd cycle. The negative values thus obtained are converted back to positive values, constituting the growing part of the even cycle. The cross- over point from positive to negative values can be readily identified as the minimum. Likewise, odd cycles can be predicted by starting with an even cycle. Standard odd and even cycles are evolved by averaging these cycles separately from the year 1749 A.D. Analyses of results arrived by this technique are too preliminary to be included in this review. There are several other simultaneous attempts being made to improve predictions of different phases of the solar cycle. One such attempt is to evolve an exponential curve to predict the declining part of the cycle. Yet another effort is underway to predict the growth part by using the peak-to-peak range rather than the minimum-to-minimum range. 124 t 1980010188-135 !' 3. IONOSPHERICPREDICTIONS Radio Science Division is responsible for advising the various organiza- tions in India on problems related to radio communications and for providing long-term predictions of propagation parameters for planning of high fra- quency communication circuits. RSD provides these predictions for the entire latitude range in the eastern region covering a longitudinal extent of 50°E - 170°E. The Ionospheric parameters viz. foF2 and MUF(4OOO)F2are pre- dicted for each of the calendar months. The predicted parameters are pre- sented in the form of contour maps and the values can be read for any latitude for all the 24 hours of the day from the map. Figures 5 and 6 show the con- tour maps of foF2 and MUF(hOOO)F2for September '77. The prediction technique is basically similar to the method developed earlier by CRPL, USA (1948) and used by several predicting agencies with some modifications. So, the method will be described only briefly; however, the data base will be described to show the bulk of observations used in the prediction exercise. ' 3.1 Data Base 3.1.1 Diurnal Variation for Fixed Solar Activity Ranges The measured ionospheric data from 27 stations (Table i) obtained over a period of two solar cycles is the basic data that Is being used in RSD pre- diction technique. Data pertaining to 13 more Russian Stations (Table II) has been added to the original data base at a later stage. The incorpora- tion of data from Russian Stations has ensured a good latitudinal coverage in the Eastern zone, especially in the high latitude region. The data comprises of monthly median values of foF2 and M(3000)F2 for all the stations for each month and hour. Median values are compi led for these 40 stations, using the observations of &bout two solar cycles. The data is LOCAL MEAN TlU[ (Mr|) -L -+- ; Figure 5. A contour map of foF2 in MHz for east zone (50°E-170°E) for September 1977 is shown. The contours are drawn _ for intervals of I MHz. 125 _' _ _" _ 1 1980010188-136 .*" i_ I 1 _"k |,,, : i ' i ktl 1 I r'_ j 1 i t 1 .4- LOCAL 011JAN?IU[ {Nrs) ..I,-. Figure 6. A contour map of MUF(4OOO)F2 in MHz for the east zone (50°E-170°E) for September 1977 is shown. The contours are drawn for intervals of 4 MHz. Table I. Ionospheric observations for two solar cycles. Stations Latltude Longitude East(0) (0) 1. Wakkana i 45.4 141.7 2. Akita 39.7 140.1 3. Tokyo 37.5 139.5 4. Yamagaw_ 31.2 130.6 " 5. Delhi 28.6 77.2 6. Formosa (Tapei) 25.0 121.2 7. Ahmedabad 23.0 72.6 8. Haringhata 23.0 88.6 9. Bombay 19.0 72.8 I0. Baguio 16.4 120.6 I1. Madras 13.1 80.3 12. Tiruchi rapal 1 i 10.8 78.7 13. Kodaikanal 10.2 77.5 14. Trivandrum 8.5 77.0 15. Singapore I .3 103.8 : 16. Townsvi Ile -19.3 148.7 17. Brisbane -27.5 152.9 ,_ 18. Watheroo -30.3 115.9 (Mundaring) (-31.99) (116.3) 19. Canberra -35.3 149.0 20. Hobart -42.9 147.3 21 Macq. Isld. -54.5 159.O 22 Clyde 70.5 291.4 23 Van imo -2.7 141.3 24 Cocos. Is. -12.2 96.8 25 Port Morseby -9.4 147.1 26 Hyderabad 17.4 78.5 : 27 Manila 14.6 121.0 _ 126 1980010188-137 Table II. Soviet ionospheric observations. Stat ions Latl tude Lon_l tude East(o) l, Heiss Is. 80.6 58.00 2. Ti khaya 80.0 52.00 3. Dixon Is. 73.5 80.4 4. Tixie Bay 71.6 128.9 ' 5. Salekhard 66.5 66.7 6. Yakutsk 62.0 129.6 : 7. Syerdlovsk 56.7 61.1 ! 8. Irkutsk 52.5 104.3 _ 9. Yuzhno Sakha] insk 47.0 143.0 I0. Alma Ata 43.2 76.9 : I !. Ashkhabad 37.9 58.3 = 12. Hirr7 -66.5 93.00 13. Vostok -78.4 106.87 i grouped into three sunspot ranges viz. 0-50, 50-100 and > lO0. Using the hourly median values for each station, a diurnal variation curve is obtained for each month and for each of the sunspot range groups. These standard diurnal curves are obtained for both foF2 and H(3000)F2. Figure 7 shows the standard diurnal variation curves for foF2 and M(3000)F2 for Delhi for the month of January. Thus, the observational data base consists of 480 (40 x 12) D[LNI JANUARY e • _RASSN 0-50 34 4 52 I , " A I $ 4 i_ -as _s= II • $ 4 IIA|IIN > *00 I0 IL 2 ; • 30 I ".1 , ,, (b) 4 OIlI I I I IO0 04 Oil II Ill |0 14 O0 0 Ise IS |O |4 LOCAL MEAN TIME (Nfll) LOCAL MEAN TIM[ (Hft) Figure 7. Standard diurnal variation curves for (a) foF2 and (b) H(3000)F2 for Delhi are shown for the month of January. Three curves represent three different ranges of solar activity. 127 .; *, 1980010188-138 diurnal variation curves for each of the three solar activity ranges for foF2 and an equal number for H(BOOO)F2. A 24-hour average value is also computed for each of these standard diurnal curves. 3.1.2 Solar Activity Variation Plots of 24-Hour Average Values of foF2 and H(3000) F2 Using the hourly medTan values, the 24-hour average of the medians is obtained for each month and station and this single value is chosen as the ionospheric Index for that particular month for the corresponding running average sunspot number (RASSN). Relationships are obtained between this ionospheric index and the RASSN for each station and month for both foF2 and H(3000)F2. Sample plots are shown In Figure 8 for foF2, and it can be noticed from the figure that a linear relationship is valid for RASSN values less than 150 and saturation effects set in for higher values. With the help of standard diurnal variation curves for foF2 and H(3000)F2 and plots of 24-hour average foF2, H(3OOO)F2_against RASSN, the .0 foF2 and HUF(3OOO)F2 values can be computed for any given sunspot number (RASSN) for any hour of the day and station. HUF(4OOO)F2 values can be ob- tained from HUF(3OOO)F2 by multiplying with appropriate conversion factors recommended by CCIR. 3.1.3 RSD Prediction Technique The dependence of ionospheric parameters on the Running Average Sun Spot Number (RASSN) is well established and this relationship is the main basis for RSD prediction technique. Radio Science Division starts its prediction activities every month by predicting a RASSN for a month six months hence. After having predicted the Sun Spot Number (refer solar activity prediction section) the corresponding ionospheric parameters ave computed using the various sets of diurnal curves and solar activity variation plots described in the data base section. The first step would be to obtain a 24-hours average of foF2 and H(3000)F2 values for each of the stations for the pre- dicted RASSN from the appropriate 24-hour average plots described in 3.1.2 above, The 24-hour average values so obtained are then plotted against geo- graphic latitude and values are read out at every 50 latitude intervals up to 400 and at every I0 ° latitude intervals for higher latitudes in both the hemispheres. These 24-hour average values are then converted into diurnal _ values using on_ of the three standard diurnal curves that are available for each of the stations depending on the predicted RASSN. The procedure involves use of tl_ standard diurnal curve for appropriate RASSN range. The station used is the one nearest to a latitude grid; for example, the standard diurnal curve of Kodaikanal is used for lOON. The hourly values of this curve are then multiplied by the ratio of predicted average at IO°N to the average of this curve to give diurnal variation at IO°N. The diurnal variation curves for foF2 and M(]OOO)F2 thus obtained at different latitudes are employed to draw the contour maps (Figures 5 and 6) of foF2 and HUF(4OOO)F2 after converting the H(3000)F2 values into H(4000)F2.I 128 t 1980010188-139 ).1.4 Post-Hortem Analysis and Further Improvement The RSD (old) method of prediction described earlier indeed has certain limitations such as i) the diurnal variation in the Ionospheric parameter within a specified sunspot range is neglected, and Ii) by choosing the 24-hour average of monthly median as an ionospheric _O index, the diurnal variation in the index is de-emphasized. A new RSD prediction method has been developed mainly with the view of reducing these errors. In the new method, plots are obtained between monthly medians and EASSN for every 2-hour Intervals resulting in 12 plots for each month, the total number of plots being 5760 (12x12x40) for 40 stations for one parameter for all the months. These large numbers of basic data plots have ensured adequate diurnal coverage and have also simplified the computational procedure considerably. Figure 9 shows prediction of foF2 for September '70 for some low latitude stations obtained by both RSD old and . new methods of predictions. They are also compared with the IF2 method (Minnis and Bazzard, 1959) where instead of RASSN, an index IF2 which is September 1977 i I _ .... I m - f I I i i i Z A_ _w_m_L_u_imum i iI v_ iii Figure 8. A plot of ionospheric index- Figure 9. Comparison between ob- 24 hour average of foF2 - served and predicted values against running average sun- of foF2 for September 1970 spot numbers is shown for are shown. These compari- Ahmedabad for three repre- sons are shown for different sentatlve months. The sat- methods of predictions, for uratlon effect can be seen low latitude stations. from a value of 150 of RA$SN in all the curves. 129 t 1980010188-140 derived from sunspot number is usod. The bold line in the figure is based on the atLual observations. A comparison of different predictions with observed values shows that for latitudes below 200 , all the prediction methods yield better value for the daytime, while at higher latitudes the nighttime predictions are better. Figures IO(a,b) also show the comparison of frequency predictions ob- tained by different methods for two stations, New Delhi (Geographic latitude 28°38'N) and Ahmedabad (Geographic latitude 23°01'N) with the observed values. The predictions are for September 1977 and the methods chosen for comparison are essentially same as the ones used for September 1970 (Figure 9) except for IF2 method; instead, predictions obtained using CCIR - 340 atlas of ionospheric c;,aracterlstics (1966) are included for comparison. The comparison of different methods for September 1977 also in general yleids results similar to those of September 1970. It can be seen from Figure iOa for New Delhi that both RSD old and new methods show better agreement with the actual observations (bold line in the figure) than CCIR - 340 atlas values. However, for Ahmedabad, which is at a slightly lower lati- tude than New Delhi, all the prediction methods yield better results during nighttime, especially at post-sunset and pre-sunrise periods. It should also be mentioned here that RSD new method gives more realistlc predi*tions in comparison to other prediction methods. Saturation effects exhibited by ionospheric parameters at very high sun- spot numbers have always been a major limitation in long-term predictions. A systematic analysis of ionospheric parameter variation with RASSN (Figure 8) has conclusively shown that satur_tio_ effect is a constan_ feature at low latitudes irrespective of season _ "_SNvalues above 150. Hence, Radio Science Division had made several _tu_._s to overcome this inherent limita- tion. One such attempt has h_en to obtain second degree relationships between the ionospheric parameters anc_8ASSN of the kind, foF2 = A + BR + CR2 H(3OOO)F2 = A' + B'R + COR2 where R represents the RASSN. Constants A, B, C and A _, B_, C_ are obtained for every two hours for all the stations and the ionospheric parameters foF2 and H(3OOO)F2 can be com- puted for any RASSN. RSD :urrently uses a IBM 1620 computer for all its routine prediction progrtmmes. Analytical studies are made from time to time to look back and evaluate the accuracy of predictions. However, it is not within the scope of this presentation to give r_rge number of month-by-month comparisons of observed and predicted values. Instead, fairly large statistical samples of these comparisons are present_ in Figures Ila and lib for the northern winter months for low (RASSN = 15) and high (RASSN = I00) solar activity respec- tively. The curves show the average of all the observed noon val._s for 130 o 1980010188-141 O• NHIVll T LIt |lli,_September 1977 -'_ ........... ,4I III Delhi - _-,,,.- .... ,.,., II _ll IIIl(_10 Ililrtllll_ llll lhllOt I " 1o.1 ,. li,iiN.ii • ii" " " _M[I_4HNJ_t _, kiTit UI_ i Iv. • el It el ¢ ,. 4 4 .e _ al N t L_I, Utl_ hilt ,m,_) N W m N eta m le ,¢) •,_MIRmmU gl ! OW&Gill 1_. Lltlt UN Figure 10. Comparison between observed Figure I1. Average of observed noun and predicted values of foF2 foF2 v_lues for northern for September 1977 are shown winter are plotted against for represen_tive stations geomagnet;c latitude (line (a) Delhi and (b) Ahmedabad curve) for low and high so- in india, These comparisons iar activities separately. are shown for different The vertica; lines bounded methods of prediction, by triangles represent the 95_ ranges o_ observed values, whereas those bounded by horizontal bars represent 95_ ranges of predicted value. 16 winter months spread over 4 years of low solar activity and 4 years of high solar activity. The vertical lines show the spread in these values, the boundaries set by small horizontal lines indicating the predicted values with 95_ probability and the boundaries set by triangles indicating the range of spread in observed values. It may be observed that in general the r_nge in predicted values is not far different from the range in observed values, especially during low solar activity. It may also be noticed tE_t at higher ;atitudes, the difference is appreciably large because the data base at these latitudes for our longitudin_l zone was inadequate ecrlier. The data is shown against the geomagnetic latitude to emphasize the point that because of large horizontal gradients associated with geomagnetic anomaly and the horizontal movements of the anomaly, the performance of our prediction system even at th_se latitudes has been satisfactory. 131 t ,i 1980010188-142 : 4. CONCLUSIONS 4 The errors and uncertainties in the finally predicted ionospheric para- : meters are due to (a) errors in the solar activity predictions, (b) errors in assuming a simplified response of the ionosphere to the sunspot number, (c) errors due to deviations in the ionosphere in response to solar and geo- physical events, and (d) errors due to large day-to-day variations apparently unrelated to any specific events. Some attempts have been made at RSD to quantify the uncertainties due to (d) above, and are presented in a companion paper. Other related activities at RSD are as follows: : i. Data services: Ionospheric and solar geophysical data obtained over India are compiled, edited and published. 2. A.R.W.C.: Radio Science Division discharges the responsibilities of an Associate Regional Warning Center (of IUWDS) and disseminates warning messages throu£h the Time and Frequency transmitter of NPL. 3. Short Term Forecasting: Based on the Indian observations and on the data received from Boulder, Sydney, Paris and Moscow, a daily i forecast is made and broadcast over the transmitter mentioned : above. However, the quality of present service is below threshold. 4. Advisory Services: A large number of communication organizations : depend on RSD for advice on frequencies, firing angles, power and antenna requirements for Foint-to-point H.F. links. Special problems arising due to location of receivers in mountainous terrain : and valleys are also attended to. LUF (ALF) and signal strength '" calculations are also done on special request. Attempts are being made to computerize all these calculations. 5. Special Propagation Studies: Research and analytical studies are _ontinuously made at RSD to solve some of the special low latitude problems arising out of the geomagnetic anomaly and rapid increase and collapse of ionization during sunrise and sunset times. Some of these problems are discussed in another companion paper. REFERENCES CCIR Report 340: Atlas of lonospneric Characteristic, Oslo 1966, published 1967. pp. 404. Cole, T. W. (1973): Periodicities in solar activity. Solar Physics, 30:I03. Kuklin, G. V. (1976): Cyclical and secular variations of solar activity. Basic Mechanisms of Solar Activity (ed. V. Bumba and J. Kleczek), IAU Symposi_ No. 71, pp. 147-190. Minnis, C. M., and G. N. Bazzard (1959): Some ind,ces of solar activity based on ionospheric and radio noise measurements. J. Atmos. Terr. Phys., 14:213. 132 i i i i 1 1980010188-143 fNBS Circular N9. 462 /1_48): Ionospheric Radio Prop_atlon, pp. 208. ._ Shove, D. J. (1955): The sunspot cycle 649 BC to AD 2000. J. Geophys. Res., 60:127. Siutz, R. J., T. B. Gray, H. L. West, F. G. Stewart and M. Leftin (1971): Solar Activity Predictions. NASAContract Report No. 1939, NASACR-i_39. Stetson, H. T. (1947): Sunspots in Action. The Ronald Press Co., USA. Waldmeier, _. (1946): A prediction of the next ma×imumof solar activity. Terr. Magn. and Atmosph. Elect., 51:270. 133 / 1980010188-144 _. A SIMPLIFIED INDEXING OF F-REGION GEOPHYSICAL NOISE AT LOW LATITUDES S. Aggarwal, D. R. Lakshmi and B. M. Reddy National Physical Laboratory New Delhi - 110012, India : The low latitude F-region, unlike at middle and high latitudes, is protected from direct particle effects as well as indirect particle effects such as ionospheric storms. In fact, during the storms the changes in the F-region critical frequency and the F-region height factors are in such a way as to retain the maximum u_able fre- quency values almost unaltered. However, the day-to-day fluctua- tions in the F-region parameters, which by themselves are rather large, are compounded by the variability in the equatorial anomaly morphology and may cause considerable deterioration in HF communi- cations, if the median MUF values predicted much in advance are used. The objective of this paper is to describe a simple method of deriving an F-region noise index that can warn the prediction users at low latitudes as to the specific months when they have to be more careful in using the long-term predictions. I. INTRODUCTION Long distance H.F. communication supported by the F-region is essentially controlled by the two parameters, foF2 and M(3OOO)F2. This is so because the Maximum Usable Frequency (MUF) is a function of foF2 as well as M(30OO)F2 which includes the information on the height of the reflection level. While prediction of the median values of these parameters for any particular month • can be made with some degree of accuracy, it has not been possible to predict their day-to-day changes. These large day-to-day variations, especially manifested in foF2 variations, are apparently unrelated to specific solar or geomagnetic events and may be termed as F-region geophysical noise. Some studies have been made on this topic earlier, especially for mid-latitudes (King and Slater, 1973; Rush et al., 1974). An excellent effort in this • direction to quantify the MUF variability is that of Barghausen et al., 1969 (also CCIR doc. 6/IO70-E, 1978), but a simpler method of indexing using essentially low latitude data is expected to be more useful because of the special nature of these problems in the low latitude zone. 2. VARIABILITY IN foF2 Several studies were made by the authors to investigate if any kind of 134 1980010188-145 relation exists between the standaed solar and geophysical indices and the foF2 fluctuations. The results of one such study using data from Vanimo (geomag. lat. 12.7°S) and Kodaikanal (geomag. lat. O.6UN) are shown in Figure 1. (Aggarwal et al., 1976). Percentage deviations of foF2, which r t0F 2 NOISE WITH SOLAR AND MAGNETIC INDICES t KOOAIKANAL ( GEO MAG LAT 0 6iN ) t_oL !.oLll I1 tl ;'i_rliJif_ n I] _ _nfl lin fl illJl _..... _. lnnllllrll _ _Ill :_ D[ C 74 JAN #_ i EB ?e) MARCH r.I. APRIL "_) MAY 75 : Figure l. Diagram shows the number of hrs/day when A foF2 is beyond the range c,J: coefficient of variation for a low solar activity period for two low latitude stations (Ap and 10.7 cm flux values for the same period are also shown). are the normalized hourly deviations in foF2 from the monthly median values multiplied by IO0, have been calculated for each station. In order to see the dispersion of the percentage deviations around the median, the coeffi- cient of variation for every hour on a monthly basis has been calculated for daytime and nighttime separately, the coefficient of variation being the standard deviation in A foF2 itself. To be more explicit, the coefficient of variation 'V' can be defined as the ratio of the standard deviation to its average value in percent (Weatherburn, 1957). For daylight hours, the reference values have been arrived at by averaging the coefficients of varia- tion for the hours 0800 to 1900 local time. Based on these daytime reference values, the number of hours with A foF2 values beyond the range of daytime reference values were calculated and it is these number of hours that are shown in Figure I as noisy hours per day. It is apparent from Figure I that no significant correlation exists between this parameter and the indices A. and 10.7 cm flux. The correlation coefficients show hardly any improvemen_ t when the time lag between the solar and magnetic'indices and the foF2 135 t ,- -_,j _.j 1980010188-146 deviations is varied between 0 and 10 days. This study only confirms that the geophysical noise in foF2 is, in fact, a random noise especially at low latitudes, because at middle and high latitudes the large variations in foF2 following magnetic storms may inject some amount of apparent correlation with A p" The morphology of the variability in foF2 with local time, season and solar activity for an equatorial station, Kodaikanai, can be visualized from Figures 2 and 3. Figure 2 shows the coefficient of variation 'V', derived as described above for the year 1968 (high solar activity) while Figure 3 is for COEFFICIENT OF VARIATION V Q v(00% • Io%6v(zo'& • ZO % 4 v (_.IOV*IIV_O_KODAIKM 4AL 19441 I i I I I OO 02 04 {_ (_1 IO 12 t4 IG III 20 22 _4 LOCAL MEAN TIME Figure 2. Seasonal and temporal changes in 'V' for a high solar activity year 1968. . COEFFICIENT OF VARIATION V DV(IO% 11_0%4v¢ZC%l_20%(v{],0% Bv _30% KOOA,,(ANAL t9"/'5 MA'_ IX¢ I O0 02 04 041 08 IO 12 _4 IQ 18 _0 22 ;)4 • LOCAL MEAN rIME Figure 3. Seasonal and temporal changes in 'V' for a low solar activity year 1975. 136 1980010188-147 the year 1975 representing low solar activity, The variations are much larger, especially for low solar actlvlty perlod, than the usual prediction accuracy for the median foF2 values, 3, VARIABILITY IN MUF A comprehensive analysis of a Ion9 series of observations (not presented here) at several low latitude stations shows conclusively that the storm-time departures in foF2 are absolutely indistinguishable from usual day-to-day Fluctuations. However, rather curiously, the variability in MUF during storms is considerably lesser than the variability on quiet days. This is obviously due to the fact that at low latitudes during magnetlc storms, foF2 values undergo marginal enhancements along with increase in the F-layer height. _ In general, whatever the reasons, whenever the changes in both foF2 and hpF2 are in the same sense, the consequent changes in HUF will only be marginal; for example, an increase in foF2 will tend to increase the HUF while an increase in hpF2 will tend to decrease the HUF. This feature at low latitudes can be used with advantage for indexing the variability, by analysing the morphology of simultaneous variations in foF2 and hpF2. 4, A METHODOF INDEXING F-REGION NOISE The simultaneous percentage deviations in foF2 and hpF2 at four different local times in September 1977 for Ahmedabad (Geomag. Latitude: 13.8°N) are shown in Figure 4. Each point in the diagram represents the value on one day in the month, with the bold points representing the coincidence of more than one point. Figures 5 and 6 show similar plots of percentage deviations in for2 and hpF2 for Ahmedabad for the months of May-June 1977 (representing summer conditions) and for the months of November-December 1977 (representing winter conditions), respectively. The data points in quadrants I and 3 present the situation when both for2 and hpF2 variations are in the same sense and hence the RUF departures will not be large. So, these two quadrants represent minimum geophysical noise conditions. Quadrant 4 represents increased HUF situation and, consequently, communication using predicted median HUF values will not be a casuality. However, the changes in quadrant 2 (with foF2 decreasing and hpF2 increasing) may reduce the actual HUF much below the predicted median and thus cause communication failure. Obviously, the number of points in quadrant 2 for a particular month and local time may be used to quantify the deterioration. For example, for Figures 4, 5 and 6, a general observation that can be made (from the number of points in quadrant 2) is that the deterioration is considerable in summer and winter compared to equinox. : Attempts are currently underway to use the data from a large number of low latitude ionospheric stations and to compute the number of points in quadrant 2 as also to give different weight factors for these points, so that their actual contribution in bringing down the MUF can be evaluated. A quality index based on this information can be derived and supplied to the prediction users, for the more the number of points in quadrant 2, the 137 1980010188-148 "_ AHMEDABAQ,SEPTEMBER-1977 40 .... I _ l T _ l i I ! i :i oooo ,zoo__ 20- • • ¢o • ! • • i • -20 - " - o O, 0600 1800 20- • oil • • e_ 0 _'" - ....• , " • •[- ._ -20 -40 I J I I I I I I 1 ! -50 -30 -I0 0 I0 30 _0 -30 -I0 0 IO 30 50 A foF 2 (O/o_ Figure 4. Plots of A hpF2 (_;) against A foF2 (_o) for September 1977 at different local times for Ahmedabad. AHMEDABAD MAY-dUNE 1977 40 0000 • 1200 • • : 20- • . • . ., . • • e e• 0 - ;'L .... •- . .•L . ,e _ : • . • • _-2( _" 0. ; 0600 ° . IBO0 20- • ;;;!,..0 .;. " • iI • o• -20- _" " • • %" " -40 i I I I I 1 J , -60 -40 -20 0 20 40 0 -40 -20 0 20 40 -60 /_ foF2(%1 Figure 5. Plots of A hpF2 (_;) against A foF2 (_;) for Hay-June 1977 at different local times for Ahmedabad. 138 1980010188-149 :. AHMEDABAD NOV-DEC 1977 , -41 0000 1200 -2( - • • • ° e • o"0. • • ._: _ "" eeoQo • j • ,'" • • , 00 _ -20 - " Q. t.- 0 - <:3 06 O0 • 1800 20_- , • • t • • • • • coo • • O . = t • _ ¢_ e f ee . . • _. • : -20 -- • • -4O -6o -40 -20 0 20 40 0 - 40 -20 0 20 40 so A foF 2 (%) Figure 6. Plots of A hpF2 (%) against A foF2 (%) for November- December1977 at different local times for Ahmedabad. k greater will be the chances of deterioration of communication unless short- term predictions are used. Acknowledgements: The computational assistance rendered by Mrs. S. Shc'stri and Mr. V. A. Chandrashekharan is gratefully acknowledged. REFERENCES Aggarwal, S., D. R. Lakshmi and B. M. Reddy (1976): Some problems of HF : communication at low latitudes and possible solutions. I.J.R.S.P., 5:302. Barghausen, A. F., J. W. Finney, L. L. Proctor and L. D. Schultz (1969): Predicting long-term operational parameters of high-frequency sky-wave telecommunication systems. ESSA Tech. Rpt. ERL ]tO-ITS 78, U. S. Govern- ment Printing Office, Washington, D. C. L King, J. W. and A. J. Siater (]973): Errors in predicted values of foF2 and hpF2 compared with observed day-to-day variability. Telecom. Journal, ; 40:766. Rush, C. M., D. Miller and J. Gibbs (1974): The relative daily variability of foF2 and hmF2 and their implications for HF radio propagation. Radio Science,9:749. Weatherburn, C. E. (1957):A first course in Mathematical Statistics. Cam- bridge University Press, Cambridge. : 139 t Z 1980010188-150 tr %I S,o-IS474" SPIRAL SUNSPOTS AND SOLAR ACTIVITY FORECASTS Ding You-ji and Zhang Bai-rong Yunnan Observatory, Academia Sinica, Kunming, Yunnan Province, China The definition of a spiral sunspot is given. Three examples illus- trate the processes of the occurrence of spiral sunspots and the roles played by them in prominence and flare activities. Finally, we propose a method for solar activity forecasts making use of pcram- : eters such as spiral sunspots, prominences and neutral lines. i. DESCRIPTION AND DEFINITION OF A SPIRAL SUNSPOT The penumbral filaments of sunspots have a great variety of patterns and forms in the fine structure of sunspots. We roughly divide them into the typical structures such as "spiral", "comb", "net", "radial", etc. These structures show, to a certain degree, that the sunspot groups are at differ- : ent evolutionary stages. : We define a spiral _pot as follows: A sunspot is a spiral spot when either the penumbral filaments of a sunspot, or the outer filaments surrounding a cluster of small sunspots, show a tangential arrangement, either with all the filaments oriented in mutually consistent directions, or if some of the filaments take on other shapes, but they are not directed in the opposite sense. See the regions indicated by arrows in Figures I and 2. These spiral filaments generally have the following characteristics: (1) The tangentially arranged filaments are relatively long, and the contrast between the bright and the dark filaments also comparatively large. (2) The penum- bral filaments between their roots at the umbra and the outer boundary of the penumbra are usually curved into arcs, and rarely as tangentially arranged straight segments. These characteristics make it rather easy for us to draw an analogy between the spiral spots and hydrodynamical vortices. The physical picture of a spiral spot is simple and clear, name!y, the magnetic tube of the sunspot suffers a twist in a certain direction, so that we can directly infer from the morphology that the integral of the magnetic field around the - SpOt, =4 i, Ci)d_ is quite large. In other words, we can infer the probable existence, in the active region, of a strong electric current from the penumbral structure. Usually, a spiral structure lasts less than one day, so that when photographed 140 1980010188-151 ; on the next day, it has disappeared. Perhaps it already changed into a radial or a net-like structure; it is then called a short-lived (or new- born) spiral, Sometimes, it continues to exist for a number of days, its vortex intensity (that is_ the angle of Inclination) weake s gradually until it disappears and changes into a radial structure. We theh call it a long- lived spiral, For answering the question whether spirals appear in a sunspot group, we point out the following characteristics: (I) For the active sun- spot groups of types C, D, E, F, G and H, the appearance of spirals (especially_short-iived spirals) is a rather common p_enomenon. (2) It is not a general phenomenon for long-lived (lasting over two days) spirals or recurring spirals to appear. They are only dependent on the peculiarities of a minority of sunspot groups. (3) The time when a SliOt appears as a spiral pattern is much shorter than its lifetime and usually the spiral stage is at the beginning of its life (Ding You-ji et al., 1976a). We shall describe the formations of spiral sunspots, the relations between them and the activities of flares and prominence, by means of some concrete examples. 2. _PIRAL SPOT IN THE ACTIVE REGION MCHATH 11976 2.1 The Formatlon of Spiral Spots We can see from Figure 5 that the spot O_u at the central portion of the t large active region McMath 11976 on the secona"of August in 1972 shows clearly a spiral penumbra. All of its penumbral filaments appear as a clockwise arranged pattern oriented in mutually consistent directions, except that the north-east portion was connected with the umbra 01 ,. The motions of the spots in this region showed that the distribution an_ relative position of the four umbrae AN, BS, CS and DS were stable; however, the large umbra (consisting of OIN and O2N) started-an obvious counter-clockwise rotation before 040OUT on August 2. The following features are special: (!) The fibrous penumbral filaments at F in Figures 3 and 4 mostly remained in simple straight-line morphology, but broke up and bent slightly westwards in Figure 5. (2) Comb-shaped penumbrae at G extended toward the south-east in an arc shape on the same day. (3) The tendency of the spirality of the penumbral filaments at H was quite obvious. These morphological consistent changes show clearly the existence of a shearing motion of the magnetic field of the umbra 02N relatively to the solar atmosphere surrounding it. Any other kinds of transfer cannot explain these morphological changes. The associa- tion between the directions of the rotational motion of the umbra magnetic field and the arranged pattern of the penumbral filaments mentioned above, however, can be explained with plasma dynamics (Solar Physics Division, Yunnan Observatory, 1974a). 2.2 Relations between Spiral Sunspots and Prominence Activities According to many photographs of the chromocphere, magl,:tic fields and flares presented by Zlrln and Tanaka (1973), the neutral Ilne (Hit - O) penetrated through the penumbra region' of the sunspots t_ice. In the first region it was north-south oriented, situated between two umbral couples, (Bs, AN) and ¢S, OIN)" In the second region it was west-east oriented, 141 , t "_o 1980010188-152 i i t situated between the umbrae AN, 02N and Ds. : The Ha filter photograph of _uly 31 shows only one twisted sunspot fila- ment, situated over the neutral lines of the two regions, and it had no Interruption in morphology. After August I, the two portions of the filament of the regions suffered different changes. The filament in the second region lost its twisted morphology, replaced by a system of chromospheric threads that connected two umbrae AN and D$ of opposite polarities. Sometimes they came near to each other and formed a coarse and dark structure; other t_mes they separated from each other, turned into an arc system, and neve, twisted each other again. The south terminatlon of the Filament in the first region evidently pointed towards umbra 02N , (belonging t_ the foll_wlng polarity) and ended at the boundary of the umbra 02N, as if it were the source of the filament. But, inside the penumbra1 region the {llament appeared as a tightly twisted prominence magnetic rope with a small width about 3 - 8 arc seconds, and outside the penumbral region it extended orthward continuously. See Figures 6 and 7. According to the twisted morphology one can judge that the inner electric current is from north to south, but the current shown by the spiral morphology of the penumbra of O2N is from the top downward. So the prominence magnetic rope may llnk with the spot magnetic tube; its inner current might be connect_d, too (Ding You-ji et al., 1976b). 2.3 Relations between Spiral Sunspot and Flare Activity The morphology and character of the changes of the major Flares in thi_ active reoion were already discussed in detail by the Solar Physical Division of the Yunnan Observatory (1974b). From the former discussion and the paper presented by Zirin et al (1973), we draw the following conclusions: (i) For all major flares, both the maximum phase of the flare and the prominence mag- netic rope associated with O2N were very similar. Perhaps, the flares are t_e reflexes of the eruptions of the prominenc_ again and agaih. (2) The region near the second neutral line was anomalously quiet during the eruptions of all major flares. The gradient of the longitudinal Field here is about equal to that of the first neutral line, since the maximum values of the magnetic fields of all major spots in this region are approximately equal. The probabilities of magnetic field annihilations and the tearing instability of plasma should also be approximately equal in these two regions, but that did not agree with the observed facts. So we presume that it war very difficu,t to explain the flare activities in August 1972 wit the sheet pinch theory of magnetic neutral layers. The spiral spot, however, may be associated with flare activities through the prominence magnetic rope connected with it. (3) The "egions on either side of the first longitudinal field neutral iin_ were pressed by the umbra BS westwards and by the umbra OIN eastwards. The second neutral line was pressed by O2N southwestwards on one side. From the increase in the longitudinal field gradient, the motions of the spots on either side of th_ neutral lines, to a certain degree, were similar to each other. But the initial bright points of the flares were largely at the portions pressed by the umbrae BS and OIN, respectively, and there was no flare activity at the portion p;essed by O2N. Therefore, we presume that the sunspots' prts: on the neutral lines may be an o_ter factor for th_ trigger of flares, and cannot provide an energy source of flares and an inner eruptive factor of instability. 1_3 t 1980010188-154 3. SPIRAL SPOTS IN THE ACTIVE REGION MCMATH 14943 3.1 Tha Formacion of Spiral Sunspots We find easily from Figures 8 - 14 that the rotational motion of the preceding spot, which consisted of two umbrae PIS and P2S in this region, became increasingly more rapid during its passage across the solar disk, The relative positions of the major spots in the g-oup, however, were changed rather slowly. The relative position of PIS and P2S rotated about 90 ° on the solar disk during September 15-17, 1977. In response to this, a crack b (Figure II), due east of umbra P2S, rotated 90 o during the same period. It indicates that umbrae PIS and P25 rotated together rather than PIS slipping from west to east along a northern arc round the center of P2S- The arranged pattern of its penumbral filaments gradually formed an obvious spiral struc- ture in the same period (Zhang Bai-rong et al., 1978). See Figures 12 - 14. : , 3.2 Relations between Prominence and Spiral Sunspots ? There was a neutral line of the longitudinal field in this active region, It largely penetrated through the centra| portion of the region in a west-east direction. There was a sunspot prominence over the neutral line, its winding appearance was in accordance with the neutral line very well. The west termination of the prominence connected with spiral spots PIS and P2S" See the portions marked by arrows in Figures 15, 16 and 18. Its east termination extended northwards, pointed to the active region McMath 14942. A twisted structure occurred clearly during the prominence's passage across the central portion of the solar disk on September 15-17tt. See the morphologies of the prominence between the spots FIN,7F2N1_ _nd P3S in Figures 16, 17 and 18, and1, west of the spot P2S in Figures I u. According to the twisted direction of the prominence} c-q =ssert that its inner axial current was from west to east. However, the current shown by the penulnbral spiral structures of PIS and P2S was outward toward the chromosphere and the corona. So it should be presumed that the prominence and spots PiS and P2S formed a current circuit and a linking magnetic tube. The areas of other spots approximately equaled to that of PIS and P2S in the region, but there was no obvious prominence activity associat_J with them. 3.3 Relations between Spiral Sunspots and Flare Activities Flare eruptions in this active region were recorded 18 times by Yunnan Observatory and Peking Observatory. They were d;stributed on the whole on either side Pnd near the neutral line or spot prominence. When the spiral morphologies of the spots PIS and P2S were obvious, the flare eruptinns tended to occur west of the spot prominences, near the spots PIS ar,i P2S. See Figure 19. As the spot P3S continuously approached spot FIN their penumbrae connec- ted and formed a _-structure on September 19. The two spots pressed on both sides of the twisted prominence magnetic rope. This is very similar to the events that occurred over the first neutral line in August 1972 (see section 2.3). After September 19, the plages in this portion were markedly : 144 1980010188-155 points P, U, V) of the penumbra of AN occurred four days after the spinning of the umbra, but in the same direction. The rotation of the penumbra seemed to be the relaxing motion of the twisted magnetic tube of the spot AN. We suppose that the spiral structure of the penumbra of AN might result from the rotational motion of the umbra AN (Ding You-ji et al., 1977a). 4.2 Relations between Spiral Sunspots and Prominences In the first three days after the region emerged from the eastern limb of the solar disc, no evident prominence activity was reported. Whole prominence activities occurred during its passage across the solar disc. So the prominence activity in the region was able to reflect sensibly the intensity acted on the coronal matter by the sunspots which were below it. • From Figures 28 - 32, we can see that the prominences in the region were associated closely with the large preceding spiral spot AN . They pointed to this spot unexceptionally and appeared as a divergent arrangement. Ding You-ji et al. (1977b) identified that the attractive center below the largest , coronal prominence was the spiral spot AN, and pointed out that the coronal prominerce was formed by tne powerful sunspot electric current which condensed matter in the coronal space. See Figure 30. There also occurred the obvious spray activity of another smaller spiral spot EN in the fegion (Figure 31). 4.3 Relations between Spiral Spots and Flares Having studied the characteristics of the distribution and evolution of 192 bright points belonging to 68 flares in this region, Ding You-ji et al. (1977b) pointed out: (i) The initial flare activity of the region started over the preceding spot AN which had the highest energy state of twisted mag- netic field among all spots of the group. Synchronizing with the relaxing process of the twisted penumbral filaments of AN, the initial bright points of flares transposed from three places near the umbra to the outer portion of the penumbra. When the twisted penumbra relaxed at all, the flare activity adjacent to the spot decreased in general. The distances from the flare bright points to the center of AN changed with time, which was shown in Figure 35. (2) Another spot EN showed itself in spiral morphology for four days, the frequency of the flare eruption adjacent to it was 5/day, but the frequency was only O.3/day when the spiral morphology did not appear. It indicated that the spiral spot associated with rather higher frequencies of flare activity. Owing to the shortage of longitudinal field data, we could _ not study the role of the neutral line in prominence and flare activities of ; the regio_ in detail. 5. SUMMARY OF THE THREE ACTIVE REGIONS : 5.1 The Formation of Spiral Sunspots r ¢ Presumably, some dynamic instability resulting from the convective motion below tke photosphere made the rotational motion of the magnetic tube of sun- spot um a occur. The penumbral magnetic field lines rooted in the umbral region, the upper portion of them diverged in the shape of a trumpet in the 148 C _.._-_ 1980010188-159 atmosphere of the chromosphere and the upper photos?here. Because the trans- mission of the disturbance along the nlagnetic fielJ line takes time, to various extent the phase of the rotational motion of penumbral filaments lagged behind from inside to outside. Therefore, the penumbral helical struc- ture was formed, and the twisted direction of the helical line was opposite to the rotational direction of the spot umbra. The three examples presented in this paper are the observational evidence of this kind of dynamic struc- : ture. 5.2 Relations between Spiral Sunspots and Prominences I. The three examples mentioned above showed unanimously that spiral spots usually associated with the stronger activity of the spot prominences, : either of the polarity of them was the preceding (McMath 12094and 14943) or the following one (McMath 11976). On the photographic data with highJ quality, the filaments a_sociated with spiral spots usually show up as obvious helical structures, and the twisted direction is mutually consistent with the spiral direction.of spots. It means that in both the magnetic tube and current tube they are connected with each other. If the self-inductance coefficient of the prominence is L, then, the twisted free energy stored in it is + 1 12W = _-- L (2) 2. On the H_ filter photograph, most portions of the prominence mag- netic rope keep the general rule of prominence position, that is to say, it extended along the neutral line of the longitudinal field except that a small piece of prominence connected with the sunspot was in the direction of the photospheric magnetic lines. 5.3 Relations between Spiral Sunspots and Flares ; i. There is rather higher frequency of the flare eruptions near and inside the spiral structure of photospheric penumbra and chromospheric fibrils. _ 2. There is also rather higher flare eruptive frequency near and inside : the prominence magnetic rope with the obvious twisted structure associated with spiral spots in morphology. _ 3. Where there is a prominence associated with spiral spots over the neutral line with higher f!eld gradient (6-magnetic), there will be rather higher probability of erupting major flares; and if not, the major flares were rare, even small flares were very few. We presume that the points I and 2 mentioned above act as sufficient . conditions, the point 3 amounts to a necessary condition. There is close physical association between the energy source of flares, the instabiIitive factor and the twisted magnetic field (or electric current). The magnitude of the gradient of the magnetic field might show up as the Intensity of the outer trigger factor for flares. The major active region McMath I1482 in 150 1980010188-161 August 1971 might be considered as the fourth example (Figure 2). Mclntosh (1972) described the motions of the sunspots in the group, and Zlrin et al. (1975) described its twisted structure of chromospheric fibrils and flare activities. Babin et al. (1974) reported Its spray activities. 6. SPIRAL SUNSPOTS AND ENERGETIC FLARE FORECASTS 6.1 The Possibility of Predicting Energetic Flares with Spiral Sunspots During the period of 1970-1972, 50 sunspot groups were photographed at Yunnan Observatory. Figure 36 shows the frequency distribution of sunspot groups according to number of days on which spirals are seen. From the figure, we may draw a conclusion that there is close relation between long-life spiral spots and energetic flares (see Ding You-ji et al., 1976a). In addition, the lags of the date between the occurrence of energetic flares i _ and the first sight of spiral spots were researched. The result showed that _ the spiral spot precedes energetic flares by three to four days. This con- clusion is useful for the energetic flare forecast with spiral spots (Figure 37). 6.2 Prediction Technique I. When the penumbral filaments of a major spot in an active region show up as a _piral pattern on fine structure photographs, it means that the 22-- * 0 16- L 14- _- _J O I Z _5 -& -_ -z o z _ _ g _o ,z _ No. of days when a spiral was seen No. of days anticipated by spirals Figure 36. Frequency distribution of Figure 37. Date difference (date of sunspot groups according to number proton flare minus date of first of days on which spiral structure sight of spiral spot in the same was seen. Shaded part with _; indi- active region) distributions. cared refers to groups accompanied by ]arger terrestrial effects. 151 1980010188-162 flare activity may greatly develop in a few days. Preliminary forecasts should be msde_ keen joint observations should be taken and close watches should be kept on the development of the active region. 2. In the chromospheric patrol, if a prominence is observed very near or pointing toward the spiral spot, then it shows that the energy, which can be used for flare eruptions, has been transmitted to the upper chromosphere and the coronal space. The flare and prominence activities in the period will then come to a climax. 3. From the shapes of the prominences and flares, and from the types of radio bursts that occur, as well as their corresponding geophysical ef- fects_ the future energetic events can be estimated very well. 4. Magnetic field measurements may provide information of the changes of the longitudinal-field gradient on either side of the prominence (neutral line) associated with spiral spots. When the spots with opposite polarities on either side of the prominence approach each other or keep a , close distance, it greatly favours the occurrence of a proton flare. 5. When the penumbrae of two spots with opposite polarities have ; joined together and formed a 6-structure, the prominence associated with the : spiral spots over the neutral line is very unstable. Sometimes it turns : into a bright plage fibre, narrow and long, and might be extended outside the sunspot group and forms a spot filament, coarse and dark. This kind of i situation is almost the foreboding signs of producing proton flares. The necessary alarm must be sent out at once. Many proton flares, though, might occur without so sufficient conditions as mentioned above. Acknowledgements We gratefully acknowledge Comrade Ling Zhong-xiu of Yunnan Observatory and Comrade Sheen Lung-xiang of Peking Observatory for their useful help. REFERENCES L Babin, A. N., L. G. Kartashova, A. N. Koval, M. B. Ogin and N. N. Stepanyan (1974): Izv. Krym. Astrophys. Obs., 52:92. Ding You-ji, Zhang Bai-rong, LI Wei-bao, Hong qin-fang and Li Zhi-kai (1976a): Acta Astronomica Sinica, 17:60. Ding You-ji, Li Wei-bao and Hong Qin-fang (1976b): Acta Astronomica Sinica, 17:110. Ding You-ji, LI Wei-bao, Hong qin-fang and LI Zhi-kai (1977a): Acta Astro- nomica Sinica, 18:40. 152 L ]9800]0]88-]63 Ding You-jl, Li Wei-bao, Wu Hing-chan, Jiang Zhl-zhong, Hong Qin-fang and i Liu Zi-zhong (1977b): Acta Astronomica Sinica, 18:143. Hclntosh, P. S. (1972): Sunspots and H-alpha plage in McHath region #11482. ! World Data Center A. for Solar-Terrestrlai Phys. Report UAG-24, Part II, ! 303-310. Solar Physics Division, Yunnan Observatory (1974a): Acta Astronomica Sinica, 15:25. i Solar Physics Division Yunnan Observatory (1974b)' Acta Astronomica i ' ' Sinica, 15:273. Zhang Bai-rong, Zhong Shu-hua and Wang Hong-zhan (1978): Submitted to Acta : Astronomica Sinica. Zlrin, H. and K. Tanaka (1973): Solar Phys., 32:i73. : Zirin, H. and B. Lazareff (1975): Solar Phys., 4l:425. k 1980010188-164 SOLAR ACTIVITY PREDICTIONS AT PEKING OBSERVATORY z i Solar Activity Prediction Group i Department of Solar Physics Peking Observatory, Academy of Sciences ; Peking, People's Republic of China Solar activity prediction procedures used at the Peking Observatory ,are briefly reviewed. Methods for short-term prediction of solar proton events and the results for proton events of importance gr=a=er than I during the period 1974-1977 are given in detail. The characteristic wlues of solar cycle 21 are predicted. Predictions of solar activity have been routine!y carried out in the Peking Observatory since 1969. Three kinds of predictions are made. (i), short-term predictions of solar proton events and SID's (72 hours ahead)_ _2), medium-term predictions (monthly) of so]at proton events and SID's; and (3), long-term predictions of solar _ctivity, mainly forecasting of the characteristic values of solar cycle 21. In this paper, we shall describe the prlnciples and methods used for these predictions with emphasis on short-term predictions of proton events (PE). : i. SHORT-TERM PREDICTIONS For proton events, we are concerned only with the short-term prediction of those with importance greater than I according to Smart and Shea's (1971) classification, i.e. E > 10 Mev protons with peak flux _lue > IO cm"_ sec'Zster "l. Only "danger" (proLon event, yes) and "safe" (p_oton event, no) are forecast. Methods based cn th¢ proton-producing-reglon probaE ,ity and multiple regression have been developed and used in parallel. The principles and some results are described below. I.I Hetnod of Proton-Producing-Region Probability In estaolishing the criterion for prediction, we analyzed the features of the solar active regions which produced proton events of greater than importance ! in the period 1965-1970. We adopted the following four parameters which are available with the present equipment of the Peking Observatory: (I), magnetic field of the sun spot group, (ii), Zuric_ class of the sun spot group, (iii), H_ flares and (iv), 3 cm radio bursts. Each 154 t °, 1980010188-165 parameter is divided into three to four classes according to the "strength" of its activity. In the magnetic parameter, besides the Mt. Wilson classifi- cation, factors based on Lhe 6 type configuration (Warwick, 1966), "A" _ configuration (Avignon et al., 1965) and magnetic axis inclination _re also taken into account. The flare parameter Is taken as a measure of the number and importance of the flares that occurred in the past 24 hours. The parameter of radio bursts is classified according to the number and peak flux of the bursts in the past 24 hours. When an active region I_s been assigned the above four parameters, classified as i, j, k, I, respectively, the region is given the name of (i j k i) type region. We have determined the types (ijkl) for all active regions that appeared on the surface of the Sun each day in the period 1965-1970. Thus a region that appears on n days is con- sidered as n separate regions. We introduce tile idea of "proton producing region", which can be defined as a region that will produce a proton flare within the next 72 hour_. The prediction of proton events for the next 72 hours can be performed by the prediction of the occurrence of a proton pro- ducing region. By definition, an active region which has a proton event occurring in it during the day or at one and two days before is named a pro- ton producing region. In the 1965-1970 period, we found 43 proton events of importance greater than i that could be identified on the visible solar surface, the corresponding proton producing regions were 119 in number. If Nij_,:_ and. Ei._l denote the. number=,of the (ijkl) active regions and protonproauclng re_lons respectavely, then the ratio of proton producing regions among various active regions in the years 1965-1970 may be written as Pi'klj = El_k/i /N.i_kl. lhls can approximately be.taken as the probability of occurrence aT a proton producing region of the (i j k I) type active region. A table of Pijkl has been compiled. Obviot!siy, when n different active. regions appear on the solar surface, the probability of occurrence of a , proton producing region for the whole surface will be n P = I - _ (I - Pijkl)t (I)t=l We calculated each day's P values for the period 1965-1970 and found that on the day of a proton event and on one and two days before that, one gets P values greater than usual. Especially, on the day of the event, one almost always has a P value greater than the critical value Po = 0.05 (Figure I). Therefore, in principle, one may establish a criterion of forecasting "danger" or "safe" by estimating the dally P value being _ Po or < Po" The actual prediction procedu_'e is as follows: I. From the solar observations of the day, obtain the classes of the four different parameters for each active region, thus assign the (i j k l) for eac, region. 2. With the known (Ijki), look up Pijkl for each active region from the table of PiJkl' : 3. Using equation (I), calculate P, and according to whether P is > 0.05 or < 0.05, forecast the next 72 hours as "danger" or "safe". For some special occasions, such as when important active regions happened 155 t % 1980010188-166 Figure I. The distribution of the number N of P values (2005 in total) during 1965-1970. P! denotes the P v_lue on the da;, and on one and two daysbefore the proton event, P2 Is the P value on the day of the proton event. N(P), N(P_) and N(P2) are respectively the numoer of deys of P, rj and P2 at different intervals. to be on the limb (so that Lhe magnetic type and configuration were not ! clear), or that they had just turned behind the west limb (so that the proton stream might still reach the Earth), the experience of the forecaster would be called upon to give a sensible _,'edic_ion. 156 Ir-" 1980010188-167 :I 1 _, I 1.2 Rethod of Mult, iple Regression : t Let y denote the occurrence of a proton event (yes or no), i.e. the quantity to be pred|cted_ and let variables x i be the parameters representing various kinds of solar activity, which are significantly correlated to y. With the help of the "historical" data, one can construct an equation of regression m y = b° + b=×l (2) _ i=i where m is the number of variables concerned. =r ._ Making use of the standard procedure of multiple regression, one may perform, with a given confidence level, significance tests for any variable _ in the regression equation, and thus may obtain an equation with as few as possible variables which represent the most significant parameters. t : _ For proton event prcdiction, we chose as variables a number of parameters : _ including the magnetic type of the sunspot group, the H_ plage, IO cm flux density .... etc. With these historical data, we analyzed the correlation of each of the variables with the proton event, so that relative weights could } be assigned to the different classes of each variable. According to the " above _thod, from the relevant data in the period 1966-1972, we obtained the final regression equation with the help of a computer: y = 2.3 + 1.8 x1 - 3.2 x 2 (3) where xl and x2 _e the we:ghting numbers given respectively to the magnetic type and Zurich c _ ss of the principal spot group vislb_e on the solar : surface an0 y, r mentioned above, is the quantity to be predicted. The p_ocedure for using this method is: I. From the observed magnetic configuration and Zurich class of the spot group in the principal active region, find the corresponding weighting numbers xI and x2. 2. From equation (3), calculate y. : 3. According to whether y is less or not less than a critical number r Vo, forecast "safe" or "danger" for the proton event. : Both methods (l.l and 1.2) have been u_ed for the routine predictions. Some results in the period from January 1974 to October i977 are shown in table I. One can see that for proton events of importance gre_ter than I : (as speci_!ed in our case), these two methods yield the ='me results. Ouring L this period all the protnn events were duly forecast except the one of > importance 1 that occurred on August 22, 1976. In the routine work, when the t_ methous disagree with each other, which rarely happens, the fore- caster would have to decide which to choose. 157 2 _ 1980010188-168 Table i. Result of predictions of proton events importance _ I during the period 1974.1-1977.10. S = Success. F = Failure). Date of occurrence of proton event importance method 1.1 method i.2 1974 Jul. 3 2 S S 1976 May I 2 S S Aug. 22 ! F F 1977 Sep. 17 ! s S Sep. 19 2 S S Sep. 24 1 S S Total number of days when forecast was made: 1056 1076 Total number of days when "danger" was forecast: 40 70 Percent 3-7% 6.5% Besides the 72 hour forecasts, we also give "immediate" (some ten minutes to several hours ahead) warnings of proton events. For this purpose, we make t use of the duration and peak flux density of the centimeter wave bursts. By analyzing the data of 1959-1962 and 1966.9-1969.4, we have found that for proton events of importance greater than l, the duration At and flux density maximum As almost always satisfy the following relations: for 3 cm bursts &t > 15m, As > I00 SFU. (4) for lO cm bursts At > 20m As > lO0 SFU. (5) Hence, Lnese relations have been used to decide to issue proton event warnings. With this simple method, we have succeeded on several occasions (e.g. in the events of March 29, 1970, June 14, 1970, and July 3, 1976). When tested by the historical data, we found that this method gives warnings for 92% of the proton events, while the false warning rate amounts to 61%. For further improvement, it becomes necessary to consider the features of the sunspot groups and flares. It seems to us from experience that the magnetic configuration of the = sunspot 9roup plays a dominant role in the prediction of proton events. According to the analysis made on the data of the 20th solar cycle, we have found that it would be useful to classify the magnetic feature of a spot group as "normal" and "abnormal" (Shi Zhong-xian et al., 1975). More 158 .Z 1980010188-169 recently, making use of the table of Proton events compiled by Svestka and Simon (1973) and the data of proton events in 1970-1972 (Solar Activity Division, Peking Observatory, 1974), we analyzed the proton events of all classes during the period 1965-1972 and the magnetic features of 207 proton active regions in which the events occurred. As a result, we have found ' that several configurations which appeared "abnormal" had more than usual significant connections with proton events. Applications of this analysis to the work of predictions is yet to be exploited. For the short term predictions of the SID's, we distinguish four classes of SID events. Thus for the cases of strong, medium, weak and no SID, we assign respectively the , mbers 3, 2, I and O. The method adopted for the SID prediction is also based on the principle of multiple r_gression, similar to that described ,a section 1.2. The only difference is that the statistics was done separately for the ascending phase and descending phase of the solar cycle. Hence one gets two regression equations: ascending: _l = 11.6 + 8.9x 1 + 5.3x 2 + 4.4x 3 (6) descending: Y2 = 11.4 + 6.9x 2 + 4.7x 3 + 2.9x 4 (7) where Xl, x2, x3 and x 4 denote respectively the values given to the types of the magnetic field, plage, iO cm. flux density and number of centimeter wave bursts of the principal active region. Procedure for the prediction is the same as that described in section 1.2. 2. MEDIUM TERM PREDICTIONS OF SOLAR ACTIVITY In this program, we are concerned with the monthly predictions of proton events. The following factors have been taken into consideration: i. Regions of "active longitude". Many authors have already pointed out that active regions which generate proton events have a tendency to con- centrate at certain longitudes (Warwick, 1965; Svestka, 1968). We found that the proton active regions in cycle 20 were roughly concentrated near , Carrington longitudes 800 , 1300 , 2300 and 2800 (Solar Activity Division, Peking Observatory, 1974). Each region had a width of about 300-400 . Hence when these regions appear on the visible hemisphere of the Sun, the probability of occurrence of proton events will be enhanced. 2. The proton activity of different longitude regions often exhibits different behavior. For example, in the 20th cycle, the region at 800 had the greatest rate of proton event production, the majority of events originated in regions having sunspot area greater than 500 (unit: 10 -6 hemi- sphere), and the accompanying m;crowave bursts were comparatively strong; while in the region at 1300 , proton events occurred in sunspot groups with spot areas less than 500, and with only weak microwave bursts. 3. The 80-day cycle of proton activity. A study of proton events in the period 1954-1972 showed that there seems to exist a cycle of 80 days in 159 1 [ ,....- 1980010188-170 proton flare activity(Solar Activity Division, Peking Observatory, 1974). :_: During a certain phase of the cycle, which comprises about 20% of the days in the whole cycle, more than 60% of the proton events of importance 2 and 80% of those of importance 3 were concentrated. 4. Estimation of the evolution and the possibility of recurrence of importantactive regions. Such estimation is based mainly on the age, shape, rate of evolution and history of recurrence of the region. In the medium term predictions, we usually plot the Zurich type of spots on the so-<.alled LZT _hart, in which date of observation is taken as abscissa and Carrington longitude as ordinate. The "position" of large = flares and proton events are then marked on the chart. Taking a sufficiently • long period it is possible by examination of the chart to find the time and space distribution and evolution tendency of the important active regions. This kind of examination, together with the four factors mentioned above, forms thu basis of our estimation for the monthly predictions. _ The same factors are considered in the medium term predictions of SID's. However, since SID's occur more frequently than the proton events, a larger : duration is covered in each prediction. In addition, studies on the medium term predictions of the total level of solar act'vities have been conducted. The idea is to consider the active regions with respect to the photosphere, chromosphere and co, ona as a whole. For this purpose, the maonetic field of the active region, Ha plage and I0 cm radio flux have been taken as parameters signifying these three solar levels. Time sequences of these p3rameters were treated by the method based on the •: theory of n,ulti-dimensionai stochastic processes. The results have been used as a reference for the medium term predictions. 3. LONG TERM PREDICTIONS OF SOLAR ACTIVITY In this program we have been concerned mainly with the study of charac- teristic values of the 21st solar cycle. The Io,,g-term behaviour of solar activities have also been investigated (Lin Yuan-chang and Chang Chien-chung, 1977; Chow Yee-hsin, 1977). The study on the prediction of the 21st cycle was carried out in 1974. At that time we predicted that the minimum phase should begin at t m = 1975.8, which is fairly close to the actual time _" 1976.25 announced later. Recently we have estimated the time of maximum activity t M and sunspot number maximum RM (monthly smoothed value) to be tM: 1980.9 ± 0.8 yr RM: 115 ± 15 Regression analysis was used for the estimations of tm and RM. Thus, let x i be the interval between adjacent maxima, x 2 be the length n_ descend- .: ing portion and x 3 the length of ascending portion of the last ,.y_le (Figure 2). 16o t \. 1980010188-171 • Figure 2. The definitions of x 1, x2 and x 3. Making use of the xi, x2, x3, values of 1-19th solar cycles (Waldmeier, 196l), we performed-the regression calculations in steps. We obtained, with confidence level of 95%, the regression equation: x2 = 9.17 - 0.58x 3 + 1.15 (yr) (8) xI = 1.73x2 + 0.55x 3 - 2.89 -+0.8 (yr) (9) : We knew in 1974 that the length of ascending portion of the 20th cycle was x3 = 4.1 years. Substituting this in equation (8), we obtained x2 = 6.9 ± 1.5 yr and :m = 1968.9 + 6.9 +- 1.5 = 1975.8 *- 1.5 yr. Now we also know that x 2 = 7.2 yrs in the 20th cycle, so from (9) we have obtained t M = 1980.9 + 0.8 yr The estimation of RM has been made in two steps. Firstly, the mean sunspot number R from 1700 through 1974 were smoothed to get an average • curve R(t). Denoting the deviation of R w'.th respect to R(t) by AR, we found that in the period of 1700-1974the distribution of Z_R (214 in total) is nearly a normal distribution. W_: know that the distribution of peak values G(x) of a random variable which has a normal distribution cen be ' taken as A(x), thus x-b GCx)_ _(x) e-e -e a= = e (I0) 161 1980010188-172 In the 1-20th solar cycle, we have 20 peak values of AR n . Substituting in (]0), we determine coefficients a and b. With these, we estimate that the peak value R21 of the 21st cycle should lie between 30 to 77 with a probability 6f 0.60. The average value R(t) at 1980 may roughly be taken as 50 (Lin Yuan-chang and Chang Chien-chung, 1977), so that we obtain Rt__ 80 - 130. On the other hand, one can look at the problem from the viewpoint of apparent processes instead of statistical distributions. Thus, the history _f solar activity since the 1st cycle, with the exception of 4-6th cycle, has shown that, if during a certain cycle the RR value decreases with respect to that of the previous cycle, then the RM value at the next cycle will invariably increase. Even in the case of the 4 - 6th cycles, where the RM : value drops in two successive cycles, the actual amount of drop in the 6th cycle from the 5th is only 0.5. Now, we see that in the periud between 19-20th cycle, the RM values have dropped sharply, reaching 90.7 a;ready. This sharpness of dropping is only surpassed by the historically sharpest : drop of 92 during the 4-5th cycle. Holding this in mind, one may expect that i possibility of another drop in the 21st cycle would be very small, and since R. = 110.6 in the 20th cycle, the RM value of the 21st cycle would most p_obably lie in the greater half of the interval 80-130, so one may predict that for the 21st cycle, RM _ 100-130. REFERENCES Avignon, Y. et al. (1965): in R. L_st (ed.), Stallar and solar magnetic • field, North-Holland, Amsterdam, 373. Chow Yee-hsin (1977): in Huan9 Shan Symposium on Astrophysics (in Chinese, to be published). Lin Yuan-chang, and Chang Chien-chung (1977): Kexue Ton_bao, (in Chinese), 22:59. Shi Zhong-xian, et al. (1975): Acta Astronomia Sinica, 16:12. Smart, D. F. and Shea, M. A. (1971): S_lar _hy_., 16:484. Solar Activity Division, Peking Observatory (1974): Acta Astronomia Sinica, 15:34. Svestka, Z. (1968): Solar Phys., 4:18. Svestka, Z. and Simon, P. (1973): Catalog of particle events, 1955-1969, Dordrecht-Hollend/Boston, U.S.A. Waldmeier, M. (1961): Sunspot Activity in the Years 1610-1960. Zurich Schulthess and Co. AG. Warwick, C. S. (1965): Astrophys. J., 141:500. Warwick, C. S. (1966): Astrophys. J., 145:215. 162 I 1980010188-17g f• f._ • ,,qw - 18 476 _" + LONG-TERM FORECASTING OF SOLAR ACTIVITY XU ZHEN-TAO, ZHAO AI-DI, I_EIYAN-LIN, GUO QUAN-SHI I PURPLE MOD_AIN OBSERVATORY, ACADEMIA SINICANANKING, CHINAE, When the 20th solar cycle was not yet ended, we made a forecast of the basic parameters of the 21st cycle (i). Now that the trend of the ascending phase of the 21st cycle has gradually become clear, we can use the new information to make corrections to the past forecast. Besides this, in consideration of the requirements of meteorological, hydrological, and geophysical units, we shall make an esti- mate about the level of solar activity in the 22nd cycle. i. PREDICTION OF _ AND tM OF THE 21ST CYCLE The 20th cycle came to an end in June, 1976, i.e. in the year 1976.5. The corresponding smoothed monthly mean of relative numberswas R = 12.2_2). Tberefore, the solar activity at present is in the ascending phase of the 21st cycle. Under this condition, it is con- venient to estimate the peak value _and the epoch of maximum _ of the present cycle by means of the me_lod of Waldmeier (3). Acco_d-- to Waldmeier's formula, we have V = , (1) 12(18.4 - 7.14 log Ru) where V is the rate of rise of the present ascending phase. From the latest issue of S.G.D. (4), we rind that the smoothed monthly mean of the relative numbers in November, 1977, is 49.9. From this we may determine the value of V and obtain from equation (i) _ = 115, and TI = 3.7 years, where TI is the estimated value of the rise time of tSe present cycle. Because the epoch of minimum was in the year 1976.5, the epoch of maximum of this cycle must be TM 1976 5 + 3 7 = 198_.2. (2) With the yearly mean values of the relative numbers in 1700-1977 some parameters and the rates of ri_ing in the first two years of the 163 ) 1980010188-174 ascending phase of all the cycles may be tabulated in the following: cycle V1 V2 Rm R(4)m Rm R(4)m TZ T2 no. _ (minimum) (maximum) (years) years) -4 6 5 5 58 5 7 -3 2 9 0 63 5 6 -2 i0 19 Zl 122 4 6 -i ii 18 5 5.2 iii 88.5 5 6 0 6 ii 5 5,2 83.4 94.8 6 5 i 0.6122.2 I 9.6 7.6 85.9 i00.6 6 5 2 26.4 32.0 111.4 7,8 106.1 96.6 3 6 3 12.8 72.7 I 7.0 8.2 154.4 107.4 3 6 4 13.9 58.8 110.2 9.6 132.0 119.6 3 ii 5 2.7 7.7 4.1 8.2 47.5 ii0.0 6 6 6 1.4 3.6 0 5.3 45.8 94.9 6 7 7 6.7 8.1 1.8 4.0 70.9 74.0 7 3 8 4.7 43.7 8.5 3.6 138.3 75.6 4 6 9 4.3 25.1 10.7 5.2 124.7 94.9 5 8 i0 18.4 32.1 4.3 6.3 95.8 107.4 4 7 Ii 30.3 37.4 7.3 7.7 139.0 124.4 3 8 12 2.6 26.3 3.4 6.4 63.7 105.8 5 6 13 0.8 28.5 6.3 5.3 85.1 95.9 4 8 14 2.3 19.4 2.7 4.9 63.5 87.8 4 8 15 8.5 37.8 1.4 3.4 103.9 79.0 4 6 16 10.9 27.6 5.8 4.0 , 77.8 82.6 5 5 17 3.0 27.4 5.7 3.9 1114.4 89.9 4 7 18 23.6 59.4 9.6 5.6 i,151.6 3.11.9 3 7 19 33.6 !L03.7 4.4 6.4 1190.2 133.5 3 7 20 4.9 I 31.8 10.2 7.5 1105'9 140,5 4 8 21 14.9 ! 12.6 9.2 ] Where V1 and v2 are the rates of rising of the first and second years in the ascenazng phase respectively. Each of them is the difference between the yearly mean value of relative numbers of the following year and that of the preceding year. R and P_.are respectively the m .M minimum and the maximum of the yearly mean values of relative numbers. R(4)"" and --"P_4)are the running mean values in four consecutive cycles. T_.and T2 aSre the lengths of the ascending and descending phases. T_ere can be seen from the above table: (i) If VI is larger than 13, the probability for RM to be larger than i00 is 0.75. The mean peak value of such cycles is RM = 130 + 14, the mean length of the ascending phases is TI = 3.6 years. (2) If Vp is larger than 40, RM should be larger than 130. The mean length of Ehe ascending phases oT these cycles is TI 3.3 years. For the 21st cycle there has been obtained V1 = 14,9. According to the "64 1980010188-175 Ii i I observations of sunspots in the first half of 1978, we guess that the yearly mean of the relatlve numbers in 1978 will be 70. If this will be true, we shall have V? >40. Therefore, we estimate that the peak :i _ value of the relative number of the 21st cycle will be _ 130. Considering the above results, we choose _= 130_ 15, tM- 1980.2_ o.3 (3) • as the predicted values of the 21st cycle. They are in rather good accordance with the results of Xan_hakis and Poulakos (5) and also with those of Hill (6). 2. PREDICTION OF RN AND R OF THE 22ND CYCLEm From the above table, _., R.? ) and R , R(4) were used to d_w Fig. i and Fig. 2 respectively. _ In'these fi_ure_ the curves of _; I l ' : I /' I I t • /" I ! , {I l ,,, ! , | | | ', -_ o '1" 8 iZ t6 p.o Fig. 1 The long-term forPcastJ_g of the y_arly maximum sunspot number as a function of sunspot cycle number. Dashed line is the yearly mean - sunspot number. Solid llne is the running mean from four consecutive •" cycles, 165 J 1980010188-176 //2" / • ,q # .4 t ! . /I /io- ,_ f t _ /i . t _c_J tl I l I _ // IJ : _ !/ ' _,\ ! I/A,\ _ \ / . t t t A Is : ,-/ \\ , ,,4',I\N ,-J/C'.l :It# ',i ,,,i '_111 I ,, ,_/ it it i i , , 7 , , I , I •-4, o + 8 /2 #I 20 /_o. Fig. 2 The long-term forecasting of the yearly minimum sunspot number, and R (4) are the typical representations of the well-known 80-year eycle_ It is interesting to note that these two curves are almost completely parallel to each other. Therefore, we have enough reason to confirm that they have the same tendency of varlationsT4.1t can be seen from Fig. i that after the 19th cycle the curve of ._ ) rises up more and more slowly. It seems to have attained a peak in the 21st cycle. On the other h_nd, if RM continued to rise, there should be _(No.22) > R_(No.18) 151.6. "This is very improbable, because the probability or appearance of such a case isonly o.12. At the same " time, this would make the ascending time of the present 80-year cycle v too long, differing much from the past two 80-year cycles. Taking these circumstances i_to consideration, we think it is comparatively reasoziable to take the RM value of the 21st cycle as the maximum of the present 80-year eycl_. The R curve should be treated similarly. _. Therefore, beginning with the 22n_ cycle, both _. and R curves will _I m probably enter a descending phase. From the above table, the de- scending rates of the _ curves for the past two 80-year cycles are respectively: ql = 15.2, Q2 = 11.4. (4) By the same reasoning, the descending rates o£ the R curves are respectively: m , P1 = 1.5, P2 " l.l. (_ 166 1980010188-177 !! Taking these as limits, one may set the predicted values of RMand R I for the 22nd cycle to be m RM = 90 to 106, R = 3.6 to 5.2. (6)i m As for the peak value of the 22nd cycle, (7) and (8) give out very low i predicted values. This is worthy of further discussion. i 3. PREDICTION OF t AND tM THE 22ND CYCLEm We have the predicted _(No.21) = 115 to 145 and take the mean value of the lengths of the descending time of some cycles with peak values lylng in this range according with the above table. Then w- get T9 - 7.8 years. By the same reasoning, from the predicted _(No.22) = 90 _o 106 we may take its mean length of a_cending time as T1 - 3.8 years. It is known that t..(No.21) - 1980.2 _ 0.3. Hence t (N6.22_ - 1980.2 + 7.8 _ 0.3 - 1988._ _ 0.3, tM(NO.22) ffi1988.0 + 3.8m+ 0.3 - 1991.8 + 0.3. In conclusion the results of our forecasting may be tabulated as follows: tm Rm tM RM cycle No. 21 1980.2 _ 0.3 130 _ 15 cycle No. 22 1988.0 + 0.3 3.6 + 5.2 1991.8 _ 0.3 90 to 106 REFERENCES i. Xu Zhen-tao, Zhao Ai-di, Xue Yi-sheng and b_i Ya_, . ::_ta Astronomica Sinica, in print. 2. Solar Geuphysical Data, No. 4v5, part i, (1978). 3. D.[_oBHTHIICKH_, RHK_HqHOCTL _! Hporzlo3m Coa._f4:_. AMTHBHOCT_t (1973). 4. Solar Geophyslcal Data, No. 406, part 1, (1978),. 5. J. Xanthakis & C. Poulakos, Solar Phys., 56, 467, (1978). 6. J. R. Hill, Nature, 266, 151, (1977). 7. T. J. Cohen & P. R. Lintz, Nature, 250, 398 (1978). 8. K. Sakurai, Nature, 269, _01, (1978). 167 1980010188-178 PREDICTION OF LARGE ACTIVE REGIONS AT THE EAST SOLAR LIHB ® Guo Quan-Shi Solar Activity Prediction Group Purple Mountain Observatory Academia Sinlca Nanking, China I • I NTRODUCTtON The execution of space proL,a plans oftem requires forecasting of solar act;vity for a period longer than t_o weeks. However, while the forecasting for half or one day beforehand i$ easy, pr_dictlons for one or several weeks are extremely difficult. The:e are two i_in problems: (I) Because of solar rotation, it is hardly possible to predict active regions which appear about the eastern i.imb, (2) it is difficult to foresee the sudden birth of solar active regions for several days beforehand. Most of the strong active regions come from behind the east limb by rotation, :,ld only a few of them are born on the solar disk. For instance, throughout the year 1959, there were 103 sunspot groups with magnitddes > 5 (according to the classification used in the Quarterly Bulletin of Solar Activity, the 9th magnitude is the largest and the Ist is the smallest). Among them, 23 groups, or" only 224, were born on the visible disk of the sun. In the years 1970-1972, among the 78 sunspot groups v'ith _.'eas larger than 500, there were ll groups, or" only 14%, that were born on the visible solar surface. Therefore, we sha_l consider the problem of forecasting the large active regions that appear Dy rotation around the east limb. From this we attempt to estimate the probability oF proton events and propose a method for forecasting quiet periods for th_ year_ ef high level of activity. 2. FOUNDATION OF FORECAST METHOD Now the question we would like t_ study is how to foresee the large active regions that appear at the east limb. At which position on the sun should a sunspot group be born if it is co be seen at the east limb (E75°-90°) as a ripely developed group of type E or F? The answer to this question is _ndefinite, because there are many possible wavs a sunspot group may develop. But the sunspot groups that are most powerful in producing flares develop according to the curves in Figure ! (Kuiper, 1953). If what we see at the east limb is the hatched region in the figure, then spot groups of types E and F should have been born in W 60 ° - > W 90 ° of the western hemisphere of the sun in the preceding .tation. Because it 168 L 1980010188-179 ls easy to observe the birth of spots near the west limb, we may judge the _p occurrence of large active regions at the east limb after two weeks. Nevertheless, the further development of a spot af:er its birth varies greatly. Consider the conditions for it to develop into a type E or F sunspot group. Such development depends upon the degree of intensity o _ magnetic flux emerging from beneath the photosphere and by the conditions of the background magnetic fields in the vicinity of the emeroing magnetic flux. The former can be judged by the small spot groups and t:._ degree of changes of the accompanying chromospheric fiocculi. As for the latter, the conceptions proposed by many authors about the birth and development of active regions as well as the _oncep_io=_ of complexes are all very useful (Dodson and Hedeman, 1968; Bumba and Howard, 19_5; _vestka, 1976; Bumba, 1970; Hartres, 1968). With th_ conc¢:ptions of these authors, we think the following is u_eful to our method of forecasting. The emergence of new magnetic flux is not an lsrlated process. The structure of the surrounding stronger _,_gnetic fields (including spot groups and background fields) may play a beneficial role for the formation of large spot groups. The energence of a new magnetTc flux in the vicinity of existing spot groups or in residual stronger fields may give ris, u a larger probability of the formation of a spot group w;th proton flares. 3. NECESSARYCONDITIONS OF OBSERVATIONS With a routine mapptlt_g of spots (mapping "s more prof;table than photog- raphy because the latter may miss the small spots in the primary stage of .I 169 1980010188-180 " _J'lT"'7_"73"' 0 I I l I I O l II-';' -'-'Y+=" -" t -,-'-_-'- _4,¢ - " " " I@ ' - "_ .."A_"" ---.... -,.,. . _ _,.. -'-: - ""L.;, _ = °¢o_ F,'iT----':,' , , .i , , - - "' - '-' " -' .... o" I s o I I I o o o "- , I . o a.Oc'c_!111_9 b. Me,,'. 21 1969 45 U.T. 06" _ U.T. Mea_ centero£ lol_r d%s_( Figure 2. Emergence of new magnetic flux in the vicinity of a spot group gives rise to a proton active region. the formation of spot groups) and a routine Ha chromospheric patrol, one may observe the outflow of magnetic flux. But if one also possesses magnetic maps and calcium monochromatic images, he can make a better judgement about the appearance of a new magnetic flux. Someauthors think that, in two or three days before the birth of spots, the surrounding magnetic fields may be influenced. This perhaps may be measured by a sensitive magnetograph. Having no magnetograph at hand, we wish that this may be perceived from the Ha chromospheric flocculi. ,, 4. CRITERIA FORECASTING From March to December in _972, we carried out regular preliminary forecasts of the spot groups that might appear near the east limb. The criteria could be only qualitative. What could most easily be forecast was the fairly strong active regions that produce severai proton events. The criterion for forecasting such kind of active regions was that near the west limb there should suddenly appear a very bright flocculus. It must develop very rapidly, the area of the flocculus may increase in 5-6 times on the second day. Simultaneously, if there appear some small spot groups that change quickly, then when the active region of this kind rotates to the east limb, it may give rise to several proton events and short-wave fadeouts in communications. _ • The criterion for a spot group that produces one or two proton events and short-wave fadeouts is as follows: A bright flocculus appears at the west limb. It increases by two to three times on the second day. At the 17o _ -& 1980010188-181 same time, there may appear some new small spots. Besides this, sunspot groups of ordinary degree of activity correspond only to the floccull of ordinary brightness on the west limb. More attention should be paid to the case mentioned above, when flocculi appear In the vicinity of existing large spot groups. This may give rise to conditions of producing magnetic fields with more complicated structures. We considered the time interval when large spot groups may possibly produce proton events and ionospheric disturbances as being 13 days from the east limb passage to the west limb. i 5. FORECASTING RESULTS IN 1972 According to the morphology of floccuii before the formation of sunspot I groups as described above, and some other photospheric and chromosphericphenomena and estimating the rate of development of spot groups by the i changes of the fields of flocculi, we carried out experimental forecasting of the level of solar activity in the coming month; this was done twice per i month. Beginning from March of 1972, we made 18 forecasts in the nine months (on the day of forecasting we took into account the short-term forecasts of the existing spot groups on the solar surface in considering the solar activity in the coming several days). The results of forecasting are : partially given in Table i as illustrative examples. • We concluded that i,i the nine months there appeared 41 spot groups of = which the maximum areas were larger than 150 in units of millionths of thei hemisphere. Among them, 16 spot groups (39%) were forecast comparatively , accurately. The estimate of both their positions and levels of activity were generally correct. Fourteen spot groups (34%) were forecast partially . accurately. Their positions were foreseen correctly, but estimates of their levels of activity were somewhat higher or lower than reality. Eleven spot groups (27%) were forecast in vain, or nearly in vain. There were another 11 groups which were not foreseen at all, but some of them were born on the£ _ solar surface, and properly speaking, one cannot expect to see any evidence _ of them before one rotation. But it should be pointed out that this was our initial test and our experience was insufficient. Moreover, the data , used for forecasting was also not complete or fine enough. Nevertheless, as : can be seen from the results obtained, the method we used is effective and applicable for pair of the strong active regions. With the accumulation of experience and the improvement of material, it may be possible to make daily forecasts of major groups. As mentioned above, we should be more confident to forecast strong active regions because they have more evident processes of formation. This has some practical significance for forecasting both periods of low activity and periods of strong disturbances in communica- tion. We can examine as examples the three strongest active regions (all of them were active regions with proton flares) in the period of our fore- _ • casts (March-December 1972). The predictability of them are as follows: i. Heliographic location: L10°¢44 °. Date of central meridian passage CMP was August 4. Three proton events and three strong SWF's. Situation of birth of new active centers: new H_ flocculli 171 • t 1980010188-182 Table i. Comparison of predicted and observed active regions during March-June 1972. Predicted Active Regions Observed Active Regions hello- Ioca- date Comments on hello- ioca- graphic tion east degree of graphic tion to Date of Iongi- lati- limb activity, iati- lati- type miss in predic- tude tude pas- group type tude tude or predic- tion L° t ° sage or area L° .._0 area tion I (!) 259 "9 April 19 D 251 -7 J March (2) 105 -8 April 4 E,F 82 -9 D 31 90 -16 D 2 (I) 105 0-10 Hay ! _ all are 93 17 E 371 D / April (2) 128 -10 April 291 moderate 128 -6 C 23 15 (3) 206 15 April 23 or small in vain (4) 250 -(5-15) April 26 groups 251 -8 C 22 389 3 (I) E April 290-320 10-20 May 13 [ strong 300 13 D 296 Illl 29 (2) 350 -15 Hay I0 ,I groups 332 -6 F 435 371 (3) I05-120 O-lO May I 93-I05 8-17 A,E 371 (4) 20 -(15-20) Hay 7 10 -8 D 408 4 (I) Hay I00-1 IO 5-10 Hay 27 moderate 98 6 C 19 15 (2) or I00-II0-(5-15) Hay 27 small 78-91 -(9-16) C,E (3) 793 35-40 5-15 June 2 46 5 B 9 (4) 200 -12 Hay 21 J 192 -9 B 13 (5) 150-170 5-15 Hay 24 unusual 165 8 D 501 group (I proton event) 5 (1) 290 -(13-23) June IO 500 302 -13 421 E May (2) 230 5-15 June 15 300 242 8 53 569 31 6 (1) 166 lO June 20 400 165 4 184 June (2) 84 -8 June 27 600 86 -4 156 15 (3) 95 5-15 June 29 300 106 7 216 172 1980010188-183 suddenly appeared on solar surface W45°, after that its area rapidly increased and a spot group appeared. Predictability: Predicted appearance of strong proton region based on observations of rapid variation of flocculli brightness on July !1-13. 2. Heliographic location: LI64°¢8 °. CMP Date: May 30. One proton event and one strong SWF. Situation of birth of new active centers: sudden appearance of H_ bright point on the solar surface W60° during 8h-19 h U.T. May 8; the next day its area rapidly increased and passed behind the west limb on May I0. Predictability: predicted appearance of strong proton region based on observation of rapid formation of new region on May 9. 3. Heliographic location: L314°¢-12 °. CMP Date: October 30. One proton event and one strong SWF. Situation of birth of new active centers: sudden appearance of H_ br{ght floccuili and small spots, after appearance of two bright points like small flares. 6. IMPROVEMENTOF FORECASTING METHODS If we introduce the conceptions of the longitudes of activity and com- plexes of activity into the criteria of forecasting, there may be a better effect for the forecasting of proton active regions. CRITERION A: The emergence of new magnetic fluxes near the west limb. a. birth of new spots on the photospheric background or in the vicinity of spot groups. b. emergence of new magnetic fluxes accompanied with the heating of the chromosphere bright points that should i be observed in the H_ chromosphere. c. comparatively bright fiocculli. CRITERION B: Condition of magnetic fields at the place of emergence of i new magnetic fluxes. a. within the longitudes of activity (130°-2000 , 310°-lO °, i 30°-90°). b. in the vicinity of existing spot groups. : c. in the regions where before one period of rotation there existed spot groups. The above criteria have been used to examine the strongest active re- gions which appeared on the east limb in 1959-1972. The results are listed in Table 2. Predictions of quiet periods in years of high level of activity may also be deduced from the criteria mentioned above. When the above criteria do not appear near the west limb, one can reliably predict safe quiet periods two weeks in advance during each solar rotation. When only criteria Ac or Bc appeared near the west limb, one can still predict quiet periods. Besides these cases, flares or proton flares may be expected. 173 1980010188-184 Table 2. The analysis of predictability of big spot groups during 1959-1972. Numher of prot. event Max. area (A) _nd Est. ofLocation Date (i0 Y hemi- >2 SWF(B) predic- NO L ° _0 of C.M.P. sphere) _r _2 flare Aa Ab Ac Ba Bb Bc tabi|ity 1 57 17 Hay 14 959 1850 1(A)15 B + + + possib 2 336 17 June 17 959 !!09 i A,16 B + + + reliab 3 330 17 July 14 959 1847 3 A, I1 B + + + reliab 4 285 6 Dec. ! 959 2475 - 15 B + imposs 5 126 14 May 25 960 1076 - 3 B + + + possib 6 351 29 May 8 960 1750 I A, 2 B + imposs 7 32 26 Nov. II 960 1770 6 A, 7 B + + + + reliab 8 47 -7 July 14 961 1570 5 A,10 B ? + imposs 9 310 13 Sept 20 963 1241 4 A,IO B + + + possib i0 146 17 Mar. 21 966 1846 1 A, 9 B + + + + reliab 11 184 22 Aug. 28 966 1002 2 A, ? ? + + possib 12 222 28 May 25 967 2031 2 A, 6 B + + reliab 13 105 23 June 3 967 1025 I A + imposs 14 100 26 July 28 967 1774 - 6 B + + reliab 15 87 22 Aug. 25 967 956 - 4 B + + imposs 16 173 -16 Oct. 28 968 1225 I A, 8 B ? + + possib 17 344 19 Nov. 11 968 776 I A + + reliab 18 79 10 Nov. 21 969 1713 I A,13 B + + + reIiab 19 130 -8 May 27 970 532 I A + + + reliab 20 227 16 June 16 970 1487 I A + imposs. 21 225 17 Jan. 25 971 1600 I A + imposs. 22 23 8 Feb. 22 972 1049 + + + possib. 23 164 8 May 30 972 501 1 A, i B + + + _ possib. 24 10 14 Aug. 972 1179 3 A, 3 B + + + + fellah. 25 314 -12 Oct. 29 972 2299 I A, 2 B + + + + reliab. REFERENCES Bumba, V. (!_70): Large-scale magnetic fields and activity patterns on the Sun. In: Solar Terrestrial Physics/1970 (ed. E. R. Dyer), IUCSTP Sym- posium, Leningrad, D. Reidel Pub1. Co., p. 21. Bumba, V., and R. Howard (1965): Large-scale distribution of solar magnetic fields. Astrophysical Journal., Vol. 141, p. 1502. Dodson, H. W., and E. R. Hedeman (1968): Some patterns in the development of centers of solar activity, 1962-1966, IAU Symposium (ed. K. 0. Kiepen- heuer), p. 56. Kuiper, C. P. (1953): The Sun. University of Chicago Press, p. 344. 174 t 1980010188-185 Hartres, H. J. (1968): Origine des regions actives solalres anormales. In: Structure and Development of Solar Active Regions, I.A.U. Symposium35, p. 25. Svestka, Z. (1976): Development of solar active regions. Proceedings of the International Symposiumon Solar-Terrestrial Physics, June 7-18, 1976, p. 129, publ. by American Geophysical Union. 175 1980010188-186 i] 5 Chen Xle-Zhen and Zhao AI-Di Solar Activity Prediction Group Purple Mountain Observatory Nanking, China 1. INTRODUCTION Hany solar physicists have discussed and reviewed short-term forecasting ! of solar activity. Extensive observations and statistics have shown that the , outburst of flares is closely connected with the magnetic structure of active i regions, the morphology of sunspots, spot areas, the morphology and bright- i ness of flocculi, the evolution of active regions, the emission of microwave radiation and X-rays, etc. Because the physical process and mechanism of the occurrence of flares are not yet very clear, the forecasting nowadays is still a probability forecasting based on statistics. ; Our project of forecasting is based on the presently available observa- : tions at the Purple Hountain Observatory. We try to adopt a comprehensive _ project of forecasting, as simple and convenient as possible. Its aim is to foresee the level of activity of every active region on the solar surface in the coming I-3 days and thereby to estimate the possibility of the occurrence of ionospheric disturbances and proton events. 2. SOURCE OF STATISTICAL DATA This project was formulated in 1972-1973. In order to fit the descend- i ing phase of the solar activity cycle, we chose the data of 1959-1963, , belonging to the 19th cycle, and those of 1970-1972 in the 20th cycle for separate statistical study. We chose the type, area, magnetic classification and longitude of activity as four parameters. The levels of activity are defined as follows: Oth level of activity: without flares of importance i or larger. Ist level of activity: with flares of importance IB or disturbances of class !. 2nd level of activity: capable of producing flares of importance 2 or disturbance of class 2. 3rd level of activity: expressing the possibility of occurrence of flares of importance 3 or disturbances of class 3. • 176 1980010188-187 -_ i ...... i iiii • - =, ,-=, . _•.... k Statistical data were chosen from the IAU _uarterly Bulletin of Sola_ Activity, Solar Data of the Soviet Union, Solar-Geophysical Data o'f the United States, the coordinated observation in our country and the records of disturbances supplied by radio communication organizations. Active regions on the solar limb are in general not taken Into account in our statistics. Sometimes more reliable forecasts of the quiet periods that are safe from disturbances are necessary. In order to ensure the safety of space flight, we were asked for definite predictions rather than false-alarms. Thus when the level of activity attains or surpasses the 2nd magnitude, we forecast that proton events may occur. 3. FUNDAHENTALSOF TECHNIQUES First of all we choose K parameters. To fit the conditions of our i observations, we have chosen 4 parameters, i.e. (I) type of sunspot groupaccording to the Zurich classification, (2) area of spot group, (3) magnetic classification of spot group, and (4) longitude of _tivity. Let each parameter have (i) k factors. For example, the magnetic _lassification con- , sists of six factors, i.e. 6-structure, abnormal magnetic structure, magnetic axis _450 , y, By and B. The parameter of spot type has six factors: F, E, G, H, D, C. Spot areas are divided into three classes. For the cycle No. 19 we choose 80°±15 °, 130°±15o, 330°±15 ° as active longitudes. The total range is e4ual to 90o . For the cycle No. 20, we added 230°±15 o in addition to the three belts of longitude mentioned above. They occupy one-third of the longitude interval on the solar disk. We did not Include in the statistical analysis those spot groups which are magnetic type, _, or Zurich type B, J, or their area < 200. To simplify the project, we did not make separate statistics for differ- ent levels of activity. Instead, we first calculated the probability (Pi)k of producing a flare above the importance IB by every factor of forecasting. ; To compare the correlation of each parameter with activity, we determined ; the weighting factor Wk of each parameter from the probability of maximum activity (Pmax)k corresponding to the forecasting factor of the parameter I (emax)k Wk" 4 . (I) l_ _ (Pmax)kk=lL The weighted probability of activity corresponding to each forecasting factor is (Fl) k Pi( )k . (2) F (Fi) k = Wk (Pmax) k t ) r 177 F 1980010188-188 i ..... The probability of appearance of activity in each active region on the visible solar surface Is 4 (3) Wp - _ (Fi) k As may be seen by the statistics, the larger the value of Wp, the more fre- quently strong activity occurs. Therefore, from the magnitude of the value 1 of Wp, we can estimate the possibility of the appearance of activity of various classes. +_ Besides optical data, we have radio observations at 3 cm and IO cm. But both are full-disk measurements, and it is impossible to specify the flux of each active region. In order to utilize every kind of material available nowadays, we can also estimate the development of active regions : with the rate of Increase of radio flux In three days. But this should be considered together with optical observations. Especially when there exist ,; simultaneously many active regions on the solar disk, and they may recur- rently rotate off the west limb or turn onto the east limb behind the sun, then the changes of radio flux cannot signify the development of some one active region. It may only characterize the level of activity of the whole solar surface. It is similar in meaning to the total area of spots on the solar disk or the relative number of spots. Hence the rate of change of radio flux ls not a quantity that may be used every day. It is taken into account when it can definitely represent the variation of flux of a certain ; active region. 4. FORECASTING TECHNIQUES Statistics r,f the data in 1959-1963 give the following results: Table I. Weighted probabilities of activity corresponding to various Zurich types of spot groups. ZUrich Sunspot Group Type __LE_ 0_.2_Weighted Probability of Activity Fi: 0.28 0.14 0.05 0.11 0.08 0.05 Table 2. Weighted probabilities of activity corresponding to various arees of spot groups. (The unit of area is 10 -6 Solar Hemi- sphere.) Area of Sunspot Group > I000 500-1000 200-500 Weighted Probability of Activity Fi: 0.25 0.13 0.05 178 1980010188-189 Table 3. Weighted probabilities of activity corresponding to various magnetic types of spot groups. Hagnetic Type abnormal magnetic magnetic axls __6 structure _ 450 __Y I_y __B Weighted Probabil ity of Activity Fl: 0.38 0.34 0.33 0.29 0.27 O.10 Active longitude: When a sunspot group appears on an active longitude, the weighted probability of activity is 0.09. . Additional parameter: When the rate of increase of radio flux of certain active regions in three days can be definitely i evaluated to be _ 5.5t, the weighted probability of = activity Is 0.10. According to this proJect, + (Wp)max = 1.1. Criteria of forecasting: + w < 0.3 yields class 0P 0.] _ W < 0.6 yields class I + p 0.6 _W < 0.8 yields class 2 W _0.8 yields class 3p _ In the first two cases, a safe or quiet period may be forecast. In the last two cases, proton events may be forecast. ! 5. FORECAST EXAMPLES i , Example I: On August I, 1972 there was observed a large active region with HcHath number 11976. Its position on the solar disk was 13°N, IO°L. _ The parametric factor of forecasting is magnetic type 6, spherical area of the whole group was !197 (10 -6 area of hemisphere). The Z_rich type was E. Rate of increase of radio flux in three days > 5.5¢. According to the above7 project, one can find from the tables that Wo - 0.38 + 0.25 + 0.14 + O.10 - 0.87. Then activity of class 3 could be far,cast and there might be a proton event. Actually, there were strong activities on August 2, 4, and 7; proton¢ _ events happened on the last two days. _ • '3 Example 2: A large active region appeared on October 24 1972. HcHath , number 12094. Position: II°S, 314°L. From the observationa_ data of ; October 27, magnetic type was 6, area 1746, type E, rise of radio flux in 1980010188-190 recen£ three days < 5.5_. Therefore, we have Wa = 0.38 0.25+ 0.14 + 0.09 • 0.86. Then an activity level of class 3 could 6e forecast and there mlght be a proton event. The outcome was that the proton event took place on October 29. Example 3: An active reglon which turned out on September 9, 1974. Posltlon: 8°N, 268°L. Judging from the observational data of September 12, it had magnetic type 6, area 935, type H, increase of radio flux in three days > 5.5_. Then according to the project w_ = 0.38 + 0.13 + 0.II + 0.I0 = 0.72. One should foresee activity level of chss 2, and there might be a proton event. In fact, there occurred a 2N flare on September 13. 6. ESTIMATION OF THE RESULTS OF FORECASTING The above project was formulated after a statistical analysis of the data in 1959-1963. It differs not very much from the statistics of the data in 1970-1972. The former period is longer and its material more complete; therefore, it was taken as the basis. As a test we applied this project to the material belonging to the period I_lrch - July of 1972. When the data were complete, we have for 75 active regions made daily estimates of the !evels of activity in the coming l-3 days. There have been performed alto- gether 694 forecasts. Among them 532 (77_) were accurate, 98 (14_) were false-alarms, and 64 (9_) were missed. As for proton events, 70-80_ of them can be accuratel_ forecast with the project. But the false-alarms were fairly often, occurring on the average 20-30_ of the total number of days of forecasts. 7. IHPROVEHENT AND SUPPLEMENT TO THE PROJECT OF FORECASTING The project described in section 3 was proposed in 1973. With the change of phase of the solar cycle, it has to be revised adequately. It has been found from practice that merely by correcting the standards of areas and active longitudes, this project still _olds for quiet years or the ascending phase. For example, near quiet years spots of large area occur quite seldom, and most flares appear in small spot groups with complex mag- netic polarities. Accordingly, the areas may be divided into three cla[ses: (I) < 200, (2) 200-500, (3) > 5OO, but the corresponding Fls value is unchanged. Active longitudes are Incessantly shifting; therefore, they should be adjusted often. After analyzing the data of activity in 1973-1976, we have ascertained four new belts of active longitudes, (I) 300-900 , (2) 140°-210 °, (3) 260°-290 °, (4) 320°-350 °. Examining the material in recent years, we find that several spot groups that appeared earliest of all in the new cycle were all situated in the above-mentioned four belts. The strong active regions that appeared in recent years were all in these longitude_. In the practical execution of the project, one may also take into account the following aspects: 180 t . + 1980010188-191 (I) When several active regions are very closely spaced to each other or when active complexes appear, sufficient estimate should be made about their activity. (2) Some related parameters cannot be accurately determined for the active regions, which are situated near the limb. In these +ases judgements may be made From the circumstances of faculae, Ha plages, limb ejections, etc. I (3) Besides magnetic types one can +.lso consider as reference thegradients of magnetic fields. They can be estimated with the degree of close approaches of spots with opposite polarities. | (4) The growth of spot areas can also be utilized in considering the development of spot groups. (5) The arrangement of the filaments in the _enumbrae of spots, the chromospheric Filaments which may shear across magnetically neutral lines, all these may be taken as references in the short-term Forecasting. REFERENCES Severny, A. B., and N. V. Steshenko (1970): Hethods for the forecasting of solar flares. In: Solar Terrestrial Physics 1970, (ed. E. R. Dyer), D. Reldei Publ. Co., p. 173. Simon, P., H. J. /_artres and J. P. Legrand (1969): Flare forecasting. In: Solar Flares and Space Research (ed. C. DeJager and Z. Svestka), North Holland Publ. Co., Amsterdam, p. 405. : Smith, J. B. (1970): Real-time prediction of solar flare locations and re_ion activity level_____s,AIAA Paper No. 70-1373, From the AIAA Observa- tion and Frediction of Solar Activity Conference held at Huntsville, Alabama, Nov. 16-18, 1970, 799. Smith, J. B. (1972): Predicting activity levels for specific locations within solar active regions. In: Solar Activity Observations and Predictions (ed. P. S. Hclntosh and H. Dryer), HIT Press, p. 429. Svestka, Z. (1970): Particle event forecasting. In: Intercorrelated Satel- lite Observations Related to Solar Events (ed. V. Hanno and D. E. Page), D. Reidel Publ. Co., p. 90. 1980010188-192 ,m_ • • ' ]8°0-18478 FORECASTSOF GEOHAGNETICACTIVITIES AND HF RADIO PROPAGATIONCONDITIONSHADEAT HIRAISO/JAPAN K. Harubashi, Y. Miyamoto, T. Kidokoro and T. Ishii Hiraiso Branch, Radio Research Laboratories 3601 Isozaki, Nakaminato, Ibaraki, 311-12 Japan The Hiraiso Branch of RRL forecasts the geomagnetic activities for a 24-hour period through the IUWDS network. Our prediction techniquesare summarized separately for the 27-day recurrent storm and the flare-associatedstorm. The storm predictions are compared with the actual geomagnetic activities in two ways. The first one is the comparison on a day-to-day basis. In the second comparison, we evaluate the accuracy of the storm predictions during 1965-1976. In addition to the storm prediction, short-term predictions of HF radio propagation conditions are conducted at Hiraiso. The HF propagation predictions are briefly described as an example of the applicationsof the magnetic storm prediction. I. INTRODUCTION The Hiraiso Branch of the Radio Research Laboratories conducts several kinds of solar-terrestrialpredictions. They include the predictions of geomagnetic activities and energetic solar flares under the program of the InternationalUrsigram and World Days Service (IUWDS) as the Regional Warning Center/Tokyo,and short-term predictions of HF radio propagation conditions averaged over the long-distancepropagation circuits from Japan to various directions. This review concentrates on the forecasts of geomagnetic activ- ities and short wave propagation conditions. More weight is given to the ; magnetic storm prediction,because the propagation prediction largely relies , on the storm prediction. ' The daily prediction of geomagnetic activities is issued in the form of ; either of three key words: bIAGALERTfor the disturbed period, _GNIL for the end of disturbvd conditions,and b_GQDIET for the quiet period. Our present criterion for the issue of bIAGALERTis that the daily sum of Kp indices is expected to exceed 25 or the commencementof such a magnetic storm is ex- pected. The decision of daily prediction is made around 0100 UT and sent through the IUWDS network. In addition, this prediction message is broadcast for the users in the Western Pacific region by Japanese Ursigram Broadcasts, JJD (10,415kHz) and JJD2 (15,950 kHz) together with other forecasts and all collected Ursigram data• The prediction of magnetic storms provides a basis for the prediction of HF radio propagation conditions. In addit_,n, when an unusually severe magnetic storm is expected, the Hiraiso Forecast Center defines the Special Interval for Ionospheric Observations in which ionograms 182 1980010188-193 are to be obtained every five minutes from five ionosonde stations of RRL. Recently, the magnetic storm predictions are used by satellite tracking, control and telemetry stations, because it has become increasingly evident that the ionospheric radio scintillations associated with magnetic storms can affect satellite communications systems at frequencies from HF and VItF even u_ to SHF. For this purpose, the storm prediction is communicated directly to some operational stations. Two types of forecasts of radio propagation conditions are issued from the Hiraiso Forecast Center, which apply to the overall characteristics of HF skywave propagation averaged over the various long-distance propagation cir- cuits around Japan. One is the Weekly Radio Disturbance Forecast (or briefly Weekly Forecast) which predicts the daily propagation quality and the possi- bility oc SWF occurrence for the coming seven days. The propagation quality is defined by deviations both in the time interval during which HF radio signals are received and in the received field strength from monthly average values statistically estimated from long-term observations. The predicted qualities are expressed by the quality figures: 1 = very poor_ 2 = poor, 3 = slightly poor, 4 = good, 5 = very good. The expectation of SWF is expressed in three grades from 0 to 2. This Weekly Forecast is sent to the registered users by mail and also sent in RATEF code through the IUWDS network. The other forecast is the Short-Term Radio Disturbance Forecast (or briefly Radio Warning) which predicts the average propagation quality during the next twelve hours. The quality is expressed by the letter code: W = severely disturbed, U = unstable, N = normal. This radio Warning is broad- cast on the Japanese standard waves, JJY, in the Morse code every ten minutes. 2. AVAILABLE DATA The real-time monitoring of the solar-geophysical phenomena is of " fundamental importance for any kind of solar-terrestrial predictions. The observations made at Hiraiso and their use for the monitoring and pred.; tionpurposes are briefly described below: (1) Observation of sunspot regions in white light. Sunspot regions are ._ sketched around 0000 UT every morning. The sketches are used to assess the probability of flares from each active region by considering day-to-day variations in the sunspot area, the spot number, and the structural complex- i ity of the spot group. (2) _leasurements of solar radio flux and polarization at 100, 200, and :, 500 bgtz. The flux measurement is made also at 9500 MHz. The spectral _ features and time variations of solar radio outbursts in these frequencies ,-_, are useful factors for prediction of subsequent occurrence of magnetic storms. (3) Riometer observation at 30 _z. The riometer detects particularlylarge X-ray flares as the sudden cosmic noise absorption events. The i spectral type II solar radio bursts are also detected sometimes by the t riometer.[ (4) Measurements of short wave field strength. The measurements are t made for various long-distance propagation circuits, which include standard waves, commercial communications, overseas broadcasts, etc. The data are used i to decide the actual propagation quality. In addition, we can detect S_F i phenomena for a considerable part of day by these measurements. (5) Amplitude and phase recording of pulse signals from Loran C (100 kHz) i 183 1980010188-194 ]i stations in Japan. This providcs us with a very sensitive detector of SID { effects. (6) Measurements of geomagnetic field variation and earth current. These provide the real-time geomagnetic data which are indispensable for the forecasters to see the progress of magnetic storms. The pattern of geo- i magnetic variation is an important factor to presume the cause of a magnetic ! storm and also to infer whether the disturbance will further continue or i subside soon. i In addition to these real-time monitorings, near real-time data are provided by some observatories in Japan as listed below: i (1) Reports on solar active regions observed by 3-cm and 8-cm wavelength ] radio interferometers from Toyokawa Observatory of Nagoya University. They i report the east-west positicn, 5-cm flux, and 3-cm to 8-cm flux ratio for each active region together with their expectation of proton flare in the URALRcode of Ursigri. ,i (2) Geomagnetic K-indices from Kakioka Magnetic Observatory. t/hen a 1! magnetic storm or other special events occur, they supplement description !i• of the events such as the start time, the end time, the maximum range in the ! horizontal component, etc. (3) Reports on SPA phenomena detected by VLF recordings from Inubo Radio Observatory, as an indicator of flare activity. i (4) The daytime and nighttime typical values of fmin from Wakkanai Radio I_ Observatory of RRL, as an indicator of the ionospheric absorption, i: ! The Ursigram is another important data source, which we rely upon in pursuing the evolutionary changes in the solar-terrestrial environment on a , i day-to-day basis. Although the Ursigram data provide only simple descriptions ilof physical phenomena, the full time coverage affor d by the world-wide data _ exchange is the most important thing. We pay special attention to (1) de- !i scription of solar active regions by sunspot groups (USSPS) and calcium + plages (UPLAK), (2) flare reports (UFLAE), (3) reports on solar radio events ! (URANJ), (4) reports on solar proton events detected by radio propagation (UABSE) or by satellite measurements (PRESTO messages), and (5) SID event reports (USIDS). Furthermore, the reports from Culgoora in plain language provides us with detailed description of solar active centers and activities ° of each region obtained from both optical and radio observations. The report on the Culgoora radioheliograph observations is particularly important for the magnetic storm prediction. In addition to these real-time monitoring and prompt reports cited above, we use later observation reports such as weekly reports and monthly reports from many observatories to identify the cause of each magneti_ storm. This identification of the causes of storms is important to understand the .,j, recurrence tendency of geomagnetic disturbances correctly. For this purpose, _, we particularly rely on the coronal hole data from He I 10830 A spectrohelio- . grams of Kitt Peak National Observatory, which is provided by the Preliminary Report and Forecast of Solar Geophysical Data published weekly by ERL/NO_ in Boulder. This is the only data source on coronal holes which is regularly and pro_tly availabe, and plays an important role in our decision of storm prediction. 184 t 1980010188-195 q3. TECHNIQUESFORMAGNETICSTORMPREDICTIONAND THEIR BASIS I_ is well known that there exist two types of magnetic storms, namely, 27-day recurrent storms and flare-associated storms. The prediction tech- niques for these t_o types of magnetic storms are quite different because of the d_ference in the origins of solar wind disturbances which cause magnetic stor_:.. The recurrent-type magnetic storms are caused by high-speed coro- t:_tba_ streams which emanate continuously from a certain region on the sun, _,ne1¢_s the flare-produced shock waves trigger the flare-associated storms. Jur techniques for the prediction of recurrent-type magnetic storms are mainly based on recent studies which attempted to identify coronal holes as the origin of the corotating fast streams and their associated geomagnetic disturbances (Krieger et al., 1973; Neupert and Pizzo, 1974; Bell and Noci, 1976; Nolte et al., 1976; Hansen et al., 1976; Sheeley et al., 1976). At present, however, the coronal hole data are used only to identify the _auses of re, urrent-type magnetic storms and thus confirm the recurrence tendency. t Lhe factors and criteria used in prediction of the recurrent-type storms ! are l_sted below: (1) Recurrence tendency as viewed from 27-day recurrent patterns <:onstructed by K indices from Kakioka. If significant magnetic disturbances were observed in the past two rotations or more, the strong recurrence tend- eacy i_ assumed for the period concerned. In this case, consideration is given to the rotation-to-rotation trends in the start time and magnitude of _/ magnetic disturbances.(2) As ociation between coronal holes and magnetic disturbances in the ./i pa_t rotations. The first priority is given to the association in the last I_: rotation. When the association is evident, the possible recurrence is as- slued even if the magnetic disturbance occurred only in the last rotation and _ not in previous ones. i (3) Evolutionary changes in the coronal hole regions. The return of cor_'naJholes which had caused a magnetic storm in the last rotation is inferred from the _istribution of active regions on the present solar disc. If there is a reg±on void of calcium plage or sunspot group at the right place, the return of coronal hole is assumed. Similarity in the magnetic neutral line and active region distributions between the rotations is also taken into consideration. (4) The return of regions which had produced energetic flares on the + previous rotation. This factor is rarely used positively to predict a magnetic storm, but is particularly important with regards to the evolution of coronal holes. Tl'_re are several examples which show the development of coronal ho.lesin co.relation with the occurrence of proton flares. Such i ,nples will b_ discussed later in some more detail. ' i (5) Near ,'eal-timecoronal hole data. The existence of coronal holes is sometimes reported from Culgoora. In such a case, the east-west position ; of the coronal hole is compared with the coronal hole distributio_ in the previous rotation to see whether the reported coronal hole is ne_ one or returning one. For the returning coronal hole, the association between the corr'al hole and magnetic disturbance in the past rotations is checked. In s_me cases, large near-equatorial coronal holes are seen which do not cause i any significantmagnetic disturbance. The relationships between magnetic storms and solar flares have long been stud_ed particularly aimed at identifying the characteristics of the , 185 1980010188-196 iI i 'i storm-producing flares (see Cook and McCue (1975) for a historical review). I Such works provide the basis of the prediction of flare-associated magnetic i i storms. Among others, our prediction techniques are based on the studies on : i the solar radio bursts of spectral type II and type IV (DeFeiter et at., 1960; Bell, 1965; Dodson and Hedeman, 1964), the radio energy emitted at meter wave- : lengths (Sinno and Hakura, 1958; Nishida, 1965; Pinter, 1972), the shock wave propagation from solar corona to the earth's environment ( Pinter, 1973; Sakurai and Chao, 1974; Dryer, 1975), and the solar proton events (Hakura and Lincoln, 196.';Castelli et al., 1967, 1977). When solar flares have occurred, the occurrence of a magnetic storm is predicted considering the following factors: (I) Ha importance and duration of the flare and the structural complexity i of the flare region. The £1ares of Ha Imp _ 2 in magnetically complex sunspot : groups are considered to be possible to cause magnetic storms. (2) Position of flare on tilesolar disc. Flares in the central region of the disc most probably cause magnetic storms, and flares in the western hemi- sphere tend to cause magnetic storms with higher probability than those in the eastern hemisphere. Flares near the east limb are normally disregarded in magnetic storm prediction. (3) Occurrence of radio type II and type IV bursts. Greater importance is attached to type IV bursts particularly at meter wavelengths. The highest _ probability of magnetic storm is expected in case of combined occurrence of • type II and type IV bursts. (4) Intensity and duration at meter wavelengths. The probability of _ storm occurrence is assessed for each of flare-associated radio events with : peak flux around 200 MHz greater than 200 s.f.u, and duration longer than 15 • minutes. The highest expectation is given to those events of which the : product of peak flux densit[ around 200 MHz (in s.f.u.) and duration (in minutes) is greater than I0". The existence of the second part of major radio burst related probably with type IV is taken into consideration when the event is locally observed at Hiraiso. (5) Frequency coverage of the radio event. The full frequency range coverage from about i00 to I0,000 }_z is an important factor to estimate the flare-released energy. Although we pay more attention to meter wave emission in connection with magnetic storm prediction, the increase in the flux density with frequency in the centimeter range is an important controlling factor of solar proton events. (6) Occurrence of solar proton events. The occurrence of magnetic storms can be predicted with higher probability of success by knowing that solar protons were emitted from the flare under consideration. (7) Delay time from flares to magnetic storms. We normally assume two days delay from solar flares to magnetic storms. Shorter delay time is as- sumed for flares just after fast streams, e.g. in case of successive flares or flares just behind coronal holes. As is evident, none of the factors cited above is conclusive to predict the occurrence of a magnetic storm for either of recurrent-type or flare-associated storms. The final decision is made subjectively by com- bining the degreee of expectation judged from each factor. When a magnetic storm occurs, we must forecast how long the storm would continue. Although it is normally assumed at Hiraiso that the duration of a magnetic storm is about three days on the average, the end times of magnetic storms are actually predicted from some features noticed in the [ 186 1980010188-197 magnetograms. The techniques are quite different for recurrent-type storms and flare-associated storms. The end of a flare-produced magnetic storm is soon expected when the storm proceeds to its recovery phase as noticed in the | H-component variation and short-period fluctuations are small. On the other i hand, in some recurrent-type storms, no substantial main phase develops, but ii bay-type disturbances are repeated in the nighttime for several days. The I end times of such events are predicted by considering the following factors: !I (I) the day-to-day trend in magnitude of bay disturbances, (2) the intensity of short-period fluctuations in H-component, and (3) the definiteness of the i! regular pattern of diurnal magnetic variation around the morning hours. Thu_ it is important for the prediction purpose to know whether the magnetic storm in progress is flare-associated or recurrent-type. Furthermore, to identify the causes of magnetic storms in the past it is necessary to understand the recurrence tendency correctly. The above-cited _i factors and criteria for prediction of the storm occurrence provide the ways I_ to discriminate between the recurrent-typestorms and _be fl_re-associated storms. Features of magnetic field variation are also tskL_ into consider- ation. The prompt and tentative identification is later su,_plementedby _" examining various data such as those in Solar Geophysical 9ata (SGD: published monthly by ERL of NOAA, Boulder). It is particularly hellJfulto examine the variations in the solar wind velocity, density and temperature, and to com- pare the coronal hole magnetic polarity with the interplanetarymagnetic field polarity. Finally, we must mention the crucial role of the interplanetarymagnetic field (IblF)in controlling the _evelopment of magnetic storms. Burton et al. (1975) showed that the main pha_:eof a magnetic storm is initiated when the north-south component of IMF becomes negative (southward)and its development i is strongly controlled by the product of the north-south INF and the solar• wind speed. This result indicates that the storm prediction based solely on the prediction of high-speed solar wind from coronal holes or flare regionsis not sufficient. Fortunately, however, the most intense fluctuations in IMF are generally associated with the region of interaction between the high- _ speed solar wind and the ambient solar wind, and thus the high-speed solar ! wind streams are usually accompani_.dby the southward INF (Burlaga, 1975). g i 4. SUI_dARYOF MAG_'ETICSTORM PREDICTIONS DURING BARTEL'S ROTATIONS 1961-1980 The magnetic storm predictions issued by Hiraiso and WWA are compared _ with the actual geomagnetic activities for the Bartels' rotation periods 1961-1980 in Fig. 1. Geomagnetic activities are exhibited by ZKp (thick solid line), ZK from Kakioka (dashed line) _nd the lowest Dst level in every 12-hour interval (thin line with small dots). Tb_ magnetic storm intervals and SC's _ taken from SGD are also shown by horiz_._talbars and triangles, respectively. , In particular, the storm intervals and SC's reported from no less than five _ observatories are indicated by thick bars and solid triangles. The intervals for which MAGALERT's were issued are shown by black shadings. The shading !_ with a white dot denotes the btAGALERTissued on the day to which the storm prediction is to be applied. The hatching indicates that the MAGALERThaving been issued for that day before was cancelled by the issue of b_GNIL. For major magnetic storms and some additional minor disturbances, their possible 187 1980010188-198 -IO0. I_1101¢rlCMII_ ' --_- ,, ,":"_.• _.... G'..'-,-r,-,-,-,nr_ ml0' OIL'IP'I d_. IIOIIIIIIL_I_Nlqlloll_'r loll m,=,,.,&"+,f _ : . "+.. , =--'!+_• IO., IIIID4CTIONlq ' '' "Ir'''"Jll' 'i'' ;'II .... _"='='o.'+__Ira'°"" " ® i -+_,n,.. •-+.•. . .,..•..-._...,..., ,-'0+,,,+ ' ' ' "IS' ' ' ' '20" ' ' ' "_' ' ..... S =-_++-.. -:--. . ....... :_ pl(llOlCTiOtI li I"" ' _ .... '_.... _ ,, , _,," '_':'_',, ,+o'-"T net.I -0OO J_0 +Vl ...,+e O_ ,,4 . #.o+ IP<_UIINI_TIC pmmo'l'_I_ ,,*--+il+tll_CflOII I'Ul_ Ill l Ilet i ".._21Z-'___:_ Fig. 1. Comparison of magnetic storm predictions with actL, al magnetic activities for Barrels' rotations 1961-1980. 188 1980010188-199 REPRODUCIB_,ITY OF'THE ,. I_e"lm"_'! " i ' ' ' 11'1' ' ' 'S' ' ' ilk'tO' ' '-- ,__ i-'_ _ CF I_ ii " " I_O._-:=_ _ _i.. I_ _ :-._.."_:':_"__.'""..... _ "_,,/ '4_,_ , . _,. ,. • __ I ,:l___ ^ IIIIIDICTI _q _,Im _ l_t,I ! z_" 'W }' _ ' '_S.... _ ' ' ' "li' , 1 I / ,-so._m..k'.L._=_:l_ _':11_-_."_ _ ./--.=_x/ i _ ] ..... -r--:-7_rl=_-'TQ.'_-"-_'_"_---=L: L o. ,o.."._._ "-.'-_ "-; -_ " llm_oecT_o_ N, ---_- ._ _-,_,,_IL__I;_.I__ _._ _ 189 1980010188-200 association with coronal holes or solar flares is indicated by attaching the letter I! or F. The association with the unstable or questionable coronal holes is indicated by Hu. The magnetic polarity of the coronal hole is also shown. Here, the major magnetic storm is defined as the storm reported from no less than five magnetic observatories whose Dst fell below -50y (nT). Such major I magnetic storms are numbered in Fig. 1. I_ During the rotation periods 1961-1970, no flare-associated magnetic storm 3 occurred until September, and therefore alI storm predictions were made by #, consideration on the recurrence tendency. It is found that most _GALERT's ji were issued after disturbances had started. This is attributed to the fact that the recurrence tendency is not so evident during these pcriods. The existence of a quiescent period over 40 days f._om rotation 1966 to rotation ,: 1967 particularly made it difficult to predict magnetic disturbances there- after. Although weak recurrence tendency is found from rotation 1962 to 19C4, the association between these disturbances and coronal holes is not very clear except for the storm No. 2. It is evident that the storm No. 4 is not the return of this recurrent series but is rather to be associated with another coronal hole with positive polarity which disappeared in the next rotation. There are five intervals for which MAGALERT'shad been issued before disturbances were observed. The V_GALERTfor January 4-7 was issued on Janu- ary 3, which was based on the distinct recurrence tendency observed in the I_ previous five or six rotations. Th_s prediction resulted in false alarms, because of breaking-up of the recurrence, though the minor disturbance : occurred. The perplexity of the forecasters at Hiraiso and also at _A is clearly seen by comparing the actual geomagnetic activities and predictions during March 5-10. The MAGALERT's issued beforehand resulted in false alarms ! in this case, too. The MAGALERT's for blay29-51 and for September 1-3 are , other two examples which were issued beforehand and resulted in false alarms. Both of these two predictions were made by the consideration of the apparent4 i association between the coronal holes and magnetic disturbances in the pre- ceding rotations. For the latter case, the magnetic quiescence could have been predicted by closer examination of data, because the UPLAK code of the ; Ursigram had reported the existence of a calcium plage region at the position of the former coronal hole. The fifth interval of _GALERT_s issued before the occurrence of magnetic disturbance is September 11-13, and this prediction agreed satisfactorilywith the actual disturbance. As is discussed above, there are a number of magnetic storms which are apparently associated with short-lived coronal holes. These storms cause the poor accuracy of our recurrent storm prediction which relies upon the coronal hole data in the previous rotations. However, the above discussion implies the possibility of much improvement by using the near real-time data of coronal holes. It is clearly seen in Fig. i that more _GALERT messages were issued before the occurrence of magnetic disturbances during the rotation periods 1971-1980 than before. There are two reasons for this change. The first reason is that there occurred storm-producing flares after the major activi- ties of b_c_lathregion 14943 during September 7-22, 1977, and therefore HAGALERT's were issued with higher degree of confidence for those events. This is particularly true for the great solar flares such as those at Mc_,_ath region 15266 in April-_lay. The second reason is that there were distinct series of recurrent magnetic storms in these periods together with many flare- associated storms. 190 1980010188-201 ................ __ I m iii [1_ .... I]lk I IIIII ii _. We can point out three series of recurrent storms for these periods. The first series began with the SC-storm having started at 0027 UT on July 29, 1977 (-to__ ,,u._'-6), and continued to the gradaal storm on December I0 (storm No. 22), though we attribute the SO-storms on September 19 and 21 to flares in McMath region 14943. The increase in _Kp on January 8, 1978 before the storm No. 24 may be regarded as the continuation of this series. The evolution of the coronal hole responsible for this series of recurrent storms seems to have been correlated with the development of active centers in the vicinity, that is, the growth of the coronal hole was seen in September and January. The second series began with the SO-storm at 1853 UT on January 28, 1978, in rota- tion 1975 and continued to the storm No. 32 in rotation 1977. These storms can be associated with the extension of the southern polar hole with negative polarity. The storm No. 37 in the next rotation seems to be the continuation of this series, though the coronal hole had disappeared in this rotation. This could be an interesting example which exhibits the return of a magnetic storm without a coronal hole. Although the storm No. 41 in the next rotation (1979) appears to be another return of this series, it does not seem true because the magnetic polarity of the solar wind was positive during the period of this storm. The storms, Nos. 8, 10 and 13 (rotations 1969-1971) seem to constitute another series of recurrent storms which are associated with the extension of the northern polar hole. In addition to the above-discussed storms which are relatively easy to predict, there are number of storms which are associated with sporadic coronal holes in the rotation periods 1971-1980, too. Table 1 shows further details on the association between storms and : coronal holes or solar flares foz major magnetic storms. In the table, the storm commencements of uncertain type are indicated by brackets (Nos. 32 and 41, see also Fig. I). The magnitude of storms is represented by the sum of Kp indices for the 24-hour period which gives the maximum value and the lowest Dst level. The coronal holes listed in the table are those within + I0° lon- gitude from the estimated source location of the solar wind which could cause the magnetic storm and within _ 30° latitude from the sub-earth point on the sun. The estimation of the source location was made with the solar wind speed listed in the table, when the HIT experiments on IMP 7 and 8 are available. Otherwise, the time delay of 2-4 days are assumed between the CHP of coronal holes and the ar_,,_l of corotating streams. The longitude ranges of coronal iholes are shown as defined within _30° latitude range of the sub-earth point. The coronal holes are classified into three types: near-equatorial (E), ex- tended from the northern polar region (PN), and extended from south (PS). Unstable or questionable coronal holes are indicated by attaching U. The coronal hole data taken from the Preliminary Report from ERL/NOAA are brack- eted. The polarity of coronal holes are determined by comparing coronal hole boundaries with Ha synoptic charts. Coronal holes crossed by neutral lines are indicated by the X polarity. The polarity of the mean solar magnetic field is also listed, which is taken from the Stanford measurements for the date when the solar wind started from the source region near the sun, the constant solar wind speed being assumed. The IMF polarity inferred from mag- netograms at Vostok and Thule is shown for three 12-hour intervals, in the first of which each magnetic storm commenced. It is seen that the coronal hole polarity generally coincides with the mean solar magnetic field polarity and the IHF polarity in the second 12-hour interval. The solar flares were taken from the Ursigram data. We listed those flares which seem capable of 191 i 1980010188-202 _o 5 _ o o_,oo)qlo _ • !o _ • oo q_ 0 . i % ° it!!| _ ! = 0 ! ou ! = ! .,.," _ _|, _ _, ._i_s__._... ."°_I_°_!. • ) 4..I I I •+ 4.= I + + + i I+ t .... ._1 _ _.>, _ ¢ m + 4" + _ 4' +4. + + + ! 4. + ).4 Jn_ I 4" @ o _, - ? ,_, _, ..... . : , , - . , 4,_ o . !o I_ _ SO _ gO f_ (h ql'JN ql' i I'_ 0 N • F. bl : : : _ _' : :_' : : I 192 1980010188-203 • _ __ " -- I I II ill lUg .......awli,,.=.=,..=._..... ~ ,_ , . ,n _ _ ,_ -*. 1980010188-204 producing magnetic storms. Therefore, it is not intended to say that all of the listed flares actually caused magnetic _torms. For exact!e, the flare on December I0, 1977, cannot be the cause of the storm on December I0 (No. 22). It is supposed, in this case, that the flare-associated stream was just behind the high-speed corotating stream and thus no significant effect could not be detected. The flare-associated type II and type IV events are taken from reports in the Ursigram from Culgoora and Nancay. In the last column of Table I, the association between these 44 major magnetic storms and coronal holes or solar flares are shown with the degree c_ confidence in three grades: triangle = possible, circle = good, and double circle • excellent. These grades are determined by the consideration of the coincidence of magnetic polarity between coronal holes and IMF, the delay time from the coronal hole CMPto the storm commencement, the delay time from flares to magnetic storms, and some characteristic features in the solar wind such as the stream interface and the shock structure. The question mark attached to the storm No. 18 indicates that this storm is probably caused by the coronal hole-associated corotating stream though no coronal hole data are available. The storms of Nos. II, 27 and 38 are wo;thy to be s:_cially men- tioned, because the times of the flare occurrence nearly coincide with th_ times of coronal hole CMP for these events. After we concluded t,.t these storms are more likely to be associated with flares as listed in the ta_e_ we noticed that the flare position was just on the west side of the coronal hole for the former two cases. It would be reasonable to suppose that the flare-associated shock will arrive at the earth earlier than the corotating stream, because the shock will propagate ahead of the stream. Finally, it is interesting to note that 40 storms out of 44 major mag- netic storms could be explained by coronal holes or solar flares. It car Le expected that the causes o_ the remaining 4 storms may become ciearer with more detailed data. The present result provides a strong basis for the storm prediction by observations of coronal holes and solar flares including radio events. S. SEPTEMBER1977 EVENTS During 7-22 September, 1977, McMath plage region 14943 was very active and most solar-geophysical activities were associated with this region, thvugh there was short intermission of activity during 12-15 September. Hiraiso issued bIAGALERT's for two intervals, 11-13 and 18-23 during this period as is seen in Fig. I. In this section, we will show how the magnetic storm predic- tions were decided and how the predictions agreed with the actual geomagnetic activities during this period. We will discuss the sho_t-term predictions of HF radio propagation conditions made at Hiraiso in association wi_h the activ- ities of _Math region 14943. Fig. 2 illustrates the solar phenomena closely related with magnetic storm prediction (upper diagram), and the Kp variations for the period con- cerned (lower diagram, of which dates are two days shifted). The solar phenomena were taken from the Ursigram-interchanged data source, and the following are shown in the figure: (1) The Ha importance, the starting time, and the longitude of all reported flares which occurred in McMath region 14943 and its vicinity. (2) The frequency spectrum of peak flux density and the start, maximum, 191_ 1980010188-205 ...... z -----_ 11111i, k.-- . and end times of selected solar radio bursts with wide frequency range cover- age and relatively long duration. (3] The _tar_, maximum, and end times, and the maximum absorption of PCA events observed by riometers at Russian Arctic stations. Proton events and electron events are indicated by letters P and E, respectively. The first major solar event occurred at 2227 UT on September 7, when McMath region 14943 was just behind the east limb at about 10°N latitude. We observed a major radio burst event at 100, 200, 500, and 9500 MHz starting at 2228 UT (See Ohbu et a1., (1978) for the radio data obtained at Hiraiso during the September 1977 events), and the SWF event of importance 3 at Hiraiso. At 0100 UT on September 8, we did not know the position of the flare which was responsible for above mentioned phenomena yet. Therefore, we postponed our decision on the probability of a flare-produced magnetic storm, and issued * NAGQUIET. Later, Culgoora reported the IB flare at 2227 UT (start time) at _ the IO°N east limb followed by the 2B flare at 2252 UT. Culgoora also reported a stronF type II at 2227 UT and a weak type IV at 2340 UT. The¢ occurrence of a L,icrowave burst event of 1900 s.f.u, was also reported by thes .i PRESTO message from Boulder. The radio heliogram at 3750 _g_zsent by facsim- _ ile from Toyokawa also informed the fla,'eposition. Although HAGALERT 09/10 and MAGALER_ 09/11 had been issued by _YA and RWC/Paris, respectively, proba- bly from this solar event, our conclusion was that this flar_.J_as too far east to cause a magnetic storm. R_ROgUCI_I"_t_,OF 1"tt_t _- oRIGI_A. ?_\GL IS pOOR r sw 1977t , 14 , IS '- POSITION (4C _**a *i IMP_q_rANC[ o . mifl_* i o o imp. I _10 oIM z w .0 _ 0,_,. s ,o I LJ (P ,o I , 4 IMXIMUM Iil I0" _. minE) ..._ I ,,__30mln _----.-'*) ,u | IN lie SIN( IPCA] _Z l 1[i.1[¢1'119#_([) _ _ I ' I0 ' II ' Is_&' l]i ' 14 s 18 _ ill * I? _ IO _ _1 _ 20 ' S_l&' 22 z 23 Fig. 2. Solar activities associated with McMath region 14943 during 7-21 September, 1977 (upper), and geomagnetic activities during 9-23 (lower). Note that dates for two diagrams are two days shifted. 195 1980010188-206 The first interval of MAGALERTfrom Hiraiso (II to 13) was decided on September 9 from consideration of the recurrence tendency as seen in Bartels' rotation periods 1968 and 1969 in Fig. I. We issued MAGALERT 11/15 on Sep. 9, intending to express our negative expectation of a magnetic storm by the September 7 solar event, and also our expectation of a recurrent magnetic storm. The 2B flare at 10°N 80°E at 1604 UT (start time) on September 9 was the second major event which accompanied the major radio event and the PCA event. However, little attention was paid to this flare, because MAGALERThad already been issued for 11 to 15. The values of EKp for September 11, 12 and 13 were 27, 11_, 51+, and the SC-storm occurred at 2113 UT on September 12. Thus, our prediction, _GALERT 11/13 appears to have been correct judging from Hiraiso's criterion for MAGALERTissuance. However, we are not quite certain whether the reasoning for the prediction was correct or not. The September 9 solar flare could have caused the SC-storm on September 12, though the flare posi- tion seems too far east. On the other hand, this storm could have been caused by the fast stream from the coronal hole which passed the central meridian of the solar disc on September 11, though the delay time between the coronal hole CMP and the magnetic storm for this association seems a little too short if not unthinkably short. The second interval of MAGALERTfrom Hiraiso (18 to 2,) can be derided into three epochs according to our decision processes as follows: (1) The flares on September 16 around 22 hours UT were accompanied by very large bursts of meter waves (Hiraiso's observation available), and Culgoora reported very strong type II and type IV emissions. (The flares were : reported as 1N and 3N flares from Culgoora, and as a 2B flare with two maxima from Boulder.) Thus, a magnetic storm was expected to commence on September 18 or later. Therefore we issued MAGQUIET on 17, and bIAGALERT 18/20 on Sep. 18. The geomagnetic activity was low on 18, but this prediction was contin- ued as MAGALERT19/20 on September 19. (2) The SO-storm commenced at 1138 UT on September 19, we concluded that the disturbed state would continue till September 21. Thus, MAGALERT 20/21 was issued on September 20. The importance 2 solar flare on September 18 was disregarded because the associated radio burst was rather weak at meter wave- lengths and its duration was short (Hiraiso's observations available). (3) On 21 September, we expected one or two magnetic storms to occur on 21 and/or 22 September, judging from the major solar flares on 19 and 20 and the associated radio phenomena. (The latter flare seems more favorable judging from the meter wave intensity; Hiraiso's radio observations being available.) Thus, MAGALERT 21/25 was issued on September 21. This prediction wa:_ extended till 24 on September 23, and denied again on September 24. (b_GNIL was issued on 24). The agreement between the prediction and the actual activity was satis- factory during the second bIAGALERT interval. The bIAGALERT for 18 was the only false alarm, and no dis_urbace was missed. It can be said that the solar events in this interval were sufficiently large so that we could predict the occurrence of magnetic storms with high degree of confidence. Finally, it should be mentioned that these £1are-associated storms occurred just in the period when a recurrent magnetic storm could be expected. Two important suggestions can be derived from these events. The flare effects may have arrived at the earth before the corotating stream because the flare position was on the west side of the coronal hole, and that the growth 196 1980010188-207 J I I lll!l 'I_ of the coronal hole may have been associated with the development of this active center. In Fig. 3, two kinds of short-term predictions of HF propagation condi- tions made at Hiraiso are compared with the observed HF field strength data ! from representative propagation circuits for the period September 9-23. The i diurnal variation of monthly median values are shown by the s_ooth curve for each circuit. The dashed curve is drawn for the time interval in which the signal reception rate of this month is lower than 50 _. Values of observed signal strength (in dB; relative value) are indicated by dots, and values affected by interferences are indicated by v-marks. The large (Importance = 3) SWF's detected at Hiraiso are shown by solid triangles. The hatchings in the lower diagram indicate the intervals for which unstable HF propagation conditions were predicted. _, In the Weekly Forecast for September 10-16 issued on September 9 (date in Japan Standard Time) the unstable propagation conditions were predicted in accordance with our expectation of a recurrent magnetic storm during September 11-15. The Warning was changed from N (Normal) to U[Unstable) with expecta- tion that the propagation would soon become unstable, when we observed the minor magnetic disturbance of September 11. Although the HF monitoring showed '- _. 197? © -I© : II_tZ I0 -IO 20 IOMI_ 5o i0 I0 50 40 lOII4I "I/30 2O Fig. S. Comparison between the short-term HF propagation predictions and the HF field strength data during September 9-25. 197 t 1980010188-208 'i t that this disturbance did not affect the propagation conditions so much, we concluded again that the propagation conditions would become unstable, when the SC-storm on September 12 was observed. Thus, the Weekly Forecast for ! September 14-20 issued on 15 predicted tbe disturbance in the HF prc2agation ", conditions for September 13. These HF propagation disturbance predictions t] resluted in false alarms as is evident in Fig. 3. _ It can be seen in the figure that the HF signal strength was rather weak during September 19-25 in association with the magnetic storm condition for _ this period. The Warning was changed from N to U again at 0200 UT on Septem- ber 19, because we expected a magnetic storm would commence soon. Thus, the Warning seems to have been correct in this case. However, we could not pre- dict the unstable propagation conditions in the Weekly Forecast for 17-25 issued on September 16, because we could not predict the occurrence of the storm-producing flare on September 16. The difficulty of the short-term HF propagation disturbance predictions is well demonstrated inthese two examples, one success and one faiJure. In the filst case, the magnetic storm was not so severe to affect the propaga- tion conditions, whereas in the second case the storm was enough to affect the HF propagation. Thus, the difficulty is that it is needed to predict the magnitude of the magnetic storm, which is strongly controlled by the north- ! south component of IMF. 6. STATISTICAL EVALUATION OF MAGNETIC STORM PREDICTION The accuracy of magnetic storm predictions can be statistically evalu- ated by comparing MAGALERTmessages w_th actual disturbances. Strangely enough, there is no _greed definition of geomagnetically disturbed days among _A and RWC's according to which MAGALERTshould be issued. Therefore, we adopt here our own definition of disturbed days that ZKp is no less than 25 to compare the disturbance predictions with actual disturbances. In Fig. 4 showing histograms of ZKp for 1965-1976, the shaded portions present the number of MAGALERT's issued from Hiraiso and _A vs. ZKp. The hatched portion indicates the number of MAGALERT's issued for those days on which magnetic storms commenced in late hours and therefore ZKp is small. These comparison between predictions and ZKp, particularly those for 1974-1976 suggests that our definition of disturbed days provides a reasonable criter- ion with which predictions are evaluated. The accuracy of these storm predictions are statistically evaluated by using the measure of correctness, R, which was originally defined by Sinno (1958): g = ::{D (x + y)}-l/2 (1) where x is the number of disturbed days correctly forecast, D is the total number of disturbed days, and y is the number of quiet days for which storm forecas_ _ere issued (i.e. false alarms). If we define two quantities A and B as A = x / D (2) S = x/ (x+y) (3) 198 1980010188-209 NIIAIIO _ MlPlAIIIO WWA .............. f ;iI,.,/ _-;,,_ .-.;--;. ,_,,_. .-.;-_, d -"%..- -,:,. . • mm m m .-m @ m . m din--@ @ (0) } ,o m . "gg_3 i_F " " " " _" ) • m _ _ m _ m m ,m, Q m a " '" ,o I10 I0 in lid 410 80 gO 0 4_ I10 K) 80 lad 410 !0 OIO Fig. 4 Number of bIAGALERT'sissued by Hiraiso and _A (shading)° superposed on histogram s_of EKp for 1965-1976. 13CI_I_'vrq OY_ ; then A is the rateO_ [ icted disturbed days to the total disturbed days, • and 8 is the rate of correct disturbance forecasts to the total disturbance *, i forecasts; R being simply the geometric mean of these quantities. We will call A and 8 the storm prediction rate and the correct alarm rate, respec- tively, and _ the measure of synthetic correctness hereafter. Table 2 shows the values of D, x, and y during the period of 1965-1976 for Hiraiso and _A. ; Although the storm predictions are evaluated only in terms of (1)-(3) in this ,, report, evaluation by other methods is possible with this table. Fig. 5 presents the variations in the storm prediction rate CA), the correct alarm rate (B), and the measure of synthetic correctness (R) calcu- lated for the MA6ALERTmessages issued by Hiraiso and _A from 1965 to 1976. It is noticeable that the storm prediction rates are particularly high in 1974-1976 at both Hiraiso and WlqA. This is caused by the sudden increase in the number of MAGALERT'sin 1974 (see Table 2). As is evident from (2), the move bIAGALERT'swere issued, the higher storm prediction rate is obtained. The main reason why many MAGALERT'swere issued in 1974 and thereafter is that there occurred many recurrent-type magnetic storms in those years, and thus forecasters could easily predict the storm occurrence. The correct alarm rates are also high in 1974-1976 and the correctness measures are at the 199 1980010188-210 Table 2. Numbers of disturbed days (D}, correct storm _! predictions Cx}, false alarms CY), and total 1 storm predictions (x+y) for MAGALERT'sissued ! from Hiraiso and WWAduring 1965-1976. HIRAISO WWA ii! Year D x y x+y x y x+y ,, 1965 21 6 10 16 1 S 6 1966 47 14 17 31 12 19 31 _ 1967 55 lb 11 27 12 S 17 ,_, 1968 73 1S 17 32 14 1S 29 1969 52 11 24 35 15 18 33 1970 56 11 19 30 19 24 43 1971 55 7 11 18 20 41 61 1972 59 5 10 1S 15 26 41 1973 131 16 13 29 69 41 110 1974 167 129 49 178 130 54 184 1975 89 68 $9 127 71 70 141 !1976 84 61 32 93 61 48 109 highest level of the twelve-year period of this study. The recurrent tenden- cy of geomagnetic disturbances was similarly stron_ also in 1975 (Ondoh and Nakamura, 1977), and correspondinglyWWA issued many MAGALERT's. As a result, the MAGALERT's from WWA in 1973 accomplished r_ther high storm prediction rate and measure of synthetic correctness. In contrast to this, the number of MAGALERT issuance from Hiraiso does not show any increase until 1974. • , , • , • • o Z _ Fig. 5. Year-to-year variations in the storm prediction rate e ........... (A), the correct alarm rate LO ..... ... ...... (B), and the measure of sy- ---A WwA thetic correctness (R) for -o-a n _ MAGALERT's issued from Hira- o --'Am"_.t_0 _ iso and WWA during 1965-1976. _ • , .| L i _ i • • ° • |(B Ill g7 m gll 7071"/_71 "N 70 "_ YEAR 200 1980010188-211 ........... , .,i IIII I -- - ,.... In particular during 1970-1973, the number of MAGALERT'swere very small in comparison with those from _A (see Table 2), and thus the storm prediction rates were at the lowest level in those years. It is also interesting to note in Fig. 5 that the trend of the measure of synthetic correctness (R) is roughly in parallel with the trend of the storm prediction rate (8). This result probably implies the measure of syn- thetic correctness of storm prediction is more strongly controlled by the storm prediction rate than by the correct alarm rate. The definition by (3) indicates that the highest correct alarm rate is accomplished by minimizing the number of false alarms (y), or in other words, by issuing MAGALERTonly when the storm occurrence is expected with high degree of confidence. However this way of MAGALERTissuance will result in the low storm prediction rate and probably in the low correctness measure. It can be concluded from above dis- cussion that the number of HAGALERT's_ere too small in the earlier times of the twelve-year period of this study, although it depends on the way of appli- cation of forecasts which is more important, the storm prediction rate or the correct alarm rate. Minnis (1977) showed the assessment method of economic values of fore- casts by comparing the actual prediction merit with the merit obtained by a simple hypothetical prediction method. Here, we will compare the above sta- tistical evaluation of MAGALERT'swith a hypothetical prediction under the assumption that the following day will always be the same as the preceding day. Suppose that there occurred a magnetic storm of the S-day duration, then this hypothetical method yields S-I correct storm predictions (x = S-I), one false alarm for the next day of the storm end (y = 1), and one miss of alarm for the starting day of the storm. By counting the numbers of correct storm predictions and false alarms for a given interval, we obtain D = x+ y and therefore A = 8 = R. The correctness meaures were calculated for this hypo- thetical prediction with the actual variations of EKp, and the result is shown in Fig. S by the dashed line (Note that this one line exhibits the variations of A, B and R). It is really disappointing that the measure of synthetic correctness for the actual prediction is no better than that for the hypothet- : ical prediction. Parti=ularly the results of actual predictions by Hiraiso during 1970-1973 are Much worse than those by the hypotnetical method, while the results for WWA are comparable to the hypothetical results. Reversely speaking, the preseut comparison indicates that the correct and early detec- tion of events i_ important for the purpose o£ magnetic storm prediction, and it may be concluded that forecasters at Hiraiso had failed to understand magnetic storms in progress. Fortunately, the trend of improvement can be seen in 1974-1976 at both Hiraiso and _A. It is also to be noted here again that the hypothetical method considered here can never predict the starting day of magnetic storms nor the isolated storms of one-day duration. Table 3 shows the results of similar statistical evaluation of MAGALERT messages for the Barrels' rotation periods 1961-1980. The accuracy of the prediction for the first half period is much lower than that for 1976. The statistical accuracy of the hypothetical prediction is also low during this period. These results seem to be similar to those for 1965, and probably caused by the breaking recurrence tendency during the sunspot minimum year as was previously mentioned in Section 4. During the rotation periods 1971-1980, the statistical accuracy of the prediction becomes higher again with increasing flare-associatedmagnetic storms. The measure of synthetic correctness is much higher than that of the corresponding phase of solar cycle 201 1980010188-212 Table 3. Evaluation of MAGALERTmessages from Hiraiso and _qA in terms of the storm prediction rate (A), the correct alarm rate (8), and the meas- ure of synthetic correctness (R). (a) Rotation periods 1961-1970 Forecaster D x x+y A 8 R Hiraiso 44 15 41 0.34 0.37 0.35 WWA 44 18 63 0.41 0.29 0.34 Hypothetical 44 16 44 0.36 0.36 0.36 {b) Rotation periods 1971-1980 Forecaster D x x+y A 8 R Hiraiso 69 44 88 0.64 0.50 0.56 WWA 69 49 97 0.71 0.51 0.60 Hypothetical 69 38 69 0.55 0.55 0.55 (c) Rotation periods 1961-1980 Forecaster D [ x x+y A B R Hiraiso 1131 59 129 0.52 0.46 0.49 WWA I13] 57 160 0.59 0.41 0.49 Hypothetical 113[ 54 I13 0.48 0.48 0.48 20 or the solar maximum years. This result probably manifests the improve- ments of techniques for magnetic storm prediction which have recently been brought about by enhanced data collection and exchange. REFERENCES Bell, B. (1963): Solar radio bursts of spectral type II and type IV: their relations to optical phenomena and to geomagnetic activity. Smithonian Contr. Astrophys., 5:259. Bell, B., and G. Noci (1976): Intensity of the Fe XV emission line corona, the level of geomagnetic activity, and the velocity of the solar wind. J. Geophys. Res., 81:4508. Burlaga, L. F. {1975): Interplanetary streams and their interaction with the earth. Space Sci. Rev., 17:527. Burton, R. K., R. L. McPherron) and C. T. Russell (1975): An empirical relationship between interplanetary conditions and Dst. J. Geophys. Res., 80:4204. 202 1980010188-213 Castelli, J. P., J. Aarons, and G. A. Michael (1967): Flux density measure- ments of radio bursts of proton producing and nonproton flares. J. Geoph_s. Res., 72:5491. Castelli, J. P., and W. R. Bare,on (1977): A catalog of solar radio bursts 1966-1976 having spectra] characteristics predictive of proton activity. J. Geoph_s. Res., 82:127S. Cook, F. E., and C. G. HcCu_ (1975): Solar terrestrial relations and short- term ionospheric forecasting. Rad. Electr. Engineer, 45:11. DeFeiter, L. D., A. D. Fokker, H. P. Th. Van Lohuizen, and J. Roosen (1960): Solar radio events and geomagnetic storms. Planet. Space Sci., 2:225. Dodson, H. W., and E. Hedeman (1964): Problems of differentiation of flares i with respect to geophysical effects. Planet. Space Sci., 12:393. ." Dryer, H. (1975): Interplanetary sho k waves: Recent developments. Space Sci. Re____v.,17:277 Hakura, Y., and J. V. Lincoln (1963): A suggestion for improving forecasts of geomagnetic storms. J. Geophys. Res., 68:1563. Hansen, R. T., S. F. Hansen, and C. Sawyer (1976): Long-lived coronal i structures and recurrent geomagnetic patterns in 1974. Planet. Space Sci. ! 24:381. Krieger, A. S., A. F. Timothy, E. C. Roelof (1973): A coronal hole as the source of a high velocity solar wind stream. Solar Phys., 29:505. Hinnis, C. M. (1977): Ionospheric disturbance forecasts: assessment of their economic value. Telecomm. J., 44:328. Neupert, W. H., and V. Pizzo (1974): Solar coronal holes as sources of geomagnetic disturbances. J. Geophys. Res., 79:3701. Nishida, A. (1965): Storm sudden commencements as solar blast waves. Rep. Ionos. Space Res. Japan, 19:201. Nolte, J. T., A. S. Krieger, A. F. Timothy, R. E. Gold, E. C. Roelof, G. Vaiana, A. J. Lazarus, J. D. Sullivan, and P. S. Hclntosh (1976): Coronal holes as sources of solar wind. Solar Phys., 46:503. Ohbu, K., K. Huranaga, S. Isozaki, A. Mitobe, T. Isobe, and E. Ohuchi (1978): Solar radio emissions of 100, 200, 500 and 9500 MHz observed at Hiraiso. Solar Terr. Environmental Res. Japan, 2:82. Ondoh, T., and Y. Nakamura (1977): Solar cycle variation of 27-day recurre_,t tendency of geomagnetic activity during sunspot cycle 20. Solar Terr. Environmental Res. Japan, 1:90. 203 1980010188-214 ?Pinter, S. (1972): Flare-associated solar wind disturbances and type II and IVm radio bursts. Bull. Astron. Inst. Czech., 25:69 Pinter, S. (1975): Close connection between flare generated coronal and interplanetary shock waves. Nature Phys. Sci., 243:96. Sakurai, K., and J. K. Chao (1974): Motion of sources for type II and type IV radio bursts and flare-associatedinterplanetarydisturbances. J. Geophys. Res., 79:661. Sheeley, N. R. , Jr., J. W. Harvey, and W. C. Feldman (1976): coronal holes, solar wind streams, and recurrent geomagnetic disturbances: 1973-1976. Solar Phys., 49:271. Sinno, K. (1958): Hit rates of radio propagation disturbance warnings and SWI warning. J. Rad. Res. Labs. (Japan), 5:109. Sinno, K., and Y. Hakura (1958): On the relation of solar eruptions to geomagnetic and ionospheric disturbances, II. On the types of solar radio outbursts. Rep. Ionos. Space Res. Japan, 12:296. 204 1980010188-215 "11-B,O-18479 A REVIEW OF SHORT-TERM FLARE FORECASTING ACTIVITIES AT TOYOKAWA Shlnzo Enome Toyokawa Observatory Research Institute of Atmospherics Nagoya University 13 Honohara 3-chome REPRODUCIi]Ii.I'IT OF THE Toyokawa 442, Japan ORIGINAL PAGE IS POOR Short-term forecasting of solar activity conducted at Toyokawa is briefly reviewed. The forecasts are based on microwave observa- tions of the slowly varying component of solar radiation associated with active regions. Forecasting started in 1966 while partici- pating in the Proton Flare Project on an experimental basis. Since May i_69, daily forecasts have been issued in the URALR Code through the UP.SIGRAMnetwork. L INTRODUCTION The _lln item of prediction at Toyokawa is the occurrence of proton flares. Predictions started in May 1966 to take part in the Proton Flare Project, and were applied, on an international basis, to the selection of a solar active center, which was to be observed in the extensively coordinated studies of solar and geophysical _ctivities, such as Proton Flare Projects (PFP) in 1966, CINOF (1972, 1973). On the occasion of the first PFP in July 1966, the prediction from Toyokawa was entirely successful (Simon, 1969). We will Join the coming SMY in 1979 through 1981. ; Forecasts from Toyokawa are also used, on the regional basis, for fully- prepared observations of active regions and as fundamental data for the warning of shortwave communication condition. Application was made of forecast for the determination of balloon launching timing to observe solar neutrons and gamma rays from flarihg regions during the level flight in 1968 in the International A_tlve Sun Year (Takakura and Ito, 1970). AVAILABLE DATA Lecal source of radio emission above an active region is most prominent at microwave frequencies. This radio emission is associated with coronal condensations, with typical density and temperature of 10 8 - 10 9 cm -3 and 2 x 106 K, which are permeated by strong magnetic fields of sunspots. At Toyokawa Observatory, we have lambda 3.2-cm (9400 MHz) and 8-cm 205 1980010188-216 (3750 I_I¢) radio Interferometers. Both sets of radio Instrllments are almost identical in geometrical configuration to give the _ame spatial resolution of 1.1, 0.4 and 2.2 arc minutes for adding, compound and T modes respective- ly at 3.2 and 8 cm. Host useful observational data with these instruments are one-dimen- sional brightness and polarization distributions at 3.2 and 8 cm with 1.1 arc minutes resolution. The term brightness and pol_rization, more strict-. ly, correspond to the sum and the difference of right-handed and left- handed circular polarizations. Active regions usually appear in the middle latitudes of the northern and southern hemispheres of the Sun. When s_veral active regions are visi- ble on the disk, it is in most cases possible to discriminate the contribu- tion of radio emission from each region, so long as the fan beam does not drift across more than two active regions at tile same time. If such geo- metrical alignment of active regions h_p_ens, which is not frequent but not entirely rare, we can still discriminate "one-dimensionally overlapped" active regions with pencil beam (T-mode) observations. After these procedures we have microwave radio flux of each active region, which is called slowly varying component or simply S-component, with information on structure of the region in brightness and polarization at 3.2 and 8 cm. An example of fan beam and pencil beam observations is shown in Figure 1 at both wavelengths. Fig. 1. Fan-beam and pencil-beam observations of the Sun at 3.2 and 8 cm. Contour level and interval are roL:ghly 15000 K at 3.2 cm and 40000 K at 8 cm. D!orth is top and east is left and geographical north and south are indicated by tick marks. Polarization (top) and brightness (middle) are shmm for drift curves. 200 % 1980010188-217 PREDICTION TECHNIQUES Parameters used in prediction are 3-cm (round number for 3.2), 3-cm to 8-ca flux ratio and spatial polarization distribution at 3 cm of each S- component. For each active reg,,on with 3-cm flux greater than 5 s.f.u. ! (10-22 _-2Hz-l) we report on a message of east-_est position, 3-cm flux, 3-cm to 8-cm flux ratio together with observational conditions and observed i spatial features. When 3-cm flux is greater than 10 s.f.u, and 3-cm to 8-cmflux ratio is greater than unity, comments are supplemented, which include ,, fl,_x ratio in percent with three digits and proton flare expectation in four i grades from 0 to 3. This last expectation number can not be strictly ob- i Jective, since it depends on the forecaster's personal experience, indi- vidual feeling and so on. }_ These pieces of information are coded according to URALR code of URSIGRA_ and sent through its network. An example of UP_LR message of i March 28, 1976 is as follows: URALR 44301 28096 757XX 20048 Xl161. In plain language, this reads URALR message from Toyokawa that an active region at 57 g east in solar radius has 3-cm flux of 20 s.f.u, with 3-cm to ; 8-cm flux ratio of 116 %, and is eruptive with spatial polarization feature favourable for proton flares, with proton flare expectation of one, in four grades from 0 to 3. , There was a proton flare on March 28 at 1936 UT, but this prediction was not successful in a strict sense, since March 28, 1976 was Sunday and this message was issued about 0200 UT on March 29 from Toyokawa. BASIS FOR TECHNIQUE It was found that the J-cm S-compouent is a useful indicator of flace activity by French group (Kundu, 1959; Pick, 1961). In 1964, Tanaka and Kaklnuma pointed out that active regions with 3-cm to 7.5-cm (4000 HHz) flu3; ratio greater than unity are closeZy 'elated to proton flares. Their work was developed later to investigate spatlal structure of S-component in brightness and polarlzat_on with respect =o proton flare productivity (Tanaka and Enome, 1975). As for the emission mechanisms of microwave S-component, it is generally accepted that they are frea-free emission of thermal electrons and gyro- resonance emission at second and third harmonics of thermal electrons both in a coronal condensation. The former mechanism mainly contributes to un- polarized halo component and the latter to polarized core component. For intense active regions the brightness temperature of core component is as high as 2 x 106 K at 8 cm and 5 x 105 K at 3.2 cm. Flux of these core component is as much as 29 s.f.u, or more both at 8 cm and 3.2 cm. Our interpretation of these data is that ar 8 cm optical depth is thick, where degree of polarization is pretty low, but not thick at 3.2 cm, where degree of polarization is high. This means that electron temperature above an intense active region is as high as 2 x 106 K, where magnetic fleld is 450 207 1980010188-218 Gauss, if 8-cm radiation is emitted at the third harmonics of electron gyro- frequency. Possible stronger field of 1100 Gauss emerges into corona, in- ! ferred from 3-cm radiation. _ Tanaka and Enome (1975) have analyzed 259 active regions in the period 1967 through 1971,and have shown that active regions with 3-cm flux greater than i0 s.f.u, and 3-cm to 8-cm flux ratio greater than 80 % have Jlose cor- relation with the occurrence of proton flares from them. It is also shown that very powerful active regions, which release energetic proton_ into interplanetary space several times during their disk paesage, are associated with an S-component of more than 20 s.f.u, at 3 cm and 8 cm. It it inferred from this that it is neccessary but not sufficient that powerful proton centers are accompanied by elect_on temperature of 2 x 106 K and by magnetic field of 1100 Gauss permeated into the corona. Tanaka and Enome further have shown that active regions with a definite type of spatial polarization dis- tribution across them are closely related to proton flare production. This polalization distribution at 3 cm is called "P-configuration", which con- sists of polarized central part with sub-regions on either side of the cen- tral part, which are polarized oppositely to the polarization sense of the central part, as seen in Figure 2. 1972 AUG I J_j S400 MHz E W P-configuration Fig. 2. Strlp-scan curves of the Sun in brightness (bottom) and polari- zation (top) at the wavelength of 3.2 cm on August i, 1972 (taken , ; from Tanaka and Enome, 1975). 208 L- 1980010188-219 As the gyroresonance theory teaches us (Kaklnuma and Swarup, 1962; Zhele- znyakov, 1962), right-handed (R) and left-handed (L) polarized waves orlgl- , nate in a hot plasma magnetized negatively and positively respectively. A P-conflguration region, therefore, corresponds to field alignment of - + - or + - +, which has close correlation with anomalous magnetic polarization regions at photospheric 1eve! such as collidlng active regions, shear or rotation of sunspots within single penumbra. From optical and/or magneto- graph observations it is well known that those magnetlcally anomalous regions produce intense flares (e.g. Smith and Howard, 1968). EVALUATION OF FORECAST During the periods of July 1974 and of August 1975 through September 1977, fifteen proton event days were recorded by earth-orblting satellites NOAA 2 and 3 and IMP 7 and 8 (Mangls, 1975; Solar-Geophyslcal Data) as listed in Table I. Table 1 List of Proton Events in July 1974 and in August 1975 - September 1977 Date Start Max End Location McMath Comment 7/ 3/74 0828 0836 0908 SI5E08 13043 Proton Alert (i)* -- 7/ 4/74 0641 0648 0700 S15W04 13043 Proton Alert (i) 1353 1354 1500 S16W07 13043 2048 2058 2126 S16W12 13043 7/ 5/74 1509 1514 1522 S15W26 13043 Proton Alert (I) 2135 2139 2153 SIOW26 13043 7/ 6/74 1842 1905 2028 SI5W40 13043 8/22/75 1027 1028 1029 N26W87 13811 No Alert 1230 1230 1232 N26W87 13811 11/21/75 0610 0619 0710 SOTW21 13937 No Alert 3/23/76 0840 0843 0909 S05E90 14143 Behind the Limb 3/25/76 1308 1318 1352 S05E69 14143 No Alert | 3/28/76 1843 1938 2348 S07E28 14143 Proton Alert (i) 4/30/76 2101 2109 2125 S08W46 14179 No Alert 8/22/76 1158 1212 1230 S02W90 14366 Behind the Limb 9/ 7/77 2225 2258 2420 NIOE90 14943 Behind the Limb 9/16/77 2219 2308 2500 NO7W21 14943 No Alert 9/19/77 1027 1101 1147 N08W57 14943 No Alert 9/24/77 14943(?) Behind the Limb * Proton Flare Expectation Grade in 0 to 3 Range. Three of these eve.-s were certain to be a behlnd-the-llmb event and one is suspected to be it, therefore, it is absolutely not possible to forecast their occurrence with our technique, since it _s based on the spectrum of S-component visible on the disk of the Sun. Fifteen events in the above Table were produced by powerful proton centers of McMath 13043, 14143 and 14943, including the above-mentloned four invisible hemisphere events. Among twelve disk event days0five event days were predicted correctly on 209 ]9800]0]88-220 24-hour basis. If we take longer predicti_ _uterval of, say, three days, prediction is true for seven cases out of t: lye.. During these periods we have issued 27 messages of proton alert, of which thirteen were positive with expectation grade 1 and one expectation grade 2. The rest of forecasts were negative with expectation grade 0, all of which turned out to be true. As for the accuracy of the zorecast, it may be very difficult to define, since the view point of forecasters may be quite different from that of users. Forecast accuracy of our proton alerts, when they are issued, is evaluated to be 5/27 + 14/?7, which is about 70 % on 24-hour basis. Mean probability of proton event day during these periods is about 15/800, which is about 2 %, while the accuracy of positive proton alerts is 5/13, which is about 38 %. This is about twenty times the mean probability of proton flare day. These values of accuracy may be compared with other forecast center's values (Simon and Mc_ntosh, 1972). From Table 1 it is readily seen that for very sporadic proton events such as those on August 22, 1975, November 21, 1975 and April 30, 1976, we could not even issue proton alert. It is extremely difficult by the present technique to predict their occurrence, since they are produced in a small active region or they are associated with an active region with very rapid evolution. For reference we list radio status of active regions on the above days in the Table 2. Table 2 Radio Status of Active Regions on Unpredictable Proton Event Days 8/22/75: No S-component greater than 5 s.f.u. 11/21/75: McMath 13937 at 26 % west with 5 s.f.u, and ratio _0 %, eruptive, low brightness. 4/30/76: McMath 14179 at 80 % west with 13 s.f.u, and ratio 80 %, burst in progress, double structure. CONCLUDING REMARKS For proton flares which take place in intense active regions, it is possible to predict their occurrence with certain accuracy, but for certain class of proton flares which occur in a small active region or which are associated with an active region, which evolves very rapidly, or for those which occur behind the limb, it is almost impossible to forecast their occurrence with our present technique. REFEKENCES Kakinuma, T. and Swarup, G. (1962): A Model for the Sources of the Slowly Varying Component of Microwave Solar Radiations. Astrophys. J., 136: 975. Kundu, M. R. (1959): Structure et proprletes des sources d'actlvlte solalre sur ondes centlmetrlques. Ann. Astr., 22:1. 210 1980010188-221 Mangis, S. J. (1975): July 1974 Solar Activity and Related Geophysical Effects. NOAA Technical Memorandam ERL SEL-38, Space Environment Laboratory, Boulder CO, U.S.A. :- Pick, M. (1961): Evolution des emissions radioelectriques solaire de type IV et leur relation avec d'autres phenomenes solalre et geophysiques. Ann. Astr., 24:183. ? Simon, P. (1969): Introductory. Annales of IQSY, Vol. 3, Proton Flare Project (The July 1966 Event), Ed. A. C. Stickland, The MIT Press, Cambridge MA, U.S.A., p. 3. {: Simon, P. and Mclntosh, P. S. (1972): Survey of Current Solar Forecast , Centers. SolarActlvity Observations and Predictions, MIT Press, Cambridge MA, U.S.A., p. 343. Smith, S. F. end Howard, R. (1968): Magnetic Classification of Active Regions. IAU Symposium No. 35, Structure and Development of Solar Active Regions, Ed. K. O. Kiepenheuer, D. Reidel, Dordrecht- Holland, p. 33. Takakura, T. and Ito, K. (1970): High Energy Radiations Associated with ,! Solar Flares (in Japanese). Ka_aku (Science), 40:340. Tanaka, H. and Enome, S. (1975): The Microwave Structure of Coronal : Condensations and Its Relation to Proton Flares. Solar Phys., 40: _" 123. Tanaka, H. and Kakinuma, T. (1964): The Relation between the Spectrum of Slowly Varying Component of Solar Radio Emission and Solar Proton ; Event. Report lonosph. Space Res. Japan, 18:32. Zheleznyakov, V. V. (1962): The Origin of the Slowly Varying Component of Solar Radio Emission. Soviet Astronomy-AJ, 6:3. 211 $ 1980010188-222 J, N8'O-1848,0. PADI 0 PROPAGATI ON PREDI CTI ON SERVICES I N JAPAN Riklo MAEDA Radio Research Laboratories 4-2-1Nukui-Kita Koganei, Tokyo 184, JAPAN Japanese prediction services for HF radio communications are outlined in relation to prediction method, performance and evalua- tion. The current prediction is based on the CCIR interim method with some m_difications matching to a computer system. The princi- pal service is the monthly median prediction issued regularly three months in advance for various communication circuits. A daily pre- diction for short-distance circuits is being prepared by using real time ionospheric sounding data. It is reportable that an evaluation theory and practice will be introduced in the future prediction ser- vice. !. Introduction The purpose of this review is to sketch a barest outline of Japanese radio propagation prediction services to professional and amateur users of HF radio communications. It is limited here to the frequency/time and path- loss predictions and excluded from the so-called radio disturbance warnings. The Radio Research Laboratories have been operating the Western Pacific Re- gional Warning Center of the International Ursigram and World Days Service, the executive facility of which is maintained by the Upper Atmosphere Re- search Section of the Hiraiso Branch. On the other hand, the regular ser- vice on radio propagation prediction is now made by the Ionospheric Radio Prediction Section of the Radio Wave Division which has been conducting the routine ionospheric soundings at Wakkanai (45°N, 142°E), Akita (40°N, 140°E), Kokubunji (35°N, 135°E), Yamagawa (31°N, 131=E) and Okinawa (26°N, 128°E) in Japan for applications to ionospheric radio communications. The Japanese prediction service on ionospheric radio propagation began with issuance of the Radio Propagation Prediction in 1947. This prediction had been based on a manual method before 1971, and at present is similar to a CCIR interim prediction modified for the TOSBAC 5600/160 electronic computer. Objects and forms of the prediction have been considered from a social and economi- cal standpoint for long-term planning and short-term operating of interna- tional and domestic communication circuits. There are two monthly predic- tions published three months in advance, one of which predicts usable fre- e quency-time ranges for various communications among main cities over the world, and the other is a prediction of time ranges appropriate for assigned 212 1980010188-223 operational frequencies on ship-to-shore communications between foreign har- bors or fishing banks and Tokyo. These prediction services are regularly offered to hundreds of domestic facilities and a few foreign organizations. In the following sections, the current prediction is described about its method, performance and evaluation, i 2. Method of predictionIn order to plan and maintain an effective communication system, it is desirable to know the upper and lower limits to the range of frequencies within which usable operational frequencies should be chosen. As these fre- quency limits are generally found for any radio communications via the iono- sphere, the radio propagation prediction is intended to provide information about effective usage of assigned frequencies. HF radio signals for communications are absorbed by and reflected from the ionosphere, therefore the ionospheric propagation is seriously affected by the density distribution of ionization in the D, E and F regions. The quality of communication depends also on the noise field which is notably _ controlled by man-made noise rather than atmospheric noise at the receiver. It is necessary for radio propagation predictions to take models of ioniza- tion and noise in the radio wave medium varying with environment parameters, i.e., the geographic position, time of day, month of year, levels of solar and geomagnetic activity and so on. The modeled ionization and noise can be numerically represented over the world by means of function series with the environment parameters, where the ionospheric characteristics are indi- cated with monthly med:an values and the level of solar activity is repre- sented by a predicted t,velve months smoothed running average sunspot number. If the normal radio propagation is assumed to be confined within the medium : bounded between the ionosphere and the earth's ground, the signal attenua- : tion, direction and mode will be calculated by empirical formulae derived from practical experiences Accomplishment of the prediction requires• ? further information about circuit parameters such as the transmitter power, the transmitting and receiving antenna characteristics, and the minimun re- ceivable field strength. AFter the above preparation, the highest and low- est usable frequencies (MUF and LUF) are determined and also the pathloss is ; calcu]ated for any operational frequency on a communication circuit. ! The computer software for the current prediction in Japan is derived from the CClR Reports 252-2 [Jones and Gallet, 1962] and 340 [Barghausen et al., 1969], with some program conversion to a TOSBAC 5600/160 computer. As an example of the prediction calculation, a computer output is reproduced concerning radio communicative features of the Hawaii-Hiraiso circuit as noted in Table 1 [Ichinose and Yamaoka, 1973]. The monthly median predic- tion is useful not only for operation plans but also for specification de- signs of communication circuits. For an actual operation of a communication circuit, a daily prediction service is urgently needed because the day-to- day MUF and LUF are hardly estimated by the monthly prediction. 213 1980010188-224 Table I. Output of monthly median prediction. M_qt_ _ _$N= J4 ||_E LOCAL TIME TI,. 'tAMNII _O*BON l_6*_Oa _Et- MIRAIbU $0,40_ 14U,_OL 8E&_|_ T*q |O0 N.I 8? DISTANCL 6ZbTK_ P_*ER L,O_ ANT|N*_A _A|N TN, O RE, 0 3M_( N3JSE=-laODBw T|'4E* _0 PEkCE_T _E_,SIt.=2OD_ GMT LT MuF _*U_ _.UO ,.00 q.UU |_.UU lb.UU |iQO0 22.0G gb.OO JU=OO FHL_. IF,HZ) 2 II 31.b _L _L ,.E _F j_ _f 2F 2F 2F Z_ _O_b -!_. !18 ,o| L,, o"'* t!l-_60 - 67 -_? -l_ -LO2 ._ 22 3e 47 _9 b_ S/_ D_,60 _.DA_b ,_ ._, ::_ ._'_' ._,:_ :_ _ ,_UU cub *93 _ _b_ _JPN_D ,UO ,UU ,UO ,OU ,6L .93 .q8 .g3 , * 13 3Z,I :_ .oo."oo'_".o_.,.,,"".,,.'".,." .,,""::_ .,,.6'_"°"_:"_0,_0 *OU ,UO ,35 ,94 ,Y9 .gB g4 15 _4_3 / ;_; "|' " _" tH !}_ -_-,.. t_!-*'_ .|_ ,1i -,," ._ ._ .,_ ._ ._, ._ 3, ._ ._ I_["_§ _7 _6.b _ 3F JF 3P #F 3! 3F 3F U O _ODL AN_L£ _a_g -9_ -IO_ .|O6 *IC_ _ -It3 -lie 0 _ bl_,D_ ,00 lO 19 13._ _ _ _F _F _F 2_ ZF 0 0 0 MODE I, 4 , _ 4 4 _ _ 0 ANGLEi .oo,_9 .9_ ._9 ._8 .7, . _0o .o0 _oo ,o° _E_ 3. Performance of prediction Radio propagation prediction services in Japan are remarkably supported by computer systems which have made major contributions to development of prediction calculations and outputs through large data filing and classifying. An important regular service is the monthly radio propagation prediction issued three months in advance with two different output forms. The flrst form illustrates diurnal changes in the monthly median HUF and LUF on a frequency-time graph, where the time is denoted by G.H.T. and J.S.T. (Japan Standard Time, 135°E Mean Time). Objective circuits of this , prediction are listed in Table 2, where the propagation azimuth is measured in degrees c]ockwise from north at the communications center and the propa- gation distance is along _he great circle from the center to a terminal in kilometers. For short-distance domestic circuits laid within one hop area over Japan, some predicted MUFand LUF monthly median curves are shown in Fig. 1. These are northern route predictions for special paths with lengths of 500, 1000, 1500 and 2000 km. Fig. 2 illustrates examples of point-to-point predictions for long-distance international circuits between Tokyo and foreign cities. t 1980010188-225 Table 2. Circuit list for monthly median prediction. ii i Canter Terminal Great Fig. Circuit ..... (;trcle Let, Long, Azlm, Let. Long. Azlm. (lm) I.I North from Halu_ate 41.IN 140.7[ 0 59.8N 140.7[ 180 2,000 I.| East from Makodate 41.k 140.7[ 90 39.3N 164.3[ 285 2,000 I.) South from Hekcxlate 41.8# 140.7E 180 23.8N 140.7[ O 2,0OO 1.4 West from MakoNate 41.8N 140.7E 270 39.3N 117.2E 75 2,000 2,1 North from Tokyo 3S.7N 140.8E 0 53,711 140.8[ 180 2,000 2.2 East from Tokyo 35.7N 140.8E 90 33.8N 162.6E 282 2,000 2.3 South fr_ Tokyo 35.7M 140.8[ 180 14.8N 140.8[ 0 2,000 2.4 W,st fr_ Tokyo 35.7N 140.8[ 270 33.8N 119.0E 77 2,000 3.1 Nouth from Nagasaki 32.4N 130. IE O 50.4N 130.l[ 180 2,000 ).2 East from Nagasaki 32.4M 130.lE 90 30.7N ISI.2E 281 2,O00 3.3 South from Magasaki 32.4M 130.1[ 180 14.5N 130.IE O 2,000 3.4 Melt fro, Nagasaki 32.4N I)O. IE 270 30.7N IO5.1E 78 2,000 4 Phil ipplnes 35.0N 140.OE 224 14.5N 121.OE 36 2,970 5 Jakarta 35.0N 140.OE 224 6.0S IO6.8E 34 5,730 6 Malaya 35.0N I40.OE 236 3.ON IO2.0E 43 5,290 7 Southern India 35.ON 140.OE 257 7.ON 80.OE 53 6,_0 ; 8 Nouthern India 35.0N 140.OE 281 28.6N 77.2[ 66 5,880 : 9 Beirut 35.0N 140.OE 305 34.0N 36.0E 53 9,000 IO.A Southern Africa 35.0N 140.OE 255 30.0S 25.0E 66 13,990 : _Short Path) IO.II Southern Africa ;. (,LongPath) 35.ON 140.OE 75 30.0S 25.OE 246 26,020 II Ghana 35.0N I40.OE 308 5.6N O.0 39 13.850 12 Iberla 35.0N 140.OE 335 39.0N 9.ow 25 11,190 13 Southern Europe 35,0N I40.OE 323 42.0N 13.0E 40 9,890 : 14 Central Euro_ 35.ON 140.OE 335 50.ON O.0 31 9,770 _ 15 Scandinavia 35.0N 140.OE 333 60.ON 20.0[ 47 8,120 16 East Coast U.S.A. 35.0N 140.OE 26 42.0N 76.0w 331 IO,700 17 Vast Indies 35.ON 140.OE 41 23.0N 82.4W 324 12,160 : 18 Par_wm 35.0N 140.OE 48 9.ON 80.O_/ 321 13,520 19 Lima 35.0N 140.OE 64 12.0S 77.OW 310 15,490 _ 20 Alaska 35.0N 140.OE 37 60.0N 145.OW 277 5,880 21 West Coast U.S.A. 35.0N 140.OE 54 37.6N 122.4W 302 8,300 2Z 14m_len Islands 35,ON 140.OE 86 21.3N 158.0W 298 6,170 • SyN Star Ion 23.A k_|frctlca (Short P0_h) 35.0N 140.OE 205 69.05 39.6E 85 14,010 |{.I lyo_ Stationk_tarctlce (LongPath) 35.0N 140.Ol 25 69.0S 39.6E 265 26,O00 24 MuSh list Aultralla 35.0N 140.OE 175 37,8S 145,O[ 355 8,110 i 2S Mosco_ 35.0N 140.OE 323 56.0N 37.5[ 59 7,550 l_'v'__'O IGINAb !"\'_;'_';IS POOl1, 21 5 1980010188-226 IsEP,z9"/1 -, - FIO X.X.ol _OKIt NORTH FROIt HAKODRTE F|O 1.1.2 IOOKII NORTH PROfl HItK(IDIliTE I G.M.To I G.fI.T. ! I ! I i L " _ F ._ / ,(. .. __ .!.-._ r r \ X! //jr/ _X'_ J.I.T. J.I.T • FIO %.|.qJ |OOl_...ll NORTHFROR.HFROOIRTE FIG 1.|. 4 _OOOKPINORTH FROR tlAKOORTE : -.. -.. . . I F I // -. "=_. m // \ _, _Ib_ - // -- _ l ",_ I __ .# \ X I J I \_A I I II, fl "_1_, -- 1 I, I III" I,.,... _. _X ] II _XI iiij v "_ XX - l J rl ,_I _-L-. I L d.$.T. J.I.T. , Fig. I. Frequency-time range prediction for domestic circuits. 216 t 1 1980010188-227 1980010188-228 In each of Figs. I and 2, a heavy curve means the MUF diurnal change and fine curves numbered I, 2, 3 and 4 denote respectively the LUF diurnal changes In cases of I0 kW, I kW, I00 W and I0 W transmitting powers. The predicted sunspot number (SSN) obtained in twelve months smoothed running average is shown on the right top of each page. It is assumed that the radio wave is propagated along a great circ:e through the normal ionosphere, the minimum receivable field strength is 3 dB, the transmitting and receiving antenna gains are 0 dE for a radio propagation with a distance less than 2000 km and I0 dB for a radio propagation more than 2000 km long. The other monthly prediction is for ship-to-shore communications. Commercial and fishery ships located on sea areas require customarily point-to-Oistrict pre- dictions. Fig. 3 shows a world distribution of communication circuits for commercial ships, where a propagation path with cross marks is a longer route around the great circle. Fig. 4 shows examples of this monthly prediction, where heavy lines parallel to the abscissa denote the time ranges within which the assigned frequencies are operational. Circular marks on the lines indicate critical times, when the operational frequency equals the HUF. At another ending of the lines with no marks, the operational frequency coin- cides with the LUF. The prediction calculation is carried out under condi- tions that the transmission type is A1 (telegraph in carr!er amplitude modu- lation), the transmitting power is I kW, and the ship and shore antenna gains are respectively 0 dB and I0 dB. In making the fishery communication pre- Fig. 3. Circuit map for commercial ship communications. 218 t +_.r. J 1980010188-229 ISEP.t9 71 diction, it is assumed that the transmission type is AI, the transmitting power is 200 W, and the ship and shore _ntenna gains are equally O dB. The outputs of the above-mentioned predictions can be designated to sheet, graph and 35 mm film/microfiche formats by computer devices. The monthly median predictions are made on the normal quiet conditions as the ionosphere responds to the daily, yearly and sunspot cycle variations of the ionizing radiation incident on it. Nhen an ionospheric disturbance occurs, the ionospheric characteristics are far from their median values ex- pected in the _onth, and the HUF and LUF may be depressed or raised and will be different from the predicted medians. It is reasonably required to esti- mate quantitatively the probable changes in the HUF and LUF. At present, a daily prediction issued one week ahead is being made into reality by using the K-modulation method [Davis, 1970], where any pseudo-objective predicting must be tentatively prepared for the K-index. The daily prediction will be soon applied to only a few hops radio propagation over Japant by means of the 219 REPRODUCIBII,ITYOF THE ORIGINALPAGE ISPOOR 1980010188-230 top-side sounding data telemetered real-time from the Ionosphere Sounding Satellite b (ISS-b). 4. Evaluation of prediction There are frequently found some differences between predicted and actual circuit conditions, though con_iderab|e efforts have bee, put into the task of improving radio propagation prediction. It is inferred that deviations of the prediction are caused by heterogeneity of source data, defect of model radio media or propagation and incompleteness of empirical formulae. Com- parisons of calculation results derived from past data with actual facts will clarify if the models and formulae used in the prediction are valid or not. For practical survey of the monthly median field strengths, the JJY 5 and I0 MHz waves transmitted from Tokyo (36"N, 140°E) were measured at Wak-. kanal (45"N, 142°E), and the WWV15 MHz wave from Washington, D.C. (39°N, 77°W) or Fort Collins (41°N, lOS°W) and also the k_/VH 15 MHz wave from Maul (21°N, 156°W) were received at Hiraiso (J6°N, 141°E). The measured field strengths and predic:ed MUFs are shown with solid lingo in Fig. 5, where broken lines denote usable operational frequencies or calculated correspond- ing field strengths [Ichinose and Yamaoka, 1971]. The result is fairly satisfactory for short-distance circuits, and the lack of coincidence found be- Fig. 5. Observed and calculated field strengths on standard r jdio wave propagation circuits. _.20 I 1980010188-231 tween observed and calculated fields on long-distance rircuits must be caused from incomplete formulation of night ionospheric propagation. It Is necessary for the prediction services to know the degree of their value. Objective methods evaluating differences between predicted and actual propagation conditions were discussed in two papers refered to by the CCIR Report 153 [Sinno, 1958; Minis, 1962]. Both methods are probably said to be _ based on the statistics of predictions, missing events and false alarms, but are not neccssarily applicable to evaluation of the rad!o propagation | prediction services. Another more general method should be developed that _ considers prediction probabilities [Maeda, 1971]. : 5. Conclusion Radio propagation predictions have been issued in Japan for the last thirty years. During the 1970's, che prediction services have been remark- ably improved in various respects by using an efficient computer system. The current prediction was briefly reviewed in the preceding sections, where the following important questions were left to be _olved in the near future. 1) Development of solar and geomagnetic activity predictions. 2) _nendment of model ionospheres by using top-side sounding data from the ISS-b. 3) Inquiry of man-made noises. 4) Treatment of ionospheric propngation with irregular modes. 5) Performance of short-term predictions by using real time ionospheric : sounding data. References _ Barghausen, A. F., J. W. Finney, L. L. Proctor, and L. D. Schulz (1_69): Predicting long-term operational parameters of high-frequency sky- waw telecommunication systems. ESSA Tech. Rep., ERL llO-ITS 78. CClR (1966): C.C.I.R. atlas of ionospheric characteristics. CClR Rep., 340. CClR (1970): C.C.I.R. interim method for estimating sky-wave field strength and transmission loss at frequencies between the approximate limits of 2 and 30 MHz. CClR Rep., 252-2. Davis, R. M. (1970): Short-term prediction of F2-1ayer maximum usable fre- quencies from local magnetic activity. ESSA Tech. Rep., ERL 157-1TS I00. Icl,inose, M., and M. Yamaoka (1971): Methods of calculation of MUF, LUF and HF sky-wave field strength by electronic computer (Japanese). Rev. Radio Res. Labs,, 17(93), 480. Ichinose, M., and M. Yamaoka (1973): Method of calcuiation by computer of various operational parameters for HF radio circuit designing (Japa- nese). Rev. Radio Res. Labs., 19(102), 213. _. 221 RJ_PRO_UCJBIJ,I_ 0_" T_ 0RIGINAI_ PAGE IS POOR 1980010188-232 Jones, W. B., and R. M. Gallet (1962): Representation of diurnal and geo- graphic variations of ionospheric data by numerical methods. J. Res. NB__S.S,66D(4), 419. Maeda, R. (1971): Weekly forecast of radio propagation conditions (Japa- nese). Rev. Radio Res. Labs.., 17(93), 499. Minis, C. M. (1962): Ionospheric storm forecasts. Electronic Technology, (May), 172. Sinno, K. (1958): Hit rates of radio propagatior disturbance warnings and SWI warnings. J. Radio Res. Labs., 5(20), 109. 222 1980010188-233 N8,0 -1848 1 RADIO DISTURBANCE WARNING ISSUANCE SYSTEM Rikio Maeda and Hisao Inuki Radio Research Laboratories 4 I 4-2-1 Nukui-Kita Koganei, Tokyo 184, Japan Twenty or more years have elapsed since the start of radio dis- _ turbance warning services in Japan. In the past decade, many devel- opments in telecommunication techniques have made some radio commun- ications more stable than before, and the efficiency of present radio propagation prediction service has been reconsidered. A radio distur- bance warning issuance system was introduced in the Hiraiso Branch of the Radio Research Laboratories in 1972 to reconstruct the current radio disturbance warning service as a social information service. :! The main theme of this report is a description of the new ideas which were experimentally systematized by means of an electronic computer. !. INTRODUCTION During the 1950s HF radiowaves had become a very important means of domestic and international communication among ground facilities and sea-goingi ships. The bearing and pathloss of radiowave propagation through the iono- sphere are always affected and frequently disturbed by variations of ionospher- ic conditions which depend to a remarkable extent on solar and magnetospheric a_tivities. The Radio Research Laboratories (RRL) have carried out observa- tions of solar and magnetospheric effects on terrestrial and ionospheric en- vironments in order to issue radiowave propagation predictions. At that time, radio forecasting was as useful for the operation plans of practical communications and schedules of scientific observations as weather forecast- = ing for transportation and industries. Communication is an indispensable function in human life as well as in transportation. • A radio propagation prediction service aims at HF radio communication circuits in a normal ionospheric state and issues monthly or weekly medians ; of usable operation frequencies and probable reception strengths three months or a few weeks in advance. On the othe: hand, a radio disturbance warning service forecasts indirectly abnormal ionospheric variations, i. e., ionospher- ic storms, sudden ionospheric disturbances, and polar cap disturbances caused by solar and terrestrial events. Therefore, this service necessarily is main- tained on routine monitoring of solar-terrestrial phenomena, e.g., solar _ flares and geomagnetic storms, which are frequently accompanied by ionospher- ic and propagational disturbances. Daily and weekly information about these events is coded in the standard ursigrams, and distributed over the Western Pacific region and to warning centers of the International Ursigram and World Days Service (IUWDS). 223 $i t 1980010188-234 In the 1960s, radio telecommunication by satellites and submarine cables was comprised, for the most part, of long-distance international communica- tion, except in developing countries. In the 1980s, ship-to-shore communica- tions will also depend on VHF radio waves, but HF radio communication circuits may remain important for broadcasting various information on social news, time-frequencies, navigation, entertainment, and so on. For that reason, the radio disturbance warning service must be improved to adapt, for future radio communication, to circumstances in which the service objects and means will be changeable. 2. RADIO DISTURBANCE WARNING SERVICES In order to operate current radio communication systems more efficiently, radio propagation predictions concerning abrupt solar-terrestrial events have routinely been offered by the RRL in Japan since the 1950s. The practical services have been conducted by the Upper Atmosphere Research Section of the Hiraiso Branch, as shown in Table i. The world days service announces infor- mation about solar flares and geomagnetic storms daily in GEOALERT (combined data and forecasts including advisory information). Actual abnormal conditions of solar radio emission and ionospheric wave propagation are occasionally reported as URANO (radio noise) and USIDS (sudden ionospheric disturbances). These ursigrams, used for scientific experiments as well as for communication practices, are distributed over the world through Telex networks of the IUWDS and broadcast by the JJD (10.415 and 15.950 MHz) transmitted from Sakura be- ginning at every 08 UT. The IUWDS has been globally maintained; the world warning center is situated at Boulder and the regional warning centers are located in Tokyo, Sydney, Paris, Darmstadt and Moscow. The Western Pacific regional Center Tokyo i_ administered by the RRL. The short-term forecast de- scribes ionospheric radio propagation conditions as normal, unstable, or warning twelve hours in advance. This is broadcast twice every hour from Koganei by Table i. Radio disturbance warning services of the RRL. Type Ti%le Descriptio. i:ore_st Issuance Distributmn _,_odI World P¢ys A Ocu/ Telecom.uni ceZi o.Service OE_AIFRT 3olr Turer:t,t_l Faci!il_ieS ( JJD) Actual URANO DIsT.urtances _ D_lly WWA_ RWC ( TLX ) ConditiOl_S US I D3 Report 3hort Term WARNIN_ H_l_ A D.y Occllsion_lly Telecomeunical:mn Fommst HF P,op_t to. Fac;l it ies ( JJ Y) Dlsturb_lces T_lecoNunlc_;o.W_.EKLY FORECAST Weekly Facilities (Ma;I ) W,oklx Aw.,k (Mo.d.y. Forest _Th.rsd_/ WWA & RWC (TLX) •AD3__ 224 t 1980010188-235 the JJY (2.5, 5, 8, i0, and 15 MHz) with the Morse code of N, U, or W. The weekly forecast assigns a condition index of radio propagation and an occur- rence grade of Delinger phenomenon to each day a week in advance, where the indices 1, 2, 3, 4, and 5 indicate, respectively, very disturbed, disturbed, unstable, normal, and good conditions. Furthermore, the grades * and ** de- note fairly expectable and very expectable occurrences, respectively. This message is mailed twice a week to several tens of domestic and international communication facilities. There are generally two types of forecasting methods. The analytical meth-od is commonly used on astronomical problems which are solv d by predicting I the future from the present by means of _K)tionand state equations on certain initial conditions. The other method is often applied to aeronomical pre- dictions by means of stochastic relations among environmental phenomena. The radio warnings have been based mainly on the latter method because of easy i phenomenological understanding that requires a steady monitoring of the sun for ionospheric radio propagation predictions. Since the data from observato- ries in only a limited region cannot thoroughly reflect the solar activity, some observations are relayed at least among the three regions of the world, i.e., Asian, American, and European areas, which promise to be able to issue the regional warnings most effectively. For this purpose, the IUWDS devel- oped its function to a remarkable extent during the International Geophysical Years (195_58) and has operated effectively in Japan for world-wide daily ser- vice. Figure 1 shows radio information flows viewed from a side of the Tokyo Warning Center and illustrates an outline of the radio disturbance warning service conducted by the Hiraiso Branch. When an observatory detects outstand- ing solar-terrestrial events, the preliminary reports are sent as soon as pos- sible to each communication center, which transmits the coded information to respective warning and data centers. The warning center compiles the phencmen- ological evidence into synoptic forecasts and distributes them to the commun- ication facilities connected with the IUWDS. Although the task has been car- ried out in Japan by technical forecasters with ten or more years of experi- 8oI_" Actwv;t;I$ I ! I ITok,+A,,,_.o ,.j c_ L _l,,r .,,d o+.,.. I re+, 1980010188-236 ence, the issuance of objective radio disturbance warnings is nowadays made smoother by steady observation of phenomena, real-time transmission of data, automatic production of information, and practical evaluation of predictions. 3. AN EXPERIMENTAL SYSTEM Recent activities of the warning and data centers in Japan were rather backward compared to the remarkable developments in the observation and com- munication networks. This fact has caused some concern about systematic op- eration of the present radio disturbance warning service, the e_fi__iency of which is principally based on objective, automatic, steady, and speedy data management. In answer to the demands for more efficient prediction, a radio disturbance warning issuance system (RADWIS) was conceived by the Hiraiso Branch of the RRL in 1972. Radio information, compiled by the Hiraiso Branch for the radio disturbance warning service, is composed of analog data obtained sequentially from its own observatory and digital data accepted randomly from other observatories, as shown in Figure 2. It is desirable for these data processes to be real-time and on-line 24 hours a day. The RADWIS has an electronic computer with a magnetic core memory of 16 kW (16 x 24 kB) which is almost assigned to occupied areas for system programs and overlaid areas for application programs stored in a magnetic drum unit. This system requires a one-way consecutive time slice automatic operation controlled by a non-stop external crystal clock and no manual operation unless unanticipated computer emergencies cause normal process jobs to cease. System breakdowns caused by hardware or software troubles are avoided principally by automatically cutting off the process in the disturbed part and skipping to a consecutive process which admits only a few after-effects into a limited small part. Power out- F i i I I __ j 1 I i _ _.._L_, - , j LF.,," --- | C_J..ol ....... 1 _.,,.,'o _ CIo&I L ....... J Z I L 1 " C_,...,u.;_t,o,,T_.,,;..I Figure 2. Hardware of the RADWIS. 226 1980010188-237 ages due to artificial or natural events make a transient system breakdown after preparing the system for the future; a system recovery is then auto- matically carried out by a computer self-restart function when the power be- comes normal. If the system breakdown lasts fairly long, operators may achieve an earlier recovery by the use of interrupting manual switches for data com- pensation--an inevitable task in an on-line and real-time computer system. Figure 2 illustrates hardware of the RADWIS, where data input is through a paper tape reader, a communication terminal typewriter, or analog-digital (A/D) converters, and the output is through the communication terminal type- writer, D/A converters, or monitor typewriters. The software is shown in Figure 3. In the data acquisition phase, the 24-channel analog data on HF and LF radio signals, solar radio emissions, and geomagnetotellurics are always sampled once a second after data adjustment i and calibration to the input apparatus. The data conversion job makes several indices and ursigrams on radio information subject to statistical arrangements. It is most essential for the radio warning to produce forecast information from _ actual observations. The public telephone circuit between the central proces- I sor at Hiraiso and the terminal typewriter at Kokubunji is used mainlyfor Telex code transmission. Unless commercial power breaks down, the computer i is waiting or working for these service jobs scheduled according to the calen- dar year, month, week, day, hour and minute designated by non-stop clock sig- _ nals. The computer constantly monitors the sequentially and randomly accessed " data received through observation and communication input, constantly fore- casts radio propagation conditions 12 hours in advance, occasionally issues a short-term radio disturbance warning if it is necessary, and also regularly reports to the IUWDS whether solar radio emission and ionospheric radio propagation were severely disturbed or not. The world days service is issued at ten o'clock every day and the weekly forecast is issued at noon every Mon- day and Thursday by compiling the source data stored in the magnetic drum unit. The results are coded in respective fixed formats, input in the magnetic tape unit, and output to the typewriters at the designated issuance times. ../_ 7 ! I_..J., Iz_ I _ I " -- [ Tum,..l I ....... I - Figure 3. Software of the _DWIS. 227 t > 1980010188-238 Radio disturbance warning services are required to monitor constantly the interplanetary radio waves and plasma generated from the sun or galaxy and modulated through interplanetary space or the magnetosphere. The RADWIS is interested in ionospheric storms, sudden ionospheric disturbances, and polar cap disturbances, which are detected directly through steady monitoring of HF and LF radio propagation circuits. The sunspot configuration and solar radio emis- sion are observed at Hiraiso, but information on the solar magnetic field and solar plasma flow is received from the IUWDS, and these data are effectively used for solar flare forecast. The magnetospheric structures, e.g., the mag- netic cusp and tail aspects, are only inferred from ground geomagnetic obser- vations, and information on the radiation belts, interplanetary magnetic fields, and cosmic rays is limited, but real-time utilization of the polar and sat- ellite data is expected in the future. Occurrence of solar flares in an ac- tive region is considered over abnormal changes in the sunspot number, area, type, solar radio intensity, temperature, spectrum, and over causal relations between geomagnetic storms and ionospheric storms. The RADWIS forecasts aver- age radio propagation conditions of middle latitudinal circuits, where radio propagation fadeouts and storms are frequently fatal for good HF radio commun- ication. The system was introduced as a trial for automatic routine radio disturbance warnings, but several years of indulgence are still necessary be- fore this will be accomplished. By its nature, this system has aimed at objectification and gradation of the forecasting method, and intends to pro- gram, evaluate, and improve the radio disturbance warning service experimen- tally. 4. CONCLUDING REMARKS The radiowave propagation forecast has changed its objectives, methods, purposes, and means according to developments in radio communication techniques to which the RRLhave consistently devoted themselves throughout recent years. The radio disturbance warning service began early at about the 19th sol3r min- imum in the 1950s, before rockets and satellites opened the new space epoch, and ionospheric physics seemed to dissolve into space research. Thereafter, ionospheric radio communications tended toward space radio communications, losing its technical usefulness for ionospheric experiments and answering to the scientific need for ionospheric studies. A similar reversal in the re- lation between ends and means is found in earth environmental science. As radiowave propagation depends on ionospheric and magnetospheric conditions, radio environmental information is useful for radio communication circuit predictions, and vice versa. Toward the end of the 20th solar cycle, main frequency bands for long-distance radio communications turned to the VHF from the HF, but HF radiowaves are still important in ship communications and over- seas broadcasts. During the 21st solar cycle, the service area of radio dis- turbance warning will cover the radio communication media between the earth and the sun. To meet the need for space disturbance warnings, the IUWDS net- work of observation, communication, data, and warning centers must be re- arranged. The RADWIS is also intended to be an improved forecasting system for solar-terrestrial phenomena and radio environmental events, a goal which • will be attainable with the program revision rendered possible by another large-scale computer. 228 1980010188-239 NS,0- 18482 A REVIEW OF THE _TION OF THE IUWDS REGIONAL WARNING CJRTRE AT THR IONOSPH_tlC PREDICTION SESVICE, SYDNEY F. E. COOK and P. DAVIES Ionospheric Prediction Service Department of Science and the Environment P.O. Box 702 Darlinghurst NSW 2010, AUSTRALIA The operation of the Sydney RWC is reviewed. Routine data reports are issued daily, while disturbance warnings (short-term forecasts) are issued as necessary. An outline of the methods used in fore- casting solar activity and geomagnetic disturbances is given. 1. INTRODUCTION , The Sydney Regional Warning Centre (RWC) of the International Ursigrsm and World D_ys Se_ice (IUWDS), operated by the Ionospheric Prediction Service (IPS) of the Australian Department of Science and the Environment, has responsibility for the Australasia and Antarctica Region. Routine solar activity reports and forecasts are issued daily, and Disturbance Warnings , (short-term forecasts) as necessary. Daily reports consist of the Sydney Geoalert, giving a forecast ofA expected solar end geomagnetic activity for the next 24 hours, coded data from the solar optical and radio observatories at Culgoora and Fleurs, and a plain langus_e report on solar active regions and activity from Oulgoora. Disturbance warnings are issued whenever major solar activity or geomagnetic disturbance is expected to cause the ionospheric radio propagation parameters to deviate noticeably from their predicted quiet-day values. Data input from the Culgoora and Fleurs observatories end the methods used at Sydney RWC for predicting solar activity and geomagnetic disturbances are described. 2. DATA 2.1 Australian Data The Ionospheric Prediction Service operates an H-alpha solar patrol _t Culgoora, about 600 km northwest of Sydney, where the IPS Observatory is co-located with the C81RO solar radio and optical observatories operated by the Division of Radiophysics end National Measurement Laboratory. The Z_3 m.. 1980010188-240 observatory As manned daily from sunrise to sunset by a team of three observers (at least one on duty each da_) who have access also to da%a from the CSlR0 instruments. Besides visual observations, 35 mm photographs of the sun are taken once per minute. Sunspot data are obtained from a hel__graph. Real- time flare data are coded from optical observations, but finL flare data are scaled from film records and sent to the World Data Centres. The CSIR0 operates a spectrograph (6-3800 MHz) and the radioheliograph. From the spectrograph real-time display, information on the type of solar radio bursts can be obtained (IUWDS URASP code). The radioheliograph, a unique instrument operating at discrete frequencies of 4_.25, 80, and 160 MHz (227.4 MHz to be added), can give information on the position and movement relative to the sun of radio burst sources, for about six hours per day centeredon local noon. Optical and radio data from 0ulgoora are coded in the IU_S format and telexed daily to the RWC in Sydney at 0000 UT and after sunset, along with plain language reports describing optical features on the sun and radio bursts observed with the radloheliograph. Major events are reported by telephone or telex as they occur. One II_Sobserver is stationed at the Fleurs Radio Astronomy Station of ' the University of Sydney, about 50 km west of the city. By using the E-W arms of the synthesis telescope, the sun is scanned daily at 0000 UT at :, frequencies of 692 and 1415 MHz with resolutions of _ and 2 minutes of arc. These scans are scaled and coded into the URALN code : they =re also used to indicate the presence of coronal holes. Also at Fleurs is a radiotelescope ; operating at frequencies of 1415, 2695 and 4995 MHz (200 and 400 MHz to be : added) : the 0000 UT quiet sun flux and data on any radio bursts are coded i into the URANJ code. In addition, the station monitors the H-component of the earth's magnetic field, and several radio circuits to detect SID's. With the exception of the F_W scan, all the data from Fleurs are tele- metered to the RWC where they are displayed for the forecasters' use. Fleurs is not usually manned Qn weekends but E-W scans are carried out automatically and the data reported in the Monday ursigram. One other local data source is a direct line to the IPS ionospheric research station at Camden, 50 _ southwest of Sydney, by means of which one can trigger the ionosonde. A telemetry system transmits the ionogram to the RWC when it may be studied on a long-persistencedisplay or photographed by a polaroid camera. Thus one can know immediately the state of the local ionosphere. The R_'Citself is manned normally from about 2200 UT to 0700 UT on week days and for some hours during the local morning on other days, wi+.hpro- vision to extend these hours when the sun is unusually active or to eupport special programmes. 2.2 Overseas Data Data are received in Sydney as ursigrams from Tokyo and Moscow during the night and from Boulder in the early morning local time, so that much of 230 I 1980010188-241 the information is at least 12 hours old (unless it is a PRESTOm_ssage). From Boulder comes coded data on calcium pla_ee (NoMath-Wulbert Observatory), radio emission (Sagamore Hill and Ottawa, discrete frequencies ranging from 245 to 15600 MHz), sunspots (Boulder) and the AYREDand T_CM forecasts. Flare and radio data from the USAT S00N-RSTN sites are also received. The daily 2200 UT Joint USAF/NOAAPrimary Report of Solar and Geophysical A_t_ity is useful as it contains a timely update of the "dark hours" :in plain language on solar L:tivity as well as information on X-ray flares and occasionally on coronal holes observed at 10850 A. The Boulder Region Summary provides information on sunspot magnetic classifications. The last two reports are re- transmitted to the Culgoora Solar Observatory. From Boulder we receive also the Meudon 1400 UT geoalert and the WWAgeoalert issued at 0400 UT. From both RWCMoscow and RWOTokyo we receive data on metre wave as well as microwave radio flux and events and both Centres send sunspot and geomagnetic data. The 5-8 cm fluxes and flux ratios from Toyokawa are useful in evaluating large active solar regions, while coronal data from Mt Norikura provide some- times our earliest indication of the approach of active regions round the eaBt limb. The Tokyo cosmic ray data also allow us to check on the passage ofsolar plasma clouds ('_nagnetic bott es") in the absence f definitive space- I craft data. The data coverage clearly could be improved I in particular, we receive °_ no decimetre or microwave radio spectral information from stations other than Oulgoora (although many such report in "Solar Geophysical Data"), and metre wave spectral data only from Oulgoora and some USAF-RSTN sites (25-75 F_z). 5. OUTPUTFROM_'I)ITEY RWC 3.I Routine Daily Reports Daily at about 0100 UT the Sydney RWC sends a message (ursigram) to the _ Centres at Boulder, Delhi and Tokyo from where the data are passed to Meudon (Paris) and Moscow Centres. This message consists of a geoalert, giving a forecast of expected solar and geomagnetic activity for the next 24 hours, : followed by coded data from the Australian observations (URANJ, URALN, URASP, UFLAE, USSPS, PTROL, USIDS). Then follows a plain language report from the Culgoora observatnry describing the solar active regions and summarizing the solar radio observations. After sunset, a similar message, but without a geoalert part, is sent direct from Culgoora to Boulder and Tokyo and to RWC Sydney. Information on major events is received in ne_,_real time from Culgoora and Fleurs and is sent direct to the five Centres mentioned as soon _ as the data have been _sembled. 5.2 Disturbance Warnings Disturbance Warnings are issued by the Sydney RWC "_henevermajor solar activity or significant geomagnetic disturbance is expected. These Warnings are sent to about 60 users, mainly in Australia and New Zealand and neighbouring countries and to Antarctic stations. The users fall into three main categories, 23! 1980010188-242 oosnunioators who are interested in effects on their H_ radio circuits, geo- physicists, mineral exploration companies and pipeline authorities who are interestedmainly in geomagnetic _torms, and researchers who need information on expected disturbed or quiet conditions or solar events. In the Warnings, the probability of flares and storms is given in three grades viz. possible, probable, or expected, and three kinds of warnings are issued, a primary warning to all the users, warning "A" to HF communicators giving more detail about MUF and SID effects, and warning "B" to users con- cerned with geomagnetic activity. Further details are to be found in the IPS Handbook (IonosphericPrediction Service, 1577).Special forecasts are issued for certain users (for example,rocket experimenters)as requested. 4." FORECASTING SOLAR AND GEOMAGN_fIC ACTIVITY It is no_ proposed here to discuss solar-terrestrialrelations in detail. The subject was reviewed in relation to short-term forecasting by C_ok and McOue (1975), and in CCIR Document 6/177-E (1976) which also liste, "_e services available from the operational forecasting centres _Iound t._,eworld. Recent developments have centred on the role of solar sector boundaries, the interplanetary field direction and solar wind parameters, on the rels _ion between coronal holes and solar wind streams and attempts to monitor the passage of coronal holes by ground-based methods, and on the "possible final resolution of the M-region problem". Much effort has also been put into the identificationof possible precursors of flares, especially proton flares. There has been some progress with the "classical" problem of flare distance east or west of central meridian and storm probability, bu_ little new information on the role of metre-wave emission from flares, and the delay time between flares and storms. Meanwhile the solar observers at Oulgoora and the disturbance forecasters at RWC Sydney find it useful to begin with a few rules of thumb and these are listed here. 4.1 Flare Forecasting Forecasting solar activity for the daily geoalert is left to the observers at Oulgoora as they have direct access to the sun in H-alpha and white light and to the solar radio data. Points which they bear in mind and which derive from a variety o_ sources ere: I. Mo=t flarcc occur nc_r _ ncutrml !_ne : check for complex neutr_! lin_ embedded in plage. Where the spot magnetic classification is gamma or delta and the region is compact, the likelihood of large flares is greatly increased. 2. Check for large magnetic gradients, complexity of spot group, spot rotation and p_er motion of spots. Large spots of opposite mag- netic polarity close together indicate steep magnetic gradient. 232 1980010188-243 3. Rapid growth of spot area, 50 to 100% in one day. 4- Presence of active dark filsment. 5. Past history of the region with respect to flare production. 6. Region with strong coronal line emission, especially yellow line, i at east limb passage (data mostly from Mt Norikura, Japan). i 7. Rapid development or decay of a delta configuration. 8. Relation of the 10-cm _adio flux value to the 90-day mean : • i major flares are likely when the current flux is higher. 4.2 Proton Flare Forecasting S I. Most proton flares occur in spot groups with delta configuration. _ 2. Polar cap absorption events (PCAs) favour the westerr_ half of the disc, and particles from western flares travel more quickly, _ particularly from longitudes 0 to 15 degrees west approximately. 3. Check for umbral coverage and parallel ribbons or Y-shaped flares. 4- Most proton flares are associated with regions where the ratio of the flux densities at 3.2 cm and 7.8 cm is greater than one. 5- Check for U-shaped radio emission spectrum. Proton flares typically give high flux den s_ties at metric and centimetric wavelengths withminima in the decimeter wavelength region (around _ 1400 MHz). Ri o time of microwave burst > 5 minutes and type IV dekametric emission observed. The 10 cm flux maximum tends to occur later than H-alpha maximum. 6. Generally, for proton events, at 19 GHz the peak flux increase should be more than 50% of the preburst value and the flux enhancement more than 10% above it for longer than 5 minutes (but RWC Sydney does not receive any 19 GHz data). 7. PCA events tend to occur in regions on their second or later rotations. Proton events seem to require a mature region. Successive prcton events during one disc transit are not uncommon. 8. Large calcium plage (>8 000 millionths) and large spot area : ( >1 000 millionths) correlate with intense radio bursts and major proton events. _. 9. Check for moving type I_rbursts with the flare. 10. Check size of optical fla_e, and microwave burst. 11. Check start _imes of radio events at different frequencies: 233 _ #._ I .... I 1980010188-244 I, indicatin_ movement of plasma out into the corona. A rough rule is that old regions produce protons, new regions magnetic storms, although storms are produced by mature regions &s well. Note also that metre wave flux values can be misleading because metre wave emission cannot escape easily in the sun-earth direction from above the solar limb (the directivity effect). The sources of most of these "rules" were given in the 1975 review by Cook and McOue and are not referenced again here. 4.2 Magnetic Storm Forecasting 4.2.I Recurrent Storms Since the difference between the Mustel' and the Allen-Saemundssonviews on the cause of recurrent storms, also mentioned in the 1975 p_per, appears to have been resolved (Gulbrandsen 1975) in favour of open magnetic regions or coronal holes, forecasters are now concerned with the detection of these holes and their relation to geomagnetic disturbances,using the data available. The "cone of avoidance" idea in relation to closed magnetic regions and centres _ of activity now seems firmly established. Coronal holes have been related to white-light coronal structures (Hanson et al. 1976) and to regions of low 284-A emission (Neupert and Pizzo 1974, Bell and Noci 1976) measured from satellites, and their relation to solar wind streams and disturbances is well documented (Nolte et al. 1976, Sheeley et al. 1976). RWC Sydney however has available only the longer-wavelength coronal line emission data from Mt. Norikura in Japan, from which CMP of weak-emission regions are inferred : and occasional He 10 830 A data from Kitt Peak mentioned in SDF Reports. More recently Ferguson (1979) has shown that the presence of coronal holes may be inferred, on a regular basis, from the solar east-west scans at decimetre wavelengths, in which the holes appear as stable troughs in the scan data. The Bartels 27-aay chart of magnetic activity continues to be the basic datum for monitoring recurrent storms, supplemented by east limb coronal data and helium line spectrograph data, when available, and now the Flours east- • west scans. A rough rule is that the presence of a spot group immediately preceding the supposed hole or M-region delays the disturbance a day or two past the usual CMP + 3 days period, and it may begin suddenly. Emergence of a centre of activity within the region will probably disrupt the recurrence altogether. These disturbances are noticeably longer and more severe near the equinoxes. They are also evident throughout the cycle, though most pro- minent in the years approaching sunspot minimum. A notable feature of the Bartels chart is the regular recurrence of quiet periods : warning of these is important for some users. 4.2.2 Sector Boundary Disturbances and InterplanetaryMagnetic Field L The dependence of magnetic disturbance indices on sector boundary crossings, 234 1980010188-245 I | the interplanetary field direction both parallel and perpendicular to the l ecliptlo, and solar wind parameters, continues to be investigated by manyworkers, leading to significance of "away" sectors, south-polntlng field, solar wind velocity and number density, and interaction between slow and fast streams in relation to disturbances. In general, the number density first increases, then the field changes direction, and finally the velocity is observed to increase, before the magnetic disturbance occurs at earth. _L_ny of the published results relate to minor disturbances • for this reason, but mainly because RWC Sydney does not receive solar wind (spacecraft) data and does not have access to very high latitude magnetic observations from which we might infer boundary crossings from the Svalgaard-Mansurov effect, not much attention has been paid to this aspect of disturbance fore- casting. • Svestka et al. (1976) list 44 proton events associated with sector : boundary transits during the years 1957-1969. Some 17 of these were accom- _ pauied by magnetic storms (including two recurrent storms) for which fore- casts would have been useful, the rest by minor activity or quiet periods, but advance warning from spacecraft observations would have been required, for forecasting. In about seven cases, the particle event began before the ? , magnetic disturbance. 4-2.3 Flare-induced Storms : While there has been progress in the association of recurrent storms with i coronal holes, the rules for forecasting flare-induced storms remain much as they were outlined in the 1975 review already mentioned. For example, it is generally accepted that a major flare near central meridian or west of it, • i_ated in a suitable active region and with large radio bursts, especially a spectral If-IV event, is likely to cause a magnetic storm. The same event t _ near east limb (noting that metre wave emission may not be observed from so far east) generally will not. But for users who want a "yes or no" forecast, : at what longitude do we change our prediction? And if there is a storm pre- dicted, what will be the delay time from flare to storm commencement? These are the two basic questions for which we still have no simple answers, as is clear from observed differences between forecasts issued by different Centres, based on the same data for the same event. The paper of Yoshida and Akasofu (1965) gave useful information on the j _ magnitude as well as time-delay distribution according to flare location east or west of central meridian and presence or absence of fast or slow particle _._ • events. Pint6r (1975) has continued his comparison of travel times of shock : waves from type II bursts. Results presented by Sakurai and Chao (1974) are disputed by Lanzerotti (1974) on the grounds of disagreement between different _ sources of data. 4 From these and many other sources there emerges another set of "rules of thumb", giving the following criteria for predicting storms following flares: I. Large flare, generally importance 2 or greater, from about 45 degrees east of CM to far west (especially if cosmic ray data indicate that the earth is inside a "magnetic bottle" connecting z35 1980010188-246 the earth to the emitting region on the sun). Note however that the great storm of February 1956, for example, was caused by a IB flare : it was near west limb but was accompanied by a very large metre wave event. 2. Spot group usually complex (see also oriteria for proton events). 3. Large radio burst with type IV longer than 15 minutes and spread over cm and metre wavelengths and preceded by type II. In the absence of spectral data, which we receive only from Culgoora obbervatory, the II-IV event may be inferred from "major burst with second part". Note possible directivity effect of metre wave emission from far east or far west events. 4. The type I_ should be a moving type IV, not stationary continuum. Again this information is available only from Culgoora, for fore- casting purposes. Sometimes a moving IV will be observed, headed east or west into space and not towards the earth : this will not cause a storm (example, the "Westward Ho" event, the first one ever recorded at Culgoora). 5. A rough rule - metre wave emission equates with magnetic storm, microwave with proton events. The largest events across the spectrum produce the major storms. 6. The first flare in a previously quiet region may not produce a storm, but successive events may do so, with the delay time gradually decreasing with each storm but this is not a reliable criterion. 7. SC storms superimposed on recurrent disturbances tend to result in very high Kp values. 8. CMP of metric radio-noisy region is associated with minor dis- turbance about two days later. 9. Disappearance of a large filament near central merid" is believed to be the cause of some disturbances. 10. Events, even large ones, in small calcium plage areas have a longer delay time than similar events in large regions (up to about 5 000 millionths area). 5. FUTURE DEWELOPMENTS, AND PROBL_NS In Australia, the IPS observatory at Culgoora is being equipped with a time-lapse video recorder so that H-alpha events which might have occurred during absence of the observer at one of th6 other sites (or at the telex, for sending data) can be played back and the data supplied to RWC Sydney and the IUWDS network. Telemetry of spectrograph and heliograph observations to 236 I 1980010188-247 the IPS building is also to be installed. The Fleurs single-frequency radio observations will continue but the east-west scans (for coronal holes as well as active regions) depend on University of Sydney observational requirements • at present, we enjoy regula_ use of the antennas but this may not always be SO. The USAF SOON-RSTN observatory at Learmonth, North-West Cape area of Western Australia should be in operation by the summer of 1979 and the Ionospheric Prediction Service will station two physicists there. Data from Learmcnth will be available to IPS and will be included in our ursigrams, and will be especially useful for proton flare warnings. Some data problems remain. We receive solar spectral data only from C_Igoora (and for some events from Sagamore Hill in the U.S.A.), but several stations at other longitudes supply spectral data to Solar Geophysical Data. _ It would be very useful to have data from these stations also, in the IUWDS i data exchange. A code is being devised for reporting the Oulgoora radio- heliograph observations which at present are reported in plain language or _ in the additional groups in URANJ code. Sydney also receives very little data from Western Europe, under existing arrangements, and the transmission to Centres beyond Delhi, Tokyo and Boulder could be improved : the cost of telex messages is the main difficulty. Another problem is that of PCA data : we have only radio communication with Antarctic bases, with no recourse to land lines or other methods when PCAs and storms interrupt the data flow. 6. IONOSPH_IC DISTURBANCES Ionospheric disturbances are not considered in this brief review. As mentioned earlier, some indication of expected MUF and absorption changes is given to users, in the Warnings. One problem which should be studied further is that of severe ionospheric disturbance with only moderate magnetic dis- turbance (example, 1978 October 50), and severe magnetic storm with only slight ionospheric effect, at least at middle latitudes (example, 1978 April 11). The ionospheric predictions of IPS are described elsewhere in these _ Proceedings. REFERENCES Bell, B. and G. Noci (1976) : Intensity of the Fe XV emission line corona, • ; the level of geomagnetic activity, and the velocity of the solar wind. Journa- of Geophysical Research, 81 : 4508. CCIR Study Group 6 (1976) : Short-term prediction of operational parameters for ionospheric radiocommunications. CCIP Doc. 6/177-E, p. 64. 237 1980010188-248 • 4. Cook, F. E. and C. G. McCue (1975) : Solar-terrestrial relations and short-term ! i ionospheric forecasting. The Radio and Electronic _k_ineer, 45 : 1I. Ferguson, B. G. (1979) • Coronal holes inferred from the "Fleurs" east-west solar scans. Paper submitted to ISTP Proceedings, Boulder, 1979. Gulbrandsen, A. (1975) : The solar M-region problem - an old problem now : facing its solution? Planetar_ and Space Science, 23 : 143. Hansen, R. T., S. F. Hansen and C. Sawyer (1976) : Long-lived coronal structures and recurrent geomagnetic patterns in 1974. Planetar 2 and Space Sci_ce, IPS (Ionospheric Prediction Service), Department of Science (1977) : Handbook IPS H-6 for use with ionospheric prediction services. Lanzerotti, L. J. (1974) : Comment on (Sakurai and Chao 1974) : Journal of Geophysical Research, 79 : 4823. Neupert, W. M. and V. Pizzo (1974) : Solar coronal holes as sources of geo- magnetic disturbances. Journal of Geophysical Research, 79 : 3701. Nolte, J. T., A. S. Krieger, A. F. Timothy, R. E. Gold, E. C. Roelof, C. Vaiana, A. J. Lazarus, J. D. Sullivan and P. S. McIntosh (1976) : Coronal holes as sources of solar wind. Solar Physics, 46 : 303. Pint'r, S. (1975) : Deceleration of flare-generated interplanetary shock waves. Bulletin of the Astronomical Institutes of Czechoslovakia, 26 : 169. Sakurai, K. and J. K. Chao (1974) : Motion of the sources for type II and type IV radio bursts and flare-associated interplanetary disturbances. Journal of Geophysical Research, 79 : 661. Sheeley, N. P_, J. W. Harvey and W. C. Feldman (1976) : Coronal holes, solar wind streams, and recurrent geomagnetic disturbances : 1973-1976. Solar Physic_s, 49 : 271. V Svestka, Z., L. Fritzov_-Svestkov_, J. T. Nolte, H. W. Dodson-Prince and E. R. Hedeman (1976) : Low-energy particle events associated with sector boun- daries. Solar Physics, 50 : 491.\ Yoshida, S. and S.-I. Akasofu (1965) : A study of the propagation of solar particles in interplanetary space. The centre-limb effect of the mag- nitude of cosmic-ray storms and of geomagnetic storms. Planetary and Space Science, 13 : 435. 238 t % 1980010188-249 i. i A DISTURBANCE I_0RECAST_'S VIE_ OF THE SEPTE_ER 1977 EVENTS i F. E. COOK Ionospheric Prediction Service Department of Science and the _vironmentP.O. Box 702 Darlinghurst NSW 2010, AUSTRALIA _, The solar-geophysical events of 1977September 7-26 presented a variety of disturbance forecasting problems, viz. disc transit of a centre of activity with high microwave flux but relatively low metre-wave flux, flares in this region near east limb, central i" meridian, and west limb with major radio events, and the presence ' of two coronal holes. The Sydney daily GE0SYD message and IPS : Disturbar_ce Warnings issued during this period are related to the solar-geophysical data available, at the time, about these events. 1. INTRODUCTION The Ionospheric Prediction Service (IPS) of the Department of Science and the Environment operates the Australasi_ and Antarctica Regional Warning Centre (RWC) of the International Ursigram and World Days Service (IUWDS). The RWC issues a GEOSYD message about 0100 _ each day. The SOLALERT part of "- this message, relating to expected solar activity, is prepared usually at the Culgoora Solar Observatory where the observers have access to the optical and radio sun, while the MAGAL_T part, relating to expected geomagnetic dis- turbance, is prepared at the R%C in Sydney, based upon the solar-terrestrial relationships outlined by Cook and McCue (1975) and on more recent develop- ments described in the literature, using data from Australian observatories and from the IUWDS ursigrams (IUWDS 1973). Disturbance Warnings (IPS 1977) are also issued as necessary to EF communicators and other users, mainly in , Australia and New Zealand. _ The solar-geophysical events of 1977 September 7-26 associated with ', _ McMath Region 14943 (Boulder 889), are already well-douumented : in particular, the monitoring of these events by means of the ursigram data received at the Japanese RWC is discussed by Nozaki et al. (1978). While much information h&s now been published, the GEOSYD messages and other forecasts were issued on the basis of information available from da_ to day, and it may therefore be useful to set out on a day-to-day basis the reasons why particular MAGALE_T messages and other Warnings were issued, from an operation_ Warning Centre, " during the period. From this it may be possible to consider what changes, if . any, should be made in the selection of data to be exchanged by ursigram, and whether we are correctly applying the data we receive already. .. 239 1980010188-250 : 2. THE DATA AVAILABLE At Sydney, the daily GEOSYD is issued at about 0100 UT which is 1100 hours Standard Time (1200 in Summer), and the data and information usually available at that time are: (I) The Culgoora visual patrol in H-alpha and white light and radio upectro- graph data (8-3800 MHz) for the previous day and up to 0000 UT : the spectrograph data (CSIRO Solar Radio Observatory) are read from a facsimile machine which does not necessarily reveal all the details which are scaled later from the film record. (2) Location of sources and of some events, at 43, 80, and 160 MHz, from the CSlRO radioheliograph. (3) Fleurs (IPS) microwave data for the same period (but not usually 0_ : weekends and holidays), the E-W scans (University of Sydney), and sudden ionospheric disturbances. (4) Chart record of the horizontal component of the earth's magnetic field recorded at Fleurs and telemetered to Sydney. (5) Ursigrame received during the night from Tokyo (data to 0600 UT), Moscow (1400 UT) and Boulder (2000 UT). (6) The Paris GEOAI_T (1400 UT), and usually the Boulder report of solar and geophysical activity issued at 2200 UT. (7) The World Warning Agency (WWA, Boulder) GEOALERT issued at 0300 UT the previous day. 3. DIARY, SEPTEMB_ 8-28, 1977 The Sydney GEOA_T messages are identified by the code word GEOSYD. In these messages, MAGQUIET indicates that no significant geomagnetic disturbance is expected on the day of issue or on the dates listed : at some Centres, because of local time differences, the GE0_T is issued some time on the previous day. MAGNIL means that a MAGAL_T which has been in force should be ended. In some cases the Sydney Centre used MAGQUIET when the correct advice should have been MAGNIL since a world-wide MAG_T had been in force the previous day; these are r_corded here as MAGQUIET (MAGNIL). Similarly, SOLAI_T, 80LQUIET and SOLNIL refer to solar activity (IUWDS 1973). All times mentioned a_.ein U.T., the GEOSYD message being issued at 0100 each day. SEP 08 DATA 8EP 07 : 2B flare (Culgoora) north 10 degrees, east limb, SEP 07/2252, maximum at 2308, preceded by IB 2227/2253/2250 : short type II and V bursts (no type IV) followed by dekametre continuum and bursts on the heliograph frequencies (43, 80, 160 MHz I. Fleurs, complex microwave burst (720, 1415, 2695, 4995 MHz) and major S-SWF. 240 AI 1980010188-251 NOTES = Data published later reduced the 23 to IN, with no mention of the IB at 2227, and they list also a type IV burst 2230-2245 at metre wavelengths, but these details were not known at the time. The short duration (15 minutes) of the type IV would have made it a borderline case according to Bell (1963), but east llmb location makes occurrence of a geomagnetic storm unlikely in any case. The probability of more than a slight disturbance from this location, even if one did occur, is also small : see e.g. Yoshida and Akasofu( 965). GEOSYD : SO_T MAGQUIE_. WARNING • issued SEP 08/0100 "... a solar flare from a region beyond the east limb of the sun, start time 07/2227Z, caused a major short wave fadeout lasting over 50 minutes ... present available data on this event do not indicate that it will cause a magDetic storm." SEP 09 DATA SEP 08 : The flare SEP 07 was IB according to Voroshilov Observatory. Hiraiso reported major metre-wave burst, starting first at 500 MHz, then 200, 9100, then 100 MHz. No reports of any PCA. i, NOTES : WWA MAGAL_T SEP 09/10 in force, Tokyo RWC advised MAGQUIET - SEP 08 (Tokyo includes this information in daily ursi6Tam). .Gmmm, SO T I S_ 10 DATA SEP 09 : IUWDS ADAL_T message lists 23 flare north 10 degrees east 80 degrees 09/1604, in McMath Region 14945 (Boulder 889) the site of the SEP 07 flare, and mentions parallel ribbons and umbral coverage which are characteristic of proton flares (K_ivsk_ 1965). Sagamore Hill report large complex microwave burst but metre-wave burst not large. Again the 210 and 415 MHz bursts occur before the higher frequencies. NOTES : Region 14945 is large, complex, and bright in H-alpha. Metre-wave flux is still low but this is possibly the well-known directivity effect. However the flare is still too far east of central meridian (see SEP 08 Disucssion). Also the burst start times at different frequencies are not consistent with disturbance travelling outwards through the corona. WWA MAGAL_T SEP 09/II in force. GEOSYD: SOLALERTMAGQUIET(MAGNIL). SEP 11 DATA SEP 10 : Microwave flux increasing, from Region 14945 according to the E-W scans, but not metre-wave. No activity SEP 10. NOTES : As for SEP 10. Slight recurrent-type geomagnetic activity up to K = 4 : Tokyo RWC recommended MAGAI_T SEP 10 probably for this reason. WWA _L%GAL_T 10/12. 9 omm, SO Ja T (INGOT). SEP 12 DATA SEP 11 : Region 14945 is now magnetically beta/gamma/delta, type E spot _coup about 50 degrees east of C.M. Microwave flux increasing but metre-wave steady. No significant activity SEP 11. NOTES : No change. The slight geomagnetic activity continued. MAGAL_T 11/15. G_OSYD: SOLAUmT_GQ_ET (MAG_L). 241 Ill 1980010188-252 SEP 13 DATA 8_ 12 t No activity reported SEP 12 which was also geomagnetically quiet. Microwave flux still increasing (Ottawa 2800 MHz 106 flux units SEP 12), metre-wave flux still low (Sagamore Hill 245MHz 9 fluxmats). NOTES : WWA MAG_T continued SEF 12. Fleurs magnetic record indicates possible start of recuxTent-type activity. GEOSYD s S0LALEST MAG_T 15. SEP 14 DATA SEP I_ : No activity reported SEP 13 in Region 14943which is still a large E-type spot group. RecuxTent-tyDe geom_etic disturbance : occurred SEP 13. (Later published data show a coronal hole CMP SEP 11 probably associated with this disturbance, but this was not known at the time). NOT...__: WWA MAG_T in force SEP 13/15. Tokyo advised SOLNIL MAGQUIET SEP13. GE0SYD : SOLQUIET (SOLNI.L)MAGAL_RT MINOR 14. WARNING z issued SEP 14/0200 "... recurrent-type activity 13 September. ... The region (near central meridian) has begun to decline, there is still some chance of flare activity causing minor daylight fadeouts." SEP 15 DATA SEP I_ : Boulder 14/2200 report mentions a coronal hole visible : -_ at I0830A about 20-30 degrees east of Region 14943,that is, about ; two days east of CMP. Metre-wave flux still low, microwave flux no longer increasing. NOTES : The coronal hole may cause geomagnetic disturbance about 3 days past CMP, that is, about SEP 19. Unless there are some more major events, CMP of Region 14943 should not produce any significant disturbance SEP 17/18 as it is a "Q" (Radio quiet) region according to the classification of Denisse etal. (1951). Tokyo RWC SEP 14 advised SOLAI_T MAGNIL, WWA SOLAIaRT 14/15 MAGAI_T 14/16. GD3SYD : SOLQUIEP MAGNIL. SEP 16 DATA SEP I_ : Only minor activity reported. Region 14945 retains delta configuration according to Boulder 15/2200 report, with microwave flux steady. NOTES : No change from previous day. WWA SOLAL_RT 15/17 MAGALERT 15-_/_. Tokyo RWC advised SOLAI_T MAGQUIET. o os, soLQ I (sou T,)MAGQ r(MAGi,.). SEP 17 DATA SEP 16 : Culgoora reports IN flare north 08 degrees west 19 degrees, in Region 14943,SEP 16/2120 to 2250 with 2N flare starting 2230 still in progress at report time, Y-shaped with umbra/ coverage, also very large metre-wave bursts and spectral type III and II-IV event, also still in progress. The type IV source was reported as moving to the west from a position (in projection) on the NW limb, at report time. These events occurred too late for mention in the Boulder 16/2200 report. NOT___ : Disturbance Warning and MAGAI_T should be issued, with some reservation as Region 14943is not a metre-wave noise source. The Y-shaped flare with umbra/ coverage is also characteristic of a proton flare (see SEP 10 data). If a magnetic storm does result from this event, the time delay from a region with calcium plage area of about 5000 millionths of the solar hemisphere as reported by McMath- 242 0 1980010188-253 b"f o Hulbert Observatory, should be at least 24 hours and possibly longer (CGok 1970). The coronal hole disturbance is also due to start about 19. The SOLNIL advice issued S_ 16 has been proved incorrect. | 18/..W_ : issued 17/0200 "confirm severe fadeout _ 16/2220Z due to major flare in the region now one _ past central meridian ... available data on this event indicate severe magnetic storm possible starting late 17 to early 18 September through 19/20 ..." S_ 18 DATA BEP 17 I Hiraiso I00, 200, 500 MHz data received for the 16/2230 t event indicate disturbance travelled outwards through the corona, ! but the 9500 MHz time is in apparent disagreement (later published I data list the burets at 100, 500, 9500 MHz all starting at 2226). i Iknulder17/2200 reports energetic particle event started 17/0300 but subsiding by 1800. Oulgoora heliograph observations to 17/0515 describe ! the type IV source moving from north-west limb to near equator at west : limb (later published data show the radial distance from disc centre decreasing, i.e. source mo',ingnearer to sun-earth llne). NOTES : These data support increased probability of geomagnetic storm _-the 16/2230 event, but give no indication of the delay time. WWA SOLAL_T 17/18 MAGALERT 17/20 PROTON ARRIVAL AL_T. G_0SYD : SOLAL_T MAGALERT 19/.. S_ 19 DATA SEP 1_ ; Culgoora 2N flare north 08 degrees west 37 degrees (Region 14943), in progress 18/0050 with umbral coverage, spectral type III and weak type IV events. Eiraiso data list minor bursts 9500, 500, 200, 100 MHz with start times progressively later with decreasing frequency. Other minor events reported in the ursigrams and in Boulder 18/2200. Solar microwave flux decreased today. NOTES = 'lV_isevent may contribute to disturbance already e_ected. The coronal hole disturbance is also expected to start today. WWA 16/23mCAI,mT18/19, RWCS0I,AI T 18/.. OALm 18/20. GEOO'YD: SO_T MAG_T 19/.. WARNING : issued 19/2330 :'moderatemagnetic storm began 9/ 143ZSEP 1 1 much later than expected end should continue through SEP 20/21 possibly severe at times. Since previous warning daylight fadeout was observed 18/0022Z and another major solar event occurred 19/I031Z in the same region now four days past central meridian. Further daylight fadeouts probable next three days." SEP 20 DATA SEP 1_) : Geomagnetic storm began about 1140 SEP 19. Boulder ;i 19/2200 reports magnetic SC 1305 (later published data list SC 1143 at some observatories). The storm has recurrent-type appearance on + • Fleurs ma_mtogram, that is, substorms but not a well-developedn_An phase. Boulder 19/2200 also reports _B flare .orth 09 degrees _est 59 degrees (Region 1494_) at 19/1045, followed by proton event at 1145 still in progress. Region 14952 (south 23 east 40) now has a D-type spot group with beta/gamma magnetic configuration and is growi:_E. Sagamore Hill data show large complex burst 8800 and 1415 MHz but small at other frequencies including the metre waves. _ , NOTES : .'Fne storm has the appearance of recurrent-type (coronalhole) disturbance with possible SC storm superimposed. If the SC 19/I_05 is 2_3 =. al 1980010188-254 • due to the flare 16/22_0, this represents a long dela_ of about 63 hours. If the 19/1045 flare also causes a storm, this is likely to be also superimposed on the coronal hole disturbance to produce a severe disturbance with about the s_me delay, !ate on SEP 21. MAGAL_T and SOLAL_TS continue. G_DSYD = SOLAL_T MAGAL_T 20/21. SEP 21 DATASEP 20 s Oulgoora 3N flare, north 15 de_rees west 55 degrees (Region 14943) in progress at 20/0321, ended after U748. Strong type II 0315, wide-band type IV 0257-0520.Complex microwave burst recorded at Fleurs. Htraiso major metre-wave event. _oulder 2200 report describes this flare as 1N. Other minor events reported. NOTES s This flare may also contribute to the disturbance. If an identifiable increase in disturbance occurs late today as a result of the 19/1045 event, this will increase the probability of another increase about 15 hours on SEP 22. WWA SOLALERT and MAGALERT continue SEP 20. GEOSYD : S0_T MAGALEST 21. SEP 22 "DATAEEP 21 : Magnetic SC 21/2045 with storm apparently superimposed on the continuing recurrent-type disturbance. No significant solar eventS• NOTES : This SC may now be plausibly attributed to the 19/1045 e_ent with a delay of 60 hours (see Discussion SEP 20), and strengthens the probability of another increase late SEP 22. WWA SOLAL_T and MAGALEST continue. G OS, SO LUmTMAGA RT22/23. W.ARNING: issued 22/0415 '9_agneticstorm ... is expected to continue through SEP 23 •.. Region producing major activity is now rotating off the sun's visible disc but possibility of major flare activity still exists through SE_ 23." SEP 23 DATA SEP 22 : No significant solar events reported. Fleurs magnetogram shows disturbance ending late SEP 22 with only slight increase in amplitude during the middle of the UT day. : No evidence of increased disturbance 60 hours after the 20/0321 flare : Fredericksburg dat-,available only to 1200 hours S_ 22. (On the Bartels Kp charts, the planetary magnetic index does increase on SEF 22 from 4 at 09-12 hours to 6+ at 15-18 hours, but the Kakloka and Fredericksb,,rgindices received next day did not indicate any new storm at this time. Later data from high-latltude stations however do show a marked increase in K index starting with the 12-15 hours interval.) WWA SOLALERT MAGAL_T MAJOR SEP 22. GEOSYD : S0_T M_.n.tL_T23. SEP 24 DATA SEF 22 : No significant solar events reported. Disturbance has subsided and the magnetometer trace was quiet up to 1300 hours SEP 23 when the mag_,etometerceased recording because of a power failure at Fleurs, NOTES : The Barrels Kp chart for the previous rot_.tion shows a _aiet perio--'--'d(at end of 26 August) followed by renewed slight disturbance. This quiet period appears to have recurred early SEP 23. W_A SOLAL_T 23 MAGALERT 23. Z4_ 1980010188-255 G_0SYD I S3LQUI_T MAGALE_T 24. i S_ 25 DATA. SEP 24 , Culgoora observed a north-west limb surge abo_t SEP 24/0556 accompanied by strong type III and II bursts. H_raiso reports short major burst 200 MHz at this time, minor at 500 MHz, no event at 100 MHz. Boulder 24/2200 reports minor PCA event 0600 _till in pro- gress. The Fleurs magnetometer beg_n recording again at about 0600, [ _howing no disturbanc. _ NOTE3 : Further significant geomagnetic disturbance unlikely. The PCA is apparently associated with the north-west llmb event (later published data amended the Culgoora type II to type IV). WWA G_DAL_T SOLNIL MAGNIL. GEOSYD : MAGQUIET SOLQUIET. S_ 26 DAT_ SEP 2_ : Minor activity in Region 14963 near north-east limb. Boulder 25/2200 reports PCA ended 25/0300. NOTES : No change from SE_ 25. WWA SOLQUIE2 MAGQUIET. G_DSZD : SOLAL_T 26/2_ MAGQUIET. SEP 27 DATA SEP 26 : M/nor flares in north-east (Region 14963) and south- west (Region 14952) with SWF and type IIi bursts anl minor microwave and metre-wave activity. Fleurs magnetometer sho_s a possible SC at 26/1418Z but no significant disturbance. N__ES : The possible SC is delayed 56 hour_ from the behind-the- limb event S_ 24/0556, consistent with delaj times from earlier events in Region 14943. WWA SOLALERT 26/28 MAGQUIE_. SEP 28 DATA SEP 27 : Minor activity continues in the north-east and north- west. Boulder reports mention a proton event observed by satellite, beginning SE_ 26/0920, now ended. The @eomagnetic field was more disturbed during the early part of the day. NO__TES: The recrldescence of geomagnetic acti'zit_ is possibly related a_so to th_ S_ 24/0556 e_ent. WWA SOLALE_T 27/28 _GAL_T 27. GEOSY_____DD: SOL@UIET MAGNIL. 4. DISCUSSION In these notes we have c3nsidered the events _uring the disc transit of one centre of activity, uncluttered by significant activity from other regions, but complicated by the presence of one or two coronal holes - see Table I. For_asting problems Drought out, and which still await resolution, include: (I) Dependence upon meridian distance : the events at east limb and _ast 80 degrees apparently caused no geomagnetic disturbance, although some Warning Centres expected them to do so. D_sturbance can b_ attributed to all but one of the west hemisphere events iz_oluding, possibly, one from behind the west limb. (2) Del_v between solar event and geomagnetic storm. All Warning Centres expect _d a storm to start much earlie_ after the SEP 16 event. A remarkable feature of the delay times is the consistent value of about 245 1980010188-256 60 hours for each event, reminiscent of homologous flares (Ellison et al. 1960) and homologous figure-associated radio events (Fok_ker 1967), but homology of flare or radio events is not apparent here. On the other hand, the reported McMath calcium plage area of Region 14945 remained stable during its disc tra_zs_t. (3) The recarrent-type disturbance starting about 1140 SEP 19 is c_nsistent with the reported CMP of a coronal hole SEP 16 : a coronal hole CMP SEP 11 was not reported at the time. Earlier separate studies of a coronal hole CMP NOV 24 1970 (Krieger et al. 1975 and Kundu and Liu 1976) identi- fied it as the source of high-speed solar wind streams but no signifi- cant geomagnetic disturbance followed its CMP on that occasion. (4) The occurre_.ce of PCA as a pointer to geomagnetic distarbance. The three flares associated with recognizable geomagnetic SC were all followed by fast PCA events according to the classification of Sinno (_961) while those associated here with recrudescence of activity without SC were not. ? (5) No information wa_ received about possible sector boundary crossings, or changes in the _irectLon of the interplanetary magnetic field, during thi_'period. (6) P,_itive identification of spectral type events, and interpretation of z_dioheliograph events is not al_ays possible when the observer is necessarily preoccupied with the optical flare, the solar and radio observatories at Culgoora being located in separate buildings. T_is results in _ocasional discrepancies between the events reported at the time (read from facsimile and heliograph display) aod the _ata published later in Sol -Geophyszcal Data (scaled from spectrograph films and heliograph tai_). _7) The weoaagnetic dzsturbauces beginning SEP 19 had the appearance, on the Fleurs magnetogram, of typical recurrent storms, that is, positive and negative excursions but with no obvious main phase. We are therefore not certain of the reality of the apparent SC storms in this sequence. It has long been k_wn (Allen 1944) that a recurrent storm can begin with a sudden commencement if the supposed M-region is preceded by a centre of activity, in this case by Region 14943. (8) A centre _f activity producing flares with conse aent geomagnetic storms woule be expected to be a stron_ metre-wave source as well as emitting bur_ts with the flares, but Region 1494_ was not. The significance of this is not clear. 246 @ 1980010188-257 RE_CES Allen, C. W. (1944) : Relation between magnetic storms and solar activity. Monthly Notices of the Royal Astronomical Society, 104 : 13. Bell, B. (1963) : Type IV Solar Radio Bursts, Geomagnetic Storms, and Polar Cap Absorption (PCA) Events. Smithsonian Contributions to Astrophysics, 8 : 119. Cook, F. E. ($970) : On the time delay between solar flares and associated geomagnetic storms, i_aper 28 in AGARD Conference Proceedings, No. 49. Cook, F. E. and C. G. McOu (1975) : Solax-terrestrial relations and short- term ionospheric forecasting. The Radio and Electronic Engineer, 45 : 11. Denisse, J.-F., J. L. Steinberg, and S. Zisler (1951) : ContrSle de l'activit_ • . . . • . . • geomagnetlque par les centres d'act_vlte solaires dlstlngues par leurs propriet@s radioelectriques. Comptes Rendus, 232 : 2290. Ellison, M. A., S.M.P. McKenna, and J. H. Reid (1960) : Cape Lyot Heliograph Results I. Light-curves of 30 solar flares in relation to sudden ionospheric disturbances. Dunsink Observatory Publications 1.1. Fokker, A. D. (1967) : Homology of solar flare-associated radio events. Solar Physics, 2 : 316. IPS (Ionospheric Prediction Service), Department of Science (1977) : Handbook IPS H-6 for use with ionospheric p_'edicLion services. IUWDS (International Ursigram and World Days Service) (1973) : Synoptic Codes for Solar and Geophysical Data, Third Revised Edition. I_rieger, A. S., A. F. Timothy, and E. C. Roelof (1973) : A coronal hole and its identification as the source of a high velocity solar w_nd stream. Solar Physics, 29 : 505. Krzvsky, L. (1963) : Y-shaped stage of cosmic ray flares development as phase conditioning ejection. Bulletin of _he Astronomical Institutes of Czechoslovakia, 14 : 77. Kundu, M. R. and S.-Y. Liu (1976) : Observation of a coronal hole at 85 GHz. Solar Physic_s_s,49 : 267. Nozaki, K., K. 1_rubashi, Y. Miyamoto, and K. 0hbu (1978) : Overview of Solar- terrestrial physics phenomena associated with McMath Region 14943 based on uzsigram data. Solar Terrestrial Environment Research in Japan, 2 : 61. Sinno, K. (1961) : Preliminary Study of Prediction Aspects of Cosmic Ray Events. Journal of _he Radio Research Laboratories (Japan), 8 : 17. Yoshida, S., and S.-I. _2.asofu (.1965) : A study of the propagation of solar particles in interplanetary space. Planetary and Space Science, 13 : 435. 248 J 1980010188-259 {;f" _._ 2.-" _'_ REAL- AND NEAR-REAI_TIMEFORECASTSOF foF2 AROUND AUSTRALIA I L. F. McNAMARA Ionospheric Prediction Service Department of Science and the _vironment : !_ P.O. Box 702 Darlinghurst NSW 2010, AUSTRALIA This paper describes techniques using ionospheric indices which have been developed to make real- and near-real-time forecasts of foF2 at points around Australia. It is divided into two parts, corresponding to two time scales and the corresponding computer programs. Part I covers forecasts I day ahead using daily ionos- pheric indices (program DALYPRED), while Part 2 covers forecasts 0-3 hours ahead using hourly ionospheric indices (program HOURPRED). Both programs provide extensive tabulations of errors in the fore- cast values of foF2, but only heavily compressed results are included here. The present analysis differs from a previous one (Reference 2) in that it is a more general approach, not relying, for example, on the presence of an ionospheric station at the position where ; the forecasts are required. It also takes advantage of the ability of the IPS world maps of foF2 to inter-relate values of foF2 at different locations once the appropriate ionospheric index has been specified. It is found that for forecasts one day ahead the most accurate results are obtained using the past variations of foF2 at several stations to determine a latitudinal-average daily ionospheric index. Forecasts made with l._ad times of 0-3 hours do not offer significant improvements over the se made one day ahead. Both programs may be used to provide answers to specific questions as they arise from customer requirements. PART1- D_YPRED I. INTRODUCTION DALYPRED is a program which makes forecasts of fcF2 at a test station one : day ahead, calculates the errors in the forecast values and analyses them in various ways. It is not an operational program in that it is not designed to 249 I 1980010188-260 : make forecasts in real time. Rather, it is a developmental program, aimed at investigating the exTors inherent in the techniques adopted to make the fore- casts and using only "old" ionospheric data recorded over recent years. Basically, the procedure which hss been adopted in making the forecasts is to monitor the ionosphere using fiw_.mainland Australian ionospheric stations, project this observed behaviour forward one day at each station and then calculate the expected behavio_u_ at a test station. The five reference stations are Mundaring (Perth), Townsville, Brisbane, Canberra and Hobart. The _est station was Norfolk Island although others could just as easily be considered. (In practical applications, this station could also be used as a reference station and the method described here modified appropriately. However at the present s_age it is necessary that it be reserved as a test station.) The principle underlying the present approach is that the behaviour of the ionosphere can be characterised for each day by an ionospheric index "T" in terms of which all calculations may be effected. Several different types of index are investigated, as described in later sections. DALYPRED requires various forms of pre-calculated input and is in fact the sixth step in the overall problem. The preceding five steps are discussed very briefly below. The logical seventh step is to apply the same techniques to operational situations, provided that the errors involved are acceptably small. 2. PRELIMINARY PROGRAMH Several of the preliminary programs are used to determine for each stat- j ion the relationship between a T index and the 24-hour average of the monthly median values of foF2: T = a + b * _ foP2 The T index used initially was the world-wlde average monthly index deduced from the calculated indices at 16 particular stations distributed around the world (reference 3). The above relationship is assumed to hold also for a time scale of one day, so that the average of 24 hoiLrly values of foF2 for a day yields an index which can be used to characterize the behaviour of foF2 throughout that day. This particular index is called IT in the program DALYPRED. The residuals of T for each month may be used to estimate the size of the error in the forecast values of foF2 caused by the scatter of the points about the fitted line. For Brisbane 1970-74, for example, the average residual in T is 2.2 and the standard deviation is 5.6. The corresponding • uncertainties in foP2 can be estimated from the rough figures that a ch_-_ in SO in the T index changes foF2 at Brisbane by about 5% or 0.5 MHz. A second index, TF2, is obtained by averaging the 5 individual values 250 °, 1980010188-261 for each day. The underlying philoso_ here is that such an approach might yield an index less affected by the scatter about the regression lines. A third index, 'IrFX,is obtained from the daily value of 10.7 cm flux using the empirical observed relationship between monthly T indices and monthly average flux values. (Figure 1 of Reference I.) Note that this index makes no use of foF2 values, and therefore negates any need to monitor the ionosphere. It has been found that, on average, the discrepancy between TF2 and TFX is almost zero, suggesting the TFX may be an adequate index, even without further analysis. 3. PROGRAM DALYPRED This is the program which makes the forecasts and aualyses the errors. Forecasts awe made for March, June, September and December for 1970-74 and compared with the observed values of foF2 at the test station. For a given T index, forecasts of foF2 are obtained by linear inter- polation between the values predicted by the standard IPS area MUF program for low (T=0) and high (T=IO0) solar activity. Forecasts are made using both the correct daily values, TC, _d the pro. : jected daily values, TP, of three indices. The three basic indices are as described in section 2:- ' IT - the separate indices at the 5 reference stations, TF2 - the average index for these 5 stations, and TFX - the index based on the 10.7 cm flux. TC(1) is calculated for the test station assuming that the index IT for a given day is a function of latitude only. TC(2) and TC(3) are set equal to TF2 and TFX. For later reference, indices derived from IT, TF2 and TFX are referred to as TI, T2, and T3 respectively. The projected indices TP for day I, say, are calculated using the correct indices TC for days I - I, I - 2, ... I-NDAY and a least-squares fit subroutine, w_h all data being given equal weight. The subject of weights will be referred to again in Section 4. Initially they are set to unity. This simple projection process was found to be unsatisfactory when ionospheric storms were present, the recovery phase yielding unrealistically high values for several days after each storm. The projected indices are therefore modified in a physically reasonable way. A storm is defined as a TTOL fractional decrease of TC from the value on • the preceding day (presently set at 0.3), and may exist for either one or two days. After a storm, if the projected TP for day I is greater than TC on the day preceding the storm, it is set equal to the average of that value and of TC on the preceding day, I - I. In this way, TP is forced to recover to 251 t _L_ 1980010188-262 the TC on the day preceding the storm. TP is also forced to recover mono- tonically. Forecast values of foF2 are calculated for the test station for each hour 7 for a particular day using the three predicted and three correct T indices, and the T vs. foF2 relationships (actually fxF2) defined by the T = 0 and 100 values of fxF2 des,_ribed in Section 2. These values are compared with the correct, observed, values and the discrepancies expressed as a percentage of the correct value. IDEL = (predicted - observed)/observed x 100% The errors are then analysed in various ways, since no way seems to have any prior advantage. Median and decile values are calculated for the 30 or _ 31 values at each hour; average v_Llues and variances are calculated for the 24 hours of each day, and for the 30 or 31 values at each hour. The program then works out which is the most successful index, according to the number of times each index gave the smallest (absolute) values of the median, average and variance at each hour. In order to compare the errors with some absolute standard, rather than just with each other, the error distributions are calculated for the case when the monthly median values are used as the forecast values. Operationally, this would correspond to the a_e of a correct predicted monthly median. These errors are characterised by interdecile widths and variances and the errors for the most successful index comp._red against them. Two success counts (_<24) are then made, depending on whether the interdecile width or the variance at each hour is less than or greater than that corresponding to the use of the median value. 4. MODIFICATIONS TO DALY_R_D In Section 2, the T index used for the regression _tualysis was the world- wide average T index. An alternative approach, which tuzu.sout to yield smaller errors, is to use the previously established relationships between a station T index _ud monthly median values of foF2. This relationship is available for each month for each of the 16 stations used to calculate the worldwide T indices. Ur_ortunately Townsville is not one of these stations, so DALYPRED needed to be slightly modified. Further possible modifications to DALYPRED, which have not been imple- mented, are _ 1. The lead time could be changed to two or more days. 2. In the projection process, the weighting scheme could be changed in order to give, say, greater weight to more recent data. • (Note, however, that the weights had little effect in the fore- " casts of foF2 a few hours ahead at Brisbane and C_nberra - see Reference 2). The length of the reference period, 'NDAY' days, 252 : I 'l m 1980010188-263 . couZd also be changed, t_erhaps going as low as 1. 3. Rather than projecting the T index for each station, and then calculating its value at the test station, perhaps the values should first be calculated for all necessary preceding days at the test station and then lhese values projected. 4. The criteria for the "best" T index could be changed, for i example, by considering only the variances. 5. The sacoess counts for the best T index could be replaced by a reference to an externally defined acceptable error limit. 5. DISCUSSION OF RESULTS Although DALYPRED provides detailed listings of the errors, we shall limit the discussion here to which is the most successful T index and how often the interdecile widths and variances are less than those obtained using a correct median value as the predicted value. The program has been used to make forecasts for March, June, September and December of the five years 1970-74. Usingworld-wide T indices (Section 2) leads to the T2 index being the most successful for zero lead time:- TI : T2 : T3 -6 : I0 : 4 For a lead time of I day, which means that additional errors are being intro- : duced by the projection process, T2 _uldT3 are equally successful:- TI : T2 : T3 -=4 : 8 : 8 Using station T indices (Section 4), on the other hand, results in T2 being clearly th._most successful index:- Zero lead time TI : T2 : T3 -0 : 17 : 3 One day lead TI : T2 : T3 --2 : 12 : 6 The question of which is the better, the station indices or worldwide indices, can be answered by comparing the success rates for the interdecile widths and variances listed in Table I. This table shows that the use of the station T indices leads to somewhat higher success rates on average (the highest "score" possible each month is 24), although its overall success rate is still less than 50% (467/960). The use of the station T indices is also indicated by the clearer superiority of one index (T2). From the practical point of view, it would be necessary to know which index to choose, TI, T2 or T3. Although the obvious choice is T2, since it is the best on most (12/20) occasions, other indices are more successful at other times, and no way has been found to predict under what conditions each index will be the best. The best index for each of the months considered is 253 u I 1980010188-264 llsted in Table 2. There seems to be no definite dependence, for example, on the presence of ionospheric storms during the month. For real-time forecasts (lead time of zero), T2 is clearly the best index to use. Interdecile widths Variances M J S D Total M J S D Total 70 11 13 2 1_ 39 17 17 6 13 53 71 19 5 15 1 40 20 17 20 8 65 72 5 5 11 11 32 12 6 18 15 51 73 7 7 13 lO 37 11 10 22 12 55 74 9 7 12 12 40 16 19 11 9 55 Total 51 37 53 47 188 76 69 77 57 279 Equinoxes 104 Solstices 84 Equinoxes 153 Solstices 126 70 10 12 9 11 42 18 11 14 11 54 _. 71 _5 4 15 o 34 17 9 17 8 51 72 8 2 11 11 32 12 6 18 15 51 73 3 4 5 6 18 8 5 16 8 37 74 4 ? 12 8 31 9 19 10 3 4/ Total 40 29 52 36 157 64 50 75 45 234 Equinoxes 92 Solstices 65 Equinoxes 139 Solstices 95 Table 1 The success counts (_24) for each month according as the inter- decile widths or variances of the forecast errors are lees than the corresponding parameters obtained using the post-facto observed median foF2 as the forecast value. The top set of results corresponds to the use of station T indices (Section 4) and the bottom set to the use of world-wide T indices (Section 2). The total success rates (out of 960) for the two are 467 and 391, respectively. 1970 2(Q) 2(Q) 2(Q) 2(Q) 19'71 2(Q) 2(Q) 2(D) 2(Q) 1972 2(Q) 3(Q) 3(Q) 3(Q) 1973 2(Q/D) 2(D) 2(Q) I(Q) 1974 3(D) 3(D) I(D) 3(D) Table 2 The T index, TI, T2, or T3, which yields the smallest forecast errors in each of four moSths for 1970-74. The index T2 is the most successful index, in the ratios T3 - 2 12 6I : T2 : : : The letters Q and D denote subjective classificationsof each : month as being magnetically quiet or disturbed. 254 I i 1980010188-265 i 6. FUTURE POSSIBILITIES FOR DALYPRED i Since it would be necessary for an operational program to use only one • index, presumably T2, the success rates for forecasts using T2 may be required. These could be obtained by slight modification of the program, which in its present form calculates the success rates for the best index. i Additional progressing could also be added (as was done for HOURPRED) to record the number of times the error in the forecast value was less than some (acceptable) level, say 10%. 7. CONCLUSION-DALYPRED It has been found that forecasts of foF2 made one day ahead for Norfolk 7sland are most accurate when made on the basis of the average behaviour of the ionosphere over mainland Australia i.e. using the index T2 defined in Section 3. Details of the errors for this and the other indices, T2 and T3, are available as computer listlngs. It has not been found possible to develop an index which is always the best, nor to determine under what conditions T2 is not the best. However there does seem to be a tendency for T3 to be better than T2 when both are small, probably because T_ varies more smoothly than T2 and is thus less susceptible to projection'errors, especially those cau§ed by rapid changes in T2 due to ionospheric storms. PART 2 - HOURPRED I. INTRODUCTION The basic approach in HOURPRED is similar to that of the program DALYPRED, the major difference being the time scale which is now hours ra_her than days. Predictions of foF2 at two test stations (Norfolk Island and Mundaring) are made for lead times of 0 to 3 hours, using the observed hourly values of foF2 at five reference stations (Port Moresby, Townsvill_, Brisbane, Canberra and Hobart). Data manipulation is done in terms of a T iI_dex; it is assumed initially that for a given hour the T index is a function of latitude only; and projections ahead in time are performed using a least-squares fit regress- ion line. Percentage discrepancies in the predicted values of foF2 are analysed in terms of medians, deciles and variances, and can thus be compared directly with the error distributions obtained using DALYPRED. It is assumed that the two test stations are at 165° and 120° E longit1_r and that the five reference stations are all at 150 E. For an operatlona] program the values of fxF2 at e required location would be obtained by 255 t 1980010188-266 _ interpolation in the appropriate world map of fxF2. 2. D_TAILS OF PROGRAMH_ 2.1 The pro.lected T indices For each hour of the given month, hourly values of a T index corresponding to the observed fxF2, T, are calculated, assure/r- ,at the relationship defined by the T = 0 and 100 values of monthly median fx_2 is also valid for hourly values of T and fxF2. The projected T indices are then calculated as in DALYPRED, except that no allowance is made for storms since it is expected that the use of hourly data would be one of the most reliable ways of making this allowance. 2.2 T indices at the test stations As mentioned previously, it was assumed initially that for a given hour the T index is a function of latitude only. The T indices for the test stations were therefore obtained by a simple interpolation in latitude. The interpolation was bypassed and no prediction made if no data weze available for Port Moresby or Hobart. Each interpolated predicted T index for a given local time at the ref- erence stations was then associated with the same local time at the test stations. This option still exists in the program but the T indices so derived have been overwritten by latitudinal average of all (up to 5) available station T indices, since it was found that the use of the average value significantly decreased the variance of the errors. 2._ Predicted values of foP2 and errors The predicted value of foF2 corresponding to each predicts- _ is calculated as described _n Section 3. The percentage discrepancies between px_dicted and obserlec ;Lupo foF2 are then analysed to yield medians, deciles, averages and vari_ _-_' In order to facilitate the discussion of errors, the individua' . _ have been tested to see if they are less than some particular value , _ned by the parameter TOL_. For the present report, TOL_R has been set at 10% and the occurrence rates of errors less than this calculated for eacl_ hour and then averaged to yield one value per month for each lead time. Since it was found in practice that the intercomparison of HOURPRED and DALYPRED variances was too sensitive to just how each program handled missing data, ROURPRED was modified to cycle _wice, the first time using some daily T index for all forecasts and the second using the hourly indices. Although any set o: daily T indices could be used, the most logical choice is the pre- dicted T2 index as used by DALYPRED for the same month. 256 t 1980010188-267 3. DISCUSSION OF RE_LTS Individual errors, median errors, upper and lower deciles, average errors end variances, and success rates for errors less than I_ are all printed out by HOURPFED. However for present purposes it is sufficient to consider average values oi th_ success rate for each month. Table 5 shows the average success rates for March, June, September and December 1970, for Mundariug (No. 2961) and Norfolk Island (No. 4260), for both pz.edicte_daily T_ indices and for latitudinal averages of the predicted hourly station T indic_s. The important point to note is that tb_ success rates are about the same for both daily and hourly indices, indicating that the effort of updating forecasts on an hourly basis is not worthwhile. (The December 1970 results are suspect b_'_.se _ small sample sizes, _ 10, due to missing data). However the use of the dail> _ndex T2 still requires the collection of hourly values of fc,F2. 2961 MARCH 70 4260 MARCH 70 0 I 2 3 0 I 2 5 Daily 60 60 60 60 Daily 72 72 72 72 Hourly 61 62 65 57 Hourly 72 69 70 70 JUNE ?0 JUNE r0 Daily 56 56 56 56 Daily 62 62 62 62 Hourly 60 64 65 57 Hourly 65 64 60 56 8EPT_B_ 70 SEFTEMB]_ 70 Daily 61 61 61 61 Daily 66 66 66 66 Hourly 71 75 75 64 Hourly 69 65 58 58 DECDmm_70 DECEMB_7: Daily 51 51 51 51 Daily 42 42 42 42 Hourly 59 63 65 54 Hourly 61 56 55 51 Table _ The monthly averags success rates for forecast errors of less than 10%, The rates are given _or forecasts made using projected values of the daily index T2 as calculated by DALYPRED and for forecasts made usinghourly indices projected 0-5 hours ahead. (_und-ring, No. 2961; Norfolk Island, No. 4260). 257 % l 1980010188-268 _CES MoNna, L. F. (1976). "The correlation of individ_,l values of foF2 and _"_O00)F2 wlth the eclat 10.7 cm flux under ma6netically _u/et conditions". i. " Series R Reporti_,IPS-R30. McNamara, L. F. (197(). "Short-te:_ forecasting of ,oF2". IPS Series R Report_ IPS-R33. Turner, J. F. (1968) "The development of the ionospheric index-T". IPS Seriee R Reports, IPS-R11. 258 @ ! 1980010188-269 P. J. WILKINSON Ionospheric Prediction Service Australian Department of Science and the Environment P.O. Box 702 Dar!inghurst NSW 2010, Australia Some of the factors limiting the prediction of foF2 are investigated i. using median and hourly foF2 values recorded at stations in theAustralian ionosonde network. Different time and spatial scal s together with different levels of solar and geomagnetic activity are considered. The unique feature of this presentation is the display of all data used in the analysis in the form of a series of scatter diagrams which provide an overall view of the accuracy limits for . foF2 predictions. f : i. INTRODUCTION ' Basic to the problem of making ionospheric predictions is the prediction of the F2 region penetration frequency, foF2. The factors limiting the accuracy of foF2 predictions are therefore fundamental to the overall problem ; of HF prediction. For a given location, foF2 can have a wide range of values dependent, to a great extent, on the sun's declination (diurnal and seasonal effects) and on solar activity. While the ionospheric F2 region response to these factors is not simple, because of their cyclic nature it is possible to produce empirical models of the ionosphere on _he basis of recorded past behaviour. These "models" can then be used for prediction purposes, the only variable requiring prediction being the level of solar activity. Most systems, such as that described, predict average monthly conditions on the basis of a predicted monthly solar index. In this paper the range of average monthly foF2 values is displayed for various stations and the effectiveness of predicting this behaviour is assessed. This 'effectiveness' is demonstrated by comparing _ recorded median values of foF2 with predicted values, the comparison being ! made by subtracting the predicted value from the observed values and the results being presented as a scatter plot. Comparison of predicted median values with recorded hourly foF2 values demonstrates the limitations in using monthly predictions to anticipate daily ionospheric conditions. _ . Some sim_le methods to reduce these limitations are investigated. These methods involve the use of previousl._'recorded hourly values to project forward both in time and space in, essentially, a real-time prediction system. 259 • C4 - i ................... =........... 1980010188-270 Because the same data presentation is used throughout it is possible to compare these methods with the use of monthly predictions. The prediction method used to obtain the monthly prediction is the standard IPS system, but the comparisons presented in this paper are not intended as a test of the IPS predictions. (It is necessary to explain the presence of a systematic offset in the comparisons. The IPS predictions give fxF2, not foF2, as the expected top frequency for the F2 region. Thus, the difference between observed and predicted values will be too low by ½fB, the geomagnetic split. It was convenient, computationally, not to remove this offset from the data before plotting). To make useful comparisons of F2 region predictions the known ionospheric index, rather than the predicted index, was used, thereby eliminating errors in predicting the correct level of solar activity. 2. MONTHLY PREDICTIONS In general, prediction systems based on median maps are only useful for predicting average undisturbed conditions during a month. In making a monthly prediction, three components are allowed for - diurnal, seasonal and solar cycle. Figure i gives all the recorded median values for three stations. The format of the diagrams is given in section 6. There is a clearly defined range of values that may be anticipated for each station and it is also possible to identify seasonal and solar cycle components in the time series. However, the diurnal variation makes further interpretation quite difficult. When only a single UT hour is plotted (figures 2, local midnight, and 3 local midday) the diurnal variation is removed and the relative importance of the solar cycle and seasonal components can be readily assessed. It is apparent that during daytime the solar cycle component is generally more important than the seasonal component. Most prediction systems can adequately account for these factors. Figure 4 _ives the difference between observed and predicted values of the monthly medians for all recorded foF2 data. Of the three stations, Brisbane is clearly the best behaved. The greatest deviations from predicted values are obtained for the high solar acti_rity in the 1957-59 period. The prediction system in this period gave a false impression of the available frequencies for communicators. If, as has been suggested (Sargent, 1977) cycle 21 is similar in intensity to cycle 19, some thought may have to be given tc improving the system for high solar activity. Seasonal plots (Figure 5) were obtained by folding all data into a one year period. The only significant effect that appears is the diurnal variation in deviations for Vanimo which is more important iu Winter than other seasons. From the small differences shown it can be seen that the nonthly predictions are effective to within a megahertz, though for Vanimo there is still a large residual component. 3. PREDICTING HOURLY VALUES It is the hourly foF2 values which are indicative of the actual HF 26O 1980010188-271 lull I_ ""lip eO I• :_ Vanimo i, Vanimo : ' '_ " A n= .... • gr.P.t_,_';;{,5.t.;., . '" ' " • • _'_'__...,_.._I.-_ ._... _ • . , _ az _ . • . t .._ ......i.::,,.. • _ : • _ •,-.. ,_,'...".-, . ,.,. OI,' THE -, .._._.,._ ...,., ..- _ o_. .,O?R_-._ .. ... .;. .', • a _ "_a-: -., • IU_R -' • ""_ " •4"1_._ .,_,,_ •._, "_'"':_""4_'_.a. _: • • • _,,:_; " ".'"*.:".P.:t',",'_-,-,. __¢-"-]_.1;.. _,-.._-.-................ _... • • O_G _': .- •,. .-...... . ;. :....: ,,,..._37.., ;.,.,",i_,, •"_ • • . • _ "s MI _"" • 4 • " v.." I Fig. la. Monthly foF2 medians for all Fig. 2a. Honthly foF2 medians for io- 24 hourly medians a_ Vanimo. cal midnighg at Vanimo. r*sbaoe I--•, Brisbane _:1 .-" I" •!=0t :-IMp." _ik_iit_ ,. : : i- '" ,,] _ - .. ,.., "- .,. "2 • _ • ." " • " '" "'" ' "" s -'..". • " . "',- " " :"" " •4 >.j -.; , .."__:.,:, : : :.,.'".=- .:.'"i';"_- ;' ;'" " Fig. lb. Same a_ Brisbane. Fig. 25. Same at Brisbane, I• n, Hawson =o Mawson I_11 *- I_1 II . I,I I', ._ I] I • 5- ".,: i z • • P* ilII _,e ._-! • .-.... .;..,:.,IL.. I.. • - _-_i '....._,.__" .. ,.,:.,_'_._.: ;. ,., ,.:..-,_.-=:.- .__=,_ • .-.. :_ -' . : ..; = • ..'. .-.. .. . ._... .• . ..._,.:-:.. "_,'_..,':_" .'" 5. • _ " _ " • %, _._ '._ '._ ',',',', ',_ _ "_ _m " _ cam" _ _-__s '--_-a"_ ,_ ",,¢" " Fig. ic. Same at Mawson. Fig. 2e. Same at it_wson. 261 1980010188-272 Vanimo • q ** . " "" ""'":" " " " 11 '. •. ,:::,-,;" ..':• • • -" t.', ._ " " • .'": _ I " "•-". ." -" • ; _':_:_ .'1:'J.._ '-.*:'." .._,k. ,.-. .... - . •, ."... -,F_,::___ ._."-.,:._:_ ' " " :iI_!_-?._."<.":"'_ "" -44 z , , ! --lP 4 •=• q --I .I MKAIIS - '" _ _ f,q'¢_ ", ....... '.". '.'.. - " _ • lus " Qp • 4iQ " 19 " _ " J'i J r,_ _ ri " _'4 " PS " TI I• Ia 6_ I_, " &r " 6I " ' " ' Fi::. 3a. Monthly foF2 medians Fig. 4a. Difference between observed £or local noon at Vanimo. and predicted foF2 monthly medians for all recorded data. • ' Brisbane m Brisbane . ,|r ii • 1 is • 1 .. Z • 1 .J. _.. . • --, _Z ,.'","- "':-'-.".;. -. " : • • "# -@-* .*%. .% "-*-* . • i It ..... :. ......, ...... -. ,,..,.,...., :.,,,,, ,,.., I....,. _ • .... .: "*Jr•.. . , ,: : :.: it,.__._,..! , ......"-':l • ," . •'-.- . % .'.. . ", : "';:-. -• _. ..'.",..., •..../: • • "_ "4r "4_ _," 'Jib"%4 %M-A-4 _ _,' g_'g'_ g'_ is" _ _" _" _i" _]" _s" _ _" _0" _]' _" _" _ Fig. 3b. Same at Brisbane. Fig. 45. Same at Brisbane. i llawson * Mawson 5 • 4 ._ t : ". • ,.' : .': • I • ._.. _.'......'.. •" / ; -'.:. -. .. -_ i • • -_, i -I _e_s -To _s Fig. 3c. Same at l_wson. Fig. 4c. Same at Mawson. 262 1980010188-273 lglRor) lr) 1FIFI_97,d operating environment. As with the median values, the three principal components_ diurnal, seasonal and solar cycle, are readily observed in mass plots of all hourly values recorded. Two years of data, 1970 and 1976, for three stations, Vanimo, Brisbane and Hobart are displayed in Figures 6 and 7. It is evident that Vanimo is strikingly different from Brisbane and Hobart, which are representative of all the Australian mid-_atitude stations. While the various components are contained in these plots, they are rather more clear, as before, when the diurnal component is removed (Figures 8 and 9). The day-to-day changes in foF2 are superimposed on a longer period seasonal variation which decreases in amplitude with the change in solar activity from 1970 to 1976. The seasonal component is clearest for 00LT Brisbane and Hobart. Night-time behaviour (Figure 8) at Vanimo is clea_'ly quite different from the mid-latitude stations, showing considerable variability. None of the stations shows very clear seasonal effects for the 12LT data. The day-to-day component is more evident after the predicted monthly medians are removed. A mass plot for all twenty-four hou" z for 1970 and 1976 is given in Figures i0 and ii. For Hobart and Brisbane the time series in Figures i0 and ii show considerable structure much of which can be associated with ionospheric storms. Vanimo, as with the median predictions, shows the greatest spread in deviations by a factor of two. Figures 12 and 13 show, for the mid-latitude stations, that the 12LT spread is greater than for 00LT while the opposite is true of Vanimo. While there is a diurnal effect, there is no obvious tendency for the deviations from predicted median behaviour to be dependent on the solar cycle level. Figures 14 and 15 give seasonal plots for 00LT and 12LT for the seven years of data available. Here some structure is apparent, Vanimo showing •slightly less spread in summer and Hobart, 12LT, showing increased spread in the equinoxes. Brisbane also shows a marginal summer to equinox increase in spread. Thus there is a diurnal deviation from the monthly median predictions, a_seasonal effect is also present, but is less evident, and there is no obvious evidence for a solar cycle dependency. These residual deviations indicate the general variability in foF2 that most prediction systems must accept as the 'noise level'. The 'noise' limits can be indicated by statistical limits_ normally the upper and lower deciles are predicted. Figures 16 and 17 shows plots of the predicted upper and lower deciles and all data that falls outside these decile limits for 00LT and 12LT. For the mid-latitude stations some of the data falling below the lower limit can be attributed to ionospheric storms. However, for Vanimo over 30% of the data appears both above the upper decile and below the lower decile. This is especially true of the OOLT data. If the deciles were completely reliable, 80% of the observed data would lie between them and for Vanimo this is not the case, suggesting the deciles are not wide enough. For Brisbane, on the other hand, almost 90% of the OOLT and 12LT data lies between the decile limits suggesting that they are too wide. Evidently, while reasonably adequate for mid-latitudes, the decile limits could be made more reliable. 4. REDUCING THE 'NOISE LEVEL' To improve the prediction of foF2 further, two basic methods can be _. 264 | 1980010188-275 IO ,, Hobar _: " Itobar t m_ m] II lit II I - • • .I 4i_ s " . _. l* : . -. • ;, ".. , ... , ; ?- ,: Time During 1976 _'_"'_'_'_'_-_ Fig. 7c. Same at Hobart Fig. 8c. Same at Hobart. 265 1980010188-276 I,_LI I Ill * .'. " " "" 14 ,I,_rl_ .__ { ' - .. . . 4 " '-_._:.._..'.':_' _.:-. ." • . .. " • .':..... .,, ;_:,;_..9.[,J[_'_ :&_ ,zo "...". :. • .:, : IS " _ ... _ "#" ," . , . .j} . • - . "" • "'1.." .- "., _ -'J'; , . ; ,. _- .. ., Z¢."= , .,. _.._.,_:.:_';&: "q__: • .4%" • r ;." • • ,. _,.,v-_. "_'_'__'_Z'_" -' " _' * _oR_5._d_._ • '" _" '_ • " "_.__';_" .1- " " t ''-' "_ :'""_ : ": ::_'_ _.a_ ,;_;_ 4 " ' ": " _" .. '. _. '_;'. ;'"_1, . l' : .,1 " _" f - t".. , • _ :-LL-1 : ii -_ :0. .......... Time During 1970 Fig. 9a. Noon .hourlyfoF2 values Fig. 10a. Observed hourly foF2 at Vanimo. values minus the predicted monthly foF2 medians. I_ _ Brisbane ,.I [ II . • .: . . : . .: ,o • " ;'_ " ':: . : _,""_i_.,,;_l_ii. ._:d_,i,_,,...:.,n,_itilii_ -_ d _.• . ..._ • . • -:. .-- _ ' , _ .,"._',-;[ _| :" i''' '[ _%L. . . , Time During 1970 Fig. 9b. Same at Brisbane. Fig. lOb. Same at Brisbane. obartHobartII • "'* .._. I_" ;. -, ' ' ' • , ...., """ '"'_'"" "" " --l ._'"_'.':_!"",,!":I Time During 1970 Fig. 9c. Same at Hobart. Fig. 10c. Same at llobart. 266 1980010188-277 P o • ," _l • ' . • o_ • ,, o , _o _.I. ", e'* ° • , __,j_ • ,.>,._-._=._ ,_, ._.._' ' "_'- ' "';' ' "' ..... " ' _. v").:.'.,_.,_, _., ._;_._._,-._ " :" +'c.'_ : _...- .. Tlme During 1976 Fig. lla. The same as Fig. i0, Fig. 12a. Difference between the except for 1976. observed and predicted foF2 values for midnight at Vanimo. i Brisbane Brisbane ":I : • " '... ".:... ".":. "_-.....'. 4." :.., -_ I} _'• It :t -,o ........ ""_ """ ;'i "'''_-'"_"i................." ........................... Tlm_ During 1976 Flg. lib, same for _risbane. Fig. 12b. Same for Brisbane. :, HiM _ ' ....... ., .., , _;.____ olll | • • " _ | a Time During 1976 Fig. llc. Same for Hobart. Fig. 12c. Same for Hobart. 267 1980010188-278 _t t t _ t t t t t 4Ptttvtt t t t t ttt_flttl_..._..;;, ; ...... ;.._..;_; ......-. ...... ... ...... ... • V+ni.+ " ! . . Wni._I III _ IP • . . " , • - .; : ..-...;;._......... •4 . ." .''" • '..,+_|." _, . +','_ _" ". • ..... . .P ˆ 4p., .,, . .. ., .: .-. : ,. -I " .,1 • . ,. '_. .0¢..., _ "" ' " -I ,... , o . . ,0, , . ". _,.i. ._. _. *..,, 41, -_lql _II- "............... '" _'_ ............................. • +-,- - _ - )lllq11+l ....... PI " fl " PI " Pdll " PII " IPll I • • il _I • P I • II II " I, Fig. 13a. Difference between the observed and predicted foF2 FiB. 14a. Season plot of the data values for noon at Vanimo. in Fisure 12 for m/dnisht at Vanimo. .., ',_'J ff._._ .. "" _.__' "_,V.. ., -, ., .- . ---,_--- _,- ...... ._. • :. ,-_+_ _ • -. I...... "_+.... . ,,+_,+,*-.-,_,-,+:; __-W_..T-...++.+,,,:, _ ,_..+.,+, :,+_++..+, .,.._. +. 1 I • • 4 I li • "O "l " IO " Ii " Ii Fig. 13b. Same at Brisbane. Fig. 14b. Same at B_'isbane. _Ir +" +-+................................. '...........'l ........+....... I _ - ' --+-.-----.--..---,r+-+, • + r " - I__LI " LL,_ 1 .... Hobart . . l]obart • 4 - t _ " - "+,,;.'N..N+_.._:.2....+...J;..'--. '.. , ++...-..._...+,.... ;:,..... ,. +.,..._._ -, .-_.._,....-+ ._ --_+'+._.._ +. '_"..-_..-i_,,-_.".++-+,__.t_P'_.:_ ,.;_,..-."_;. ._'-i_"++"_ -, t:.'w_r..,PPA_..,.:, ,,:_._u..,,,.--,, __-. +_..-_ _;'_ •# ._. ," i_ . • .+] _" _' ._, "IFI'.-_. " 1,- -." " • "" +. . .." ..',' • , • . • ;'_ '..._'w- ,x,: o,o ... +....,., , +; . + , ::l,,+- :-- . , ,o , • ,. , _o + +" I-+I t • " _ " ,i""" ,_ --: .;._+---:--_----: _ -_o _ " • ' + ' " " • "_+_ , .....• • .e ,. ii Fig. 13c. Same at Hobart. FiG. 14c. Same at llobarC. :.',8 1980010188-279 C;.'( • . . vant._ . - .,,_ ,. , -_ _m :" ".',." _ . .'." . ".' • • • • _ • ...,_:. .. _ -_ ;'-.",'. • .... ,.. • '!_l.'... _.," l.., .... _X. .' " •4 • . , . dP ".' '•' : '.'''" I _'-_ _-J_ r_-_'__._" "."_;¢""°-"'-_'-" • _ . ,_,i, .... ,,'..:_"_..'.'__.'_$. • • " " '_' • . ._ __-_._ ..,. • _..,,.:_;_,.,_ ,..,,.,.: ., _,_._,:,.,_,,::," '._._.._ • _---------_----_ -, _,,:,.-.:.-.;_....:.-..,....._,--,.,,,:,...,. . ... • *, t,,,,_'-._",'r,_. "'" {',;." ..-':, !9_,,.am.,ir_.- "" - ";'-"" "_ ". .." • .... • "*I."'"':."'" '..':,.v. . , _...,....... ..• ..! , -_ -. .'..'"' :I FiB. 15a.Seasonalplo_ of _he Fill.16a.P1o_o£ foF2predlc_1onerrors ou_slde _he upper and 1o_er data in Fig. 12. deciles for _dnil;hc ac Pani_o. • .... • . . ._:. .... .;_,. ., ... ".. ..• :=...,....-.:..,,,...II i Fill. 15b. Same for Brisbane. FSg. 16b. Sa_e for Br_.sbane. • Rt'_Pllt._!t,t _,,I ..,, , _u, T • ORYGINAL PA(;I.._IS POOR--_ - ' • I_'_._.-.,_r,_.:_..,.,,,._• . o : ......... Fig. 15c. Sa_e for llobart. Fig. 16c. Same for Hobart. 269 198f1C11_1_R_9_n • :'[-_- H a • I .qM||I i Time During 1970 Fig. 17c. Same for llobart. Fig. 18c. Same for Hobart. 270 "19800"10'188-28'1 _ applied. I I. New informationmay be added to identify periods when observedbehaviour will deviate from predicted behaviour. Two solar parameters will be considered here: the 10cm flux, indicative of solar electromagnetic radiation incident on the ionosphere and the K index, indicative of the I effects of solar particle fluxes observed at the earth's surface. 2. As mentioned earlier, the time series in Figures i0 and ii show i onsiderable structure. Knowing past structure, it should be possible, to a _: certain extent, to anticipate future structure. F_therm_re, the Brisbane i and Hobart time series are sufficiently similar to make projections from onelocation to another reasonable. The second method therefore consists of _ projecting pa_t data to a fUture time, or current data to a remote location in an attempt to reduce the 'noise level' in the predictionz. _ 4.1 Introducing new iuformation L It is well known that large depressions of foF2 may occur as the high , and mid latitude ionospheresrespond to changes in the solar wind arriving at the earth. These changes are paramaterized here by the K index for Toolangi. Figure 18 shows deviations recorded when the Toolangi K index was less than 3 during 1970. These can be compared with Figure I0 where all the available data was used. In general, the overall noise is reduced slightly with many extreme values being removed. This is more effective for the mid-latitude stations tha_ Vanimo. The K index effects are more obvious in Figure 19 where only data for K_3 &re plotted. For Brisbane and Hobart the main effects of ionospheric storms, causing depressed foF2 values, are very clear. For Vanimo the K index ! information is rather less useful as an indicator, the range of deviations in foF2 being both positive and negative. Thus being able to anticipate the K index in advance may enable the large : excursions of foF2 from predicted values to be anticipated. (This is a convenient method of warning communicators of the likelihood of an iono- spheric storm and the extent to which it will affect communications circuits.) Variations in intensity of the ionizing radiation from the sun may also cause variations in foF2 (McNamara,1976). This hypothesis is hard to present in the diagramatic fashion used in this paper as no completely reliable method exists to convert a solar flux parameter (the Ottawa lOcm flux, in this case) into a daily ionospheric index that is compatible with the IPS prediction method. Figure 20a shows a plot of daily changes from the monthly mean 10cm flux value (vertical axis) compared with changes in the 12LT foF2 values from the predicted median values for ]970 for midday. While there is a correlation of 0.3 for this data, it is not significant. A similar comparison for 1976 data (Figure 20b) which gave a correlation of 0.02 supports this hypothesis. Thus, while long term changes in lOcm flux can be related to changes in foF2 medians, there is little apparent evidence for day-to-day variations being highly correlated with the changes in foF2. This does not preclude any effect, though it suggests that 10cm flux departures would need to be quite large before associated foF2 departures would be observed. Such large departures are responsible for the presence of any correlation in the 1970 data. It can be concluded that while some additional information may be 271 0 _ 1980010188-282 • + m • (a) 1970 • m ". " i " ii "'" ""''• . " . • . ' 80 ' .. . :j..': .... • ;. ._ ; o " • -_ _'.. • ",¢,,r._ " " " ". i, _ 80 , . .... - • .... _._ . ,b_¢'-.:_ !,:_: _ ,. .... : .... . : . -I " "" • *:" i: ",;, "._-:..r " " ..":._' ;'.V;.;': . , ".' ,. _ ."" ._' ,_ , i',_ ""'c : . ;, . : . -am "'. 4: .," ":'..;'..• - " . t ... i= " _" " .... " :'" ;";';'" ,_ • _1| * i * ,i • ' "__st -Z ! O O • .... Time During 1970 Fig. 19a. Like Fig. 10 except foF2 Deviations, MHz. data, recorded when the Toolang: K '.ndex was < 3, were excluded. Brisbane HI (b) 1976 '+'I"- ; •. :-. i_. _ {:i ._.:,,:.. :.,;;!-:..._,.'.". "...... ", ' "_ I' : :_".*:1 ,;,_:, !; :.! -_: q D _ - . .;t.,.. ., -, _ .'_ ;. I; " ;";"'_ |; ,1_ .'_ : _] : .._" . _ : • % . . "GO --_l "e -o | -in • .... * • _lP_ j -_' O le ! ,, Time During 1970 foF2 Deviations, MHz Fig. 19b. Same for Brisbane. Fig. 20. Plots of changes in the daily Ottawa 10 cm solar flux with radio flux units) against devia- tions in foF2 from the predicted median behavior (horizontal axis, . range +-3 [H{z) for (a) 1970 and • i .. • ". i :_ (b) 1976 -i': r,:'.. ,o;[. +_, ."'" '" ,,,'..'-,, .+:,r _ •,, ....+ _ 't "4 i ;.',+ i "• ''" ''' I: ,nil Time During 1970 Fib. 19c. Same for Hobart. 272 t "19800"10"188-283 obtained using the K index and 10cm flux observations, in general they will not greatly improve the overall noise level in daily predictions. 4.2 Projecting Data forward in Time and Space The second approach utilises previously recorded data either at the locations being studied or at some other remote location. The average condition of the ionosphere at a particular time (and/or place) will vary in the known predictable median fashion. This effect can be removed and predictions made using the deviations from the predicted monthly median foF2, rather than the actual recorded foF2 values. It is not reasonable to assume that the cause of a deviation from predicted median behaviour at one location (in time and space) will have exactly the same effect at some other location. For instance, the observed ionospheric response to travelling disturbances depends on the electron density gradient which changes during the day and from place to place. This evidently results in travelling disturbances having less effect on foF2 at night compared with daytime. While some scaling factor should be introduced to compensate for this, it is not currently possible to set sensible limits i on the proposed effect. In the present work, the ionospheric response is assumed constant. The simplest case for projection forward in time is considered. The recorded deviation from median behaviour as observed at some particular time is assumed the same as that which will be observed at some future time. Two lead times are illustrated; a one hour lead in Figure 21 and a 24 hour lead in Figure 22. In both Figures, all the data for 1970 is displayed. Figures 23 and 24 give the same data, only now it is folded into 24 hours so that the diurnal variation in spread is more apparent. It could be argued that projecting percentage departures from median behaviour would be more satisfactory, but as there was no clear diurnal variation in the size of the departures this was not considered necessary. While a lead of 1 hour (Figures 21 and 23) reduces the sprea( in values compared with those in Figure 10, its effects are possibly not as L "eat as might have been expected. In fact, for Vanimo the spread is still greater than experienced for the mid-latitude stations when deviations from the predicted medians are considered. The results for a lead of 24 hours (Figures 22 and 24) are rather disappointing, the spread now being only slightly less than for the monthly predictions. This is partly due to a failure to deal adequately with ionospheric storm data, the time series in Figure 22 being quite spikey (for Brisbane and Hobart) because of this. The effect shows up more clearly in Figure 24 as a halo of extreme deviations for daytime hours superimposed on a relatively structureless background. For Vanimo the day-to-day spread in observed differences is quite large, giving a further indication of the difficulty in predicting the behaviour of foF2 at this location. It also helps explain some of the difficulty experienced in predicting the monthly medians for Vanimo (see Figures 4 and 5). Different lead times (3, 6 and 12 hours) are illustrated for only one station, Brisbane, in Figure 25. It is aoparent that for lead times greater than one hour there is only a slight advantage over the median prediction, the effects of ionospheric storms becoming more evident, while the background z73 i 1980010188-284 19800101RA-gR.r--, r"t. i ::_i_[ " :- _:i_E.ll • " . • " ,, "* " _ •• ....,:::";:-- '' • -_-,_;"_._.: i_._:-._._._.. !: ._...::...:.. _."_ -.---"':._-.,:':7•..::.....•' _-_;i,..,,:_....-.__ .,,.!!,vy_, -II .... . i,lqum_ • • • • . _ L -I0 . 8 . 12 16 20 24 _ 8 12 16 20 _4 ?Ig. 23a. Diurnal plot of data Fig. 24a. Diurnal plot of data _, in Fig. 21. in Fig. 22. _:., _ Brisbane ._ : : ..:.. .& •. .... ". • , , • - . •.... _ .- ......... "L, ....... .." ........., °_" .''# ,. I • • _,.t, : o..,,!Fs- o" 'h_.._."'" ;" '_ 8 12 16 20 24 4 8 12 16 20 Pig. 23b. Same for Brisbane. Fig. 24b. Same for Brisbane. 4 • , | , -_ , 4 _ 12 16 20 2_ _ 8 12 1"6 20 2_ Fig. 23c. Same for Hobart. Fig. 24c. Same for Hobart. : 275 1980010188-286 • I .... I • • • • °• • . • .. U Q ,_e '" _" "" e:_.",o" .% • 4-... •..._:;- _| tlours _ u tl ......... I 2 4 6 8 10 ][2 lk. 16 18 20 22 2k. 2 U 6 8 10 ],2 tU |6 18 20 22 24 Fig. 26b. Comparing Brisbane and Fig. 25a. Projections with three Norfolk Island data with the hours lead. same local time. Ill Ilta _'___ ,,Jk _._". ... _...._._ .'_¢ • _ • • = _,.*'.i_-. I ml . • . . 4 4 Hours 4 --II " , , , . ,. 2 4 6 8 I0 12 14 ].6 18 20 22 24 Fig. 25b. Projections with six Fig. 26c. Projecting Norfold Island hours lead. to Brisbane with a i hour lead. • \ \ 2 b, 6 8 10 12 l_ 16 4\ 2 4 6 8 10 ].2 14 16 ],8 20 22 2b, Fig. 26d. The same as FIR. 26c Fig, 26a. Projecting Brisbane data wlth a 2 hour leadto Norfolk Island with an hour lead. 276 1980010188-287 spread in deviations also increases. Using this simple projection model it is evident that while previous data will eliminate gross errors in the predicted index of solar activity, (errors that are not discussed in this paper), it is not until the lead time is reduced to less than 3 hours that this technique can be considered effective. For such short lead times the effects of ionospheric storms and, ' to some extent, travelling disturbances will be allowed for. However, as ! many travelling disturbances have periods of the order of one hour or less, probably insufficient information will be available to greatly improve on this even if more careful analysis techniques are used. It could be more necessary to obtain data sampled more frequently than once per hour before _ significant improvements on this result are obtained. Most prediction systems requiring improved local predictions also require these predictions to be extended to more remote regions. As with time projections, the simplest approach is to project deviations from the predicted median behaviour at the known location to the required region. This _ will be adequate for improving on the predicted solar index and allowing for ionospheric storms. The latter will be most effective for regions with the same geomagnetic latitude (McNamara, 1977). As an example of this, the deviations recorded at Norfolk Island are projected to Brisbane. For behaviour that is constant in local time, the deviations at Norfolk Island should lead Brisbane by just over one hour. (Because Norfolk Island data is recorded on 180°F_ the lead is two hours for the data used here. ) For greater lead times both time and space projections are mixed together. Figure 26 gives a series of different lead times for predicting Brisbane using Norfolk Island and also Norfolk Island using Brisbane. The best results are recorded for data at Norfolk which leads Brisbane data by I hour in local time, as anticipated. The spread in deviations is equivalent to a two hour lead time projection for Brisbane data, which might be expected because the projection technique will not allow for travelling disturbances. Thus, using simple projection methods the 'noise level' in the predictions is reduced, though no substantial improvements are obtained. 5. CONCLUSION In general, for mid-latitudes, monthly predictions of median foF2 values can be accomplished quite successfully for medium and low solar activity. The spread in recorded hourly values is adequately indicated by the predicted deciles and this spread is not greatly reduced by utilizing past data. Evidently, to significantly reduce the 'noise level', rather more skilful techniques must be applied. One example of skill is the prediction of ionospheric storms, which if successful in both timing and amplitude, will reduce the large depressions observed in foF2. The low latitude (Vanimo) data was particularly variable and the spread was greater for all categories discussed. 277 1980010188-288 6. THE DIAGRAMS In addition to the time series plots, for all comparisons a histogram with the same resolution as the vertical axis is plotted in the top right hand corner. It is normalised by the total number of entries thus giving a probability of occurrence of deviations between two scale divisions. The amount of data that was not included because only a descriptive replacement letter was scaled is shown at the side of the histogram. The complete histograms m&y be obta/ned from the author if required. Table I gives some information on the stations used in obtaining the diagrams in this paper. TABLE I ION0_C STATIONS DISPLAYED IN DIAGRAMS STATI ON GEOGRAPHI C GEOMAGNETI C GEOMAGNETI C COORDINATES COORDINATES SPLIT Vanimo 2.7S 141.3E 12.6S 211.1E 0.4 Brisbane 27.5S 152.9E 35.4S 228.6E 0.6 Norfolk Is. 29S 168E 34.8S 243.2E 0.5 Hobart 42.9S 147.2E 51.7S 224.3E 0.7 M_wson 67.6S 62.9E 73.1S I02.9E 0.8 Acknowledgements: I wish to thank Mr. D. Carrington, University of New South Wales, for his assistance in copying the data used in this paper from magnetic tape to RK05 disk for use on the IPS PDPII/10 computer. REFERENCES McNanara, L. F. (1976): The Correlation of individual values of foF2 and M(B000)F2 with the solar 10.7cm flux under magnetically quiet conditions. lonospheric Prediction Service Series R 30. McNamara, L. F. (1977): Ionospheric storm effects on maximum usable frequencies on circuits around Australia. Ionospheric Prediction Service Series R35. Sargent, H. H. (1978): A Prediction for the next solar cycle. IEEE Vehicular Technology Conference, 278 1980010188-289 ,% 17 •| 8,0 -18486A WEEXLY IONOSPH_II0 INDEX J. F. Turner and P. J. Wilkinson Ionospheric Prediction Service iv Australian Department of Science and the _vironment il Darlinghurst, 2010, N.S.W., Australia i'_, ABSTRACTI: For some years the Ionospheric Prediction Service has used a :_- weekly T index as a monitor of the global ionosphere and an aid for predicting the monthly T index. Both the monthly and weekly T indices are discussed and some possibilities for improving the reliability of the weekly T index are considered including extend- ing its applicability to daily variations of the global ionosphere. u I• INTRODUCTION _ For the purposes of H.P. prediction the Australian Ionospheric Prediction Service (IPS) uses a monthly ionospheric index produced from the average of the 24 hourly median foP2 values for selected ionospheric stations and the : Zurich twelve-month smoothed sunspot numbers. This index, the T index, is _ the basis of the IPS prediction maps. For prediction purposes, there is usually too @@eat a lag between the calculation of the monthly index and the period for which the next prediction is required. To bridge this gap a weekly index was produced and has been found very useful over the past 10 years. This weekly index is based on the IPS prediction maps and hence on the monthly T index. As groundwork for discussion, the prediction maps and the monthly index are reviewed briefly before the weekly T index is discussed. More details are available in Turner, 1968. 2. THE MONTHLY T INDEX Using an ionospheric parameter (the average of the 24 hourly median foF2 values for a month) and a solar parameter (the 12-month smoothed Zurich Sun- spot number), an ionospheric index is derive_ by an iterative process. At each station, given in TABLE 1, a first order relation is found. For the ith station it is of the form, 279 i 1980010188-290 s(t) = miFi(t) i where S(t) is the solar parameter, R12 Yi(t) the ionospheric parameter, all the average of the 24 hourly median foP2 values for a particular mon'h mSoi regression coefficients. These relatlons are used to derive a hemisphere T index by finding the station T index for each data point, tJi = miFi + ci where tJi is the station index for the jth iteration. The hemisphere index, TJH, is then found by averaging the appropriate station T indices, N I TJH = _ Y. tJi.i For the first iteration the solar parameter used is RI , but on success- ive iterations this is replaced by the new solar parameter,2the hemisphere T index, TJH. Iterations are continued until a stable TH is obtained. Finally, the global T index is obtained by averaging the northern and southern hemisphere T indices. This is now the monthly T index. Usually three to five iterations are required. Allowance is made for high levels of solar activity, where the first order relation is less reliable, by weighting against this data in the first two iterations. The averaging in the final step minimises the seasonal dependencies in the index resulting from differences between the two hemispheres and from any imbalance in the number of stations selected from each hemisphere. The process used is similar to that used for calculating IF2 (Minnie and Bazzard, 1960) with three basic additions. The ionospheric and solar para- meters are chosen to reduce the scatter, both being less variable than their IF2 equivalents. Rather than a single cycle, iterations are continued until the new solar activity parameter remains constant. Finally, whereas a global index is calculated in one step for IF2, differences between the two hemis- pheres _e minimised in calculating the T index by averaging indices cal- culate_ for each hemisphere. '_e index derived by the above method is ca,led the monthly T index and is uEed in constructing the IPS prediction maps. 280 1980010188-291 TABLE I. STATIONS USED FOR THE MONTHLY T INDEX Station Geographic Coordinates I Ad__k. N51.9 W176.6 2 Fairbanks (College) N64.9 W147.8 3 Frelburg N48.0 E?.8 4 rl.hari(t,m,ui) N20.8 W156.5 ' 5 Moscow N55 E37 6 0klnawa N26.3 E127.8 7 Tokyo N35.6 E139.7 8 Wallops Is./ Washington N37 W75 9 Bri,bane $27 El53 I0 Canberra $35 E149 11 Capetown S34.2 E18.3 ! 12 Christchurch S43 E172 i 13 Falkland Is (Pt Stanley) $51.7 W57.7 ;_i 14 Hobart $43 E147 : 15 Johannesburg $26.2 E28.0 16 Mundaring/Watheroo S32 El16t The process used is similar to that used for calculating IF2 (Minnis and Bazzard, 1960) with three basic additions. The ionospheric and solar para- meters are chosen to reduce the scatter, both being less variable than their IF2 equivalents. Rather than a single cycle, iterations are continued until the new solar activity parameter remains constant. Finally, whereas a global index is calculated in one step for IF2, differences between the two hemis- pheres are minimised in calculating the T index by averaging indices cal- culated for each hemisphere. The index derived by the above method is called the monthly T index and is used in constructing the IPS prediction maps. For all available ionosonde station data, the first order least-squares- fit line is found between the monthly T index and the median F region critical frequency for each hour of the day and month of the year. Using the fitted line two values are recorded, the median foY2 value for T=O and T=IO0. (These points are roughly equivalent to sunspot numbers 0 and 100.) These values are then used to construct the IPS F region prediction maps - there being 576 xn all (24 hours x 12 months x 2 levels of solar activity) which, together with maps of M(3000)F2, are used to make monthly HF predictions for F region modes. 281 1980010188-292 _. THEWE_LY IND_ of the main reasons for developing the T index and the IF2 index was data recorded in different solar activity cycles to be combined. While monthly T index is found valuable for resolving such problems, allowing to be made of almost all of the recorded ionospheric data, it is not satisfactoryfor prediction purposes. The time lag resulting from the 12 month mean Zurich sunspot number together with delays in obtaining data result in unrealistic lead times for predicting the monthly T To reduce this lead time an index was developed based on recorded values at a number of stations. data consist of the foF2 values for the four hours (0, 6, 12 and 18 UT) ionosonde stations (TABLE 2.). This data is made available through the data exchange system usually within a few days of being recorded. TABLE 2. STATIONS USED FOR THE WE_Y T INDEX. Station GeoEraphic Coordinates I Ouagadougou N12 W02 2 Slough N51 W01 3 Lannion N48 W03 4 Dakar NI5 W17 5 Wallops Is N37 W75 6 Boulder N40 W105 7 Poitiers N46 E00 8 Lindau N51 EIO 9 Kiruna N68 E20 10 Djibouti N12 E43 11 Moscow N55 E37 12 Alma Ata N43 E77 13 Irkutsk N52 E104 14 KokubouJi (Tokyo) N36 El39 15 Ra_atonga S21 W159 16 Mawson S68 E63 17 Mundaring S32 E116 18 Vanimo SO3 E141 19 Townsville 819 E147 282 1980010188-293 @ TABLE2. continued. i Station Geographic Coordinates ill i i i 20 Brisbane $27 E153 r 21 Canberra $35 E149 22 Hobart S43 E147 j 23 Norfolk Is 829 E168 24 Christchurch _ S43 E172 25 Campbell Is S52 E169 J _ 26 Scott Base S77 E166 ! Using this data base and the maps described in the p_vious section a daily ionospheric index can b. derived. For each station mo-thly predictions for T=O and T=IO0 are obtained. These are used to produce two time series for which the fourier components are found. The fouxier component_ of the time series are then used as a convenient interpolation for finding the T--O : and T=-100 predictions for any day of the year for each of the UT times of interest. Using these fourier components the recorde _ data is then converted into a T index by interpolation between the T--O and T=IO0 levels of activity, assuming a first order relationshin. Figure I gives a plot of T in_'ces derived in this way for 1976. As can be seen values range from -100 to +180, the range being marginally less in the equinoxes. Figure 2 gives an indication of the relationship between magneti: con- ditions, represented here by the average of the eight daily K indices for Toolangi, solar electromagnetic radiation, represented by the 10.7 cm flux level observed at Ottawa and the daily T index. The T index shown is the average of the station T indices for a day with the length of the bar being twice the standard deviation of the mean for the sample. During 1976 the geomagnetic disturbances were due to recurrez_t storms and the solar electromagnetic radiation output was reasonably constant, the daily 10.7 om flux varying between 60 and 70 flux units during the year. Gross features in these two parameters can be related to the daily T index, increases in the 10.7 om flux causing increases in the daily T index and increases in the K index occurring at the same time as depressions in the daily T index. For operational use the daily index is averaged over seven days to pro- duce a weekly index. Experience has shown that the weekly T index, as des- cribed, is a useful measure of solar effects on the ionosphere, variations in the index being fairly easily related to Known solar variations, as shown by figure2. 283 1980010188-294 Figure I. A/I hourly T indices recorded for all Stations during 1976. T l"ol:'...! "" "-:"""'"" ": "'" " "";:4 ";.T'v. "" " *"" • . ". " :'- "'" jz=oi',-_'.,.... ..., .... .: . ---.,._,/ oo"'_=.:':_"-.;,"f '":--'""-."": : "'_..:_:.+.".'..';-'_ .''"- ". " L': _ ,'.._;!.'r;--:,:".:"_:._ --'.., gO 20 0 -2 0 • . • t _,. '_ -_o__ ,_i_.. "B0 ..:'::" _": .'.._- .* ., ..', ";,-.r.'. • ,_;_;"; ;"-. MONTH..:':'..',._..,;-a>:-..;_'[ -10v- I = = 4 _ $ _' "--" O'-''_ tO II IZ Fisure 2. Dally T index, 10 cm flux and Toolar_i K for 1976. .11 8 lrlOOLRNG | ok , 0 , ' ' 'o |NDE >', , , , ,_-"..Jk _", _igl_,,,k_,a,,_ a ,,_'_ | iK.FLux ........................... L_/ . : _.,-_:.-_ : \ .:_',:,,.,,',-_:- _._ _sO _ • _ • : , oo,,1-_--,, ............. 50 . . . , , 25 L_" . , ._'_ :);-..0"-.." ". • . ..''- "-'. [-.':1 ^['. ,;'::_..."_;/.._.".'".- "..""::.._,.,"."":.","<"::,;,:.:.-..,.<:.:::'.J | . I Figure 2(_). Weekly T index for same period. 75l'r '-x l ot. f ! " J%1%,,_ e " ar ,w , • : _'_,,... : • / ".F" -..,,", . ..: : I I I t _ 3 4. _ _' z B 9 1¢1. AL *_' 284 1980010188-295 4. IMFROVEMENTS TO THE WEI_(LY INDEX The weekly index, as described in this paper, was initially conceived to improve predictions of the monthly index and also to meet the needs of some of IPS customers who wished to receive prediction updates with as short a lead time as was reasonable. For these purposes the index as described has proven valuable. However, in the course of preparing this paper it has become apparent that various improvements to the index may make it even more useful. Experimentation has shown that for the 1976 data, provided about half the station data required is available, the daily index is as reliable as the seven day average, with the added advantage of potentially greater resolution. For this reason, all the diagrams presented in the paper have displayed the daily index rather than the weekly index. The results using the weekly index are the same with only s small decrease in the scatter of data points. However, it has not yet been possible to fully investigate the relationship between the two indices (daily and weekly) to decide wbgther the apparent improved resol- ution in the daily index is real or related to statist_oal fluctuations. There is some evidence to suggest that for higher levels of solar activity and with the addition of flare induced storms, as being experienced by the end of 1978, the daily index becomes very variable. The data ba_e used consists of four data points per day per station, the data being recorded at the same universal time for all stations. Provided the distribution of stations is adequate, this should produce a global average of the ionospheric effects for a particu" _r change in solar output. While the evidence in Figure 2 m_gests this is accomplished, it will be necessary to establish that the final result is representative of a global condition. Using data derived from night-time foF2 values is also questionable as the variation due to solar activity is much less then, giving rise to greater errors. Figure I is repeated in figure 5c along with the same plot for only night-time data (figure 3a) and daytime data (figure 5b). The daytime plot shows less scatter suggesting a better index may be obtained either by care- ful screening of the data averaged, or by using only daytime data. In calculating the monthly index allowance was made for differences between the two hemispheres. A similar approach is not adopted in calculating the weekly index. In figure 4a, a plot of the difference between the daily northern hemisphere T index and southern hemisphere index is given. To allow some estimate of the significance of this result, the difference between two daily indices constructed by dividing the available station data per day in half by selecting stations randomly is shown in figure 4b. No clear structure appears in figure 4b, though the scatter of points reduces near the equinoxes. In contrast, figure 4a shows a consistent difference between the two hemis- pheres which is least in the equinoxes. The source of this difference is clearly a seasonal difference between the two sets of data, but it is not clear why it still exists in the data as it should have been eliminated in • calculating the station T indices. Further investigation will be required to clear this point up. 285 "19800"10"188-296 Figure 3(a). Daytime hourly station Figure 3(b). Night-time hourly station T indices. T indices. t_ • o |,4B Nil' .: :. " ' ' " el _ :." . ; . .... ... .. i .did •., . ......_.. ".'...'.: ... . • • MONTH" :_, ...., MONTH '.,, c_..:.'-_'... I 2 3 k., 5 6 7 e 9 10 11 12 1 2 3 _ 5 6 7 8 9 10 11 12 Figure 4(a). Daily T index. Northern Hemisphere-Southern Hemisphere T. g •" : '"";:i; "-. . -;'.. • .-" ]i" -.-'".-" . "."-. 4" "- , Ill . .,, .._ .-. .. .. • D . o- .... . : - . .:.".."_-. -.: ...... .': • ..:,:-." :. - . . --II " J'" - _"" - . . ['IDNTM. , , , . . I --i 2 11 dl 5 t ;F ll 9 II I I I_ l?igure 4(b). Daily T index. Difference between two random samples. INDEX 70 50 30 . 10 • " '. " ' , , ,' ' . ',': .,,,' ':.." ..... ;.. "f. '..... ..'.. . ,, .,, .. ,,, '," • %- -10 ' : l -30 " " " " illN IM " J i J -50 I I 3 4 I • _, m • le iJ It. 286 t 1980010188-297 5. CONCLUSION The weekly T index used by IPS for some years now has proven ve_7 useful, operationally, as an indication of conditions in the F2 region of the ionos- phere. This information has enabled the lead time for prediction of the monthly T index to be substantially shortened and also enabled IPS to provide some customers with weekly predictions. Investigation of the weeklyT index using the data recorded in 1976has revealed that a dally index may be just as reliable, though there is a need for more careful analysis of the operation of the index before it can be considered completely reliable. 6. DIAGRAMS In all diagrams the horizontal axis is the time axis, the scale markers indicating 30 day intervals. Figure I. Plot of all recorded T indices derived from the hourly station data for 1976. Figure 2. Plot of the Toolangi K index the 10.7 cm flux and the daily T index. (a) The weekly T index. Figure 3. Plot of all recorded T indices as in figure I for i!I only night-time data only day-time data a repeat of figure I. Figure 4. (a) The difference between the northern and southern hemisphere daily T indices. (b) The difference between two daily T indices constructed by selecting _t_tions at random £rom the daily sample of available station data. 7• _CES Minnis, C.M. and Bazzard, G.H., (1960). A monthly ionospheric index of solar activity based on F2 layer ionization at eleven stations. Jnl. Atmos. and Terr. Phys., 4, 297. Turner, J.F., (1968). The Development of the ionospheric T index. Ionospheric Prediction Service Series R12. 287 1980010188-298 , N8'O -18 48,7 CORONAL HOLES INFERRED FROM THE "FLEURS" EAST-WEST SOLAR SCANS BRIAN G. FERGUSON Ionospheric Prediction Service Department of Science and the Environment P.O. Box 702 Darlinghurst NSW 2010, Australia The presence of coronal holes can be inferred from the "Fleurs" Fast- West solar scans produced daily on 692 and 1415 MHz. An analysis of these data for the period 1968 to 1973 confirms that coronal holes are stable features displaying significantly reduced electromagnetic emission which can persist for many solar rotations. The coronal hole observations are summarised in a series of tables to provide a consistent set of data for the period 1968 to 1973. i. INTRODUCTION Coronal holes are large-scale solar features which have much lower densities and temperatures than the typical background corona. Spectro- heliograms at X-ray (Timothy et al. 1975), XUV (Bohlin 1977), EUV (Munro and Withbroe 1972, Neupert and Pizzo 197h) and radio (Dulk and Sheridan 1974, F_rst and Hirth 1975) wavelengths reveal coronal holes as regions exhibiting a localised diminution in electromagnetic emission. Coronal holes appear to possess an open, diverging magnetic field structure and are bounded by apparently divergent coronal loop structures (Vaiana et al. 1973), Thus the characteristic appearance of a coronal hole is one of an open, dark feature with very low levels of emission at coronal wavelengths, Coronal holes have been found to be associated with recurrent, high-speed, solar-wind streams and with geomagnetic activity (Krieger et al. 1973, Neupert and Pizzo 197_, Sheeley and Harvey 1978). Thus the detection of coronal holes is important in Solar-Terrestrial Predictions. In addition to the two-dimensional spectroheliograms, coronal holes manifest themselves in one-dimensional East-West (E-W) solar scans at metric and centimetric wavelengths (Chiuderi Drago et al. 1977, Covington 1977). The purpose of this paper is to identify and classify coronal holes for the period 1968 to 1973 (inclusive) using the 692 MHz (43 cm) and 1415 MHz (21 cm) solar scans generated daily by the E-W arm of the "Fleurs" Synthesis Telescope. In Section 2 the production and presentation of the scan data are outlined, whilst in Section 3 the detection of coronal holes and the criterion for recognising them are discussed. A subjective classification is assigned to each hole depending on the degree to which the identification criterion is satisfied. 288 1980010188-299 i, In Section 4 scan data are presented and analysed for the period January to July, 1973 with the coronal holes being labelled when they appear on the scans. The holes inferred from the radio scans are then associated with the holes evident on the Fe XV (0S0-7) and soft X-ray (Skylab) spectro- heliograms. Three tables are presented which summarise the observations of the "radio" holes for 1973 with the central meridian passages (CMPs) Tor the corresponding "X-ray" holes being included for labelling and comparison purposes. The analysis is extended to include the period 1968 to 1972 with Section 5 being reserved for tables summarising the coronal hole observations during this time interval. 2. DATA The solar scan data are made possible by the "Fleurs" Radio Astronomy Station of the University of Sydney, Australia. The E-W arm of the "Fleurs" Synthesis Telescope consists of a row of thirty-two parabolic reflectors 5.8 metres in diameter with an _nterantenna spacing of 12.2 metres. The signals i from the antennas when combined at any moment produce a North-South (N-S) fan beam which is optically equivalent to a slit. As the sun drifts through the beam a one-dimensional brightness distribution is produced. The fan beam inherently integrates all the radio emissions in the N-S direction so all the latitudinal information is lost. Fan beams can be produced on two frequencies 692 and 1415 MHz with resolutions of 4 and 2 minutes of arc respectively. The sun is scanned daily at around the beginning of the UT day. The scans are traced and reduced for publication in "Solar Geophysical Data". A series of solar scans, obtained on 1415 MHz, is shown in Figure 1. The two short horizontal lines drawn crossing the central vertical line indicate the cold-sky level and the estimated quiet-sun level. The curves are not normalised to account for daily variations in equipment gain or for tilting of the sun's axis of rotation throughout the year. The scans reveal the presence of discrete emissive sources which give rise to the slowly-varying component (SVC) in solar radio flux measurements. These sources appear on the scans as peaks superimposed on a quiet-sun profile attributed to the undisturbed background-corona and to any conglomeration of broad, faint extensive sources (Covington 1977). The SVC encompasses changes over hours or days in flux measurements and is due to the development and decay of active solar regions. This gives a series of daily scans a dynamic appearance in that the amplitude of the peaks appear to wax and wane from day to day. Traditionally only the radio source information is of interest with the positions and amplitudes of the peaks being scaled daily. This information is coded and then transmitted to varicus Regional Warning Centres of the International Ursigram and World Days Service. This paper adopts the opposite approach in that it is confined to investigating the antithesis of the active regions associated with the peaks. 289 1980010188-300 est_T'o,oumxsun,'_. COLDSKY LI_[L \ _ Hay j_ \\ _ 28 I | 30 0012 UT OO03UT 0012 LIT O040 tIT O003 UT O003UT 012S UT :/kw O003 UT O008 UT O02? UT FIG. la lhl5 MHz E-W SCANS FOR THE PERIOD MAY 27 TO JUNE 06, 1973 Jun ./11| E W E W E W E W 0159 UT O001 UT O001 UT O0$3 UT 0112 UT 0007 UT 0103 UT 0012 u'r Jul. _ O_[ 01 2t UT W E W E W O023 UT 0109 UT FIG. lb lhl5 MHz E-W SCANS FOR THE PERIOD JUNE 23 TO JULY 03, 1973 290 1980010188-301 !Here attention is focused on the trough information to determine whether or not it indicates the presence of coronal holes. 3. DETECTION OF CORONAL HOLES Not every trough on a scan can be attributed to the presence of a coronal hole and it is necessary therefore to distinguish between quiet regions and coronal holes. The criterion which emerged and was used for inferring the presence of coronal holes is that they manifest themselves as pronounced minima on the scans which endure for days and generally recur on subsequent solar rotations. Other observations at X-ray (Timothy et al.1975) and EUV (Neupert and Pizzo 1974) wavelengths confirm that coronal holes have _ lifetimes which may span many solar rotations. _ The characteristics of a hole profile appear to be preserved over a sequence of daily scans with the trough shifting in position as the s,,_n rotates. Thus the apparent passage of a hole across the solar disc is marked by a deep and stable trough on the scans. On the other hand quiet regions _ not associated with holes have profiles in which the troughs are prone to _ change during a solar rotation and can disappear or even be replaced by a _ peak during a subsequent rotation. In summary coronal holes appear to be steady-state structures whose scan profiles have deep minima which are more or less invariant whereas quiet and active regions have dynamic profiles which can change from day to day. Consider the series of 1415 MHz scans for June 23 to July OB 197B which are shown in Figure lb. For June 2B a number of troughs appear on the scan. Consider the three troughs labelled "Tl", "T2", and "TB". By comparing the profiles on subsequent days it may be concluded that the elevated troughs TI and T2 are associated with quiet regions whilst the lowest trough T3 is associated with a coronal hole. Whereas T1 and T2 are both shallow, T3 is deep. The depth of TB is preserved on subsequent scans whereas those of T1 and T2 are variable. In fact T2 disappeared on June 27. In addition, the information obtained for the previous rotation May 27 to June 06 is shown in Figure la. Comparing the scan for any one day of one rotation with the scan for the corresponding day of the other rotation suggests the only stable feature common to both rotations was the coronal hole indicated by the small vertical arrow and labelled "TB" in Figures la and lb. The scans indicate the coronal hole under_ent CMP late on May B1 and on the subsequent rotation, late on June 27. The 692 _z scans for June 2B to July 0B are shown in Figure 2a. The pesks on these scans appear broader than the corresponding ones on 1415 MHz (Figure 2b) for the same period because the resolution is reduced by a factor of 2. A hole is initially detected on the June 27 692 MHz Scan and is conspicuous on June 28 and 29. CMP was mid June 28. In general the 692 MHz observation period of holes is limited to two or three days each side of CMP whilst at 1415 MHz this period is extended to 291 1980010188-302 t mT'MYD. O_LrrSUNLVL. I / COLDSKY LEVEL I i ,_n / / E- -,,-- -W E W g W E W 0159 UT 23 $8 UT 0018 UT 0009 UT E _W E_/_W E W E W O00g UT 0142 UT 0010 UT 0020 UT Jut 0048 UT O0 20UT 0152 UT FIG. 2a 692 HHz E-W SCANS FOR THE PERIOD JVNE 23 TO J_LY 03, 1973 Jun _/1R E W E W E W E W 01$9 UT O001 UT O0O1UT 00 $3 UT E W E W E W E W 0112 UT 0007 UT 0103 0012 UT E " -L _-W E W 0126 UT E Woo23 UT 01OgUT FIG. 2b lh15 MHz E-W SCANS FOR THE PERIOD JT;NE23 TO JULY 03, 1973 292 1980010188-303 four or five days each side of CMP. Also the CMP times for the same hole observed at different frequencies are not necessarily coindicent. For example at 692 MHz, CMP occurred mid June 28 but on 1415 MHz, it occurred late June 27. Once the presence of a hole is detected from a series of scans_ a subjective classification is assigned to the hole. The holes are divided into three groups (labelled "I", "2", "3") depending on the degree to which they satisfy the criterion of being low-emissive, stable features. The group "3" candidates are the most satisfactory in fulfilling the requirements for inferring the existence of a hole. The hole mentioned above was classified as a "3" on 1415 MHz on both rotations, a "i" on 692 MHz for the first rotation and a "3" for the second rotation. The classification number assigned herein to each coronal hole represents a subjective assessment of the contrast, that is, the depth of the trough in relation to the quiet-Sun background, and the degree to which the contrast is preserved over a series of scans. The class 3 holes have high contrast and the class 1 holes have low contrast. The classification number "2" is reserved for holes which have high contrast around the central meridian but low contrast near the solar limbs. The class 2 and class 3 boles are readily detected on the scan curves but the class 1 holes are more difficult to detect because the contrast is low. Thus with the present technique, difficulties arise in inferring the presence of a coronal hole when the contrast is low. One "low-contrast" situation arises when a radio source is located on the same longitude,but on a different latitude, as a hole because the N-S fan beam sums the contributions from both the source and the hole. Thus the depression associated with the hole is reduced and the contrast degraded. Low-contrast conditions are frequent at Solar Minimum because of a reduction in the number of weak, extensiv_ sources which serve to improve the contrast between the hole and the background reference level. 4. COMPARISON W_TH OTHER OBSERVATIONS Using the criterion mentioned in the previous section to infer the presence of a hole, it is interesting to compare the results with observations made at the sa_ time but at other wavelengths. For instance EUV spectroheliograms at 284_ (Fe XV line) have been published in "Solar Geophysical Data" using the information from the 0S0-7 satellite. The maps are presented as a series of contours with the number of counts nc - 5, 10, 20, 40, ........ marked on each isophote. The maps for May 31 and June 27 are shown in Figure 3. These dates correspond to the two CMP dates of the hole inferred from the radio scan data. The Eb'_ analogue of this hole is the area delineated by the isophote nc = i0. This area has been darkened in the diagram and closely resembles the shape of a coronal hole observed at soft X-ray wavelengths (Timothy et a1.1975). The hole is observed at CMP on both EUV maps which is consistent with the radio scan observations. O °Similarly observations of holes at X-ray wavelengths (2-32A and 44-54_ ) for the period May 28 to November 21, 1973 were made by the spectro- 1980010188-304 NASA GSFC Fe lrv 0S0-7 284A E W 1257UT S 31 HAY 1973 NASA GSI=C N Fe 0S0-7 2114A E W 1348 UT 27 JUNE ig73 S r FIG. 3 EUV SPECTROHELIOGRAMSHOWING A CORONAL HOIZ AT CENTRAL MERIDIAN ON MAY 31 AND JUNE 27, 1973. THE DARKENED AREA ENCLOSED BY THE ISOPHOT_Enc --l0 REPRESENTS THE HOLE. 294 1980010188-305 graphic telescope on Skylab. An atlas of the coronal hole boundary positions observed during this period has been compiled by Nolte et al. (1976). Six coronal holes were observed during this period four of which can be inferred from the radio scans. To trace the evolution of coronal holes during the first half of 197_, the 1415 MHz scans for January to July are reproduced in Figures 4 to 10. The CMP dates for these holes are then compared with those obtained using _ Skylab data when the observation periods overlapped. The notation and ! numbering of the coronal holes are the same as those used by Nolte et al. (1976) which facilitates comparisons between the two sets of observations. For example the coronal hole T3 observed in P'_res 1 and 2 was relabelled "CHI" in ,ubsequent diagrams. By presenting the radio data for January to July 1975, the history of certain coronal holes can be traced not only for a sequence of days but for a sequence of solar rotations. This set of diagrams shows the evolution of the coronal holes observed by Skylab. In addition the presentation of a long series of scans enables some insight to be gained into the assignment of classification numbers to the "radio" holes. The Skylab data indicated CHI underwent CMPs on May 31, June 27, July 25, August 21, September 16, October 14 and November I0 (Nolte et ai,1976). The scan data indicated CMPs on February 12, March 12, April 07, May 04, May 31, June 27, July 25 and August 22. The scan data indicated the hole disappeared after August although it was still observed by Skylab. Whereas Timothy et al. (1975) conjectured that CHI formed only one rotation prior to their first observations in late May, the radio data imply that it existed for fiYe previous rotations. The passage of CHI across the disc at the beginning of the Skylab mission is indicated by the labelled arrowb in Figures 8 and 9. Data were not available for May 26 and 27 but the existence of CHI is evident from 5_y 28 to June 05 with CMP occu_u'ing on May 31. Similarly Figures 9 and i0 reveal the presence of C!_Ion the scans for June 23 to July 03 with CMP late June 27, whilst in Figure i0, the observation period is restricted to July 23 to 30 with CMP on July 25. A summary of these observations is included in Table i. The results of the analysis of the scan data for CH3 are shown in Table 2. The information for the other holes inferred for 1973 is presented in Table 3. All these tables show the Skylab number (if applicable), the scan frequency, the classification, the period of observation and the Figure refe_-ence number (if observed between January and July) and the CMP dates infexTed from the scans and from the Skylab data. 5. SUMMARY OF RESULTS FOR THE PF_IOD 1968-1972 The technique of inferring the presence of coronal holes was extended to include the period 1968 to 1972. The results are presented in Tables 4 to 8, each table being reserved for the observations of holes for a 295 1980010188-306 particular year. The tables contain the scan frequency, the classification number and the CMP date for each coronal hole detected. 6. CONCLUSION Coronal holes can be inferred from solar E-W scan data. The holes manifest themselves as regions of depressed radio emissLon. The holes appear to be stable structures with lifetimes spanning many solar rotations. The holes inferred from the scans were generally the radio-wav_ analogues of the holes observed at X-ray and EUV wavelengths. The inferred coronal hole data were tabulated fo_ the period 1968 to 1973 and should be valuable to users requiring a consistent set of coronal hole observations during this period. ACK_0WLEDGE4ENTS The author wishes to thank George Daniel for assiztance in producing this paper and Eugene Molineux for help in analysing the data. 296 t O 1980010188-307 Jon I COLDSKY LEVEL A t W W E W E W i 0012 UT 01S5UT OOmLIT O0 33 UT _, O015UT 0203 UT O016u"r Og 10 11 12 E W E W E W E W 0007 UT 2320 LIT 0003UT 00 23 UT O024 UT 0157 UT O010tit O010 u'r NO DATA Jon Ig E W E W E W O030 UT 0124UT O0O0LIT 00 011UT 00 0g UT 00 4? UT 0010UT E - _--.W E W E W E W 0122 UT O010 UT O011UT O011UT 29 30 31 E-W SCA/_'_FOR JANUARY E w E W E w 1973oo',6ur ooz6 U'T O0oeu'r 297 1980010188-308 t, ,, • t, ESTM T O.QUIET SUN LVL. ' I COLD SKY LEVEL. _I73 A I I "- 'j' 02 o3 E I ''n+ _ _W E W E W E W O003 UT O00e UT 0207 UT E W E W E W E W O0 09 UT O019UT OQO0UT 0005 UT E ,.- --,- _-,., ",-W E W E W E W 0010 UT 0015 UT 0122UT 0001 UT O001UT O0O1UT IX}02 UT O0 09 UT E W E W E W E W O043 UT O011UT O016 UT O00'7 UT 21 22 23 2_1_ oool UT 0021 UT 0012 UT 0017 UT 272525 E W E W E W E W " 01Z$ UT O0 03 LIT O0 O?UT O002 UT FIG. 5 1415 MHz E-W SCANS FOR FEBRUARY, 1973 298 .£ 1980010188-309 t t t t t ESTHT D. QUIET SUN LVL. ,/ COLD SKY LEVEL _ lS73 A .! ; o, o, ;_ E_" CH3-'¢" • W E _lN E W W i 0002 UT 0O02 UT OOS_UT 0104 UT 05 O?O6g oe _. E W E W E W E W O025 UT O002 UT O020 UT 03 54 UT s E W E W E W E W O001 UT O006 UT 012? UT O010UT : E w E W E W E W O014 UT O005 UT O027 UT O014 UT O004 UT 00 2? UT O0 26 UT O0 03 UT 21 22 23 24 O0Oil UT 010it UT 00 03 LIT O02S UT E W E W E W 0119 UT O002 LIT 0102 UI" O001 UT E-W SCANS F0_ e w e w E W _R_, 19T3oo36UT O00Su'r oo04 299 1980010188-310 1973 _ 0E3_Ap,. A 01 NO DATA April 4 E W E W W O0 04 UT O0O0UT 01 011LIT 0,5 E W E W E W E ,W 02 08 UT 00 S9 UT 00 20 UT 00 $4UT E w E w E W E W oolo UT ooOSUT O00_u'r o02?u'r A z_o. quw_SUN_V'L. Jl_ I COLD SKY LEVELA ooo,, oo,, oo., 17 18 19 20 0003 UT 0002 UT 0006 UT 0020 UT E -CH3_ --W E W E W E W 0054 UT O0 23 UT IX]It UT O036 UT O0le UT O0 09UT O004 UT O017UT E W E W FIG. 7 l_15 MHz E-W SCANS FOR oI_7 UT 00I_UT APRIL, 1973 3OO t 1980010188-311 ESr'_T'D.QU_TSUN_/k. COLD SKY LEVEL MOy . \ II ' . o, E_ CH1"g"CH3_"W W E W O0 02 UT O016 UT 01O0LIT E W OS OS 07 OS E W E # W E WO0 SS UT E mZOUT W 0001UT O9 E W E W E W 00 23 UT 00 45 UT OO05UT E 0132 UT W May13 E W E W E W OO04 UT O008 LIT 01 23 LIT E W _'WE O022UT E w _ E W E w 00 22 UT 00 03 UT 00 0g UT E w 25 28 NO DATA NO DATA May 26 May 2"/ E W E W 00 12 UT oo 03UT 29 : FIG. 8 l_15 MHz E-W SCANS FOR E W E W E W MAY, 1973 O003 UT 00 IZ UT OO40 UT 30| t r 1980010188-312 I I t ESTHTD, QUIET SUN L'V'L. I COLD SKY LEVEL.1973 . I , 0 I I 02 03 04 E,'*" _M_ x.w E w E w E W 0003 UT 0003 UT 0125 UT 00 03UT O00Q UT O027 UT O0 10 UT O0 03 UT NO DATA NO DATA NO DATA Jun 9 Jun 10 Jun 11 E W 00 liUT oo. .o.Jun 14 Jun 16E W E W 0004 UT 00 04 UT 0118 UT O003 UT O015 UT O0 03 UT W W W E O0 01 UT E 0159 UT E O0 01 UT E W E W E W E W 0001 UT O0 53 UT 01 12 UT 0007 UT E 'w E w JUNE, 1973 o_03uT DO_2u': 302 1980010188-313 i! 0_/_c_ 03 04E w _ _ ,L 0126 LIT E W E W L \0023 UT 010g LIT E W ,: 00 23 UT OS 07 08 II • NO DATA ESTHTD. QUIET SUNi _/L. : E W E W E W 00 14UT 00 og UT 0010 UT JuL 0g Jul. 10 Jul 11 Jut 12 13 14 15 16 00 25 UT 02 02 UT 01 28 UT 0012 UT INOOA'AI OOA"I 'OOA'AI I O A]Jut 1"/ Jut 18 Jut 19 JuL 20 23 24 I 'NO DATA NO DATA E W E WJuL 21 JuL 22 0018 UT 0004 UT 0004 UT 0103 UT 00 04 UT 00 04 UT E w E W E w E-W SCANS FOR DOs8_r DOos _ Doos_ J-u'LY,1973 303 1980010188-314 TABLE 1 SUMMARY OF OBSERVATIONS OF CORONAL HOLE CHI DURING 1973 Figure Scan Class Period of C'MP CMP Frequency Observation from Scans from Skylab (MHz) on Scans (1973) (Nolte et al 1976_( 973) (1973) ., 4 1415 I Jan ll-Jan 20 mid Jan 15 5 1415 i Feb 08-Feb 18 late Feb 12 6 1415 i Mar 08-Mar 15 early Mar 12 7 1415 2 Apt 02-Apr 12 early Apr 07 7, 8 1415 2 Apt 29-May i0 early May 04 692 2 May 02-May 06 early May 04 8, 9 1415 3 May 28-Jun 05 mid May 31 May 31 692 i May 30-Jun 02 May 31 9, i0 1415 3 Jun 23-Jul 03 late Jun 27 Jun 27 692 3 Jun 27-J_u_ 30 early Jun 28 I0 1415 2 Jul 23-Jul 30 early Jul 25 Jul 25 692 2 Jul 23-Jul 27 early Jul 26 692 i Aug 21-Aug 22 early Aug 22 Aug 21 Sep 16 0et 14 Nov I0 TABLE 2 SUMMARY 0F OBSERVATIONS OF CORONAL HOLE CH3 DURING 1973 Figure Scan _lass Period of CMP CMP Frequency Observation from Scans from Skylab MHz on Scans (1973) (Nolte et al 1976_ (1973) e rly" "09(1973)4 1415 Jan 05-'Jan 15 692 Jan 08-Jan 12 late Jan 09 5 1415 Feb 01-Feb l0 Feb 05 692 Feb 0h-Feb 07 late Feb 05 5, 6 1415 Feb 28-Mar 07 early Mar 03 692 Mar 03-Mar 06 mid Mar 03 6, 7 1415 Mar 27-Apr 03 late Mar 29 692 Mar 29-Apr 02 early Mar 30 17, 8 1415 Apt 21-May 02 early Apr 26 692 Apt 24-Apr 30 late Apt 26 8 1415 May 18-May 28 early May 22 692 May 21-May 25 mid May 23 9 1415 Jun 13-Jun 24 mid Jun 18 June 21 692 Jun 17-Jun 20 early Jun 19 July 17 Aug 13 304 ii 1980010188-315 TABLE3 SUMMARYOF OBSERVATIONSOF CORONALHOLES (OTHER THAN CHI (TABLE I) AND CH3 (TABLE 2) DURING 1973 Coronal Hole Figure Scan Class Period of CMP CMP from Frequency Observation Skylab (Nolte et al 1976 4 1415 1 Jan 17-Jan 26 mid Jan 21 5 1415 1 Feb 12-Feb 20 Feb 16 4 II15 i Jan 20--Jan 28 ._arl_Jan 24' 92 1 Jan 29-Jan 31 Jan 30 6 1415 1 Mar 19l'Mar 28 ._arlyMar 22 692 i Mar 22-Mar 24 _arl_ Mar 23 . CH5 9 1415 1 Jun 07,08,12, June 09 13 lO 1415 1 J6A 06-Jul08 " Jul08 1415 1 Aug 02-Aug 09 ._arlyAug 05 CH4 1415 3 Sep 26-0ct 02 late Sep 28 Sep 28 692 2 Sep 28-0ct 02 _arly Sep 29 1415 2 Oct 22-0ct 29 Oct 25 Oct 25 692 i Oct 23-0ct 27 Oct 25 1415 2 Nov 18-Nov 26 mid Nov 22 No" 21 692 1 Nov 21-Nov 24 late Nov 22 N_v 21 1415 1 Aug 06-Aug 13 Aug 09 692 1 Aug 08-Aug 12 _arly Aug i0 1415 1 Sep 01-Sep 07 late Sep 04 1415 i Sep 19-Sep 27 Sep 22 _H6 Oct 20 Nov 16 CH2 May 28 Jun 22 Jul 20 Aug 16 Sep 12 Oct 09 Nov 05 3O5 t 1980010188-316 TABLE 4 CORONAL HOLE DATA FOE 1968 INFERRED FROM THE "FLEURS" E-W RADIO SCANS 1415 MHz 692 MHz CMP Class CMP Class e = early (1968) "m = mid 1 = late e Jan 02 2 1 Jan 02 3 m Jan ;6 1 1 Feb i0 2 e Feb ll 3 m Feb 16 1 e Mar 03 2 Feb 23 1 m Mar 14 1 m Mar 22 3 e Mar 22 2 No data Apr 17-May 16 Apr 08 1 May 16 3 No data Apr IT-May 16 e May 25 1 May 16 2 m Jun 07 3 e May 25 I I Jun 22 i e Jun 01 I e Jul 15 2 e Jun 07 3 e Jul 20 1 e Jun 13 1 e Aug 15 1 e Jun 17 1 e Sep 08 I m Jun 22 2 m Sep 12 1 m Jul 14 1 e Sep 18 1 e Jul 20 1 e Sep 28 1 m Jul 29 1 e Oct 09 1 e Aug 15 2 e Oct 26 1 m Aug 30 1 e Oct 30 3 m Sep 08 1 m Nov I0 1 m Sep 12 I m Nov 18 i e Sep 18 I i Dec 07 i i Oct 09 2 e Dec ii I e Oct 26 i e Dec 30 2 e Oct 30 3 m Nov i0 3 m Nov 18 2 e Nov 26 1 e Dec 08 1 m Dec .ii 1 m Dec 30 3 306 1980010188-317 TABLE 5 CORONAL HOLE DATA FOR 1969 INFERRED FROM THE "FLEURS" E-W RADIO SCANS ! lhl_,MHz ,, .6_2 ,.MHz [ CMP Class CMP Class e = early (1969) m--mid _ I = late (i 69) _ m Jan 30 i m Jan 30 1 i Feb 03 1 1 Feb 03 2 Feb 26 2 m Mar 03 i m Mar 03 i e Jim 03 I I Mar 13 1 No data Jun 04-06 e Mar 28 1 e Aug 08 i 1 Apr 09 1 e Sep 11 1 e May 29 2 e Sep 19 1 _, No data Jun 04-09 e Oct 04 2 m Jun 13 1 m Oct 08 2 e Jul 12 ! m Oct 17 I m Jul 21 I I Nov 09 i e Aug 08 3 m Dee 06 2 e Aug 20 i m Dec 22 2 e Sep ii I No data Dec 25-31 i Sep 30 2 e Oct 04 1 e Oct 08 I e Oct 17 I i Nov 26 2 m Dec 06 3 m Dec 22 2 No data Dec 25-31 307 1980010188-318 TABLE 6 CORONAL HOLE DATA FOR 1970 INFERRED FP" ,ITHE "FLEURS" E-W RADIO SCANS .,, MHz MHz '"CMP Class " CMP " Class e = early (1970) m=mld I = late (z To) • Jan Oh 2 m Jan ii i Jan ii I i Jan 24 i m Jan 21 2 m Feb 07 I I Feb 07 I e Mar Ol i e Mar 01 2 e Mar 18 I e Mar 18 3 e May 12 2 Mar 28 2 e Jun 04 I e Apr 03 i m Jun 18 I m Apr 15 I m Jun 27 2 m Apr 2h 2 e Jul 06 i m Apr 30 I e Jul 09 3 e May 08 I m Jul 26 2 May 12 3 I Aug 06 2 i May 20 I m Aug 17 2 I May 23 I i Aug 20 I m Jun 04 I e Sep 01 i e Jun 09 I e Sep 07 i Jun 19 3 e Sep 13 1 m Jun 27 2 e Sep 18 I 1 Jul Ol 1 m Sep 30 3 e Jul 06 2 e Oct i0 2 m Jul 09 3 m Oct 14 2 Jul 26 3 m Oct 27 2 i Jul 31 i m Nov ii i Aug 07 2 m Nov 20 I i Aug i0 I m Nov 23 2 m AUg 17 I I Dec 07 I e AUg 21 I m Dec 16 I m Aug 26 i Sep 01 2 e Sep 07 I m Sep 13 I m Sep 18 2 m Sep 30 3 Oct i0 I I Oct 14 2 m Oct 27 I e Nov 02 I e Nov 20 I m Nov 23 3 m Dec 07 3 e Dec 17 2 308 i9800i0i88-3i9 TABLE 7 CORONAL HOLE DATA FOB 1971 INFERRED FROM THE "FLEURS" E-W RADIO SCANS CMP Class CMP Class e = early (1971) m = mid 1 = la_e (1911) , , ,. • e Jan 12 1 m Jan 27 3 e Jun 28 1 e Feb 04 2 Feb 01 1 m Feb 22 3 e Feb 04 1 m Mar 21 1 m Feb 12 2 1 Mar 25 1 m Feb 22 2 I Apt 29 2 I Mar 02 I m May 04 i e Mar 04 I m May 26 2 m Mar 21 i m May 31 3 I Mar 25 i I Jun 26 2 m Mar 30 I e Jul 01 I Apr 03 I e Jul 24 3 m Apr 09 1 i J_1 27 1 e Apr 15 i m Aug 01 2 e Apt 29 3 i Aug ii 3 e May 04 2 e Aug 19 1 I May 30 3 m Aug 27 2 e Jun 0h 2 1 Sep 06 1 i Jun 26 3 m Sep 23 3 i Jun 30 3 e Sep 28 2 m Jul 14 2 m Oct 02 2 e Jul 24 3 e Oct 07 i e Aug 01 1 m Oct Ii 3 e Aug Oh i m Oct 25 3 e Aug ll 1 Nov 03 2 e Aug 19 1 e Nov 08 1 m Aug 27 2 m Nov 22 i m Aug 31 I m Nov 29 2 m Sep 23 3 e Dec 05 2 e Sep 28 3 e Dec 19 3 m Oct 02 2 m Dec 26 1 Oct ii 1 m Oct 20 1 m Oct 2_ 3 e Oct 29 i Nov 03 2 Nov 08 2 e Nov 22 3 e Nov 29 I m Dec lh 1 m Dec 18 3 m Dec 28 i SO9 1980010188-320 TABLE 8 CORONAL HOLE DATA FOR 1972 INFERRED FROM THE "FLEU_S" _-W RADTO SCANS 14.z5. 1 CMP Class ........CMP Class e = early (1972) m = mid i = fat .(z 72) ,,. m Jan 01 1 m Jan 01 1.... m Jan 05 2 1 Jan 05 1 1 Jan 15 1 Jan 16 2 e Jan 22 i Feb ii I e Jan 25 1 1 Feb 23 2 Feb ii 2 m Mar 05 1 m Feb 18 2 e Mar I0 2 m Feb 23 3 m Mar 22 I I Mar 09 3 m Mar 28 i e Mar 16 2 e Apr 02 1 m Mar 22 1 m Apr 24 2 m Mar 28 2 m Apr 29 2 e Apr 05 1 m May 26 1 e Apt 24 2 1 Jun Ol 1 1 Apt 28 3 e Jun 19 2 m May ll 1 e Jun 28 1 m May 16 1 m Jul ii 1 m May 26 3 m Jul 21 2 e Jun 02 2 m Jul 25 1 1 Jun 19 1 e Aug 17 1 1 Jun 24 1 e Sep 13 1 No data Jun °7-30 No data Oct 02-04 1 Jul Oh 2 e Oct 17 1 m Jul 06 1 Oct 25 1 Jul 08 1 e Nov 20 1 m J1ulii 1 m Dec 13 2 e Jul 21 2 m Jul 25 1 Aug Ol 2 1 Aug I0 1 e Aug 17 1 e Aug 28 3 e Sep 06 1 e Sep 13 2 No data Oct 02-04 Oct 04 1 e Oct 17 1 Oct 25 3 m Oct 27 i Nov 19 Dec 05 I e Dec !3 2 " e Dec 17 2 310 t J 1980010188-321 REFERENCES Bohlin, J.D. (1977): Extreme-Ultraviolet Observations of Coronal Holes. I : Locations, Sizes and Evolution of Coronal Holes, June 1973 - January 1974. Solar Physics 51, 377-398. Chiuderi Drago, F., Avignon Y., and Thomas, R.J. (1977): Structure of Coronal £ Holes from UV and Radio Observations. Solar Physics 51, 143-158. Covington, A.E. (1977): Coronal X-Ray Holes and the Quiet Radio Sun at 2800 MHz. Solar Physics 5h 393-402. Dulk, G.A., and Sheridan, K.V. (1974): The Structure of the Middle Corona from Observations at 80 and 160 MHz. Solar Physics 36, 191-202. Furst, E., and Hirth, W. (1975)" Coronal Hole Observed at 10.7 GHz with a Large Single Dish. Solar Physics 42 157-161. Krieger, A.S., Timothy, A.F., and Roelef, E.C. (1973): A Coronal Hole and Identification as the Source of a High Velocity Solar Wind Stream. Solar Physics 29, 505-525. Munro, R.H., and Withbroe, G-L. (1972): Properties of a Coronal Hole Derived from Extreme Ultraviolet 0bser_ _tions. Astrophys. J. 176, 511-520 Neupert, W.M., and Pizzo, V. (1974): Solar Coronal Holes as Sources of Recurrent Geomagnetic Disturbances. Journal of Geophysical Research 79, 3701-3709. Nolte, J.T., Krieger, A.A., Timothy, A. F., Vaiana, G.S., and Zombeck, M.V. (1976): An Atlas of Coronal Hole Boundary Positions May 28 to November 21, 1973. Solar Physics 46, 291-301. Sheelcj, N.R., and Harvey, J.W. (1978): Coronal Holes, Solar Wind Streams, and Geomagnetic Activity during the New Sunspot Cycle. Solar Physics 59, 159-173. Timothy, A.F., Krieger, A.S., and Vaiana, G.S. (1975): The Structure and Evolution of Coronal Holes. Solar Physics 42, 135-156. Vaiana, G.S., Krieger, A.S., and Thaothy, A.F. (1973): Identification and Analysis of Stuctures in the Corona from X-Ray Photography. Solar Physics 32, 81. 311 t 1980010188-322 -F" L8'O -18 48,8, PROI_ET: AN APPLICATION OF PROPAGATION FORECASTING PRINCIPLES Paul E. Argo Ilan J. Rothmuller Naval Ocean Systems Center San Diego, CA 92152 The performance of a wide variety of Naval electromagnetic systems depends on ionospheric conditions. For example, solar disturbances of the ionosphere can interrupt co_nmicatlons, render surveillance systems useless and cause navigation system inaccuracies. Real-time propagation assessment is a technique to optimize system performance by allowing knowledge of current ionospheric conditions to be used to minimize detrimental effects of an unknown or disturbed propagation envlrom, ent. An envlrom,ental prediction and assessment system (EPAS), comprised of a variety of real-time sources of solar/geophysical data and a center which collects, processes and selectively dissemin- ates these data to regional propagation assessment terminals, can pro- vide real-time knowledge of propagation conditions. The US Naval Ocean Systems Center is developing a terminal called PROPHET (for propagation forecasting), a key element of the EPAS that uses real-time solar/ geophysical data to make regional assessments which are tailored to specific system useage. PROPHET uses models to translate data from satellite and ground based sources into performance predictions for specific systems. PROPHET is presently being evaluated at a Naval Com_unicatlon Station and has proved to be useful to operational personnel. INTRODUCTION The performance of electromagnetic communication and navigation systems depends on and is vulnerable to the electromagnetic (EM) wave propagation envi- ronment. The Navy's Global Positioning (shf satellite) System, Fleet Satellite Con_nunlcatlons, the hf and vlf Fleet Broadcast, vlf and elf strategic communi- cations are all examples of systems which make use of EMwaves whose propagation environment is the ionosphere. Changes in the ionosphere, and consequent changes in wave propagation characteristics, may degrade system performance through large signal loss, rapid signal fluctuations, rapid phase changes, changes in angle of arrival, etc. Given foreknowledge of ionospheric changes and the resultant effects on systems, the user is warned of imminent problems and so may be able to take mitigating steps. An Environmental Prediction and Assess- ment System (EPAS) under development will provide the Navy user with the required foreknowledge. 312 1980010188-323 An effective EPAS contains three functional parts: (1) a number of sensors which obtain raw solar and geophysical data; (2) a data reduction, sorting, and collection function which puts the pertinent data into useable form and forwards it to a forecasting center; (3) a subsystem which digests these data and produces assessments and forecasts of system performance. It gives guidance for frequency selection, antenna selection, transmitter power - all of the factors required to optimize system performance. These outputs are tailored for the specific system and user. Besides providing real time guidance for system users, an effective EPAS will help the Navy satisfy a number of logistic, design, and operational requirements. For example, the Navy needs (1) more efficient frequency manage- ment in a world of shrinking available radio spectrum; (2) to minimize the number of transmitters and transmitter power to those actually needed for acceptable communications probabilities (an "overkill" policy is used now); (3) to take in- to account extreme environmental effects during system design and test, such as accounting for scintillation fading on FLTSATCOM reliability; (4) to develop tactical communication strategies for anti-jam and low-probability-of-intercept operations. The EPAS under development is designed for those systems whose propagation characteristics are controlled by the ionosphere and magnetosphere. In turn, the characteristics of the ionosphere and the magnetosphere are controlled by the sun's activity. Short term variations in solar activity such as solar flares, x-ray bursts, and high speed solar wind streams cause significant changes in the ionosphere and magnetosphere. Long term cycles of solar activity control the long term average characteristics of the ionosphere. Thus, the first two func- tions of the EPAS are concerned with the collection and delivery of solar- terrestrial activity data to a propagation assessment and forecasting subsystem. The third function of EPAS is to provide propagation assessments and fore- casts tailored for particular Navy systems and users. NOSC is developing a propagation assessment and forecast terminal (PROPHET) (Richter, et al., 1976; Pmthmuller, 1978) to fulfill this function. In the remainder of this paper we will describe the philosophy behind PROPHET and several of the prediction models We will also discuss the future of the PROPHET concept and those areas of solar- terrestrial physics where more information is needed to bring PROPHET to its full potential. PROPHET DESCRIPTION "Propagation conditions for the North Pacific are likely to be unsettled for the next 12 hours because of a geomagnetic storm." General advisories of this sort are about all that has been available for the Navy communications officer for the last decade. They are useful in that the communicator is warned to expect erratic system behavior, to expect the usual hf communications fre- quencies to become unusable, and to understand that when outages do occur it is unlikely they are due to equipment or operator failure. However, the communica- tions officer really wants to know the specific steps to take to minimize the message error rates whenever there is an environmentally caused upset to his system. For example, the hf communicator would like to have predictions of the 313 I 1980010188-324 lowest useable frequency (LUF), the maximum useable frequency (HUF), and the frequency of optimum transmission (FOT) for each specific circuit. The PROPHET propagation assessment and forecasting terminal is designed to provide him this information. PROPHET provides this information by ta_Lng environmental data from a variety of satellite and ground-based sensors and using these data as inputs to a variety of analytical models that output the specific propagation forecasts required by the communications officer. Figure 1 shows the information flow for the entire EPAS. Data from satellites (SOLRAD HI, SNS/COES, etc.) and ionospheric assessment centers such as Air Force Global Weather Central and NDAA Space Enviromnental Services Center pass through a general center (pre- sently NOSC, La Posta), where the data are sorted, converted, and relayed to the individual PROPHET terminals. F.c, '' KS HO_OLUL f Figure i. Information flow for the PROPHET environmental prediction terminal The PROPHET terminal, or minicomputer (presently an AN/L_K-20) contains specific numerical models stored in memory. The models quantify environmental effects on propagation and address specific system vulnerabili_ies. Initially we at NOSC identify the system characteristics and vulnerabilities, generally through inputs from Naval Communications Officers. Once each problem area is understood then we begin model development. Often the necessary information is available through scientific literature or personal contacts. When it is not available we either do the research ourselves or, when possible, we aid re- searchers at other institutions in finding support for specific research pro- Jects. Three approaches can be taken to develop models assessing the propagation environment: phenomenological, statistical, and semi-emplrical. The phenomeno- logical is based on an understanding of the physical processes relating cause and effect. The statistical approach is based on synoptic studies and assumes that predictions can be made based on historic data. The semi-empirical method 314 1980010188-325 ...... L-. ! _ (presently used most often by the NOSC modeling group) combines knowledge of the i basic physical causes vlth observational data to generate simplified models which , obviate modeling the complicated processes between the basic driving cause and :_ final effect. An illustration of a seml-empirlcal model is the effect of high energy solar proton emission on the Omega vlf navigation system. The physics of the process i is that the polar D-reglon of the ionosphere (50-90 kin)is anomalously ionized ._ by high energy (E > i0 MeV) protons. Thls increased ionization causes greater than normal phase advances on signals crossing the affected region and thus, if _ unaccounted for, would cause navigational errors. The increased ionization also causes high absorption in the hf (and vhf) band, which has given the name polar cap absorption or PCA to these events. " The Omega correction model developed by NOSC (Argo, 1975) translates the ,_ solar proton flux measured at satellite altitudes into phase error corrections _- (Figure 2). These corrections are in the form now routinely used to account for variations due to time of day, season, etc. The model has as parameters integrated proton flux (E >10 HeY), time of day, and path length through the "polar cap" defined to be greater than 63 ° geomagnetic latitude. The model will correct the phase back to provide the nominal 2 _n accuracy of Omega. The imple- mentation of the model uses an auxiliary look-up table (Argo, 1976) of normalized phase advance, with PROPHET providing a "proton flux factor" that accounts for the event magnitude. HYPERBOLIC 30 _L_EC - 3rim HYPERIOLIC RANGING (NORWAY -- HAWAII) I nm RANGING (NORWAY) 2 § CORRECTED, ACTUAL -- IOOpSEC 0 B 12 111 24 B 12 1B 24 B 12 18 24 e 12 18 24 6 12 18 24 6 MAR 7 MAR II MAR B MAR 10 MAR 1070 Figure 2. Corrected and uncorrected O_ega positioning errors 315 t ]9800]0]88-326 This semi-empirical approach has yielded other models relating solar proton : flux data to systems effects. Argo and Hill (1976) developed a model relating absorption of transpolar hf and polar region vhf/uhf satellite co-,,unications to the proton flux. These models use average loss rates (with seasonal variations) to define the D-region electron density, and then calculate the absorption directly. Another model developed by NOSC is the effect of an x-ray flare on hf com- muntcatinns. This model has two "components. '* One models the time history of the flare. The flare duration model predicts the duration oace the flare has reached its peak. The prediction technique uses the discovery by NOSC that the flare decay can be accurately described with one of a small set of defiutte decay rates. The rise time and peak amplitude are used to estimate the decay rate (Argo, etal., 1978). The model adjusts itself in case of multiple flares and can update the decay estimate. Figure 3 shows an example of the model using historical SOLRAD 9 data. -l Figure 3. Solar flare I-8A x_ray flux (solid curve) and flare forecasting results (x's) Increased x-ray fluxes cause increased absorption of hf signals, wlth the amount of absorption increasing as the hf frequency decreases. The effect of an x-ray event is therefore to increase the lowest useable frequency (LUF). This relatlonship between the x-ray flux and the LUF (Rose, etal., 1974) is the second component of the flare model. Once the flare x-ray is modeled then the time to recovery as a function of frequency can be calculated for each conmunica- tion circuit. The time to recovery will be the point in time that the x-ray enhanced LUF drops below the specified frequency. Thls output, which now specifically indicates to the co-,,unlcatlons controller how long his given circuit will be unavailable, is displayed in Figure 4. 1980010188-327 | :;] :.:: ;_.-'%.;.:L_-_ oot.M,,_Tsvs'_m'Mm'o ..................i:: ............ i i _ ,, ....._ _ ............. ,........................ OtJ_ Figure 4. PROPHET display predicting time to recovery from a solar flare for the ship DURHAM In addition to these propagation predictions during disturbed periods, the Naval Communications Officers expressed interest in having automated the routine predictions of diurnal variations in the MUFs and LUFs. To achieve this models by both the Air Force (1975) and NOSC (Rose, et al., 1978) have been implemented. The Air Force model uses hourly updates to a statistically based model - the updating is done by refitting to data obtained from an hf ionospheric sounder network. The NOSC model is extremely simple, but gives results with nearly the accuracy of the unmodified Air Force model. Therefore it is useful in long term predictions. Figure 5 is an example of the output of the MAP routine, which gives the communications controller the pertinent propagation information such as LUF, MUF, and FOT for his established circuits. The graphics display is used to allow the controller to visualize the area he is responsible for. These dis- | _ SOLPdaP_?XUZl_L__.L_L L_I -*_ ,,,,,,, -, .,da, ,,,e OOL,m,c'rlvz"rv, out["r _ _- _ ['-'i]]] _: _ :: :_: "_'-'-'#-"--1 t.* f*oe,rnD_?AT.ZNK,UP _ll_..- _-"-#_ '-_- : , _ _ 'T-- N L.......:.........,_ ........_........L_,.. ...............L .L...---A-............. : i-,_,. ) : ! I i _ : i ! I _\_: | ":"........+'"-"_...................._--"+ ......"_.......+........t '_'*" ! "! .........r_.....'.....-__---1.-._,-.--k:.,-_---.--r.... .. •,I,_ i.' ! ' : i ., -...-. ,6-.,: r_!_'_f" -I'.........:I .....' ......._i '_'i..................t_ .... t _--I. ' t ' ' | 4, Figure 5. PROPHET display showing controller the LUF, MUF and FOT for the established circuits 317 i 1980010188-328 !i I plays are also available during disturbed periods (such as x-ray flares), in which case the MUYs and LUYs are those expected at the moment of request. Another statistical model adapted for Navy use is the scintillation depth of fading model (LaBahn, 1974). This is an adaptation of the model developed by Fremouw and RLno (1973) and modified by Pope (1974). An important aspect of these models developed at NOSC is that they are tailored to a specific system application and that the output is in easily used terms (e.g. navigation correction £n miles, lowest useable frequency, etc.). Table I stmnartzes the completed forecasting models available to PROPHET. TABLR1. D_CLTPTIO_ 01 _ POP_ASTIHGMODILS Hodel Function Cauutive System _t_n sourca) so_ce Flare det:ectinu Disturbance 'l-SX z-raya All hf V_ varnin8 vlf navisation Flare duration Disturbance 1-_ x-rays All hf WarnLn8 sat:blator vsrld_ll vlf naViptlon (_OSC) SID GRID Disturbance 1-8_ x-raya All hf hf coin 010SO) _rntnl_SWF • Fraq shift • Rerouta traffic HFDP • Net impact 888U_t PCA/vlf Disturbance 10 _kV vlf Correction factor (_OSC) warninl;-?CA particles navtpt ion for transpolar (Omsa), circuits SPA/vIf Disturbance 1-8_ x-rays vlf Correction factor _qOSC) wun_ins-SPA naviption for s_it circuits (o.'sa) LOF split Tactical Solar diurnal Covert Optlm_ freq (NOSC) (reduced transition hg Byst_ selection asaLnst intercept 1-8_ x-ray8 known receivers vulnerability) Scintillation Disturbance Unkno_ vhf/uhf Advisory - dB ' srld (SRI) _rnins/ (statistical satellite fade probability tactical model) co--u_nica- based on locatiou tton8 PCA/hf Disturbance- lO MeV All _ I Adviso---ryd-B.... (NOSC) vsrutns-PCA particles • frsq shift • reroute traYftc HFDF • Net impact ass_smant quiet HUF )_F durln8 ALt hf hf co_ (ITS/_C) natal tines normal operations QLUY LUY durin8 All hf hf co_ ncrna£ tines normal operations , l_ytraca Receiver _ll h_ Ihf cons sccuatbiltty , normal operations 1980010188-329 iNOSC conducted a fleld test of the PROPHET assessment terminal at the Naval Conznunications Station, Stockton, CA. The test objectives were to test the util- ity (and reliability) of real-time tailored propagation assessments in an opera- tlonal environment. Analysis has shown that: (i) COMMSTA personnel made fre- quent use of PROPHET - termlnal access on an average of once every nine minutes; (2) using PROPHET predictions, QSYs (number of frequency shifts) and outages due ! to propagation were reduced by 15% wlth duration of outages reduced by 15%-20%; (3) CO_4STA personnel gave a qualitative general "usefulness" rating of PROPHET of 8 on a scale of 10. FUTURE PLANS FOR PROPHET AND NEEDS FOR CONTINUED DEVELOPMENT i • As use of ionosphere and transionospheric communication channels is in- creased the impact of and need for solar-terrestrlal predictions will increase ,, concommitantly. PROPHET enhances the ability to do individual specific fore- : casts, and so the Nay, will likely move towards increased use of shipboard (and airborne) prediction terminals. These can be used to take tactical advantage of situations such as anomalous intercept opportunities (e.g. P re, 1976). In addition, the FAA has expressed an interest in a tailored PROPl_-like terlrinal to ensure adequate communication capabilities for commercial airplanes. In- deed, as the use of the present PROPHET techniques become standard in the com- munications world then we expect the design of wholly new sophisticated commun- ication systems that will require a new generation of more accurate PROPHET models. Minicomputers are becoming both smaller and more powerful and so more com- plicated models (both phenomenological and statistical) can realistically be expected to opevate in a real-time environment. These increased capabilities will make tactical propagation advantages available to the user of these fore- casting systems. This is important, because presently much effort is being expended in gaining a few dB advantage for communications systems through sig- nal coding and information processing techniques. Intelligent use of a PROPHET system can presently yield advantages significantly higher than that (Rose, 1976; Argo, et al., 1976). Present modeling capabilities are limited by lack of phenomenological pic- tures of some solar and ionospheric processes. Future work must be aimed at the removal of these limitations in the following problem areas: (i) Flare prediction. It is anticipated that by 1980-81 flare activity will be high, and so flare occvrrence and flare product prediction models must take high priority. (2) Scintillation phenomenology. The increased dependence on space borne systems increases the vulnerability to debilitating scintillations caused by irregularities in the intervening medium. Thus, in anticipating outages and . to provide circumventing strategies, scintillation morphology an___dphenomenology must be better kmown and understood. REPRODUCIBILITY OF THE 319 ORIGINAL PAGE IS POOR # f 1980010188-330 (3) Ma_netlc and ionospheric storms. These disturbances affect a wide vari- ety of military systems such as space track radars, hf communications and sur- veillance systems, as well as ant_submarine warfare (ASW) and mine warfare sys- tems. An understanding of the energy inputs into the neutral atmosphere 3d the redistribution of this energy through gravity waves and winds must be attain- ed if the effects of these storms are to be adequately modeled. (4) Ambient ionosphere. If forecasting is to be developed and used as a tactlcal operating strategy, better knowledge must be gained of the llnk between real-time solar geophysical information and ionospheric behavior. Although much work has been done in this area, interactions between the magnetosphere, iono- sphere and neutral atmosphere compllcate the problem and make the development of an accurate model very difficult. Note that the difference between the iono- sphere disturbed by an ionospheric storm and the typically varying ambient iono- sphere is strictly a matter of degree, and so problem areas 3 and 4 are closely linked. CONCLUSION The effectiveness of Naval Communications is dependent upon establishing and maintaining adequate communication links. Because ionospheric disturbances may be a disrupting factor on these communication links, it is important to have accurate long and short term predictions of solar-terrestrial activity. In order to take remedial action for a specific system we have found it necessary to have real-time forecasts tailored to fit the user's needs. The tailored forecast concept has been implemented through the PROPHET propagation assess- ment and forecasting terminal. Recent field tests (Levine, 1978) indicate that for the specific systems and disturbances addressed that the PROPHET concept is extremely useful. PROPHET has only just touched the surface in the application of propagation prediction for communication systems. As computational capabilities increase and costs drop there is much more room for expansion into increasingly sophisti- cated areas such as propagation channel modeling, where actual statistics of the channel are calculated rather than the more general character such as MUF and LUF. There are also several specific disturbance phenomena (e.g. scintil- lations and ionospheric storms) that are still inadequately understood in a phenomenological sense, and these must be understood if we are to maximize the usefulness of solar-terrestrial forecasting. REFERENCES Air Force Global Weather Central (1973): AFGWC Manual 105-1. • Argo, P. E. (1975): Modeling Omega PCA Phase Advances, NELC Technical Report 1950. 320 t 1980010188-331 Argo, P. K. (1976): Omega Navigation Disturbance Correction Tables, Polar Cap Absorption Events; East Pacific, 10.2 kHz, NELC Technical Document 480. Argo, P. E. and J. R. Hill (1976): High Frequency Polar £ p Absorption Model, SOLRAD Application, _LC Technical Note 3249. Argo, P. E., I. J. Rothmu]ler and J. R. Hill (1976): Environmental Effects on HF Propagation, NELC Technical Note 3204. Argo, P. E., I. J. Rothmuller and J. R. Hill (1978): X-Ray Flare and Short-Wave Fade Duration Model, NOSC Technical Report 255. Fremouw, E. J. and C. L. Rino (1973): Modeling of Transionospheric Radio Propa- gation, Radio Sci., 8, 213. LaBahn, R. W. (1974): Development of a Scintillation Prediction Grid, NELC Technical Note 2814. Levine, P. (1978): SOLRAD PROPHET DT&E Field Test #I (HF Communications), Megatek Report #R2005-096-F-I. Pope, J. H. (1974): High Latitude Ionospheric Irregularity Model, Radio Sci., 9, 675. Richter, J. H., I. J. Rothmuller and R. B. Rose (1976): PROPHET: Real rime Propagation Forecasting Terminal, paper presented at the 7th Technical Exchange Conference, E1 Paso, TX, 30 Nov - 3 Dec 1976. Rose, R. B. (1976): Tactical Application of HF Propagation Forecasts, NELC Technical Note 3179. Rose, R. B., J. R. Hill and 5,. P. Bleiweiss (1974): Sudden Ionospheric Disturb- ance Grid, NELC Technical Report 1938. Rose, R. B., J. N. Martin, P. H. Levine (1978): MINIMUF-3: A Simplified hf MUF Prediction Algorithm, NOSC TR 186, Feb 1978. Rothmuller, I. J. (1978): Real Time Propagation Assessment, AGARD-CPP- 238, Ottawa (24-28 April 1978). 321 1980010188-332 PREDICTIGNS OF THE SPACE ENVIRONMENT SERVICES CENTER G. R. Heckman Space Environment Services Center Space Environment Laboratory National Oceanic and Atmospheric Administration Boulder, Colorado 80303, USA A list is presented of all the current types of users of the Space Environment Services Center. All the data collected by the Center are listed and a short description of each primary index or activity summary is given. Finally, each type of regularly produced forecast is described, along with a summary of the mr thods used to produce each prediction. l. INTRODUCTION - THE SPACE ENVIRONMENT SERVICES CENTER The Space Environment Services Center (SESC) in Boulder is operated jointly by the National Oceanic and Atmospheric Administration (NOAA) and the U. S. Air Force Air Weather Service (USAWS) and exists to provide predic- tions, alerts and real-time aata descrying solar geophysical disturbances to users througho,Jt the United States. In addition, it is designated as the World Warning Agency (WWA) for the International Ursigram and World Days Service (IUWDS), which is operated under the International Union of Radio Science (URSI). SESC collects data from its own sensors, from cooperating agencies and institutions and from other countries, both through IUWgS and through bilateral exchange agreements. The data collected include in,_ges and parametric measurements of the sun, the ,iterplanetary medium and the earth's magnetosphere and ionosphere. Users of the SESC service can be divided into roughly three equaA parts. _bout one-third find some form of the predictinns to be sufficiently accu- rate to be useful for planning purposes; another third wait until they receive an alert that an event has started 'erore they take action. The final third are users interested in receiving data in some form, which may be indices, activity sunrnaries or high resolution data. A summary of the types of users is given in Table !. 322 ) 1980010188-333 °! ! Table !. Users and Types of Activity Affecting Their Systems : Customer Type of Activity Producln_ Effect • Civilian satellite communication • Magnetic storms • Commercial aviation--mid-latitude • Solar radio emissions communication (VHF) • Commercial aviation--polar cap PCA, magnetic storms, x-ray communication (HF) bursts • Commercial aviation navigation (VLF) • PCA, magnetic storms, x-ray bursts • Electric power companies Magnetic storms • Long line telephone communication Magnetic storms " High altitude polar flights • Solar proton events radiation hazards • Civilian HF communication X-ray emission, U.V. emission, Coast Guard, GSA, commercial magnetic storms companies, VOA • Geophysical exploration Magnetic storms • Satellite orbital variation U.V. emission, magnetic storms military and civilian • DOD SATCOMcommunication • Magnetic storms • DOD HF communication • X-ray emission, U.V. emission, PCA, magnetic storms • DOD reconnaissance • PCA, magnetic storms • DOD navigation • X-ray emission, U.V. emission ERDA communication X-ray emission, U.V. emission : (prospective customers) magnetic storms International commun!ty • All • Scientific sateilit_ studies Optical solar flares, magnetic IMS, Solar Maximum mission, storms, X-ray emission, U.V. Shuttle, _ lal constant measure- emission, solar proton events, ments, stratospheric ozone vari- solar features ation, interplanetary missions • Scientific rocket studies • Optical solar flares, solar IMS, magnetosphere, ionosphere, features, magnetic storms, solar upper atmosphere, sun proton emission, X-ray emission • Scientific 9round studies • Optical solar flare, magnetic IMS, sun, interplanetary storms, solar proton emission, magnetosphere, ionosphere, upper X-ray emission, U.V. emission, atmosphere, stratosphere, tropo- _olar features sphere, seismological/geomagnetic 2. DATA SOURCES All of the observational data in the SESC is ]isted in Table 2. Though it is presumed that the STP workshop will not limit its coverage to prediction techniques that rely on currently available data, it is possible that some of the discussions may illuminate new and better uses of some of the current data, substitution of more appropriate data sets or elimination of some observations that began with more promise than fulfillment. It should be 323 I 1980010188-334 Table 2. Data Sources Wavelength or Time Primary Type Energy Renga Resolution Sources Solar Patrol X-rays I-B _, 0.5-4 _ I Minute Averages GOES Optical Hydrogen Alpha Continuous Several Redid 202-15,400 HHz I Hlnute Several Solar Synoptic Hydrogen Alpha Inubges Several per day Boulder White Light Images Several per day Boulder Ca K Images Daily Sac Peak He 10830 Imlges Daily Kltt Peak I_lgnetograms (Full Disk and Daily Kitt Peak, Huntsville Regional) Sunspot I_gnettc Fields Daily Ms. Wilson, Boulder Ca K Scaled Reports Daily McMath Sunspot Reports Several Daily Several Solar Mean Field Daily Stanford 10.7 Cm Radio Flux 2800 HHz 3 Times Oaily Ottawa (Algonquin) East-West Radio Drift Scan 2800 HHz Daily Ottawa (A19:_quin) Energetic Particle Patrol Synchronous Orbit Protons: 06-500 HeY I Minute A,,erages GOES Alphas: 4,0-329 HeV Electrons: _2 HeY Polar Orbit Protons 30 keY-850 ReV TIROS-N Magnetometer Patrol Synchronous Orbit 3 Components I Minute Averages GOES Terrestrial 3 Components I0 Seconds IMS North American Chain l0 Seconds Boulder 15 Minutes Air Force Real Time Chain 15 Minutes Thule 15 Minutes Vostok Hlscellaneous Neutron MOnitor 15 Minutes Thule High Latitude Riometers 30 RHz 15 Minutes Alaskan Chain, Thule Auroral Backscatter Radar 50 MHz IS Minutes Anchorage Interplanetary Scintillation 74 MHz Daily UCSD Ionosondes foF2, M3000, Fmln One hour or six hours Several noted that many of the data sets available to the SESC are provided due to the generosity and cooperative nature of the organizations that make them available. 3- DISTRIBUTION SYSTEMS In order to give STP Workshop participants an overall view of the SESC operation, a survey of the systems for distributing its produces is given in Table 3. 4. OBSERVED INDICES AND ACTIVITY SUMMARIES With the exception of research scientists conducting experiments, most : users of space environment services do not want a continuous stream of data. They typically ,refer indices or activity summaries that make it easier for them to relate solar and geophysical variations to problems in their operating systems. Though the topic of the STP Workshop is to study and recommend improved predictions, there are some instances where improved predictions may 324 1980010188-335 Table 3. Distribution Systems Telephone: • FTS (Federal Telephone Service) • WATS (Wide Area Telephone Service) • Commercial Telephone Service • Dedicated Telephone Lines (Hot Lines) • Recorded Information Numbers Teletype: • ATN (Astro-Geophysical Teletype Network) • AUTODIN (U.S. Government Teletype Service) : " Commercial Teletype Services • Secondary Networks Computer Links: • Space Environment Laboratory Data Acquisition and Display System (SELDADS) Public User Access Dedicated Data Links WWVShortwave Broadcast Mail depend on innovative indices that provide a more concise description of the environmental variations. The indices and summaries compiled in the SESC are summarized in Table 4 and described in more detail in this section. Table 4. Observed Indices and Activity Summaries Solar Active Region Summary Report Sunspot Number Flare (and Other Event) Lists : Solar Neutral Line Analysis and Synoptic Maps Ten-Centimeter Flux Solar Proton Events and Proton Flux SST Radiation Levels Geomagnetic A- and K-indices Substorm Log Sector Boundaries (at I A.U.) 4.1 Solar Active Region Summary Report Each day, a summary is compiled of all the solar active regions on the visible face of the sun. The summary includes all active regions that con- tain sunspots plus some additional spotless regions that already have or are believed capable of producing solar flares. The parameters listed for each 325 I J 1980010188-336 • _ region include the following: Boulder Region Number, location (Heliocentric Latitude and Longitude), Carrington longitude, sunspot area, modified Z_rich classification, sunspot count, and magnetic classification. All the parameter values assigned each day are a composite summary, made by the Boulder fore- caster, from the observatory reports received during the day. The accuracy in the locations reflects the variability in the observatory reports and is typically plus or minus five degrees (heliocentric) unless a special request for higher accuracy has been received in advance. 4.2 Sunspot Number Using the region summary, a sunspot number is computed each day. The number is unsmoothed and is subject to the considerable variability in the sunspot counts reported by the real time observatories. The numbers are published in the weekly "Preliminary Report and Forecast of Solar-Geophysical Activity" 4.3 Flare (and Other Event) Lists Each day, a running log is maintained in the SELDADS, of observed solar energetic events including flares (x-ray and optical) and solar radio bursts. The primary flare classif;cation system used is the C,M,X x-ray classes assigned from the SMS/GOES data although optical importance is also included where available. Spectral radio events (Types l-V) are also included in the events list. Non-flare optical events (e.g., erupting filaments, surges) are included although the reporting coverage of these types of activity is less complete. The primary SESC flare classification system is based on the peak flux @ • in Watts m-2 measured on the GOES satellite I-8 Angstrom ionization chamber. The classes are assigned as follows: Class X for I0-_ _ @, Class M for - IO-s _ @ < I0-4, and Class C for lO-6 _ @ < I0-s flux (Watts m-2). An addi- tional digit is added to the C, M, X to indicate the coefficient of the flux. For example, C8 implies 8 x I0-6, M3 implies 3 x lO-s, etc. 4.4 Solar Neutral Line and Synoptic Maps Each day, the SESC forecaster prepares a map of the visible face of the sun using procedures described by Mclntosh (1979). The map includes the location of all magnetic neutral lines (the line of separation between the inwardly and outwardly directed magnetic fields) at the solar surface. The neutral line infcFmation is compiled from solar magnetograms and inferred using the principles reviewed by Mclntosh where lines of magnetic reversal tend to be traced by filaments, by filament channels and plage corridors and where fibrils and arch-filament systems lie perpendicular to the neutral line. Magnetic field polarities are assigned using magnetograph observations and by applying the law of Hale for polarities of sunspots in active regions (leading spots positive in the northern hemisphere in cycle 21, leading spots negative in the southern hemisphere in cycle 21). Active regions are noted by their region number, and following the procedures described in the prediction section, flare probabilities are recorded for each region. The location of 326 • I _- :._.L_.... 1980010188-337 coronal holes, as indicated by x-ray images (when available), or inferred from the Helium 10830 images made at Kitt Peak, are outlined on the map. Host _ of the information on the daily "neutral line" map (excepting the flare probabilities) is extracted onto a Mercator projection map of the sun that covers ]60 degrees of heliographic longitude and latitudes from north 70 to south 70 degrees. This synoptic chart is published in the weekly "Preliminary Report and Forecast of Solar Geomagnetic Activity" and in the NOAA publication Solar Geophysical Data. i 4.5 Ten Centimeter Radio Flux t Since 1947, the solar flux at 2700 MHz has been measured daily by the ! Canadian National Research Council. Because of its relatively long history of careful observation, and because it gives at least a general estimation of solar radiation in EUV wavelengths, the ten centimeter flux has evolved as a well entrenched parameter that is used by a large number of radio communica- : tors and upper atmospheric modelers for a variety of operational applications. The SESC receives the "Ottawa" flux each day and redistributes it to the • _ operational community. In fact, th_s parameter is probably the most fre- quently requested and used parameter in the field of solar-geophysical meas- urements. It has been argued by some of the scientific community that the ' ten-centimeter flux is too gross a parameter to accurately evaluate the state of the ionosphere and upper atmosphere and that more specific measurements in the EUV would improve the operational models. However, the lack of a con- sistent and on-going EUV observational program and the expense and disruption of redoing present operational models probably precludes any rapid phase out : in the use of the models that operate with ten-centimeter Flux as an input. 4.6 Solar Proton Events and Proton Flux Solar proton flux is collected in real time from the NOAA SMS/GOES satellites at synchronous orbit and the NOAA TIROS-N satellite in polar orbit. The SffS/GOES data are processed in the SELDADS system to provide oqe-minute (or longer, if desired) (I) averages of proton fluxes above some specified energy levels (4, I0, 16, 40, 80 HEY), (2) spectral and flux parameters Ro and Jo from a least _Huares fit to an exponential-in-rigidity spectrum and (3) fluxes above arbitrarily selected energies by using the results of the minute-by-minute least squares fit. No processing beyond raw counting rate has yet been done with the TIROS-N data. To simplify communications with the user community, a proton event has been defined as an event where proton fluxes exceed IO protons cm--'c'Z-sec"l ster -1 for particles with energies _ greater than lO MeV. It should be noted that though the SMS/GOES satellites measure alpha particle fluxes over six energy ranges, these data have not yet been integrated into the proton processing and display programs. 4.7 SST Radiation Levels Using the Jo, RQ parameters calculated for proton events, an estimated radiation dose rate Is calculated for a hypothetical aircraft flying at 65,000 feet over the polar cap. The dose is calculated by using the transport [ equation o" Flamm and Lingerfalter (1964) as evaluated by Haurwitz (1971) with Jo and Ro ¢ the independent parameters. The accuracy of the calculation is 327 I 1980010188-338 , reduced by the necessity of fitting the parameters Jo and Ro to spectra for energies measured over the range i to 500 MeV (with poor observational resolu- ; tion toward the upper end) where the primary dose contribution originates. A newer series of particle sensors on future SMS/GOES satellites could improve the higher energy spectral fit but another limitation in the present system is that Haurwitz carried out the numerical integration for relatively modest values of RQ and much higher values have since been measured. As now done, the calculation gives an order-of-magnitude calculation for a quantity that can vary over five orders-of-magnitude. 4.8 Geomagnetic A and K Indices The A- and K-indices serve as the lowest common denominator for communi- cating with users whose systems are affected by varying levels of geomagnetic activity. Many scientific users would prefer to have a real time DST or AE for their use and SESC has, in response to some special requests, made real time estimates of DST using the low latitude stations in the IMS North American chain. Continual real time extrapolation of DST and AE, using the same IMS chain, have been discussed but the lack of longitudinal coverage and the uncertain future of the IMS chain has tended to slow the progress toward implementation. The problems with the A- and K-indices are of two kinds. The first is time resolution. The implicit limit in resolution in the present system is three hours for the real time K values (or 90 minutes from some of the Air Force stations). Field perturbations of a few minutes duration but large in magnitude and in rate of change are presumably of greatest concern to several users such as the electric power industry and the K-indices give only an approximate estimate of prob|ems for them. Secondly, the A- and K-indices are global in nature while the effects of magnetic disturbances can be confined to comparatively narrow longitudinal bands so that alerts based on K-indices are not universally applicable. Like the ten-centimeter flux index, a number of users have large invest- ments in models that use A- and K-indices for measuring geomagnetic activity. Unless, and until, someone redoes all the work that goes into the construc- tion of such models, these users are likely to continue to use A and K values despite the availability of more specific indices. There are some additional problems with the real time A- and K-indices. For many years, users tended to apply a real time K from a mid-latitude station such as Fredericksburg interchangeably with the planetary Ko values which are not available until some months after the fact. In fact_ Adams (1975) has shown that the Fredericksburg values vary systematically by up to 25 percent from the plan- etary values. The Air Force uses a real-time estimation of K and A from a set of about five stations, but the estimated K_ and A is frequently madeP artificially large by the effects of auroral electroje_s on their station sample. 4.9 Substorm Log One type of summary that provides positional information for geomagnetic disturbances is the Substorm Log. The log is compiled from the IMS North American chain and provides a continuous listing of substorm beginning times and longitudinal location. The catalog has attracted interest primarily among 328 $ 1980010188-339 scientific researchers although some operational type users have requested it. Essential shortcomings from the user point of view would probably be the problem ot" relating the substorm intensity to a threshold at which the user experiences problems and the failure of the equivalent current model, on which the catalog is based, at times of major geomagnetic storms. 4.10 Interplanetary Sector Boundaries Since the SESC receives no real time solar wind from in situ measurement and because there is a continuing interest among some users in monitoring the current status of sector boundaries in the interplanetary magnetic field, daily inferences of the polarity of the interplanetary magnetic field are made from the Z component of the magnetic field at Thule, Greenland, using the method of Svaigaard (1968) and of Mansurov and Mansurova (1969). : 5• ALERT CATEGORIES Some users of the SESC services find that the predictions as now made are not sufficiently accurate to use as a planning guide, but yet these customers are not interested in monitoring a full data stream at all times• For these users, an alert service is provided that allows them to receive a telephone call or teletype message when a particular kind of disturbance reaches a specific threshold. About one-third of all SESC users fall into this category of service, although a large number of them would prefer to rely on better predictions if such were available. Also, as pointed out in the section on geomagnetic indices, some of the indices are global in nature and the use of a general index at a single threshold typically means that some users are experiencing adverse effects when they receive an alert while others receive what amounts to a false alarm. A complete set of alert thresholds is given in Table 5. Table 5• SESC Alert Categories SOLAR FLARES '_ X5 (I-8A X-ray Classification) ._ Xl M5 3B (Optical Classification) 2B IB MAGNETI C OI STURBANCES • A >__50 (real time A measured at Boulder) • A>_30 • A_>20 K >__6 (real time K measured at Boulder) K 2_ 5 observed in successive three-hour intervals • K>5 329 r ,'¢ 1980010188-340 Table 5 (continued). • K>_4 • Sudden Commencement RADIO BURSTS/_ ISE STORHS • I0 CH Radio Burst Greater Than I00 Flux Units • 245 HHz or Noise Storm • Type II and/or Type IV Decametric Emission PROTON EVENTS Proton Flux (E > lO HeY) > lO cm-2sec'lsterad "l 6. PREDICTIONS All the predictions routinely produced by the SESC are included in this section. Each type is discussed separately so that the reader may refer to that section a]one. A summary is provided in Table 6. Table 6. Prediction Products (Lead Time Given in Parentheses) LONG TERN SOLAR ACTIVITY AND SOLAR RADIATION LEVELS • Smoothed sunspot number (1 month-iO years) • Geomagnetic activity and ten-centimeter flux (I month-lO years) • General level of solar activity (27 days) SOLAR ACTIVITY--SHORT TERN • Solar Flares (1, 2, 3 days) Solar proton events (i, 2, 3 days, PFP*) SOLAR RADIATION LEVELS -- SHORT TERN • Ten-centimeter flux (l, 2, 3 days) GEOMAGNETICDISTURBANCE LEVELS • A, K-indices (], 2, 3 days) Time of sudden commencements (PFP), storm size (PFP) PFP: Post Flare Prediction - A prediction of a flare consequence once the flare has occurred. 330 t 1980010188-341 6.1 Smoothed Sunspot Number The predictions are made in terms of the smoothed sunspot number, where the smoothing is done according to the standard procedure n[ o] Rn -_ Rn_6 + Rn+6 + _. 2R , (!) n-5 where R is the smoothed (over 13 months) number, R is the monthly average for month n. SESC cooperates with World Data Center A of the Environmental Data and Information Service (EDIS) of NOAA in issuing predictions of sunspot numbers. EDIS has, for several solar cycles, published predicted smoothed sunspot numbers, along with 95 percent confidence intervals, derived by the method of McNish and Lincoln (1949), for each month remaining in the cycle in progress. Sargent (1978), of the SESC staff, has published a prediction for the maximum sunspot number of cycle 21 along with an expected confidence . interval for that cycle. The two predictions for the maximum number are com- bined, along with the variance for each, to obtain a single value for maximum and a 95 percent confidence interval to go with it. The many published methods of sunspot prediction can generally be divided into four broad categories as follows: 1. Spectral Analysis: extracting important periodicities in the sunspot record and using these periodicities to extrapolate into the future is a popular method of making the predictions. The dilemma for the user is to choose among the highly variable, but apparently reputable, predictions. For example, the predictions for the present cycle, using spectral methods, have ranged from very small (Cohen and Lintz, 1974) to very large (Hill, 1977). For this reason, spectral methods have not been used in the NOAA predictions. 2. Planetary Influence Models: the published results from this method have also been highly variable and are not included in the NOAA predictions. 3. Comparison of the Rate of Rise of the Current Cycle with Previous Cycles: Figure l (from Sargent, ibid) illustrates the time rate of rise of past solar cycles for tilefirst few months following solar minimum. Some ordering of the cycles is apparent and several prediction methods take advantage of this ordering. The most widely used is the method of McNish and Lincoln (ibid). The McNish-Lincoln method accomplishes this statistically by computing an average of all the past cycles and comparing the current cycle to it. If the current cycle lles above or below the mean, that trend is carried forward into the prediction of the future, but it is balanced by a tendency of the predicted curve to move toward the mean. During the rising portion of the cycle, the McNish-Lincoln method is most accurate for the immediate future, with widening confidence intervals as the prediction is taken further and further in advance. This trend reverses later in the declining phase of the cycle when the prediction is for p,_riods when the sun- spot is decreasing. As pointed out by other authors (Lundquist and Vaughn, 1978),the principal problems with the McNish-Lincoln method are its 331 I I 1980010188-342 sensitivity to the selection of the time of minimum and the choice of the _om Sargent(1978) method of normalizing each cycle to its minimum. 150 4. The Hethod of Ohi: Originally suggested by Ohi (1976), the method has been studied by several authors, including Sargent (ibid). Though the o various authors have used different indices and techniques of data treat- ment, they have obtained relatively _ 100 consistent results. The Ohl method z 17(1192) statistically relates the recurrent geomagnetic activity at the end of one solar cycle with the peak sunspot intensity of the next cycle. Sargent_s results are shown in Figure 2. The _ 50 work is interesting because of the E possible implications in the associa- tion of the nature of the large scale i magnetic fields from one cycle to the I strength or coverage of the small M_thsofterMinimum scale sunspot fields in the next. 0_ l0 20 30 40 Figure I. The early phases of the 6.2 Long Range Geomagnetic Activity past eleven cycles and progress, to : and Ten Centimeter Predictions date, in Cycle 21. The predicted quantities are smoothed monthly averages, where the _ T T _ , , smoothing is the same ]3 month averag- _ 2_ _rgent(1978)Z ing that is done on sunspot numbers. The smoothed numbers are not neces- _ xj sarily what are needed by users, but are the limit of wha_. can presently be _ done. Even so, in the case of the _ 100 /x , )/geomagnetic activity the relationship ois tenuous. A plot of smoothed sun- _ |spot numbers versus ten centimeter flux is shown in Figure 3 (Slutz et o al., 1971). An analytical relation- ship that has been used in NOAA's _ 0 10_ 20A 30= _ 50 60 70 recent work is from Slowly (1979) and M_fi_ OhlVoriob_ is given by: Figure 2. Scatter plot showing the FlO.7 = 49.4 + 0.97 • R (2) relation of the modified Ohi variableto smoothed sunspot maxima for the past ni;,e sunspot cycles. The pre- + 17.6 exp (-0.035 • R) • dieted value for Cycle 21 is indica- ted and the 95_ confidence limits are shown. 332 I o 1980010188-343 250 I = f 6.3 General Level of Solar and Geo- ,: magnetic Activity (27 days)Slutz etal (1971) .,;' ' The ability to predict solar flares 200 i. and associated activity more than a few._ days in advance is marginal. There- '"." fore, the prediction categories are _O 150 j_, defined in general terms. For sola.___.__r , activity they are: Very Low m usually . _.'_.' * ,, x .:.. only quiet regions on the solar disk ._._." ., and no more than five of these; fewer .=. I_ _ than ten class C subflares without _ centimetric radio bursts or SID E observed or expected. Low - usually u more than five but less than ten quiet 50 regions on the solar disk; only class Q C subflares without centimetric radio bursts or SID observed or expected. ___ _ J Moderate _ eruptive regions on the 50 100 150 200 solar disk; fewer than five class M x- ZurichSunspotNumber ray events with centimetric radio bursts and SID observed or expected. Figure 3. Smoothed Mean IO.7-cm High _ active regions on solar disk; Solar Flux versus Smoothed Mean s=veral class M x-ray events with centimetric radio bursts and strongZurich Sunspot Number. SID; and/or one to two importance 2 chromospheric flares or class X x-ray events observed or expected. Very High _ region capable of producing protons on the sun; one or more chromospheric flares of importance 2 or greater with outstanding centimetric radio bursts (500 flux units or greater), class X x-ray bursts, and major SID observed or expected. For the Geomagnetic Field: Quiet - A < 7; Unsettled - 7 < Ap _ 15; Active - 15 < Ap _ 30; Minor Storm - 30 _ Ap Eo) = Fo e-P/Po, where F is flux, E is energy and P is rigi4ity. For phase b: the particle flux accelerated was estimated from the total (tlme integrated) energy of the flare in the radio spectrum at ten centimeters or from the total energy in the I to 8 Angstrom x-ray flux during the early part of the flare. An average spectral : parameter was assumed for all events and was taken as Po = 200 My. These quantities were estimated based on average values for flares where the earth, to first approximation, was directly connected to the flare location. Phase c: Burlaga (1967) described an anisotropic diffusion model to describe the propagation of the several hundred HeY protons which produce Ground Level Events. His solutions were especially useful in that the rise time to proton maximum (from flare time) and the scaling of the maximum particle intensity is correlated with the angle between the flare longitude and the connection point of the solar wind intercepting the earth. For the 30 HeY energies, a least squares fit was done from Explorer data published in Solar Geophysical Ddta (SGD, 1972). The spectral parameter at the peak of the event was found to have a similar dependence. An important feature of this model is the additional use of the equation for the flux (1) as a function of time (t) l(t) = A exp(S/t) t -5/2 (4) After differentiation, a linear relationship can be found between the flux at each point of time and the lapsed time from the flare maximum to that point. The slope of this line, determined from a least squares fit For each proton event while the event is stil! in progress, is the rise time to proton event ; maximum. Using this relationship, observations made early in the event can be used to obtain a more accurate prediction of the maximum proton flux yet to come. Experience has shown that this portion of the model works well for some events, but not so well for others. Phase d: in building real-time 342 I 1980010188-353 prediction models, the modeler must be careful to insure that the model be > computationally simple and quick and that it does not rely on parameters that are beyond the prediction capability of the user of the model. Estimation of the solar wind velocity for five to fifty hours in advance is an example of such a parameter. For this model, the forecaster assumes a single value that will be "typical" of the hours from the flare through the peak of the protnn event. The typical values are low solar wind velocity, implying a connection point near the west limb of the sun, moderate, with a connection about 45 to 50 degrees west and high, with a connection near central meridian. The results with this model, used operationally through the declining years of solar cycle 20, was that the peak flux, which can vary by five orders of magnitude, would typically be off by one order of magnitude. The time of rise, from flare to proton flux maximum, which can vary from around three to 100 hours, would be off by about 50 percent. The results of the model, as applied to the August, 1972 events, can be seen in Figure 10. This model is still used in a quick look version in the SESC. The Smart-Shea Proton Prediction Study: Smart and Shea (1978) have con- structed a model that provides a comprehensive treatment, especially of phases c and d. Their model is available in on-line computers at both the SESC and at GWC. In phase b of their model, Smart and Shea provide a variety of radio frequencies that can be used to estimate the size of the proton flux. However, they must rely on the same kind of correlative relationships that were done previously. Part of the SERF (Solar Energy Release in Flares) portion of the Solar Maximum Year is an observational program dedicated to I000 1I00 B E I .- _" "% 0.01 ; _--_- . . _ , .-T-_0' 0'6 6T 08 : August,1972 Figure IO. Predicted and extrapolated radiation doses in August, 1972. The predictions were mode on tl_e basis of the flare x-ray and radio energies. From McKinnon (1972). 343 t Z 1980010188-354 this problem, and if successful, may promise the first real progress in this !i phase in some time. In their treatment of phase c, Smart and Shea provide for two aspects of the particle propagation. First is the diffusion of the protons from the flare location through the solar corona to the magnetic field line connected to the earth. Second is the travel of the particles along the interplanetary magnetic field lines to the earth. The Smart and Shea approach appears to be a more realistic treatment of the propagation problem and may provid _ore consistency from event to event, Smart and Shea also assume a more comp,ex spectral form in order to obtain an improved match to the observed particle fluxes from low (one MeV) to high (severa] hundred MeV) energies. Smart and Shea handle phase d by allowing for continual co-rotation and varying solar wind velocity through the event. As pointed out in a number of recent papers, the coronal phase of the propagation is more complex than that modeled by assuming uniform coronal diffusion and escape. For example, Gold et al. (1977) discuss an instance where the primary particle :elease was above the flare site for two days, then apparently shifted to a location about lO0 degrees away. This effect probably accounts for much of the order-of-magnitude error in the flux pre- dictions. At present, there is no clear method of predicting the anomolous escape locations in advance. 7. VERIFICATION Verification of forecasts is a complex and controversial subject, and many of the methods used are misleading or inaccurate. Vertification of forecasts in the SESC are intended to serve four purposes: (I) Ascertain whether the subiective probability forecasts show skill in relation to other forecast methods. (2) Determine sources of error in the forecasts. (3) Determine the utility of the forecasts for operational customers of SESC. (4) Provide a baseline for evaluating the usefulness of future developments in statistical or physica] forecast n_dels. The most important restriction on verification systems is that they be "proper" (Murphy and Epstein, 1967) in the sense that the forecaster expresses his true be]ief in order to obtain the best verification score. A mathematical definition exists so that a scoring _ystem can be shown to be proper or not. An example of an improper scoring system is the one where the evaluation is done by simply counting all the correct forecasts and dividing by the total number issued--the percent of correct forecasts. Specifica]ly, if ten major flares occur per year, a forecaster should issue a "no" forecast every day and he will have a 97 per- cent score. For evaluating flare forecasts, SESC uses the Brier P-score (Brier, 1950) as modified by Sanders (1963). This system is mathematically proper. Let the verification score P be defined by I K M k)2P = _ _ Z: (fi - Oi (5) k=l i=l k 344 t 1980010188-355 where M is the number of forecasts made, K is the number of states or cate- gories on each forecast occasion, f is the forecast probability with range 0 to i in each category k (_f = I), and 0 is the observation and is 1 if the k event in category k occurs and is zero if it does not. If the forecaster is able to interpret c_=rrent conditions, he will forecast high probabilities when he expects an event and low when no event is expected. If he is correct in his evaluation, the difference (fik " Oik) will be small and the Brier P score will be small. If the forecaster is perfecLiy competent, he will have predicted _ik = 1 for every occasion when an event subsequently occurred and f:k = 0 when an event aid not follow. Then his Brier P score will be zero. I_ the forecaster is perfectly incompetent, Lhe difference (fi - Oi) will be ± I and his verification score will approach the value 1.0. In summary, the lower the score, the more skillful the forecaster. Sanders' modification is to divide the forecasts being evaluated into subsets according to the probability f. For each subset s, the score is partitioned into two terms (fs " Os) (where Os is the average for subset 5) and O (l-Os). Sanders called the first term "sorting" ability. As applied to solar flare forecasts, it is the forecaster's ability to estimate whether a region will produce a class C, M or X flare, but not when it will occur. The second term, which Sanders called "labeling" ability, measures how well the forecaster predicts the day of the flare. Figure 5 i11ustrates these two terms and shows that the forecaster, using current methods, can estimate the amount of energy stored in a flare, but he cannot predict the time of the : flare-- that is, he cannot anticipate the "flare trigger". REFERENCES Adams, H. A., III (1975): A comparison between Fredericksburg and planetary geomagnetic A values, National Oceanic and Atmospheric Administration Technical Report ERL 350-SEL 36. : Allen, C. W. (1944): Relation between magnetic storms and solar activity, Monthly Notices Roy. Astron. Soc., I04:13. Akasofu, S.-I., and S. Yoshida (1967): The structure of the plasma flow generated by solar flares, Planet. Space Sci., 15:39. Bartels, J. (]932): Terrestrial-magnetic activity and its relations to solar phenomena, Journ. Terr.. Magn. Aim. Electr., 37:1. Bell, B. (1965): Major flares and geomagnetic activity, Smithsonian Contr. As trophys., 5:69. Brier, G. W. (1950): Verification of forecasts expressed in terms of probability, Mon. Wea. Rev., 78:1. t Burlaga, L. F. (1967): Anisotropic diffusion of solar cosmic rays, J. Geo- phys. Res., 72:4449. 1980010188-356 i; Cohen, T, J., and P. R. Lintz (1974): Long term periodicities in the sunspot cycle, Nature, 250:398.k Cook, F. E. (1970): On the time delay between solar flares and associated geomagnetic storms. Paper 28 in AGARD Conference Proceedings, No. 49. Dodson, H. W., and E. R. Hedeman (1958): Geomagnetic disturbances associated with solar flares with major premaximum bursts at radio frequencies 200 MC/S. Journ. Geophys. Res., 63:77. Dodson, H. W., and E. R. Hedeman (:971): An experimental comprehensive flare index and its derivation for "major" flares, 1955-1969. UAG Report-14, U. S. Department of Commerce, Boulder, Colorado 80303, USA. Flamn, E. J., and R. E. Lingenfelter (1964): Neutron and proton dosage in the upper atmosphere from solar flare radiation, Science, 144:!556. Gardner, L. J. (1975): Geomagnetic forecasting based on EUV coronal intensity, NOAA Tech. Rep. ERL 320-5EL 33. Giovaneili, R. G. (1939): The relations between eruptions and sunspots, As trophys.. Journ., 89:555. Gold, R. E., E. P. Keath and E. C. Roelof (1977): Coronal structure of the April IO, 1969, solar flare particle event, in 15th International Cosmi_ Ray Conference, Conference Papers, Vol. 5. Harvey, J. W., and N. R. Sheeley, Jr. (1977): A comparison of He II 304A and , He I IO830A spectroheliegrams, Solar _hysics, 54:343. Haurwitz, M. (1971): Procedures of the Space Environment Services Center, U. S. Department of Commerce, Boulder, Colorado 80303, USA (Unpublished). Heckman, G. R. (1979): Verificatior. of the solar flare forecasts at the Space Environment Services Center for the years 1969-1971:. Technical Report (to appear), U. S. Department of Commerce, National Oceanic and Atmuspheric AdministrPtion. Hill, J. R. (1977): Long term solar activity forecasting using high-resolu- _ tion time spectral analysis, Nature, 266:i51. Hirman, J. W., and W. E. Flowers (1978): An objective approach to region analysis for flare forecasting. Proceedings of the Symposium on the Effect of the Ionosphere on Space Systems and Communications, Naval Research Laboratory, Washington, D. C. (in press). Horan, D. M, and R. W. Kreplin (1970): Predictions of solar flare activity based on solar x-ray energy flux measurements. Paper 28 in AGARD : Conference Proceedings, No. 49. -' Kildahi, K. (1979): Private communication. Lundquist, C. A., and W. W. Vaughn (1978): Private communication. 346 I 1980010188-357 ; w Mansurov, S. M., and L. G. Mansurova (1969): New evidence of a relationship between magnetic fields in space and on earth, Geomagn. Ae,ron., :622. Martes, M.-J., J. Rayrole, E. Ribes, M. Semel and I. Soru-Escaut (1974): On the importance of photospheric velocities in relation to flares, in Flare-Related Magnetlc Field Dynamics, Manu_cripts presented at a con- ference held in Boulder, Colorado, 23-25 September, 1974. Martin, S. F., and H. E. Ramsey (1972): Early recognition of major solar flares in H-alpha, in Solar Activity Observations and Predictions, Progress in Astronautics and Aeronautlc's 30, MIT Press, p. Mclntosh, P. S. (1969): Sunspots associated with the proton flare of 23 May 1967, in UAG Report 5, U. S. Department of Commerce, Boulder, Colorado 80303, USA. Mclntosh, P. S. (1972): Inference of solar magnetic polarities from H- alpha observations, in Solar Activity Observations and Predictions, Progress in Astronautics and Aeronautics 30, MIT Press, p. 65. Mclntosh, P. S. (1972-1979): Sunspot classification, in Solar Geophysical Data Descriptive Text, p. 25. Mclntosh, P. S. (1979): Annotated atlas of H_ synoptic charts for solar cycle 20, 1964-1974, UAG Report 14; U. S. Department of Commerce, Boulder, Colorado 80303, USA. McNish, A. G., and J. V. Lincoln (1949): Prediction of sunspot numbers, Transactions of the American Geophysical Union, 30:5. Murphey, A. H., and E. S. Epstein (1967): A note on probability forecasts and "Hedging", J. Appl. Meteor., 6:1002. ; Neidig, D. F., H. L. DeMastis, and P. H. Wiborg (1978): Flares, force- free fields, emerging flux and other phenomena in McMath 14943 (September,1977),AFGL-TR-78-O194. Neidig, D. F., H. L. DeMastus, and P. H. Wiborg (1978): Flares, force- free fields, emerging flux and other phenomena in llcMath 14943 (September, 1977), Air Force Geephysics Laboratory Environmental Research Paper, No. 637. Neupert, W. M., and V. Pizzo (I974): Solar cGronal holes as sources of recurrent geomagnetic disturbances, J. Geophys. Res., 79:3701. Newton, H. W. (1943): Solar flares and magnetic storms, Monthly Notices Roy. Astron. Soc., 90:820. Nolte, J. T., A. S. Krieger, A. F. Timothy, R. E. Gold, E. C. Roelof, G. Vaiana, A. J. Lazarus, J. D. Sullivan and P. S. 14clntosh (1976): Coronal holes as sources of solar wind, Solar Phys., 46:303. 3_7 t 1980010188-358 Oh1, A. I. (1976): A preilmlnary forecast of some parameters of cycle no. 21 of the solar activity, Soinechnaya Dannyye No. 9, P. 73. Posialdlo, R. T. (1973): An objective baseline for flare prediction, C0SPAR Symposium 15, Academle-Verlag, Berlin, p. 817. Robbins, D. E. (1964): Apollo shielding analysis, in Second Symposium on Protection Against Radiations in Space, NASA SP-71. Sakurai, K. (1967): Solar cosmic-ray flares and related sunspot magnetic fields, R_e_t. Ionospheric and Space Res , Japan, 21:113. Sanders, F. (1963): On subjective probability forecasting, d..AppI. Meteor., 2:191. Sargent, H. H., III (1978): A prediction for the next solar cycle, 28th IEEE Vehicu]ar Technology Conference, p. 490. Sargent, H. H., III (1979): A geomagnetic activity recurrence index, Solar- Terrestrial Influences on Weather and _limate, Reidei, Dordrecht, Holland, p. 101. Severny, A. B. (1963): On the behavior of ma3netic fields associated with solar flares, AAS-NASA Symposium on the Fh_sics of Solar Flares, NASA SP-50. SGD (1972): Solar Geophysical Data, No. 330 (suoplement), U. S. Department of Commerce, Boulder, Colorado. Sheeley, N. R., dr., d. W. Harvey, and W. C. Feldman (1976): Coronal holes, solar wind streams and recurrent geomagnetic disturbances: 1973-1976, Sol. Phys., 49:271. Sheeley, N. R., Jr., and d. W. Harvey (1978): Coronal holes, solar wind streams, and geomagnetic activity during the new sunspot cycle, Naval Research Laboratory Skylab/ATM Preprint. Siowley, E. (1979): Private communication. Slutz, R. J., T. B. Gray, M. L. West, F. G. Stewart, and M. Lu, L:.. iI Solar activity prediction, NASA CR-1939. Smart, D., and H. Shea (1978): Private communication. Smith, J. B., Jr. (1972): Predicting activity levels for specific I¢_cations within solar active regions, in Solar Activity Observations and Predic- tions, Progress in Astronautics and Aeronautics, Vol. 30, MIT Press, Smith, S. F., and R. Howard (1968): Magnetic classification of active regions, in Structure and Development of Solar Active Re_lions, iAU S_,mposium No. 35, Reidel, Dordrecht. 348 1980010188-359 Sutorik, J. (1978): Procedures of the Space Environment Services CenLer, U. S. Department of Commerce, Boulder, Co.ado (Unpublished). Svalgaard, L. (1968): Sector structure of the interplanetary magnetic field and the daily varlatio, of the geomagnetic field 3t high latitudes, _eophys. Papers, R-6, Danish Heteorol. Inst., Copenhagen. Svestka, Z. (1972): Several" solar aspects of flare-associated particle events, in Solar Activity Observations and Predictions, Progress in Astronautics and Aeronautics, Vol. JO, HIT Press, p..41. : Tanaka, K., and Y. Nakagawa (1973): Force-free magnetic fields and flares of August, 1972, Sol. Ph_ 3_ 187. 349 L- 1980010188-360 - 18'0 "18 49 O; :'. GEOPHYSICAL FORECASTING AT AFGWC Richard L. Thompson _ Lt Col USAF and James A. Secan Capt USAF Air Force Global Weather Central* Offutt Air Force Base, Nebraska, USA *mailing address- AFGWC/DOX, Offutt AFB, NE 68113, USA : Advanced systems that either use or are affected by the environment above 50 kilometers require forecast support. The Air Weather Service provides a worldwide network of sensors and a central facility to monitor and forecast the state of the space environment; the sun, interplanetary field, magnetosphere and ionosphere. This paper discusses forecast methods and problem areas requiring additional advances in scientific understanding to support ever increasing operational requirements. i. INTRODUCTION The Air Weather Service (AWS), _hrough its operational forecast centers of the Air Force Global Weather Central (AFGWC), provides space environmental support to the entire Department of Defense. Although the types and intensity of support are varied, the overall driving requir_aent is to minimize system effects caused by impulsive solar/ geophysical activity and ionospheric variations. The knowledge of these effects, preferably beforehand, provides the decision maker with information to utilize his resources effectively. AFGWC ° provides around-the-clock service in forecasting and specifying the aerospace environment by applying varied data to the problem. The Air ; Force has been active in space environment research and forecasting for over a decade. Many earlier papers (Snyder, 1974; yon Flotow, 1978; Thompson, 1978) have described data sources and techniques used to provide : real-time and post-analysis support. In this paper, we will describe in greater detail the method and problems of real-time space environmental forecasting and specification. Although the ionosphere will be the area of greatest concentration, uhe areas of magnetospheric and energetic partice forecasting will be addressed. 3so 1980010188-361 2. HISTORY/REQUIREMENTS The Air Force has been active in space research for many years. ! Most of this effort has either been completed or directed by the Air Force Geophysical Laboratory (AFGL). Through the cooperation of AWS and AFGL, a worldwide network of observatories was established ' to observe solar activity during the maximum years of solar cycle 20. _ This network helped to support national requirements during the Apollo _ _ and Skylab missions and direct support to operational decision-makers concerned with environmental effects on electronic systems. We are again prepared to support both national and defense oriented missions during the maximum years of solar cycle 21. Two major advances in observational equipment are the deploy- _ ment of the new solar observing optical network (SOON) and radio solar : telescope network (RSTN). These systems are partially in place and we expect to complete deployment during 1980. Both systems employ on-site 4 automation to provide AFGWC with reliable data in minimum time. During solar cycle 20 we were able to provide only limited real-time information that a flare was in progress. With the data from SOON, RSTN, and X-ray data from the Geosynchronous Operational Environmental Satellites (GOES) we can provide a real-time comprehensive analysis of the flare and its effects on the space environment. These data can also be used to predict, with reasonable confidence, high energy solar particle events. The requirement for reliable predictions of high-energy solar events was a primary driving force behind the installation of the SOON/RSTN. In the late sixties, AWS installed a series of real-time magnetometers to observe and forecast the state of the magnetosphere. This network allows real-time specification of a pseudo Ap. This _ data coupled with the 10.7 cm solar radio flux remain the main inputs to upper atmosphere neutral density _ : models. However, new requirements concerning the interaction of spacecraft and the magnetosphere dictate a more comprehensive understanding and observational database to provide effective support. The importance of the ionosphere to AFGWC supported operations increased significantly during the solar minimum years of the mid seventies. AWS has been active in ionospheric prediction and specifi- cation since 1967, but not at the present magnitude. Early efforts were generally manual analyses for special cases of limited areas, both in time and space. However, as technology advanced, previously neglected errors caused by the ionosphere reached parity with other : error sources. Interest in reducing these ionospheric errors placed new demands on AFGWC for more accurate analyses and forecasts. It i soon became apparent that manual methods would not be able to satisfy . increased demands. A major break-through occurred with the development and implementation of the Flattery-Ramsay (1975) model which coupled the data handlLng and processing techniques of meteorology with observa- ? 351 [ 1980010188-362 tions of the ionosphere. This permitted a hemispheric real-time update of the Institute of Telecommunication Services (ITS) ionospheric coefficients. This model was continuous only in three dimensions (latitude, longitude, and time) which was an operational limitation. This limitation led to preliminary development of the four dimensional ionospheric (4D} model, continuous in latitude, longitude, time, and height, with great potential for using a "build and apply" concept. Using the "build and apply" approach, we build the most comprehensive specificationforecast of the ionosphere, then apply this data base to generate automated products needed to satisfy a variety of customer requirements. 3. CURRENT FORECASTING 3.1 Solar Flare Particle Emissions : The forecasting and observation of high energy solar protons remains one of the highest priority functions within the Space Environmental Support System (SESS). Solar flare-produced protons can cause damage to satellites operating in the near-earth space environment and are an ever present danger to people in space. Rapid gathering and processing of flare data are essential in providing useful forecast services. High speed communication links between world-wide observatories and AFGWC, such as the Automated Weather Network (AWN), supply the rapid transfer of data. Automation at both the observatories and AFGWC aids in rapid data reduction. The data used in proton prediction and observation come from a myriad of ground-based and satellite-borne sensors. Most of the ground-based data is from SOON and RSTN currently being deployed by AWS. Table 3.1 gives the locat_on, data types available, and Status of the various sites in the SOON/RSTN and other sites which provide data to AFGWC. While these ground-based observations are • used primarily as inputs to various proton prediction schemes, satellite observations ate used for direct proton observations as well. Most of the satellite observations used at AFGWC come from the National Oceanic and Atmospheric Administration (NOAA) GOES satellites. These data, which include solar x-ray and high-energy proton observations, : are collected at the NOAA in Boulder and transmitted to AFGWC in real-time via a dedicated computer-to-computer data line. A small amount of solar x-ray and particle data is still received from the VELA 5 and 5B satellites, but this data source has deteriorated significantly over the last several years and its use is scheduled to be terminated at AFGWC in the near future. Other operational Department of Defense satellites provide nearly continuous proton/electronn data. Both ground-based and satellite data are transmitted to AFGWC in real-time via several data links, such as the AWN and NOAA-AFGWC lines. Once at AFGWC, these data are processed in real-time and made avallable for preparation of various prediction and observation products for AFGWC customers. _ 352 7 & 1980010188-363 i _ X X X X • x X x X x X x X x X r X X X X X X X X X X X X H X X X X X _ X X X X W 0 X X x x x 0 A W Z 0 X X X X X A H Z _" 0 0 N X W W _ A _ _ _ W W _ 3 W A _ r o 0 _. 353 $ 1980010188-364 The AFGWC prediction of solar proton events is divided into two areas: (i) Did a given flare produce _rotons? (Yes/No forecast, qualitative) (2) If protons are produced, how will the particle spectrum grow and decay near the earth as a function of time? (particle flux forecast, quantitative) Both of these forecasts are time critical; the yes/no qualitative i forecast must be transmitted within i0 minutes of flare maximum, the particle flux forecast within 30 minutes of maximum. The yes/no forecast is primarily a subjective one, based on flare signatures such as parallel ribbons on the H-alpha flare, the Castelli U-shaped radio burst, the type-II radio burst, and many others. For this forecast, the computer acts as a data collector, with the forecaster providing the i actual _'orecast intelligence. The particle flux forecast is done almost entirely by the computer. After the incoming data has been "5 processed, the Smart-Shea Proton Prediction Study (PPS) program is run to produce the forecast Further details on the yes/no forecast techniques used and the Smart-Shea PPS program can be found in Smart and Shea (1977 a, b) and Cliver, et al., (1978). Although AWS SESS has recently made several improvements in ITS proton prediction capability, most notably the SEON data and the Smart-Shea PPS program, there are still many problems and shortcomings. ' In the area of yes/no prediction the most frustrating problem is the sub- jective nature of these forecasts. Almost every worker in this field i has his own list of proton signatures and while most all lists include roughly the same signatures, the emphasis is often different. This lack of agreement within the scientific community on the importance of : various signatures coupled with the often ambiguous collection of signatures from a given event (parallel ribbon flares without a Castelli-U burst, etc.) makes it very difficult to produce a reasonable yes/no forecast within the required timelines. Ambiguity is also a problem with the Smart-Shea PPS Program. Different input parameters often produce very different forecasts. While this problem can be often resolved by forecaster experience and expertise, a far more serious problem is the lack of interplanetary medium observations. Without up-to-date knowledge of the structure of the solar wind between the sun and the earth, only a rough estimate can be made of the propagation of high energy protons from their source to the earth. This can seriously affect the timing and intensity of the proton event at the earth. Finally, flares which occur behind the west limb of the sun, producing at best only weak radio signatures are a problem. 3.2 Magnetospheric Variations Magnetospheric forecasting at AFGWC is still in its infancy. Real-time observations of geomagnetic indices and magnetospheric protons and electrons are only the beginning of a comprehensive 354 t _ 1980010188-365 i ...... _n_ forecasting and specification scheme. These observations are currently used to give the forecaster a subjective "feel" for the current state of the magnetosphere rather than to drive specification and prediction models. As in the solar proton area, magnetospheric data are collected by a number of ground-based and satellite sensors. The majority i of the ground-based data consists of magnetometer observations. These data come from a network of magnetometers operated in near real-time for AWS and from various magnetometers located around the world which I provide geomagnetic indices 12-36 hours after the fact (table 3.2) A wide range of data is collected by the NOAA/AWS High Latitude Monitoring Station (HLMS) and transmitted every 15 minutes to i AFGWC. A partial list of these data includes magnetometer observations,riometer observations, and auroral backscatter radar measurements. These ! ground observations are complemented by observations in two different magnetospheric regimes, the geostationary satellites and the low altitude i (840 KM) polar orbiting Defense Meteorological Satellite Program (DMSP) satellites. Table 3.3 lists the data available from the GOES andEP (1976-059A and 1977-007A) satellites. The DMSP satellites provide ; visual observations of the aurora (Snyder, et al., 1974) and precipitating electron data (pitch angle = 0o) in the energy range 50 eV to 20 KeV from the SSJ/3 sensor. The GOES and EP data are transmitted to AFGWC in real-time via high speed data lines; the DMSP data is collected and processed on a nearly rev-by-rev basis at AFGWC. Gross magnetospheric variations are monitored using a pseudo ap or Kp index calculated using a subset of the ground-based magnetometers indicated in table 3.2. This index, calculated within 30 minutes of the end of each three hour period, can be used as an input to models that require an integrated measure of global geomagnetic activity. The 24 hour Ap index is forecast daily for the upcoming seven days using a variety of largely subjective techniques. The K_ calculated in real-time normally has a mean difference of roughly one K_ unit from that by the Geophysikalisches Institut at the University of_ottingen, F.R. of Germany (Snyder, 1974). Auroral substorm activity is monitored using DMSP visual and SSJ/3 data. These data are used to calculate two indices of auroral activity; • an effective Q index, Qe (Snyder, 1974); and Ca, a subjective measure i : of the level of auroral activity on a given DMSP photograph (Pike, 1975). The Qe index is derived from the equatorial boundary of the diffuse aurora on the DMSP photographs or the equatorward point where auroral electrons are detected on the 0.652 KeY channel of the SSJ/3 sensor, mapped down to IOOKM. Qe is calculated from these measurements using a variation of the technique discussed by Gassman (1973), Feldstein and Starkov (1967), and Starkov (1969). This measure of the size of the auroral oval is used with Ca to monitor substorm activity. 355 J 1980010188-366 GEOGRAPHIC GEOMAGNETIC OBSERVATORY LATITUDE LONGITUDE LATITUDE LONGITUDE THULE* 77N 69W +87 + 38 KRENKEL 80N 58E +75 +144 DIKSON 73N 80E +68 +154 TIXIE 71N 129E +66 +196 GOOSE BAY* 54N 60W +65 + 21 _ COLLEGE* 65N 148W +65 +260 MURMANSK 68N 33E +64 +114 ANCHORAGE 61N 150W +61 +261 LORING AFB* 48N 68W +60 + 8 TUNGUSKA 60N 90E +56 +161 LENINGRAD 60N 30E +56 +107 i MAGADAN 60N 151E +54 +217 SVERDLOVSK 56N 61E +52 +132T FREDRICKSBURG 38N 77W +51 +353 MOSCOW 55N 37E +51 +111 WINGST 54N 9E +51 + 88 BOULDER* 40N 105W +50 +316 IRKUTSK 52N 104E +48 +175 PETROPAVLOVSK 53N 158E _47 +225 CHAMBON 48N 2E +46 + 81 KAKIOKA 36N 140E +30 +210 MANILA 15N 121E + 9 +190 KERGUELEN 49S 70E -58 +121 • *RECEIVED AND PROCESSED IN REAL-TIME TABLE 3.2 MAGNETOMETER DATA NETWORK ., 356 g 1980010188-367 GEOSYNCHRONOUS SATELLITE DATA A. GOES DATA PROTON ELECTRON ALPHA 0.8 - 4.0 MEV 2.0 MEV 8.2 - 10.0 MEV 4.0 - 8.0 MEV 10.0 - 16.0 MEV 8.0 - 16.0 MEV 16.0 - 60.0 MEV 16.0 - 215.0 MEV 85.0 - 182.0 MEV 36.0 - 215.0 MEV 156.0 - 228.0 MEV 80.0 - 215.0 MEV 326.0 - 412.0 MEV 215.0 - 500.0 MEV : B. EP DATA PROTON ELECTRON 50 - 500 KEV 0.30 - 0.40 MEV 30 - 300 KEV 0.20 -2.0 MEV 63 - 500 KEV 0.40 - 0.53 MEV 44 - 300 KEV 0.20 - 2.0 MEV 80 - 500 KEV 0.53 - 0.71 MEV 64 - 300 KEV 0.43 - 2.0 MEV 100 - 500 KEV 0.71 - 0.94 MEV 95 - 300 KEV 0.63 - 2.0 MEV 126 - 50U KEV 0.94 - 1.25 MEV 139 - 300 KEV 0.93 - 2.0 MEV 158 - 500 KEV 1.25 - 1.66 MEV 204 - 300 KEV 1.40 - 2.0 MEV 198 - 500 KEV 1.66 - 2.20 MEV 252 - 500 KEV 2.20 - 4.73 MEV 316 - 500 KEV 4.73 - 8.00 MEV 396 - 500 KEV 8.00 - 13.50 MEV 13.50 - 22.80 MEV 22.80 - 33.20 MEV 33.20 - 48.40 MEV 48.40 - 70.60 MEV 70.60 - 103.00 MEV 103.00 - 150.00 MEV TABLE 3.3 GOES AND EP DATA 357 1980010188-368 There are many problems encountered in forecasting magnetospheric activity beyond the simple seven day Ap forecast mentioned earlier. The main difficulty is forecasting the occurrence and duration of sub- storm activity prior to onset. Most of the methods of short term fore- casting of substorm onset (lead times < 1 hour) depend on solar wind measurements taken "upwind" of the earth's magnetopause. With the gradual failure of the VELA solar wind sensors, AFGWC has become _ssentially blind to structures in the solar wind which can cause substorms. This blindness affects our ability to observe newly formed fast stream structures in the solar wind which could be used by the fcrecast center to warn customers of an upcoming disturbance. Until the problem of forecasting substorms is better understood, and the data to make such forecasts available, magnetospheric forecasting at AFGWC will remain in its infancy. 3.3 Ionospheric Variations During the time between the maximum of solar cycle 20 and the start of cycle 21, AWS SESS has seen its involvement in ionospheric predic- tion and specification grow from reporting flare-induced ionosphere variations (SWF's and PCA's) to modeling and forecasting large scale ionosphere behavior. The remainder of this section will deal with techniques used in forecasting and specifying ionospheric fields. Details on the products produced from the ionospherlc parameters drawn from these fields, such as radio propagation forecasts, can be found in Thompson (1978) and yon Flotow (1978) and will not be covered here. The data used in current ionospheric work was also covered in detail in yon Flotow (1978), Thompson (1978), and Snyder (1974), and will be only briefly discussed here. Table 3.4 summarizes the ground-based and satellite (DMSP) ionospheric data available at AFGWC. A variety of methods and models are used at AEGWC to forecast and specify the ionosphere, but generally they all follow the same lines. An automated model is used to combine observations and climatology to provide a "best guess" specification of the ionosphere. The forecaster then manually, or with the aid of various computer software "tools", makes any modifications required to fill in data sparse areas of a specification. He also makes changes he believes will occur during the forecast period which are not reflected in the computer produced analysis. Currently, two models are used to provide the "best guess" base field; the Flattery-Ramsay Hybrid Model (Flattery and Ramsay, i_75), and the new 4D model (Tascione, et al, 1978). Details of these moJels can be found in the given references; here we will discuss how the_ models are used and indicate the advantages and limitations of each. 3.3.1 The Flattery-Ramsay Hybrid Model This model was developed at AFGWC in the early 1970s (Flattery and Ramsay, 1975) and is still being used operationally by SESS forecasters. It is a two stage model in which input data (foF2 and M3000) 358 1980010188-369 k. VERTICAL IONOSONDES OBSERVATORY LATITUDE LONGITUDE OBSERVATORY LATITUDE LONGITUDE MANILA 14N 121E MAGADAN 59N 151E DJIBOUTI 12N 43N OTTAWA 45N 76W OUAGADOUGOU 12N 1W ST. JOHNS 47N 52W TAIPEI 25N 121E SVWEDLOVSK 56N 61E DAKAR 15N 17W KENORA 50N 97W MAUl 21N 156W TOMSK 56N 85E KOKUBUNJI 36N 133E GOOSE BAY 53N 60W ASHKHABAD 40N 58E YAKUTSK 62N 130W VANDENBERG AFB 35N 121W COLLEGE 65N 147W ALMA ATA 43N 77E CAPE SCHMIDT 69N 179W PATRICK AFB 29N 80W KIRUNA 68N 20E POITIERS 46N 00W MURMANSK 69N 33E LANNION 48N 03W TUNGUSKA 61N 90E KHABAROBSK 48N 135E CHURCHILL 58N 94W BOULDER 40N 105W NARSSARSUAK 61N 45W PETROPAVLOVSK 53N 159E SALEKHARD 67N 67W WALLOPS IS 37N 75W DISON 73N 80E KIEV 51N 31E KRENDEL 80N 58E IRKUTSK 52N 104E GODHAVEN 69N 54W MOSCOW 55N 32E RESOLUTE BAY 75N 95W DOURBES 50N 5E QANAQ 76N 68W LINDAU 51N 10E SLOUGH 51N 1W FARADAY POLARIMETERS OBSERVATORY LATITUDE LONGITUDE OBSERVATORY LATITUDE LONGITUDE ATHENS 34N 15W SAGAMORE HILL 39N 71W GOOSE BAY 48N 62W BOULDER 37N 105W OSAN AFB 35N 128E TAIWAN 23N 122E PATRICK AFB 27N 80W PALEHUA 20N 157W RAMEY AFB 17N 70W SHEMYA AFB 49N 179W DMSP SPECIAL SENSORS SENSOR DESCRIPTION SSJ/3 PRECIPITATINGELECTRONS {16 CHANNELS, 6EV-20KEV} SSIE IN-SITU PLASMA PROBE {NE, TE, NI, TI, MI} SSIP PASSIVE FOF2 MONITOR {NOISE BREAKTHROUGH} TABLE 3.4 IONOSPHERIC DATA 353 1980010188-370 are first meshed with the ITS-78 ionospheric climatology (Barghausen, et al 1969j to produce updated fields of foF2 and M3000. These fields are passed to an electron density profile (EDP) model, a modified version of the Damon and Hartranft model (Damon and Hartranft, 1970), to produce ionospheric variations in the vertical, Z, coordinate. The type of input data will determine the nature of the output fields: single day observa- tions of foF2 and M3000 provide a specification; five-day means of these same parameters provide the base "best guess" field for a forecast. As an example of the use of this model, let us suppose the forecaster wishes to produce an EDP forecast over North America on 5° x 5° latitude/ longitude grid. The first stage of the model is run on five-day means of whatever foF2/M3000 observations are available. This stage is accomplished in two phases. First, the sunspot number (SSN) used as input to the ITS-78 model is varied to provide the best overall fit of the ITS-78 foF2 to the five day mean foF2 values (Thompson, 1978). Then the foF2/M3000 fields produced using this SSN are meshed, with the mean observed foF2 and M3000 values to produce an updated set of coefficients in ITS-78 format. This coefficient set is then used to produce a grid of foF2 and M3000 for the forecaster to modify as he requires. This modified grid is then passed to the second state, EDP model, to produce the required profile forecast. The major advantage of this model is that it provides a global iono- sphere based on actual observations. It is used to give a single hour fore- cast of foF2 at a single point, or a global grid of total electron content (TEC), depending on the forecaster's need. The fact that the output of the first stage of the model is a set of coefficients identical in format to those used by ITS-78 is another major positive attribute of this model. Any computer software which uses the ITS-78 ionosphere, ._uch as the HF propagation program for which the ITS-78 climatology was developed (Barghausen, et al., 1969), can use the output of this model. This adds what is essentially a "real-time" capability to such software. There are four major deficiencies with this modeling scheme: (i) It is mathematically continuous in only three dimensions (latitude, longitude, and time); (2) It is computationally clumsy and slow; (3) It cannot handle strong spatial and temporal gradients such as the high latitude trough; and (4) It can use only foF2 and M3000 as input observations. The combination of these deficiencies, particularly the last two, led to the development of what has become known as the 4D ionospheric model. 3.3.2 Four Dimensional (4D) Ionospheric Model The 4D model was developed in the mid-1970s to remedy some of the problems of the Flattery-Ramsay model (Flattery and Davenport, 1977; Tascione, et al., 1978). All four dimensions (latitude, longitude, height, 360 t 1980010188-371 and time) are incorporated within _ single mathematical framework which is computationally far more efficient than the earlier model. The 4D model will accept a variety of observations as input data. It is curreptly con- figured to use foF2, M3000 (or Hmax), TEC, and electron density (Ne) or plasma scale height (He) at any level in the topside ionosphere. This last feature allows the use of Ne and He data at 840 km from the DMSP SSIS plasma probe. Finally, the ability to fit sharp gzadients can be controlled through the number of wave numbers used in the latitude/ lol_gitude/time spectral analysis of the model. As with the Flattery-Ramsay model, this model runs in two stages, but both stages are within the same program. All input data are first used to produce density profiles using a modified version of the Damon-Hartranft EDP mode]. _ach profile is reduced to four coefficients by fitting a set of climatological height eigenvectors to the profile (Tascione, et al., 1978). These coefficients are then analyzed and merged into a climatological onalysis to provide the final output of the model. The coefficients output by the model can then be used to produce any macroscopic ionospheric parameter desired. This model would be used as follows to prov:de the forecast discussed earlier in section 3.3.2. All available _ata, including DMSP observations, are input to the 4D. After the model provides the "best guess" field grid the forecaster can make modifications to whatever parameter desired, including electron density at any level. The modified grids are then passed back to the 4D which provides an internally consistent analysis based o,i the forecaster's modifications. The forecaster can then use this analysis to produce the desired EDPs, or run through the modification loop again to produce another set of 4D analysis coefficients. Figure 3.1 shows the process schematically. The two primary advantages of the 4D model are its ability to use a wide spectrum of input data and the flexibility that a system built around the 4D model gives the ionospheric forecaster. The example given dbove shows only a fraction of the options available. The f recaster can selectively delete or modify data initially input to th_ 4D, use different base climatology fields including the field produced by a prior run of the 4D, and can pass a manually produced analysis of any input parameter to the 4D for analysis. For all its flexibility the current version of the 4D model has some problem areas. The model is data starved, i.e., there is not enough data currently available to take advantage of the profiling and gradient reproducing features of the model. This problem can be dealt with by acquiring more data, such as the new SSIE and SSIP (Rush, et al., 1978) sensors on the DMSP satellite, or by using the forecaster's inputs to "spread" the available data in latitude and longitude. A second problem, which is proving more difficult to address, concerns the height eigenvectors. The currcnt set was derived from a collection of EDP observations from the 361 ) 1980010188-372 TA I,,,'1 • I,_ k-,I,,,,,_VJ I,_ _1)-. I_ ,,_ I,,_ ,,.,.I O ,, Interconnect the sets of | (al,a2,a3,a_)N with surfaceFspherical harmonics to; develop s_ectral coefficientset (A,B,C,D) 4, Use (A,B,C,D)to generate (al,a2,aya_) N at any arbitrary location K Figure 2. Program logic flow chart. 370 1980010188-381 Z - h - hmax (2) hs where h is the height, hmax is the height %-_ of Ne max, and hs is the scale height. -,,, The E layer hmax is fixed at 120 km. _ -_,. The F2 layer hmax (HMAX) is computed via i "_'" a modified version of Shimazaki's equation p "_ (1955) from observed M3000. The F1 layer I"'_-. hmax is allowed to vary in a predetermined I "-......... _ 3 illustrates an electron density profile ..-- which is the sum of the E layer, F1 layer, .._ and F2 layer Chapman distributions. The " ...... next step is to reduce an electron density _aa_muNm_, profile (computed via equation i) to a Figure 3. Electron density pro- set of weighting coefficients, file as the sum of the E, F1, and F2 Chapman layer contributions. In gneral, at any height (h), an electron density profile can be approx- imated by a partial sum of orthonormal basis functions and an appropriate set of weighting coefficients. That is: N Ne(_) = [ akWk(_ } (3) k=l where Wk(_) represents an empirically derived set of discrete orthonormal basis functions evaluated at level _ and the ak's are weighting coefficients. i These basis funcitons are provided as input variables to the model.Presently 127 levels span the physical distance between 95 and 2000km. The interval spacing between each level is: ? 5 km for level 1 through 81 (95-500 km); I I I I/ i\ I I I . 20 km for levels 82 through i06 (520- - _._/ \_I1____ [ i000 km); and 50 km for levels 107 i through 127 (1050-2000 km). A detailed discussion of the derivation of a suit- • able empirical orthonormal basis set is / given in a paper by Flattery, et al. :sm \ ! (1979). Presently, we only use the first .m_; /four terms of the partial sum given by k \ i equation 3. This choice represents a ___ trade-off between model accuracy and ___._.___. computer run time The sum can easily be extended as computer resources permit and accuracy demands. Figure 4 is an / ---- /" _. example of the first three empirical _L I ....I ... lJ I ,_/ I _ ' _ orthonormal functions which were derived ._ ._ ._ i _i from Millstone _ill backscatuer radar i I observations. Figure 4. Empirical orthogonal functions. 371 1980010188-382 r_ t, The next step is to determine the values for the weight coefficients, ak• Currently we evaluate (using equation 1) the electron density at i0 different heights= 50 km below HMAX; 20 km below HMAX; at HMAX; 20, 50, 100 and 200 km above HMAX; 740 km; 840 km (DMSP height); and 940 km. These points maximlze the profile shape definition with a minimun of points. There is nothing unique about the points selectedl the number and locatlon of points can be changed as required• Whenever possible, real observations are inserted at HMAX and 840 km• Therefore, from equation 3, we have ten simultaneous equations: Ne(50 below) = alWl(50 below) + a2W2(50 below) + .... + a4W4(50 below) .... (4) Ne (940) - alW 1 (940) + a2W 2 (940) + .... + a4W 4 (940) From this set of equations we can find a least squares fat solution of aI, a2, a3, a4. Once the weighting coefficients are known, we can compute the TEC by using: 2000 4DTEC = J Ne(h) dh95 -or- (5) 81 4 106 4 127 4 4DTEC = ALl_lX k=l[ akWkCm) _L2m=82k=l[ [ akWkCm) AL3m!107k=l[akWkCm) where AL1 = 5 km AL = 20 km AL_ = 50 km The 4DTEC is then compared against any nearby observed TEC. We can minimize any differences by appropriately modifying the F2 layer scale height (hs) in equation 2. That is, F2 layer hs controls the topside thickness of the profile; we can add or subtract electrons by adjusting this h s. Therefore, our complete set of equations includes 4, 5, i and 2. Presently, we solve this set of equations by an iterative technique. In practice it is better to do the entire profile analysis in terms of plasma frequency rather than plasma density. The variance of plasma fre- quency over the profile is much smaller than the variance of plasma density. This smaller variance allows for more stable computational solutions. How- ever, conversion to plasma frequencies can cause some problems. In terms of plasma density, equation 3 becomes: n f(_) = [ akZ k(_) (6) k=l 372 1980010188-383 i where f(1) is the plasma frequency at level _ and Zk(1) is a discrete i (empirical) orthonormal basis set which spans the frequency space. Diffi- culties arise in the 4DTEC equation because it b_comes quadratic. That is, '" 81 4 106 4 127 4 _i (k!lakZk (m)4DTEC " _L_m 12 + AL_ _ I _ akZ k(m) 12 + -L' _ ( [ a Z (m) 12 i m_82 k-i 3m=I07 k=l k k r (7) where ALi .24 lO4 1 _ _L_ " 1.24 x 104 dL2 124x104 3 In equation _, the conversion of plasma frequency to plasma density is given by: Ne(£) = 1.24 x 104 f2(£) (8) The analogous s_t of simultaneous equations becumes: f_50 below) = alZl(50 below) + ... + a4Z4(50 below) (9) f(940) = alZl(946) + ........ _ a4Z4(940) Therefore, in practice, we solve for the weighting coefficients (al, a2, a3, a4) by an iterative technique using equations 9, 7, i, 2, and 8. In effect, t_is .,nalysis reduces an entire electron density profile to just four weight- ing coefficients. That is, once aI thru a4 are known, one can reconstruct an entire electron density profile from equations 6 and 8. Remember, the empirical basis functions_ Zk(_), are input parameters to the model. This analysis is applied at every observing location. The next step is to interconnect the weighting coefficients with "best fit" functions which are continuous _n time and space• These continuous functions can then be used to compute a weighting coefficient set anywhere un the surface of the Northern Hemi_l_here (i.e. we can compute aI thru a4 for any given latitude, longitude, and time). In this way, the model expands a discrete set of observations into a continuous specification of the iono- sphere. Presently, we use surface spherical harmonic functions for the spacial connections and simple trigonometric functions are used for time relation- ship. In the near future, we hope to replace the trigonometric time functions with a more representative set of empirically derived functions. These new time functions should have a basic shape which characterizes the ionosphere's very strong sunrise gradient and its weaker (but still strong)sunset gradient. A potential benefit of these new functions is that fewer terms would be needed in the temporal partial sum in order to match diurnal ionospheric behavior. 373 1980010188-384 The current equation for plasma frequency is given by: ! t I J 2(_+m-1)4 . . ! Am, k cos cos sin cosi'O]'O mZJ k-X m,k +c ,Jcos sinit Di'j sin sin p (sin,)Z m,k m,k i where ¢ is the latitude, _ is the longitude, and t is the time. The wave numbers, I,J,M are selected on the basis of ionospheric gradients. That is the steeper the gradients, the more terms we need to include in the partial sums. Presently, Z, J, and M are set at 17, 12, and i0. The terms p_(sin_) are associated Lengendre polynomials. Equation I0 is the result of sub- stituting the uolution for the ak's (surface spherical harmonics) back into equation 6. Therefore, if one knows all the spectral coefficients (A, B, C, and D), then one can produce an electron density profile anywhere in the ionosphere. The 4-D is Initialized from a data grid of foF2's and M3000's extracted from ITS78 (See Barghausen, et al (1969)). Each grid @oint will affect certain spectral coefficients and the wave numbers I, J, and M determine the _rid spacing necessary to initialize all spectral coefficients. Presently the grid is 100 of longitude by I0° of latitude. Points separated by more than 150 will have minimal influence on one another. After the 4-D is initialized, all observed data are entered. These observations will change specific spectral coefficients; those spectral coefficients outside the range of influence of a given observation will remain virtually unchanged. Therefore, if observations differ significantly from the initial field, unrealistic gradients can be Lntroduced into the ionospheric specification. Figure 5 has an example of scattere_ observations which differ appreciably from the initial field. The spectral coefficients for observations i _,I 2 are 50% higher than the initial coefficient fiel._. In figure 5a, the observation points are separated by more than 30 degrees and in figure 5b, the observation points are withln 10 degrees s one another. Notice, in figure 5a, that the observation points are too far apart to influence one another. The 4-D solution (solid line) returns to the initial (dotted line) field about 15 to 20 degrees beyond an observation. The dashed line is the desired solution. In figure 5b, the observations are close enough to affect each other; about 15-20° beyond the observation region the 4-D reverts back to the base field. Therefore, we have a two-fold problem. First, the data points ;m_st be close enough together so that they influence one another. Secondly, the data network must be extensive enough so as to avoid unrealistic gradients as the field changes from the observa- tions back to the base field. In fact, to avoid these artificial model gradients completely, we would need an entire Northern Hemisphere I00 by i0° grid. Realistically, this I0° by 100 grid need only exist in the areas of interest. Beyond these areas, we could blend back into the base field, generating _x_el gradients in areas which would not affect our solution. 37_ t 1980010188-385 _ irl • _ _IAC.I'tOFITUDE _PECTRAL "11 i C'4EFFIC IErtTS _._:om + i ++ /o o Io Zr_ _ "Io Sot ,ms I ols 2 2 t RELATIVE SEPARA1ION (DEGREES) t + P,A6NI TUDE _PECTRAL : C'_KFr IC IENTSt [ P,o /o 0 lO 30 qO! , "e ols ! ols 2 +: RELATIVE SE*ARATIOfl (DEr._tHS) : _ Figure 5. Examples of artifical gradients induced by observati¢,ns which differsignificantlyfrom the base field. Base field(.... ); 4-D analysis • ; (--); Desired solution (- - -). *_ Presently, we are developing modeling techniques which can spread out the i information from our observation network into a space/time grid of "psuedo , observations" (See Flattery, et al. (1979)). These techniques have shown great promise and they will soon become preprocessor_ for the 14-D model. : . h. CONCLUSIONS = The ]+-D model is being developed to provide a continuous and comprehen- slve speciflcation of the ionosphere. No attempt has been made to develop a dynamic forecast model. Instead we have concentrated on providi:,6 a representative re[,roduction of the complete ionosphere from disperse ; observations. f The model reduces the entire ionosphere into a small collection of coefficients which can be easily transmitted to a customer or stored for future reference. Given a coefficient set, a user can reconstruct an electron density profile anywhere in the Northern Hemisphere for the valid 24 hour _pecification period. 375 , 1980010188-386 The h-D model is currently in limited use at AFGWC as part of an ionospheric specification and forecasting system (See Thompson and Secan, 1978). While there are still some operational limitations to the current version of the model, there is a great deal of potential in the flexibility of the model. Work is currently underway at AFGWC in learning how to exploit this flexibility to meet current and future operational requirements. REFERENCES Allen, R.S. and J.D. Tardelli (1973). Interim Technical Report - Model Evaluation Program, Air Force Cambridge Research Laboratories, Hanscom Air Force Base, Massachusetts 01731. Bradley, P.A. and J.R. Dudeney (1973). A simple model of the vertical distribution of electron concentration in the ionosphere, Journal of Atmospheric and Terrestrial Physics, 35, 2131-2146. Bryon, F.A. and R.W. Fuller (1969). Mathematics of Classical and Quantum Physics, Addison-Wesley Publishing Company; Reading, Massachusetts. Barghausen, A.F., F.W. Finney, L.L. Proctor, and L.D. Schultig (1969). Predicting Long-Term Operational Par_neters of High-Frequency Sky-Wave Telecommunication Systems, ESSA Tech. Rept. ERL II0-ITS78 (U.S. Govern- Printing Office, Washington D.C. 20402). Ching, B.K. and Y.T. Chiu (1973). A phenomenological model of global ionospheric electron density in the E-, FI- and F2-regions, Journal of Atmr_pheric and Terrestrial Physics, 35, 1615-1630. Cookingham, R.E. (1972). Modelin_ the Bottomside lonospheric Electron Density Profile, AFCRL-72-0340, Air Force Cambridge Research Laboratories Hanscom AFB, Massachusetts 01731. D_mon, T.D. snd F.R. Hartranft (1970). Ionospheric Electron Density Profile Model, Aerospace Environmental Support Center Technical Memorandum 70-3, Ent Air Force Base, Colorado 80914. Flattery, T.W. (1971). Spectral Models for Global Analysis and Forecasting, Air Weather Service Technical Report 242, Air Weather Service, Scott AFB, Illinois 62225. Flattery, T.W., T.F. Tascione, J._. Secan and J.W. Taylor, Jr. (1979). A Four Dimensional Ionospheric Moc 1 (to be published). Jasperse, J.R. (1976). Boltzmann-Fokker-Planck _'odel for _he Electron Distribution Function in the Earth's lonosphere, Planetary and Space Science, 24 (i), 33-h0. ,i 376 ]9800]0]88-387 Llewellyn, S.K. and R,B. Bent (1973). Documentation and Description of the Bent lonospheric Model - Delay Model Version, SAMSO TR 73-252, Space and Missile Systems Organization, Los Angeles, California 90009. Miller, D.C. and J. Gibbs (197h). lonospheric Analysis and lonospheric Modeling, AFCRL-TR-Th-036h, Air Force Cambridge Research Laboratories, Hanscom Air Force Base, Massachusetts 01731. Nisbet, J.S. (1970). On the Construction and Use of the Penn State MKI Ionospheric Model, Scientific Report No. 355, Ionospheric Research Laboratory, The Pennsylvania State University, University Park, Pennsylvania 16802. Nisbet, J.S. (1971). On the construction and use of a simple ionospheric model, Radio Science, 6, h37-h6h.! Nisbet, J.S. (1975). Models of the Ionosphere, Atmospheres of Earth and the Planets, edited by B.M. McCormac, 2h5-258, D. Reidel, Dordrecht, Holland. Rush, C.M. and D. Miller (1973). A Three Dimensional Ionospheric Model Using Observed Ionospheric Parameters, AFCRL-TR-73-0567, A r Force Cambridge I Research Laboratories, Hanscom Air Force Base, Massachusetts 01731. !, Shimazaki, T. (1955). Worldwide Daily Variations in the Heiuht of the Maximum Electron Density of the Ionospheric F2 Layer, J. Radio Res. Labs, Japan, 2, 85-97. Thompson, R.L. and J.A. Secan (1978). Ge( _ysical Forecasting at AFGWC, (these proceedings). von Flotow, C.S. (1978). Ionospheric forecasting at Air Force Global Weather Central, Ionospheric Effects Symposium, Arlington, Va, Jan 78. (in press). 377 1980010188-388 :;i: ..-?z i &W8,0- 18 49'2 MSFCSOLARACTIVITY PREOICTIONSFOR SATELLITE ORBIT,e¢LIFETIME ESTIMATION H. C. Euler, C. A. Lundquist, and W. W. Vaughan Space Sciences Laboratory NASAMarshall Space Flight Center i Marshall Space Flight Center, Alabama 35812, USA ! i The procedure to predict solar activity indices for use in upper atmosphere density models is given together with an example of the performance. The prediction procedure employs a least- square linear regression model to generate the predicted smoothed RI3 and geomagnetic Ap(13) values. Linear regression equations are then employed to compute corresponding _10.7{13).. solar flux values from the predicted RI3 values. The output Is issued princi- pally for satellite orbital lifetime estimations. INTRODUCTION ? _: The purpose of this paper is to present the Marshall Space Flight Center (MSFC) solar activity prediction technique and the basis for the procedure used. The motivation for issuing the predictions monthly is the need for a current and systematic input for use in an upper atmosphere density model employed in satellite orbital l[fetime estimates. The mission analysis and planning for launches require estimates of orbital lifetimes up to 10 to 15 years in the future for satellites at various orbital altitudes, inclina- tions, and eccentricities. The upper atmosphere density model useo (NASA, 1973) is based on the work and models developed by Jacchia (1970, 1971) and his collc_gues of the _" ; Smithsonian Astrophysical Observatory (SAO). To compute neutral density at the pertinent orbital altitudes and times, the model requires as an input the mean value of the 10.7 cm solar flux (Flo.7) evaluated over + three solar rotations, the mean daily value of solar flux (Flo.7), and the three-hourly geomagnetic index (Ap) taken 6 to 9 hours prior to the time of desired / orbital density calculation. The behavior of the daily solar flux and geo- magnetic indices precludes the long-term prediction of the daily values with any degree of accuracy. Therefore, ]3-month smoothed, solar-activity-related indices for Flo.7 and Ap are predicted using a least-square linear regres- sion procedure. The predicted smoothed values of Fio.7(13) and Ap(13 ) are ; " " 378 1980010188-389 used as input to the upper atmosphere denslt_ model Instead of the shorter tl_w period values of Flo.7_ and Ap. This produces the necessary values of the neutral density applicable to the computation of medium and long-term satelllte orbital lifetime estimates. PREDICTIONTECHNIQUE The HSFC solar activity prediction technique uses a modified McNlsh- Lincoln (1949) method to project the remainder of a current sunspot cycle based on data for the past 20 cycles and the current smoothed Z_rich sunspot number RI3" The statistical prediction technique uses a least-square linear regression model to generate the predicted smoothed sunspot number R13 (Avarltt, 197_). Prior to being used as input data the values _¢_r each cycle are reduced so that the cycle minima at the start of the cycles are normalized to zero. A n_an cycle is computed based on all past cycles used in the predic- tion. This provides a first approximation to the predicted cycle. All 20 cycles used In the prediction model data base use the ZUrich established minima as the starting point. However, because of the behavior of the mini- mumfor cycle 20, whereby two identical values of "RI3 in March and June 1976 are given, an interpolation procedure using the shape of the descending and ascending curve for cycles 20 and 21, respectively, was employed (Lundquist and Vaughan, 1978). This interpolation gave April 1976 as the index for the starting point of cycle 21. This point and the most recent value of R]3 for cycle 21 are used in the prediction. The variance of the prediction error of the predicted deviations from the mean cycle is minimized by the model. The predicted deviations are then applied to the mean cycle to obtain an up- : dated projection for the remainder of cycle 21. The least-square linear regression procedure assumes that if the current cycle is running abov.= (or below) the ensemble mean cycle, then in the future the current cycle is likely to follow the course of previous cycles than ran correspondingly above (or below) the mean. The variance is computed for each _; updated prediction as the average squared error obtained when the correspond- ing point of each previous cycle is predicted by the same procedure as that . applied to predicting the current cycle. The est estimate_ f the standard _ deviation for the predicted smoothed sunspot number R13 is then computed by taking the square root of this variance. The resulting computer output ln- _ cludes predicted nominal RI3 and associated plus- and minus-two standard _ '_ deviation values arranged in a time sequence, Since upper atmosphere model Inputs are needed for orbital lifetime estimates used in the design of missions beyond the current cycle, the later _. entries in the output of the least-square !inear regression model are _ slightly modified, The modification involves the selection of June 1987 as _- !. the initialization point for cycle 22. The R13 values of zero, six, and twelve are then used, respectively, for the minus-two standard deviation, ¢ 379 1980010188-390 ......... . .. , , ,, • ....... .. nominal, and plus-two standard deviation values for the minimum of cycle 22, with the balance of cycle 22 being represented by the mean and standard : deviation values computed from cycles I through 20. The predicted sunspot : values for the final 3 years prior to June 1987 are interpolated from the i least-square linear regression model prediction for cycle 21 to the estimated start of cycle 22.! To compute the smoothed FIo.7(13) solar flux values, a relationship 7(,3=494,097 ,76 exp is used. This regression equation is used with each predicted nominal and two-standard-deviation value of R13. The values are arranged in a time _ sequence on the computer output. A linear correlation coefficient of 0.98 i was obtained for this fit of the data which extend back to 1947. The data base for the geomagnetic index A extends back to 1932. Itsp behavior is poorly correlated with the solar cycle, even for 13-month = smoothed values. Therefore, the 13-month smoothed values of the geomagnetic I index _AA(13) are employed in a prediction model using the least-square : _ linear regression technique. The output is an estimate of the nominal Ap{_13) and associated plus- and mlnus-two standard deviations arranged in a time _ sequence. BASIS FOR TECHNIQUE Yu. I. Vitinskii (1965) conducted a very extensive survey and analysis of solar activity forecasting methods. While recognizing the considerable importance of the problem and encouraging studies of active processes taking place on the Sun as a contribution toward the solution of the problem, he well stated what is still the current situation. Vitinskii concluded: "...we have shown that the re]lability of the results obtained using these methods stil] leaves much to be desired." However, his analysis showed that : the linear regression method gives the best accuracy for R13 predictions up to a year in advance. For predictions several years in advance the linear regression method becomes increasingly less accurate. The uncertainty of this prediction method is given by the standard deviation estimates as com- puted by the MSFC least-square linear regression method reliability proce- dure. Therefore, with the lack of any proven deterministic prediction scheme based on physical laws, the statistical predictions produced by this technique are used as a basis for developing future orbital lifetime esti- mates needed for mission planning and analysis. 380 1980010188-391 EXAMPLE OF TECHNIQUE The Marshall Space Flight Center generates a monthly memorandum which contains data on the recently determined solar activity Indices and the current predictions for the smoothed values of Ri3' F10.7(13)' and r_(13)" Part of the output given In the November 1978 memorandum is reproduced In Table I as an example of the results of the prediction technique. SUMMARYOF EVALUATIONS Obviously, no evaluation is yet possible relative to the prediction accuracy for the magnitude of cycle 21 peak Ri3 values. The least-square linear regression model technique is sensitive to the initiation date of the cycle being predicted, the ensemble of cycles used in the historical data base, etc. Figure 1 summarizes the application to date of the linear regres- sion procedure to the prediction of cycle 21. The choice of the date for the start (minimum) of cycle 21 strongly influences the results. The data base and normalization choices have significant effects in_the early phase ofcycle 21 development, relative to the predicted peak R13 values, but become less significant as the cycle develops. This figure is helpful in represent- ing the range of rational linear regression projections for the forthcoming maximum of cycle 21. The current MSFC nomlnal prediction employs the inter- _ polated April minimum and uses a normalized data base of cycles I through 20. There is a need for a systematic representation of solar activity pre- !: dictions for use in upper atmosphere density models employed in satellite lifetime studies. The MSFC solar activity predictions are intended to serve that function. ( i 381 ! 1980010188-392 TABLE |. PREDICTED SOLAR ACTIVITY SMOOTHED13-MONTH MEANS.a,b • • °liUNSPOT NUIIB[Iq*• * • • • • I0.7 ¢11 SOLARFLUX • * • • O[_IAON[TIC INDEX • • Wl= Plo.7 Zp Tm[ N_IN_. el.USYNO MINUSTNO NOMINALPLUSTNO mNUSTWO N_JN_ _.USTWO MINUSTNO SlOMA StOre StOre SJOm ;lOre S10_ 1978.3.13 MAY "/9.N I 03.01 "_.80 12?.50 130.88 tL:'_.16 16.M 16.8"/ 15.63 19"/B._17 dUN 83.51 90.1'* '/8.80 131.35 13?._HJ 125.17 fS._J 17.10 Iq.M 19'78.500 JUL 0"/.03 65.'7? '719.30 13_.65 1_2.91 1_5._9 15.79 16.91 1_.67 I9"/IJ.Jr_3 AUO 09. c_. 100.N2 "/9.W2 13/.38 11'?.33 127.53 15.09 16.5! 15.26 19"/8.667 Slip 92.88 1116.01 ?9.'/5 1_*0.18 15_.86 1_._?.8q 15.70 16.qC IL,.Se 19"/8.?_0 OCT 96.22 112.63 'FJ.8_ Iq3.3_(16) 156.99(01 127.90(31) IS.L:_ 17.60 1_.09 19"719.033NOV 99.81 118.L:_ 80.97 1N6.58(10) 16q.38(0) 128.97(20) 1F.93 18.8/. IN._J 1978,917 DEC 103.N6 123.7g 63.17 150._'_ (6) 169.71 (01 13l.O](lO) 16.90 18.51 15.N6 19"79.000 JAN 107.21 129.65 0*4.87 153.81 (0) r75.25(0) 136.63 70) i?. 15 18.60 15.63 19'79.083 f'CB 110.27 i_P,. 35 85.10 156.73 160.84 1_.92 1"7.3'7 19.0"/ 15.67 19"/9,167 MAR 112.68 139.80 85.55 159.0 _* 165. I_* ; 31.27 17.59 19.68 15.L,7 19?9. L:_r.,OAPR 115.07 IW3.03 8?.11 161.33 188.L_ 13_.73 17.86 20.22 15.k9 19'79.333 PlAY I I?.00 1q6.+6 6?.rr_ 163. 18 191 .57 135. Iq 17.63 20.08 15. I? 19"/9._ I? J.JN 118.91 151.30 98. T_? 165.02 198. L:_ 13_. 18 I?.NI 20.50 Iq,3_ 19"/9.500 ,.U.. 121 ._9 155.35 87.23 167.30 200. I? 13k.8_ 17.57 21.00 l_. 1_ 1979.583 AUO 123.15 157.31 IF3.00 169. I0 202.06 136.51 17.70 21.71 13.85 19";9.667 S(P IL_. 15 1_8.25 90.05 170.05 202.97 137.50 17.8_ 22.58 13.07 197g.750 OCT 12N.3N 157.85 go.8_* 170.2'* 202.59 138.25 I?._ 21.66 13.02 19"19.933 NOV 1_*.23 157.5_e ._0.9| 170.13 _02._J 13tt.31 16.67 _0.60 12.x_, IgTg.gi? DEC 123.53 157.61 89.N_* 16g._*S 202.35 138.92 16.38 20.78 11.99 1980.000 JAN 122.33 157.01 8"/.66 188.31 201 .?? 135.25 16.3"/ 20.10 12.6_ 1960.083 FF.B J:J1.7? 157.38 86. I? 167.78 _02. I 1 133.8q 16._0 20.21 12.59 1980. 167 MAR 121 .NO 157.06 8'3.10 167._9 202.60 13_.8'5 16.50 19.83 13.16 1980.250 Aim 120.?,?. 157.]6 8_,.og 166.76 202.11 131.69 16.65 19.31 13.99 1980.333 l_Y I |9.59 158.k2 8_.76 165,67 201.20 130.65 16.65 19.0_ 1q.28 IS80._17 J.IN 117.96 15q.B8 81.ON 16k.10 199.71 1Lx3.O_ 16.70 19,11 IW.30 1960.500 JUL 116.17 15q. 10 78.25 162.,,,9 1.qS.96 t_6._ 16.'r_ 18.9'4 1_.10 1980.583 AUO I Iq. J38 Ir-_. 39 ?W.77 160,86 199._ 123.21 16,_9 19.11 13.1_ IcJSO.E6"/SlEeP 113.56 1_.71 72._1 159.89 199.56 I_1.0q 16.75 20.0_ 13.N7 1980.750 OCT I 13._ 155.91 ?O.r'_ 159.56 200.71 119.30 16.70 20.76 12.79 1cj80.833 NOV 112.B8 157.00 68.?6 159.23 201.76 !17._ 16.71 21._ 11.80 1980.917 O[C 112.29 155.97 68.62 158.67 200.?? 117.r'_ 16.8'4 _2.q6 I 1.20 I_1.000 JAN I I 1.53 153.N3 69.63 15?.gq 198.31 ! 18.NB 17.02 22.95 I 1,09 I_1.0B3 FEB 110.L_ 151._2 69.U9 156.72 196.37 117.99 16.88 Z2.91 10.B_ 1981.166 MAR 108.39 150.38 6E.'_0 15N.93 19_. 35 115.53 16. r'_ 22.89 10.15 1981.2%0 APR 1015.I_e IN8.78 83.50 15_.79 193._ I 1_.90 16.E_ _-51 9.96 a. Because orbital 11fetime programs require a specific date to associate • pred i towlth _he input of solar flux (_]0 7) and geomagnetic (_p) ict on compute a corresponding atmospheric density, the data points are identi- fied with the midday of the given month. b. Values In parentheses beside the solar flux predictions are estimates of the near-term average values, for F10.7 " _10.7 projected 3 months into the future. The first va]ue shown in parentheses is based on the actual average value of FlO.7 for the month given. See Equation A-]O of NASA SP-802] for use of these data. 382 1980010188-393 383 1980010188-394 REFERENCES Avarltt, D. (1972): Sunspot predictlon program (sunspot deck). Program Document 3F0360, Computer Sciences Corporation, Huntsville, Alabama. Jacchla, L. G. (1970): New static models of the thermosphere and exosphere with emplrlca] temperature profiles. Special Report No. 313, Smlthsonlan Astrophysical Observatory, Cambridge, Massachusetts. Jacchla, L. G. (1971): Revised static models of the thermosphere and exosphere wlth empirical temperature profiles. Special Report No. 332, Smlthsonian Astrophysical Observatory, Cambridge, Massachusetts. Lundquist, C. A., and W. W. Vaughan (1978): Analysis of regression methods for solar activity forecasting. Symposium/Workshop on Solar-Terrestrial Influences on Weather and Climate, Columbus, Ohio, July 24-28, 1978. McNish, A. G., and J. V. Lincoln (1949): Prediction of sunspot numbers. Transac.tions t American Geophysical Union, 30:673. NASASpace Vehicle Design Criteria (1973): Mode]s of earth's atmosphere (90 to 2500 km). NASASP-8021, National Aeronautics and Space Administration, Washington, D. C. Vitinskii, Yu. I. (1965): So]at-activity fprecastin_. Translated from Russian, Israel Program for Scientific Translations, Jerusalem. NASA TT F-289, Clearinghouse For Federal Scientific and Technical Information, Springfield, Virginia 22151. 384 1980010188-395 8'0 "184 3 Examples of "Problem" Flares or Situations in Past Solar-Terrestrial Observations H. W. Dodson, E. R. Hedeman, O. C. Mohler McMath-Hulbert Observatory of The University of Michigan 895 Lake Angelos Road Pontiac, Michigan 48055 An effort is made to recall instances of solar activity in past years that probably would have presented difficulties in prediction. Unexpected flares of significance include the so called "spotless" flares in centers of activity with only small or no spots. Isolated major flares in relatively flaze- poor centers of activity also can be a surprise. Centers of activity with atypical levels of flare production with respect to their spot associations can introduce confusions in predic- tion. Flare-rlch centers of activity without significant associated geomagnetic disturbance likewise are often misleading to forecasters. At the present time, energetic particle events stemming from flares on the invisible hemisphere present almost insurmountable difficulties for those who carry the responsi- bility of prediction. i. Introduction It has been suggested by Dr. Donnelly and the Program Committee of the Solar-Terrestrial Predictions Workshop that we at the McMath-Hulbert Observatory might be able to recall certain flares or solar-terrestrial situations that would have presented unusual difficulties had efforts been made to fore- cast their occurrence. In the hope that experiences of the past may be of value to future efforts, we wiJl try to share, informally, some of the _on-typical or unexpected solar and/or geophysical circumstances that we remember from past years. 2. Flares, probably unexpected, with significant geophysical effects. 2.1 Major flares in centers of activity with small or no spots Although the majority of flares occur in centers of 385 1980010188-396 activity with well developed spots, especially those with spots with complex magnetic configurations, it has been known for many years that a small proportion of flares, including a few important flares, develop in regions with small or no spots. It seems probable, that major flares in regions with only small or no spots may present especially difficl,lt prediction clrc,un- stances. In 1970 we published a report (Dodson and Hedeman, 1970) of flares of Ha importance _2 in such regions in the years 1956-1968. These "spotless" flares included approximately 7% of all flares of Ha importance 2 and 3 in the years in question. At the time that we prepared the above paper we indicated that only a small proportion of the important "spot- less" flares were accompanied by significant geophysical effects. The identification of the geophysically important major flares in "spotless" regions thus becomes a prediction problem with double difficulty. Such flares do occur, and over the years have presented interesting geophysical circumstances. Table i lists examples of such major "spotless" flares with information relating to their geophysical associations. TABLE 1 Examples of Major "Spotless ''+Flares with Significant Geophysical Associations Ha Flare Date and Time Position Imp. CFI Geophyslcal Effects 1957 Nov.29d00h45m N41 E63 3+ 4 Geomagnetic storm Dec. i. Max 3-hr. Kp - 6. 1959 Jul.14dl4ho0 m S25 E37 3 8 SID imp.2. There were so many great flares in another center of activity in mid-July 1959 that it is im- possible to evaluate other possible geo- physical effects of this flare. 1967 Jan.lld01h37 m $27 W45 3 5 Geomagnetic storm Jan.13, max. 3-hr. KD - 8; Sat. proton ehission _90 MEV. 1967 Feb.13d17h47 m N21WII _b Z8 SID imp. i; geomag. storm Feb. 15, max. 3-hr. Kp- 8, Polar Cap Absorption. 386 0 1980010188-397 TABLE I (con't) Ha _lare Date and Time Position Imp. CFI GeophysJc_l Effects 1969 Sep.25d06h58 m NI3 WI5 3n 5 This large flare had no reported associat- ed SlD. Nevertheless, it was apparently the source of _30 MEV protons and Polar Cap Absorption. A mod- erately severe geo- magnetic storm began late on Sep. 27. 1970 Nov. 5d03h07 m S13 E35 3b 14 SID imp.3; geomag. (This was the only flare of H_ storm Nov.7, max. importance 3 in the year 1970) 3-hr.Kp = 7. Sot.proton emission >60 MEV; Polar Cap Absorption 1973 Jul.29d13h12 m NI4 E45 3b Ii SID imp. 2; Geomag. (This large two-ribbon flare storm Jul.31. max. 3-hr. KD - 5 weak occurred during the Skylab mission proton _mission andbut it apparently was not predicted weak Polar Capbecause the astronauts did not obtain observations until four or Absorption. more hours after the start of the flare. It was the only flare of H_ importance 3 in 197_.) + The term "spotless" means a region with only very small or no spots The locations in which these large "spotless" flares develop, generally include a large and relatively bright calcium plage (intensity 4'2.5 on McMath scale). More importantly, the flare is generally associated with a large, dark, previously existing filament in H_ observations. Prior to the onset of the flare, the filament often shows motion through the presence of Doppler shifts in the dark gases. The filaments associated with "spotless" flare_ either traverse a plage or lie between regions described as two plages. 2.2 Other possibly unexpected significant flares. Sometimes when the general level of solar activity is low, unexpected events occur. Were "majoE" flares with Polar Cap Absorption expected on 1964 March lOalSh53 m and 1965 February ._q7 1980010188-398 5d17h50m? It seems unlikely. Some months later, in late September and early October 1965, Plage 8005 (CMP October 3, N21 °) a bright plage with a relatively large v-type spot was crossing the solar disk and producing a modest number of flares and subflares. On October 3, a ne___wspot and plage formed on the disk south of this active reglon and near a previously existing plage. The newly formed region was Plage 8012 at S19 °. This newly formed region grew rapidly, and on October 4, to our surprise, developed a "major" flare, Comprehensive flare Index of 8 and Ha importance 2, associated with a proton event and Polar Cap Absorption. This was the first "major" flare since June 13, 1965 and it did not occur in the center of activity that was attracting our primary attention. Later, in September 1968, when solar activity was at a relatively_ level, there was a good example of an important flare event t_t probably was hard to predict. It was what we refer to as an "isolated major flare" and it _ad significant geophysical effects. On 1968 September 29d16h17 m, in Plage 9678 (CMP September 25, NI6 °) there developed a 2b flare. It was the "major" flare, and the _n__onlflare to be associated with a Short Wave Fade in the history of the plage. This unique flare event was the apparent source of very energetic protons, Polar Cap Absorption, and a very small Ground-Level Enhancement of cosmic rays. The plage in which this isolated major flare occurred included several spot groups. The largest had a maximum area of _800 millionths and was classified as magneti- cally complex (_) on September 21, 25, and z6. Relatively isolated major flares do occur throughout the _olar cycle and probably present especial difficulties for forecasters. 3. Centers of Activity with atypical levels of flare production. 3.1 Unexpectedly low levels of flare production. Centers of activity with large spots (maximum area >_i000 millionths of the hemisphere) and complex magnetic structure (_T, ?, and _configurations) are well known to be the sites of numereus Ha flares and SID-associated subflares during the transit of the region across the solar disk. Memory does not at once recall an outstanding exception to this pattern, though numerous large, spotted, plages do, indeed, traverse the disk without producing exceptionally great flares. A survey of our records shows that regions with large spots that are not mag- netically complex frequently traverse the solar disk w'l-Ehout producing more than ten flares of He importance __I, or ten SID's. It is rare, however, for the disk passage of large mag- netically complex groups to fail to be associated with at least ten flares of importance one or with at least I0 SID-associated 388 1980010188-399 flares or subflares. If our records are correct, at least two such atypical regions can be pointed out. (A) McMath Plage 9073, CMP, 1967 November 19, NI6 °, with a spot of maximum area 1085, and mean area 825 millionths of the hemisphere. The spot was classified as _? and S. According to our reevaluation of flare importance values for 1967 (UAG-19) there were only four flares of importance 51 in this region. These four flares, however, included one event of importance 3, two flares of importance 2, and one of importance i. The plage was large and very bright and was in its third rotation. The region returned in December, underwent additional growth and had a type spot with maximum area 2427 millionths, developed II flares of importance 91, and was associated with 18 SID's. Failure of a large complex region to produce many flares, such as Plage 9073 in November, apparently is not a reliable indica- tion that that region on the sun has lost its potential for increased activity. (B) McMath plage 11883, CMP 1972 May 20, N09 °, with a spot of maximum area 1262, and mean area 1020 millionths of the hemi- sphere. The spot was classified as By and E. The region developed only five flares with importance _I and none of them qualified as "major" flares. Only six flares and subflares in the region were associated with SID's. Meter wave-length radiation was enhanced during the central meridian passage of this plage and spot. Geomagnetic disturbance was not associated with this region or its flares. Over the years we apparently have not been especially mindful of the potentially active regions that failed to live up to their solar expectations. It might be worthwhile in the future to try to keep track of large, complex spot groups that do not develop a reasonable number of flares or SID's, perhaps i0 flares of He importance t l and i0 SID's associated with flares or subflares. Suggestions of such regions are given in Table 2. Should plage 14822 with the large spot group with CMP June 29, 1977 be considered in this category? The region developed only six flares with He importance as great as one and only one X-ray enhancement as great as M i. Thirteen flare associated SWF's were reported. All of the flares and SWF's took place during the days when the region was in the east and when the spot group was generally magnetically complex and with maximum area of the order of i000 millionths of the hemisphere. During transit of the western half of the disk, the spot lost its magnetic complexity, decreased in area, and the production of flares of importance fl and SWF's stopped. Perhaps the real challenge in flare prediction is to try to anticipate when the magnetic complexity of a spot group will cease, and when the spot area will diminish. 389 1980010188-400 TABJ_ 2 Possible Examples of Ap_ ently Significant Centers of Activity with below Average Flare Production _ Spot Number + Mag. Flare Plage Date CMP Lat. LonK. Area Class. l_m_. SWF 3747 1956 Nov. 9 N27 273 ° _1007(max.) _p 4 0 520(mean) 5491 1959 Dec.15 NI5 106 '1413 _p 6 I 11034 5505 1959 Dec.26 N09 314 _1282 _ 5 0 589 5556 1960 Feb.ll N24 5t 1191 _ 5 0 8609 1966 Dec.ll $23 250 1218 _ 3 3 8629 1967 Jan. 2 $22 320 1803 _p 6 3 ' 9073 1967 Nov.19 NI6 35 1085 _-_ 4 3 10242 1969 Aug. 5 S13 67 825 _ i 0 10344 1069 Oct. 3 S12 14 511 _ I 2 10918 1970 Sep. 5 NI8 239 ii00 _ 7 i0 iiiii 1971 Jan. 10 S04 5 _1255 _p 4 4 970 11137 1971 Feb. 3 NI3 55 i1320 _p, 5 0 _i003 $? ona d 11565 1971 0ct.22 NI0 208 _1571 _p 4 2 _i167 By on 2 days 11619 1971 Nov.27 S13 93 734 _p 5 2 '670 _ on 3 days 11883 1972 May 20 S14 299 '1262 _- £ 5 6 !1020 + Data not homogeneous over the years. 3.2 Unexpectedly flare-rich centers of activity. Efforts to identify in past experience centers of activity that produced r,ore flares of Ha importance _i or more flare- associated SID's than the area and magnetic complexity of the spot would suggest, encounter the effects of changing standards in flare-reporting and the more abundant modern techniques for detection of ionospheric disturbances. Regardless of these inhomogeneities, it seems clear that large Plage 4065, CMP 1957 July 20, N21 °, which included spots no larger than 500 millionths of the hemisphere and _ magnetic classification, was an above average flare producer. It apparently was the site 390 1980010188-401 of 74 flares called importance one or greater and produced 17 of the relatively rare SlD's of those years. On March 25, 1970, a very large bright plage, McMath 10641, with numerous small spot groups crossed the central meridian. Several of the spot groups were reported as having magnetic complexity on one or two days. During the course of its transit across the disk, the plage was the site of 27 flares of importance 3 i, including 6 "major" flares, for two of which the Comprehensive Flare Indices were ii and 14 respectively. These "great" flares on March 25 and 29 _:ere associated with important SlD's, well defi_ed proton enhancements, and the onsets of geomagnetic storms. The region seems to have been well above average in flare production for a center of activity with such unpretentious spots. There are probably many other such above average flare producers in the records. A list of suggestions for regions in this category is given in Table 3. TABLE 3 Possible Examples of Unexpectedly Flare-Rich Centers of Activity Spot Number + Mag. Flire Date CMP Lat. Lon_. Area Class. Imp. SWF's 4065 1957 Jul.20 N21 175 ° 465 _f 74 17 4436 1958 Feb.25 $21 144 450 _p 15 2 4634 1958 Jul. 7 N28 203 373 _ 23 5315 1959 Aug.10 NI8 342 311 _p 52 7 5627 1960 Apr.13 N08 319 804 _ 31 5 5680 1960 Jun. 4 N28 351 494 _ 34 9 5863 1960 Sep.25 S15 297 452 _p 33 2 7812 1965 May 23 N24 176 440 _p 6 0 9946 1969 Feb.23 NI6 60 897 B 16 31 9966 1969 Mar. 6 NI3 275 293 _ 15 19 10109 1969 May 26 NI3 292 397 _p 26 23 10477 1969 Dec.19 NI2 72 170 _p 19 7 10568 1970 Feb.10 NI6 94 797 _p 22 37 (&on 2 days) 10584 1970 Feb.22 S15 296 284 _p 17 8 10607 1970 Mar. 5 N08 158 604 _p 18 28 10641 1970 Mar.25 NI6 247 358 B 271 _p 27 18 (_? on 2 days) + Data not homogeneous over the years. 39! 1980010188-402 4. Flare-rich centers of activity with associated geomagnetic disturbance at relatively low levels. 4.1 Plage 3400; CMP February 17, 1956; N20 °. There are so many instances of the occurrence of signifi- cant geomagnetic storms and well defined proton increases in time association with the transit across the disk of great, flare-rich centers of activity that one tends to forget that there have been conspicuous exceptions to this pattern. We well remember the circumstances in February 1956 when McMath Plage 3400 with its large spots and ? classification crossed the disk with an impressive series of flare events starting on February I0 with a great flare and ejection at the east limb. At least 6 "major" flares took place in the region between February i0 and February 19 (UAG-14). All had severe SWF's and assorted radio frequency emission. At least one event (Feb- ruary 14d05h38 m, E32 °) was a "great" flare with Comprehensive Index of 12. Geomagnetic storms were expected but did not materialize beyond the most minor of disturbances. It was not until February 23 when this "disappointing" region was at W80 ° that another "great" flare (CFI _13) developed and initiated the severe and well known GLE and Polar Cap Absorption of that date. A significant geomagnetic storm began on February 25. We often have puzzled over the series of circumstances pre- sented by Plage 3400. Were propagation circumstances at the sun, in the interplanetary medium, or both, especially unfavor- able for the transport of particles from the sun to earth until the flare at W80 ° took place? 4.2 Plage 2325; CMP May 16, 1951; NI0 ° Another great, flare-rich region of earlier decades that crossed the disk without initiating geomagnetic disturbance that was commensurate with its flare production was McMath Plage 2325 with CMP May 16, 1951. This region, like 3400 described above, began its flare activity at the east limb (May 8). It produced many significant flares throughout its entire transit of the solar disk. Included in its record were especially important flares on May 18 and 19. A geomagnetic storm with maximum 3-hourly Kp of 7 began on May 9, lasted for two or three days but the disturbance appears to have been primarily sequential in nature. During the remainder of the region's transit across the disk, geomag- netic disturbance was most moderate, and what did occur seems sequential rather than flare associated. It was a time when forecasters might indeed have expected greater geomagnetic storms. 392 1980010188-403 4.3 Plage 8223, CMP April 4, 1966; N27 °. The difference between the geomagnetic effects of somewhat comparable centers of activity was well exemplified in early 1966 by the Polar Cap Absorption and geomagnetic storms associated with the great March region of that year (Plage 8207, CMP March 22) and the near absence of such effects with the usually forgotten April region (Plage 8223, CMP April 3). From the purely solar point of vie_ the two regions were rather similar with large, magnetically complex spots, high levels of radio frequercy emission, many flares, and many flare- associated SID's. Our evaluations lead to Active Region Indices of 14 and 12 respectively, the two highest Active Region Indices in 1966. It should be noted, however, that the "major" flares in the April region took place east of the central meridian. In addition, the two regions were located almost exactly 180 ° apart in solar longitude. Did this special geometrical relationship play any part in the probable difference in propagation patterns relative to the two regions? 4.4 Plage 8704, CMP February 27, 1967; N23 ° and Plage 8740, CMP March 27, 1967; N210. These two plages are successive rotations of the same center of activity. In both the February and March (1967) transits, the regions were flare-rich and had abundant x-ray emission as exemplified by the numerous flare-associated SID's. Their Active Region Indices were high, 16 and 12 respectively. In spite of these circumstances, geomagnetic disturbance was at a low level or absent during the time of transit of the regions. During each rotation the plage developed eight events that we identified as "major" flares on the basis of ionizing, radio frequency, or optical radiation (UAG-14) but there was only one flare in each transit for which the Comprehensive Flare Index was _Ii, indicating a "great" flare (UAG-14). Polar Cap Absorption was not confidently associated with any flare in these two plages (Svestka and Simon Catalogue). 4.5 Plage 8905, CMP July 28, 1967; N27 = and Plage 8942, CMP August 26, 1967; N22 °. These two plages again constitute successive rotations of the same center of activity. Plage 8942 in August (1967) was a return not only of 8905 but also of 8907 (CMP July 29) in latitude NI60. In July, 1967, Plage 8905 with its large spot and complex magnetic field developed large zlumbers of flares and very large numbers of associated SID's. Its Active Region Index is 18, the highest value we have assigned to any plage in all of solar cycle 20. In its return in August as Plage 8942, the spots continued to be large and magnetically complex, flare and SID production was high, and the Active Region Index was 12. Again, in spite of these circumstances, during the 393 r 1980010188-404 July and August transits of this center of activity, geomagnetic disturbance was low or absent and Polar Cap Absorption was not recorded. It may be noteworthy that during both the July and August rotations of the region, only four flares in each rotation qualified as "major" flares (UAG-14) and for none of these events was the Comprehensive Flare Index 511. We would therefore comment that although flares in these regions were numerous, a_d abundant X-ray producers, none of them was "greatl' 4.6 Plage 9184, CMP January 31, 1968; NI5 °. Plage 9184 (CMP January 31, 1968) with its very large spot and complex magnetic field was another flare-rich region with below average geomagnetic storm association. The maximum area of the spot was 2769 millionths and mean area, 2176. The magnetic classification was _? with & on January 31. Twenty- three flares of importance _i and 21 SlD's were recorded in the region. In spite of these manifestations of activity, only four of the flares qualified as "major" on the basis of ionizing, radio frequency, or optical phenomena, and none was "great" (i.e. Comprehensive Flare Index _ Ii). Did this absence of "great" flares account for the fact that geomagnetic disturbance during the transit of this significant center of activity (Active Region Index = 16) was at a relatively low level? Polar Cap Absorption recorded on February 8, 1968 has been ascribed to this region (Plage 9184) when it was two days beyond the west limb. (Svestka-Simon Catalogue). 4.7 Plage 10432, CMP November 22, 1969; Nil ° . We also should point out Plage 10432, CMP November 22, 1969 as another center of activity with large spots, complex magnetic fields, many flares and short wave fades, and enhanced radio frequency emission, but with perhaps less than anticipated geomagnetic disturbance. Its Active Region Index was 13. More importantly perhaps, it was, according to our records and reevaluations, the center of activity in cycle 20 with the largest number of flares of importance _i. There were 55 such flares in this region (Dodson, Hedeman, and Mohler, 1974). Nine of the flares qualified as "major" (UAG-14) and three of them fall into our category of "great" flares. The three "great" flares occurred on November 18, November 24, and November 27 (E40 °, W31 °, W82 °) with Comprehensive Flare Indices of 14, 12 and 13 respectively. Polar Cap Absorption was observed following the flare on November 24 and a "moderate" SC, geomagnetic storm developed on November 26. In general, however, geomagnetic disturbance during the transit of this especially flare-rich region seems not commensurate with the observed level of solar activity. It also seems noteworthy that the "great" flare with CFI of 13 at W82 ° on November 27 was not accompanied by a recorded increase in proton flux in the ne---_ghborhoodof the earth in spite of its favorable location 394 1980010188-405 in Lhe far western part of the solar disk. Even the low energy i-i0 MEV flux shows no increase and continues its steady decline following this flare. 5. Comments on geomagnetic disturbance associated with centers of activity with as many as four "great" flares (Comprehensive Flare Index __ii). The foregoing significantly flare-rich centers of activity can indeed be considered as having been "disappointing" in their geophysical consequences if one has in mind the severe geomagnetic disturbances associated with such centers of activity as Plage 8818 (CMP May 25, 1967), Plage 9740 (CMP October 28, 1968), and Plage 11976 (CMP August 4, 1972). We have often noted that these three regions were all plages in which as many as four "great" flares (Comprehensive Flare Index-_ii) occurred. We have been led to ask if the occurrence of at least four "great" flares in a center of activity provides sufficient grounds for the confident expectation of severe geomagnetic disturbance. In solar cycle 19 we identified five plages with at least four "great" flares (Dodson, Hedeman, and Mohler, 1974): Plage 4134, CMP September i0, 1957 Plage 5265, CMP July 14, 1959 Plage 5615, CMP March 31, 1960 Plage 5925, CMP November 12, 1960 Plage 6171, CMP July 14, 1961 All of the aDove regions were accompanied by geomagnetic storms of marked severity attributed to flares within these regions. In solar cycle 20, five regions with at least four "great" flares also have been identified: Plage 8818, CMP May 25, 1967 Plage 9740, CMP October 28, 1968 Plage 9946, CMP February 23, 1969 Plage 10607, CMP March 5, 1970 Plage 11976, CMP August 4, 1972 As mentioned earlier, for three of these regions, 8818, 9740, and 11976, the associated severe geomagnetic disturbance is something that many solar astronomers and geophysicists remem- ber well. For the other two regions, a few comments seem needed. In 1969, Plage 9946 with CMP February 23 was the site of 16 flares of importance-_i including i0 "major" flares for which four had Comprehensive Indices 311 The greatest of these flares (February 25d09h00 m, W37 °) was a white light flare associated with a ground level cosmic ray enhancement, proton 395 1980010188-406 increase, and Polar Cap Absorption. Geomagnetic storms are : indeed reported as starting on February 26 and 27, but these storms were not severe. A glance at plots of Kp and C9 values for the interval in question shows that the geomagnetic dis- turbance, though clear cut, was not unusually severe. Why was the geomagnetic disturbance not greater? In 1970, Plage 10607, CMP March 5, also might have been misleading to forecasters of geomagnetic disturbance. This region with 18 flares of importance _i had all of its seven l'major" flares, including the four "great" ]_l'ares,when the plage was east of the central meridian (February 27-March 2, E69°-E35°). Moderate geomagnetic storms occurred March 1-2 and March 5-6, perhaps attributable to the foregoing flares. However, starting on March 6 and March 8, two storms, moderately severe and severe, respectively, take place, long after Plage 10607 has ceased having "major" flares. It is probable that these two later storms were flare associated, but not with flares in Plage 10607. Why was this geomagnetic disturbance so great? Considerable caution seems indicated in efforts to anticipate the magnitude of geomagnetic disturbances. Sometimes it is much greater than expected as was the case with the great "problem" geomagnetic storm of October 5-7, 1900. 6. Final Comments At times the forecasting of proton enhancements and geo- physical phenomena is practically impossible. We remember especially the great events of January 28, 1967 -- ground level enhancements of cosmic rays, Polar Cap Absorption, and great energetic proton enhancements. If association of these events with a flare or flares far on the invisible hemisphere is correct, the forecaster, under present circumstances, had practically no way of anticipating these events. Lesser, but probably similar events seem to have occurred at other times. For example, did anyone anticipate the significant geophysical phenomena that occurred around May 12-16, 1969? One of the most severe geomagnetic storms of 1969 began on May 14, Polar Cap Absorption was reported on May 13. Faced with no suitabTe observed phenomena to account for the PCA, in the Svestka-Simon ,l "Catalogue of Solar Particle Events, we suggest activity on the invisible hemisphere as the particle source. Perhaps, in years ahead, satellite data will provide information about the invisible hemisphere of the sun that will be of assistance in forecasting certain solar and/or geophysical phenomena. The foregoing comments so emphasize the problem cases and disorder in solar and geophysical phenomena that they may give the wrong impression. Comments could have been prepared that would have stressed the general accord that exists between the development of large, magnetically complex spots in large bright 396 % 1980010188-407 plages, and the occurrence ef flares, ionospheric disturbances, energetic particle enhancements, and frequently subsequent geomagnetic storms. The prediction of such phenomena appears to be a possible, but difficult, undertaking. The frequency of atypical phenomena or situations suggests that notes of caution in a forecast will often considerably increase its credibility. Those who carry the responsibility of forecasting solar and geophysical circumstances deserve the highest level of support ip their difficult task from all who observe the sun or try to keep track of its effects on terrestrial circumstances. The authors gratefully acknowledge support from the follow- ing grants and contracts in the preparation of this report: NASA - NGL 23-005-275, NOAA 03-4-022-94, and Johns Hopkins Applied Physics Laboratory Contract No. 600868. References and Data Sources Dodson, H. W., and Hedeman, E. R. (1970): Major Ha Flares in Centers of Activity with Very Small or No Spots. Solar £hysics, Vol. 13, 401. Dodson, H. W., and Hedeman, E. R. (1971): An Experimental Comprehensive Flare Index and its Derivation for 'Major' Flares, 1955-1969. Report UAG-14, WDC-A, NOAA, Boulder, Colorado. Dodson, H. W., Hedeman, E. R., and Mohler, O. C. (1974): Comparison of Activity in Solar Cycle 18, 19, and 20. Reviews of Geophysics and Space Physics, Vol. 12, 329. Dodson, H. W., and Hedeman, E. R. (1975): Experimental Comprehensive Solar Flare Indices for Certain Flares, 1970-1974. Report UAG-52, WDC-A, NOAA, Boulder, Colo. Quarterly Bulletin of Solar Activity of the International Astronomical Union: Publisher Eidgen, Sternwarte in Zurich. Solar Geophysical Data: Natioral Geophysical and Solar-Terres- rial Data Center, Coulder, Colorado. 397 1980010188-408 FORECASTS OF GEOMAGNETIC ACTIVITY BY OTTAWA MAGNETIC OBSERVATORY: THEIR RELIABILITY AND APPLICATION J. Hruskd Earth Physics Branch Dept. of Energy, Mines and Resources Ottawa, Ontario, Canada Ottawa Magnetic Observatory issues a 27-day forecast of geomag- netic activity based on the forecasts from the Environment Services Center in Boulder, Colorado and on data from Canadian geomagretic observatories. To evaluate the reliability of the forecasts, the predicted activity has been compared with the actual geomagnetic activity at five geomagnetic observatories located in different magnetic zones: Ottawa (57.0°N, 351.5°E,) Meanook (61.8°N 301.O°E) Fort Churchill (68.8°N, 322.5°E), Cambridge Bay (76.7°N, 294.0°E), and Resolute Bay (83.1°N, 287.7°E). The percentage of correct predictions varies with the geomagnetic latitude of the observatory used for comparison. The percentage is on the average highest _or lower latitude stations and lowest for the northern stations. The number of incorrect predictions ranges from 4.7_ for Ottawa to 8.4_ for Churchill. The 27-day predictiens of Geomagnetic Activity are used for geomagnetic surveys and in various research projects. User evalua- tions of the reliability and usefulness of these forecasts are summarized. I. Canadian Forecasts of Geomagnetic Activity. Beginning in 1974 the Ottawa Magnetic Observatory, Earth Physics Branch, issues a "27-day Forecast of Geomagnetic Activity" every three weeks especially designed for use by mining and geophysical survey companies and by various government agencies. The predict- ions are based on the "27-day Forecast of Solar and Geophysical Activity" issued by the Space Environment Services Center in Boulder, modified to more closely reflect the characteristics of the geomagnetic field at higher latitudes [K. Whitham, E.I. Loomer, E.R. Niblett (1960)]. The records from the Canadian Observatories are used to monitor the 27-day recurrence of geomagnetic activity. The trends of activity estimated in this manner -re compared with the 27-day forecast issued by the Space Environmental Services Center. A new forecast which respects the predictions from Boulder as well as the trends of activity expected from the 27-day cycle observed in Canadian data, is then issued. 398 1980010188-409 Since the predictions are destined mainly for practical use in magnetic surveys, they cover every day of the interval for which they are issued and describe the geomagnetic activity ;n simple, easily understood terms. They also include a review of the magnetic activity during the past month as given by the K-indices from Ottawa and Meanook observatories--see e_ample of the Forecast of Geomagnetic Activity by Ottawa, Review of Geomagnetic Activity and the Boulder Forecast, for July 1978. FORECAST OF GEOMAGNETICACTIVITY FOR PERIOD: LA PREVISION DE L'ACTIVITE GEOMAGNETIQUEPOUR LA DUREE: June (juin) 28 to (_) July (juillet) 30 1978 The geomagnetic field is expected to be: Le champ geomagnetique sera probablement: active (actif) : July (juillet) 4-8, 16-19 unsettled : the rest of forecast period (agite) (le reste de la period prevue) quiet (calme) : July (juiilet) i-3 Ottawa Magnetic Observatory L'obser_atoire magnetique d'Ottawa THE REVIEW OF GEOMAGNETICACTIVITY IN JULY 1978 LA PREVISION DE L'ACTIVlTE GEOMAGNETIQUEDE MOIS DE JUILLET OTTAWA Geomag. Co-ord.: 57.0 o Lat. N., 351.50 Long. E. active (actif) : 4-5, 13-16, 18 unsettled : the rest of July (agite) (le reste du juillet) quiet (calme) : 29-31 MEANOOK Geomag. Co-oral.: 61.8° Lat. N., 301.00 Long. E. active (actif) : 4-5, 7, 13-14 unsettled : the rest of July (agite) (le reste du juilIet) quiet : I-2, 6, 9, If-12, 15-16, 19-20, 23-31. 399 1980010188-410 R 260400Z JUN 78 FM SPACE ENVIRONMENT SERVICES CENTER BOULDER COLO UNCLAS WU 27 DAY FORECAST OF SOLAR AND GEOPHYSICAL ACTIVITY ISSUED O2OIZ 26 JUNE 1978 SOLAR ACTIVITY SHOULD BE MODERATE EARLY IN THE PERIOD, BECOMING LOW AS REGIONS I170, I172 AND I176 DISAPPEAR FROM VIEW. THE ACTIVITY SHOULD REMAIN RELATIVELY LOW UNTIL ABOUT 14 JULY WHEN THESE REGIONS WILL RETURN TO THE VISIBLE DISK. SOLAR ACTIVITY MAY BECOME HIGH AT THAT POINT. THE GEOMAGNETIC FIELD IS EXPECTED TO BE GENERALLY UNSETTLED THROUGH THE NEXT 27 DAYS WITH OCCASIONAL QUIET PERIODS. SHOULD A MAJOR FLARE OCCUR, A MAJOR MAGNETIC STORM IS A POSSIBILITY. SOLTERWARN The geomagnetic activity is defined as quiet, unsettled or active. The latitudinal dependence of the activity is reasonably well eliminated from the predictions with the h_lp of an empirically determined hourly index P which effectively normalizes activity levels in three geomagnetic zones: sub-auroral (50 ° - 60°); auroral (60° - 70°); and polar cap (above 70° north). The values of P in nanoteslas are listed in the following table. The index has been based on values of K-indices for the subauroral zone and on R-indices, derived from the hourly ranges, For the auroi I and polar cap zones. geo-magn, activi_. 50 ° 60 ° 60° - 70° over 70° ._L_6magnetic fat. N quiet P_3OnT P_IOOnT P_8OnT unsettled 30>P_60 100>P_230 80>P_150 active P > 60 P "230 P _150 400 1980010188-411 2. The Reliability of Geomagnetic Predictions during 1975 to 1977. The Ottawa predictions for the period 1975 to 1977 have been evalua o ted by comparing the forecasted activity with the activity actually observed at five magnetic observatories located in different geomagnetic : zones: Ottawa -- OTT (subauroral), Fort Churchill -- FCC (auroral), Resolute Bay -- RES (polar cap), and the transitional stations Meanook -- MEA and Cambridge Bay -- CBB [E.I. Loomer (1976)]. The geographic location of these stations is shown in Fig. I. Canadian Magnetic Observatory Network Symbol Station Geographic Geomagnetic Lat. Long E Lat. Long E. ALE Alert 82.50 297.50 85.9 168.2 RES Resolute Bay 74.42 265.10 83.1 287.7 _ MBC Mould Bay _6.2 240.6 79.1 284.7 CBB Cambridge Bay 69.10 255.0 76.7 294.0 PEB Pelly Bay 68.53 270.49 78.6 320.45 BLC Baker Lake 64.33 263.97 73.9 314.8 YKC Yellowknife 62.50 245.5 69.0 293.1 : RII Rankin Inlet 62.63 268.08 72.9 321.9 : EKP Eskimo Point 61.10 265.93 71.1 321.8 FCC Fort Churchill 58.80 265.77 67.B 323.0 - FSV Fort Severn 55.98 272.35 66.B 333.0 GWS Great Whale River 55.27 282.22 66.8 347.2 GIM Gillam 56.35 265.32 66.2 323.4 TMP Thompson 55.72 262.12 65.4 319.3 MEA Meanook 54.60 246.7 61.8 301.0 ISL Island Lake 53.88 265.32 64.0 324.4 : WHS White Shell 49.75 264.75 59.9 325.3 STS St. John's 47.60 307.3 58.7 21.4 VIC Victoria 48.50 236.6 54.3 292.7 , OTT Ottawa 45.40 284.5 57.0 351.5 401 1980010188-412 402 1980010188-41 g The period from 1975 to 1977 is centred on the minimum of solar geomagnetic activity and the terms "active", "unsettled" and : "quiet" characterize the levels of activity on a chosen day with respect to the average activity during solar minimum. [E.I. Loomer, G. Jansen van Beek (1969)]. The comparison of the predictions with the activity actually observed at individual stations has been made by dividing the predictions into three groups: correct, uncertain and incorrect. For a given reference observatory the prediction is classified as correct (uncertain or incorrect), when the observed and the predicted levels (i.e. quiet, unsettled, active) are the same (differ by one level, or differ by two levels, respectively). The results of this comparison are summarized in Figures 2a, 2b and 3. Figures 2a and 2b show the annual variationTHE AVERAGE PERCENTAGE OF %F60 CORRECT PREDICTIONS of the percentage ofcorrect and incorrect | /_,_ 1975 -1977 predictions for the Mean _50 / _ , individual observator les / / _ p_ / It is interesting to note/ \ / .s _ /\ / L40 -_/_ %# \ / that the number of /A_ ,4/ correct predlctions was greatest in the period March to May and least in November. The dependence RES of the percentage of correct predictions on 3C) 50 i / _/_ geomagnetic latitude isf sh wn in Fig. 3 (curve 2). It is evident that the CBB 40 level of activity is more accurately predictable at _ :U _50"1130 " V f"_ VI lower latitudes. The percentage of uncertain predictions (curve I) does not vary substantially with latitude. In summary, o.retowere incorrect, 47.4% wereMEA 50 uncertain and 45.5 were: correct in the three-year " 60 L40 I / I _ _ k / period 1975-1977 as 5 reference observatories. OTT Illllll Figure 2a. 403 1980010188-414 THE AVERAGE PERCENTAGE OF AVERAGE PERCENTAGE INCORRECT PREDICTIONS o/. 1975-1977 I'. I"- _0"/" _'-_' ) ] ]1975-1977' ' ' -;_ 10],.Mean-'" """ ""/ \'` I ]I'N'-" ,t',_°_ll ,=, cO''EcT' ' ]%%'_.._,/°J/ 40 _=) PREDICTIONS PREDICTIONS Ca o¢° 1_ _ _o _;o"_Eo..LAT _o]FCC____=.. Figure3. oJ 10. _ //\ 10 L0 _ _ A I I I I I | I 1 ,t • t I " J F M A M J J A S O N D Figure 2b, 3. Application of the predictions. To evaluate the usefulness of the predictions and to find out the needs of our users, a survey was made of 70 users in January, 1978 [Hruska, J. (1978)]. The predictions are being used by the private sector (geophysical surveys and mining companies, etc.), by government agencies (observatories, Geological Survey, etc.), and by universities. Sixty-flve percent of our users are in the private sector, 32% in government agencies and 3% in universities. The predictions are used on a continuing basis by 32% of users, regularly in certain periods by 43% of users, and occasionally by 25% of users. The predictions are used over a wide range of geographic latitude as seen in the following table: Geographic Lat. N. % of Users 30 - 40 o 7% 41 50 25% z 51 60 27% ; 61 70 IB% 71 - 80 15% over 81 8% 404 O 1980010188-415 They are used for various types of work as is described in the follow- ing table: % of Users Ground survey (general) 2_% Mineral exploration 22% Observatory work 20% Airborne survey 19% Research 8% Telluric methods 3% Education I% Other (information, calibration, radio communications-citizen bands, etc.) 17% Users classify the value of predictions for various purposes as follows: very useful useful not useful no. of replies Planning and organizing 25% 70% 5% 53 Reduction of operational expenses 15% 59% 26% 46 Time saving 17% 74% 9_ 54 On the basis of their experience only 6% of the users estimated that pre- dictions are correct in less than 50% of cases. Forty-eight percent found the predictions correct in 51 to 65% of the cases and 26% estimated that the probability of correct predictions is 66 to 80%. Seven percent of users estimated that the forecasts are correct in more than 81% of the cases. Twelve percent of the users stated that they did not have enoujh experience to estimate the reliability of the forecasts. Acknowledgments: We would like to express our thanks to the Space Environ- ment Services Center in Boulder for their provision of the data which are basic for our forecasts of magnetic activity. The author extends special thanks to P. H. Serson and E. I. Loomer for their interest and helpful comments. REFERENCES Hruska, J. (1978): Consulting Services Provided by the Ottawa Magnetic Observatory. Internal Report of Earth Physics Branch, DEMR, Ottawa. Loomer, E. I. (1976): The Canadian Magnetic Observatory Network. Seminar of Earth Physics Branch, DEMR, Ottawa. Loomer, E. I., G. Jansen van Beek (1969): The Effect of the Solar Cycle on Magnetic Activity at High Latitudes. Publ. of the Dominion Observatory, Vol. XXXVII, No. 6. Whitham, K., E. I. Loomer and E. R. Nibletc (1960): The Latitudinal Distribution of Magnetic Activity in Canada. J. Geophys. Res., 6_55, 3961. 405 ( _ % 1980010188-416 PPS76 - A COMPUTERIZED "EVENT MODE" SOLAR PROTON FORECASTING TECHNIQUE D. F. Smart and M. A. Shea Air Force Geophysics Laboratory Bedford, Massachusetts,01731, USA A procedure has been developed to generate a computerized time- intensity profile of the solar proton intensity expected at the earth after the occurrence of a significant solar flare on the sun. This procedure is a combination of many pieces of inde- pendent research and theoretical results. It is not a compre- hensive self-consistentanalytical method, but is a construc- tion of selected experimental and theoretical results from the entire domain of solar-terrestrial physics. 1. INTRODUCTION Any proton prediction system intended to be operationally useful must conform to the available data and operational constraints of the environment in which it is to be utilized. The prediction system de- scribed in this report is designed to operate with the information likely to be available in real time or near real time to the Air Weather Ser- vice Space Environmental Support Branch of AFGWC and the NOAA Space Envi- • ronmental Services Center. The information available to these organiza- tions is currently limited to the following: a. H-alpha flare reports (onset time, position, time of maximunl, flare importanceand end time); b. Radio observations at discrete frequencies (606, 1415, 2695, 4995, 8800, and 15400 MHZ); c. Soft X-ray fluxes from satellites such as SMS/GOES or SOLRAD; d. Very limited solar wind data from spacecraft; e. Real time solar particle data from satellites such as SMS/GOES; = these particle data are needed for verification of events and prediction updates. o 406 1980010188-417 The overall objective of the computer program is to be able (at the time of the H-alpha maximum) to make a prediction of the solar proton flux at the earth for the next 300 hours. The program must be able to operate on the data available at any time the forecaster selects. Since some of the desired input data may not always be available, algorithms are included to calculate nominal values for essential data that may be missing. The program also has an update capability so when additional data are available they can be included for new prediction attempts. In addi- tion if the solar particle data available in real time has features grossly different from those initially predicted, then the update capabi- lity can be utilized to include these gross features in calculating the future behavior of the event. This procedure is intended to be a tool for forecast personnel in making decisions for warnings and alerts. Users are cautioned that the computer merely executes algorithms included in the program. The closer a predicted profile conforms to actual obervations, the more confidently the prediction may be utilized. 2. CONCEPTS INVOLVED 2.1 General Concepts Solar protons are assumed to be accelerated in solar active regions during solar flare events. The X-ray, radio and optical emissions during the solar flare event are the indicators that proton acceleration is occurring. We interpret the "U-shaped" spectral signature of the radio emission (Castelli,et al., 1967) as the indicator that solar protons are being released from the solar active region. This assumption is based on a comparison of those flares having a "U-shaped" characteristic signa- ture in the radio power spectra (Castelli and Guidice, 1976) with the proton events given in the "Catalog of Solar Particle Events, 1955-1969" (Dodson et al., 1975) along with those of Castelli and Tarnstrom (1978). For operation of this program no minimum radio flux is required. Any "U-shaped spectral signature in the power spectrum of the radio out- burst in time coincidence with an observed optical flare is assumed to release solar protons from the accelerating regien. The computer pro- gram algorithm will compute a predicted maximum proton intensity expec- ted at the earth; this may range anywhere from a very large event to a solar proton flux so small it cannot be discerned in a particle sensor channel above a cosmic ray induced background. After the solar protons are released from the confines of the ac- celerating region, they will travel through the solar corona and some will propagate outwards along the interplanetarymagnetic field lines. 4O7 1980010188-418 Those emitted from the inner solar corona at a "favorable" position may : intercept the earth. From examining the solar proton data acquired during the past two solar cycles, we can generalize to the following phases of a solar proton event observed at the earth as illustrated in Fioure l. a, There is a propagation delay from the time of the solar flare until the first particles are observed at the earth. b. After the initial onset of particles, there is a rise in the solar proton flux until a maximum flux is observed. c. After the time of the maximum solar proton intensity, there is a i general exponential decay of the particle flux to background levels. This decay is usually much slower than the rising phase of the event. We employ the same _eneralizationsto predict the time-intensity profile to be expected at the earth after the occurrence of a solar proton producing flare on th¢ sun. Specifically,v_eknow there will be some propagation delay time necessary for the particles to travel from the sun to the earth. If we specify the maximum flux and the time of maximum, we can construct an exponential curve from the onset of the event to the maximum of the event. Then by specifying an exponential decay and appropriate decay constant, w,,ecan complete the calculation of the time-intensity profile expected fo_* _ particular solar flare. The actual profile may be quite different from the idealized curve illu- strated in Figure l, but the generalizations orovide an adequate de- scription of the time history of a solar protop event. STARTOF : INCREASE SOLAR TIMEOF FLARE MAXIMUM :_ MAXIMUMFLUX __TIME FROMRELATIVETIME PROPAGATION STARTTOMAX • DELAY Fig. I. Illustration of the general characterstic_ of solar proton events. : 408 .T 1980010188-419 ii 2.2 Solar Flare Generalizations ' The sudden release of large amounts of energy in localized regions of the sun results in the acceleration of some of the ambient solar plasma to high energies and in characteristicelectromagnetic emissions • from the site of the energy release. The optical phenomenon viewed when this occurs is called a solar flare, and there are concurrent emissions in the radio and X-ray wavelengths. At times the radio and X-ray emissions are detected without a corresponding optical emission, a phenomena that can occur when the solar flare is located behind the visible portion of the disk, particularly near the eastern or western limb. The exact pro- cess by which particles are accelerated is still a matter of s,cientific debate with a number of mechanisms currently proposed (see Svestka, 1976); however, these mechanisms are not germane to this report. The optical emissions, usually viewed in the H-alpha absorption line, are well suited for defining the position of the solar flare, although the dynamic range available in the actual emission above the solar background is somewhat limited besides being a somewhat subjective measure. There ar_ statistical trends betvmen the brightness of a flare and particle emission; however, the variance in the data between indivi- dual flares is approximately the same size as the total range of the flare brigJtness scale. In the radio emissions, the amount of energy released at various frequencies may increase by many orders of magnitude. Consequently, radio patrols of the sun are an excellent method of iden- tifying solar activity, although the resolution of the location of the emitting source on the solar disk may be very poor. The soft X-ray emission may have a dynamic range of many orders of magnitude above the quiet solar background; however, as with the radio emission, there is generally no resolution of the position of the emitting source. For the proton prediction program described in this report, we uti- lize the optical emission to provide the position of the accelerating region on the sun, and use the radio and/or X-ray emission to determine the size of the particle event to be expected from this sudden release of ' energy into the solar corona. If the "U-shaped" spectral characteristic signature is present in the radio data, we assume that some of the accel- erated particles were able to escape the confining magnetic fields of the active region into the solar corona. If the "U-shaped" spectral characteristic signature is not present, we assume that the energy re- lease was insufficientto allow these particles to escape from the region. 3. PROPAGATIONOF SOLARPARTICLES We will consider the propagation of solar protons from the flare site to the earth as being separable into two distinct and independent • phases. The first phase is diffusion from the flare site through the 409 19800] 0188-420 tI _ I solar coronato th_ "foot"of the Archimedeansprialpath formedby the interplanetarymagneticfield line betweenthe sun and the earth. The secondpart is the propagationin the interplanetarymediumfrom the sun to tF_ earth alongthe interplanetarymagneticfield lines. Figure2 providesan illustrationof the propagationconcepts. 3.1 Propagationin the Solar Corona The conceptswe have usedfor the propagationof solar protonsin the solarcoronaare similarto those advancedby Reinhardand Wibberenz (1974). We will make veryfew assumptionsas to the mannerof coronal transportexceptthat somestochasticprocessesdominateand the parti- cles havemany interactionsbetweentheir sourceat the flare siteand their releasepointalongan interplanetarymagneticfield line. In this contextwe takethe fundamentalelementsof solarparticlediffu- siontheoryas developedby a numberof researchers,someof the original resultsbeingthose of Axford(1965),Reid (1964),Burlaga(1967)and Krimigis(1965),and assumethatalmostall of the diffusiveeffects occur in the solarcorona. The distancethe solarparticlestravelin the solarcoronafrom the presumedsource(i.e.the solar flaresite)to the foot of the Archimedeanspiralpathfromthe sun to the eerth is designatedby the symbolO • For measurementof this distancewe willuse a he_iocen- : tric coordinatesystemwhere0 degreesis the longitudeof the plane containingthe earth-sunline (positivein the senseof solar rotation), and latitudeis measuredeither"up"or "down"fromthe solarequator. The solar flarewill hwe coordinatesdefinedas solar latitude, XF, :HIMEDIES SUN SPIRAL PATH - - ..... O eAeT. FLARE Fig.2. Illustrationof the propagationconcept. The coronal propagationdistance _ is illustratedby the heavyarc on the sun. Interplanetarypropagationproceedsalongthe Archimedean spiralpath fromthe sun to the earth. 410 1980010188-421 and hellocentrlclongitude _F" If we make the assumptionthatthe "foot"of the Archlmedeanspiralwill be at the solarequator,it wlll haveheliocentriclongitudeof _A. The heliocentriclongitudinaldis- tance fromthe flaresiteto the "foot"of the Archlmedeanspiralwill be _A - @F. If we alsoconsiderthattheremay be , heliocentric latitudeeffect,and assumethat the simplifiedgeometryof a right sphericaltriangleIs adequatefor simplecalculations,then cose - cos ( _A- _F ) cos x F. (I) Weassumethat coronal propagation is a function of _, Fromdiffusiontheorywe wouldexpectit to be proportlonalto _2. (See Wlbberenz(1974)for a contemporarydiscussionof diffusiontheory.) For largevaluesof _ the propagationdelaytime is dominatedby the dif- fusiontime,and so we selecta normalizationcontentsuchthat we can obtainsomeof the verylongdelay timesreportedfor Easternhemisphere flares. Typicaldata setscontainingonsettimesof particleeventsat the earthare thoseof Baroughet el. (1971),and Lanzerottl(1973). As a resultof diffusionin the solarcoronafromthe flaresiteto the "foot"of the Archlmedeanspiral,and the Inhe, t assumptionthat some stochasticprocessesare operating,we would_ect that thereis a solarparticlegradientexistingIn the solarc,,,_asuchthat the protonintensitydecreasesas a functionof distanc_fromthe flaresite. ThereIs someobservationalevidencefor the existenceof sucha gradient (Goldet el., 1975;HcCrackenand Rao, 1970;Roelofe( al., Ig7G;HcCrocken et el., 1971;Roelof,1976). The observationalevid.n_¢ suggeststhat the gradientmay varyfromcase to case;however,for this prooramwe have assumedthat thisgradientfrom the presumedparticlesource(i.e. the flarelocation)to the releasepointof sole,_._tonsobservableat the earth {i.e.the "foot"of the Arc),imedeanspiralof the interplane- tarymagneticfieldlinebetweenthe earth and the sun) is 3e per radlan. Therefore,an observerat cne astronomicalt,nltwho is connectedvia the interplanetarymagneticfieldlineto the hc,iographiclongitudeof the flaringregionwouldobservethe maximumpossibleparticleintensity, IF. An observerwhoseinterplanetarymagneticfieldconnectionis at a distanceof _ fromthe flaringlocationwould observean intensityof IF exp (-3_ ). 3.2 Propagationin the InterplanetaryMedium Afterthe particlespropagatethroughthe solarcoronaand are re- leasedintothe interplanetarymedium,theyessentiallypropagatealong the interplanetarymagnetlcfieldline. Duringthis phaseof theirpro- pagationwe assumethattheirmean free pathlengthis of the orderof 0.I to 0.3 AU. Theseassumptionsappearto be consistentwith someof the V_4 more recentstudiesof solarparticlepropagation(Roelof_no .,,_Igis: 1973; Roelof, 1975; Palmer et al., 1978). While these ass;Jmpttons are net \ 411 1980010188-422 "scatterfree"propagatlon,the numberof scatteringcentersthe partlcle may encounterbetweenthe sun and the earth Is relatlvelysma11;there- fore,much of the earlywork regardlngdiffusivecosmicray transportIs appllcableunder"thesecondltlons. For thlscomputerprogramwe havemade the slmplestposslbleassump- tions regardln_transportIn the Interplanetarymedium. Theseassumptions are as follows: a. The particles travel essentially along the interplanetary magne- tic field ltnes with a veloctty which is a function of the parttcle energy. b. Diffusion perpendicular to the interplanetary magnetfc field ts assumedto be negligible. c. The mtntmumdtstance to travel from the sun tu the earth is the dtstance along the Archtmedeansptral path. The mtntmumpropagation delay wt11 be for particles that essentially travel along the interplanetary magnetic field lines with very little scattering, so for scatter free onsets the propagation time from the sun to the earth will be the distance traveled divided by the particle velocity. After the initial onset it ts reasonable to expect that some scattering has taken place and that someaspects of diffusion theory are applicable. Almost all theories involving differential transport show that the timeof maximumis proportionalto the squareof the distance traveled. (SeeWibberenz,1974). We will furthermake the assumptionthat the averageparticlepropa- gatesthroughthe interplanetarymediumwith a bulk velocitywhich Is half the speedto be expectedfromthe particleenergyof an average pitchangleof 60 degrees. The polarformof the equationof the Archimedeanspirali_ p : . (2) If we normalizeto the earth'sorbitthena = I AU and # = CA in our heliographicoordinatesystem. The distancealongthe Archimedeanspi- ral path fromthe sun to the earth is the integralalongthe arc length CA 2 d P D = f [ P + (d-'-O--)d# ]. (3) -U The resultsin our system,norma]izedto I AU is ' 2 + I)]/¢A} (4)2 + I + [In (_A+_ OAD = I/2 V OA The solarlongitudeof the "footpoint"(CA) of the Archimedean 412 ), c " | 198001O188-423 spiral path on the sun can be computedby considering that a plasma ele- ment leaves the sun at the earth-sun line and propagates recttltnearly to 1 AUwith an awrage velocity v. Whenthis plasma element reaches 1 AU, the sun wtll have rotated such that the solar longitude of the starting point on the sun will be to the west of central meridian and equal to CA, If the plasma carries "frozen in" magnetic fields, then +.he "foot point" of the interplanetary magnetic field ltn-. that extends from the sun to the earth can be given by CA = "--_r (5) where _s is the solarsynodicrotationr_te (13.333degrees/day)and r is l AU. Ifwe expressthe Archimedeanspiralequationin the polarform normalizedto l AU, then 404 " (6) where V sw is the measuredsolarwindvelocityin km/secat the earth (I AU) and 404 is the nominalsolarwindvelocitywhen the solar rotation is equalto one radianduringthe timeit takesthe plasmaelement o travelI AU. The protonvelocityis givenby yrB = l - [(E/mcZ) + i]Z (7) where E is the kineticenergyof the parLiclein MeV, _ is the velocity of the particlecRmparedwiththe velocityof light,and the rest energy of the proton,mc_, has the valueof 938.323MeV. (Rote: 19=V/c). The minimumtime in (hours)requiredfor a protonto propagatealong the Archimedeanspiralpathfromthe sun to the earth witha zero pitch angle is 0.1333D/_ • 4. EVENTPROFILECONSTRUCTION 4.1 ParticleIncreaseOnsetTime Reviewarticleson solarparticlepropagationnotethatthere is an apparentanomalousdelay of about 15 minuteseven for relativisticsolar particlesto propagatefromthe sun to the _arth. Beginningwith this constantthe propagationdelay time is the sum of the coronalpropaga- tion time plusthe interplanetarypropagationtime alongthe lengthof the Archimedeanspiralfr_ the sun to the earth for the fastest(highest e_ler_v)protonsdetectedin a specificso1-rprotonsensor. We assume that for a givenchanne__his is Onset= 0.25 + 0.1333D/_ + 4 0 2 (8) 413 1980010188-424 J_ _ where D is the distance along the length of the Archimedean spiral from the _ _ sun to the earth and _ is the velocity of the most energetic proton measured * by the channel. This calculation is carried out for each energy channel to _ compute the particle onset time. * _ The distribution of onset times expected for 30 MeV protons for nominal solar wind speeds (taken as 404 km/sec) is shown in Figure 3. The data points shown on the figure are taken fr,_mBarouch et al., (1971) and _ndicate i typical variations that may be expected. The minimum in the figure corre- sponds to a flare at the "foot point" of the Archimedean spiral path between the sun and the earth (57.2 degrees west of central meridian). Variations in the solar wind will result in changes in the computed "foot point" of the Archimedean spiral path. ' When dealing with actual data, the onset cannot be recognized until the additional flux is larger than the background of the channel. The onset time that will be apparent from the program output is the time when the particle flux increases above the background. 4.2 Time of Maximum The time of maximum in each channel is computed by assuming the parti- cles have an average bulk velocity that is about half of the sp_ed of parti- cles of the same energy traveling rectilinearly. Most theoretical models _ find the time of maximum proportionalto the square of the distance along the field line from the sun and so this _actor is included. zsoo : _ _oo , /48: | 200 __. ,o,, I ]i _ ,. . i _ i , 90" 70" 50" 30" IOeO'l_ 30 e 50 ° 70" 90" SOLAR LONGITUDE OF FLARe" Wut Fig. 3. Distributionof onset time of 30 MeV protons as a function of solar longitude. The data points are the measurements of Baro_ch et al., (1971) 414 1980010188-425 The maximum number of particles in a specified energy range is assumed - i to be at the lowest energy of the channel. The time of maximum is computed , by Tm (in hours) : 2.0 (0.1333 D2/_ ) + 80 2• (9) The distribution of the computed time of maximum as a function of solarL longitude is illustrated in Figure 4. The data points are from Reinhard and Wibberanz (1974) and show the typical range of variations that can be expected. The minimum in the curve corresponds to a flare at the "foot point" of 57.2 degrees for the Archimedean spiral path between the earth and the sun computed from a nominal solar wind of 404 km/sec. Similar curves for 30 MeV protons and 60 MeV protons cannot be shown here for lack of space, but they are not significantlydifferent from the curve given in Figure 4. Figure 5 illustrates the computed curve for our nominal solar wind speed (heavy solid line) as compared with the extensive data set of Van Hollebeke et ai.(1975) showing the distribution of the time of maximum of 20 to 80 MeV protons with solar longitude. Variations in the solar wind speed will shift the curve to correspond to the position of the "foot point" of the Archimedean spiral path. As an example a fast solar wind velocity of 808 km/sec results in the foot point of an idealized Archimedean spiral to be at 28.6 degrees west of central 80 . 70 " PROTONS=, I0 MeV E_' "_' •E_' ' b ' 'W_' 'W_' 'w_' '*,_, SOLARLONGITUDEOFFLARE : Fig. 4. Distribution of time of solar proton event maximum as a function of solar longitude. The data points are those of Reinhard and Wibberanz (1974). 415 C. 1980010188-426 3o _/eO" . ) EAST 120" 80 ° 40" 0 40" 80" 120" 160" WEST i" HELIOLONGITUOE WtTH RESRECT TO THir CIENTRAL M_R(DrAN _ Fig. 5. Distribution of the measured time of maximum of 20 to 80 MeV protons obtained by Van Hollebeke et al. (1975) as compared with our coronal propagation parameter. meridian while a slow solar wind velocity of 323 km/sec results in _ "foot point" at 71o6 degrees west of central meridian. Thus, the variation intro- duced by changes in the solar wind velocity can result in large displacements from the idealized curve of maximum times. The sun's rotation introduces i other effects that modulate the shape of the idealized time of maximum curves shown in these figures. (Co-rotationeffects are discussed in more detail in section 4.4.2 of this paper.) 4.3 Maximum Intensity 4.3.1 Maximum Intensity Along the Archimedean Spiral Path From the Flare • The maximum intensity at some energy is predicted by converting an elec- ; tromagnetic emission parameter (usually radio flux or x-ray flux) to a proton flux. Many different algorithms can be used but we have followed the proce- dures developed by Castelli and co-workers (private communication) and have used relationshipsbetween integrated or semi-integrated radio flux and peak proton flux. We have modified the ,esults obtained by Straka (1970), Newell (19/2), Castelli et al. (1973) and Cliver (]976) to develop a set of forecast parameters for the peak flux of protons > lO MeV or protons > 5.2.MeV from . radio burst data, and we have modified the results of Kuck et al. (1971) to develop forecast parameters for peak proton flux > lO MeV from integrated x-ray data. A summary of some of the algorithms used is given in Table I. 416 1980010188-427 :. We solve for a differential energy spectrum such that f K E• + 1P(>E) : K E_ dE :_II (lO) where _ is the slope of the differential energy spectrum and K represents the flux at l MeV. For example, if we are using an algorithm for the > 10 MeV peak flux, the above equation is solved by an iterative technique so that E = lO00 MeV : P(>IO) (11)KE? E = lO MeV where • is the variable slope described in section 4.5. The radio data input to the program is assumed to be in SFU (solar flux units); however_ it can accept a large variety of different units (SFU, Joules, watt sec) as long as the units are identified during the input. The software makes an internal conversion to the correct units for use in the algorithms. The X-ray data input can be in either units of watts per square meter-second : or ergs per square centimeter-second. The various algorithms given in the following tables are formated so the term inside the parenthesis is the riometer absorption. The conversion from riometer absorption to proton flux follows the conventional form of J : CA2 where C denotes a conversion constant. (See Section 5 for a further discus- sion of these conversions.)% i 4.3.2 Maximum Intensity at the Earth After the maximum intensity along the Archimedean spiral path from the flare has been computed, the maximum intensity at the earth is derived by attenuating the flux through the coronal gradient over the heliocentric dis- tance in the corona between the flare position and the solar equatorial longitude of the foot point of the Archimedean spiral path from the sun to the earth. (See Section 3.1 for a previous discussion of this concept.) If the coronal gradient can be specified as a simple exponential func- tion in terms of the heliocentric angle between the flare site a,d the foot of the interplanetarymagnetic field line "connecting" the earth to the sun, then the flux at the earth's Archimedean spiral "foot point", Ie, will be re- duced from the flux at the flare site If by Ie = If exp - (g _ ) (12) where g describes the coronal gradient. 418 -\ r_ _ 1980010188-429 There is another factor that attenuates the earth-sensed solar flare proton flux. We contruct a pseudo time-intensity profile for the proton flux at one AU along the interplanetarymagnetic field line "connected" with the : flare site. We assume that the decay rate discussed in section 4.4.1 com- : mences immediately after the particle acceleration. This, coupled with the time delays associated with the coronal propagation effects to other hello- longitude positions, (See Section 4.2) implies that the flux along interplane_ tarymagnetic field lines removed from the flare site will have an additional flux decrease by a factor of exp -(tm/td). For practical applications of : proton flux prediction at the earth, tm is the time of maximum intensity for the specified channel and td is the l/e decay constant in the co-rotating reference frame of the interplanetarymagnetic field line "rooted" to the flare site (See Section 4.4.2). 4.4 Event Decay 4.4.1 Decay Constant The decaying portion of the event is modeled after the principles of collimated convection (Roelof, 1973). After making a number of simplifying : assumptions (some of which are that the particle flux can be represented by a simple power law, the anisotropy of the particle flux is small, the inter- planeterymagnetic field falls off as r-2, and that the particle flux gradient is field aligned and small), the decay constant, as derived by Roelof (private communication) can be represented as : T _3D/[4V(I_I + I)] (13) ? where T is the I/e decay constant, D is the distance along the Archimedean spiral path, V is the solar wind velocity, and • is the differential spectral : exponent. 4.4.2 Effect of Coronal Gradient on Apparent Decay Rate. By assuming that the source of the particle flux is the active region of the flare, and that there exists a particle gradient in the corona as de- scribed earlier, and that after the particles are released from the corona the particle propagation is along the interplanetarymagnetic field with negligible diffusion perpendicular to the field lines, then the rotation of the sun affects the apparent decay rated observed at a fixed location in space (such as the earth at l AU). An example of this effect is shown in Figure 6. In this case the decay rate on the Archimedean spiral path co- rotating with the solar particle source position is constant. If the particle : flux along the interplanetarymagnetic field lines is considered as a co- _ rotating structure being convected out of the system, and if the particle source is rotating toward the "foot point" of the Archimedean spiral from the sun to the earth, such as is the case for flares on the eastern hemisphere or near central meridian, the apparent decay rate will be slower than if the particle source is rotating away from the "foot point", such as is the case for extreme western hemisphere flares. Also, the solar wind velocity , 419 C 1980010188-430 w¥ - SMSIGO[S _ ,,,=_.--.-,_._,_,, > 0 '_m4t. _-*-C - 3_.C t?b :._ l ) _) i I _,% , '_ ,IG 'cite i ' CE ) i .C'_ n _ ' , 6") ,_ ! (...) C2_ "--.......---i .........i.....................: ............ . _ _'_t_.........................____ __ L_ ......._ .__:.__........_ _.' X __.1 :, a, '_ ....... _ ....... '- ............. __,,. ........................................................... "I I:I _,'_L S "- - 4 3 _ I .7" • -]'- ............................. . ............................. _'--- _L ..... _ T-:--_-_--_-_-_ ........ > | 8 ._,_'.. _." ¢. - _3 C (Zb Z: Or_ ) 0 ';_'., '_,G - JOU ' Y?2t, ' _..... _ ....................... : ......... i ............... "77":',::':'_.:., ..,_-_.,_........ ............._-_ ---_ ---,.......--_ ' _.J- _ ......._.........._i_.........__L_..........._- ........ •. rl _ ,_._.__...__._.___.....__.......................... __... .................................... _< Fig. 6. Predictedsolarprotontime intensityprofilesfor the earth from a : solarflareat 0000 relativetime (positionedat the left axis). The t%panel illustratesthe profilefor a westernhemisphereflare at 30°N, 60 W. : The bottompanel illustratesthe profilefor an easternhemisphereflareat 30°N,60°E. For both casesa nominalsolarwind velocityof 404 km/secand a differentialspectralslopeof -4 was used. Eachhorizontalline is one • day and each verticalline is a factorof lO in flux. . 420 I y--- 1980010188-431 , T t t| will affect the decay rate so the effects of co-rotation and solar wind velocity changes must be considered in constructing time-intensity profiles. 4.5 Differential Energy Spectra , We assume that at the time of the maximum flux of l MeV protons it is possible to construct a differential energy spectrum that is representative of the entire event. Our representativespectra have an average reference value of T R between IO and 200 MeV; however the energy dependent slope be- comes slightly shallower at low energies and slightly steeper at higher energies. The slope of the spectrum is modulated as a function of the parti- cle velocity (B = v/c) so that at any energy slope of the differential energy spectrum Y_ is YR + _ " _ where _ is the normalization constant necessary to give the proper average value between IO and 200 MeV. For a predicted reference slope we prefer to utilize the techniques de- veloped by Bakshi and Barron (1974, 1975, 1979) to derive the slope of the integral solar proton spectrum (which we convert to a differential energy spectral slope) when appropriate maximum and minimum flux values of the U-shaped signature in the radio power spectrum at radio burst maximum are available. If a prediction of the spectral slope is not available, and no explicit specification of the spectral slope has been entered as input data, a default value is computed. The default spectral algorithm is based on the work of Van Hollebeke et al. (1975) shown in Figure 7. -VI. , . , . , . , . , • ,w. , • I . , - , . , , _ . , . , . _ . v , _ A \ ZO-O0 Uev too'toNs _ '_\ . i . | . i , i , i , I . i , i , I , i . I , i _ i . I . i . i . i EAST IZO" I10" 40' _. 40" 80" i_O• tGo"V_ST H[LIOLONGITUD[ WITH Ht[SPECT TO TN[ C[NTmAL MERIDIAN Fig. 7. Variation of the spectal index _ in the 20-80 MeV range as function of the heliolongitude. The dashed contour lines enclose 92% • of the "normal" events. The solid line is a least square fit that can be represented by --2.7[I + 0/2]. r 421 1980010188-432 We havemodifiedtheir functionby assumingthat the hardestspectra(the ;: minimumin the solidcurveof Figure7) shouldbe alongthe interplanetary magneticfield linesleadingawayfrom the flaresite,and that it will vary at otherheliolongitudesapproximatelyas the Van Hollebekeet al. (1975) function. The defaultdifferentialenergyspectralslopealgorithmused is mR = -2.5(l.O+ _/2) (14) t 5. PREDICTIONOF POLARCAP ABSORPTION q Afterthe eventtime intensityprofileshavebeen computedfor each energy rangeof interest,it is possibleto convertto specializedoutputssuchas the predictionof the profileof the ricmeterabsorptionin the polarionosphere. We have usedthe conceptsand relationshipsdevelopedby Sellerset al. (!977) to derivesimpleequationsthat in theiropinionwere relativelyindependentof the slopeof the solarprotonenergyspectrumto convertsatellitemeasured protonfluxesto equivalentriometerabsorptionthatcomparedfavorablywith the outputof more elaboratecalculationsemployingthe solar protonenergy depositionand ionizationratesin the atmospherewiththe reactionrates of atmosphericionsand electronsin the ionosphere.Theseauthorsselected ; the integralprotonfluxgreaterthan 5.2 MeV to be a good predictorof the riometerabsorptionin the sunlitpolarionosphereand the integralproton fluxgreaterthan 2.2 MeV for predictingthe expectedriometerabsorption in the nighttime polarionosphere.Thereequationsare:L_ A (day)= _ 0.115_J(>5.2) (15) A (night)= _ 0.02_J(>2.2) (16) By includingthe integralfluxabove 2.2 and 5.2 MeV inthe solarproton time intensityprofilesit is possibleto generatedisplaysof the predicted riometerabsorptionexpectedin the polarcap duringa solarprotonevent. The generalapplicationof equation15 was used in convertingthe origi- nal algorithmsbetweenradio fluxand riometerabsorptionin the polarcap developedby Strakaand Barron(1969),Straka(1970),Newell(1972)and Cliver (1976)to equivalentintegralprotonfluxesso a relativelystandardized set of algorithmscouldbe employedfor conversionof radiodata to peak protonflux, (see Tablel for listingof the variousalgorithmsused in PPS76). h22 -o _, 1980010188-433 6. MULTIPLE FLARE INJECTION OF PROTONS PPS76 makes calculation of the predictions of the proton flux for up to 15 different proton energy ranges in addition to calculations of the predicted day and night riometer absorption in the polar cap. The program computes the projected time intensity profiles for the next 300 hours, and if there are subsequent new flares acceleratin_ solar protons, these new solar proton injections are added to the residual flux from the previous flare(s). Varia- tions in the amplitude of the new events, variations in spectra, and changes in the solar wind velocity provide for a fascinating variety of possibilities. 7. OPERATIONAL USE OF PPS76 PPS76 has been in operational use at both of the principal forecast centers in the United States since 1976. This composite prediction procedure appears to be a very useful tool for the prediction of "normal" solar proton events and polar cap absorption events. It is more limited when it is used for high energy (more than a few hundred MeV) or relativistic solar proaton events. Cliver et al. (1978) have discussed the prediction generated by this technique during the September ]977 series of solar proton events and the comparison of these predictions with the actual observations available in near real time at one of the forecast centers. An example of some of the capabilities of this prediction technique is given in Figure 7. These time intensity profiles (or would be predictions from the input data listed at the top of the figure) may be compared with the actual measurements of the IMP 8 spacecraft shown in Figure 8. The prediction illustration in Figure 8 does not allow for enough high energy protons to produce a gound level solar cosmic ray event when in fact the 30 April 1976 event was a relativistic solar cosmic ray event detected at the surface of the _arth. A somewhat similar event occured on 7 May 1978. The forecast information available predicted only a small PCA with a hard energy spectrum. A small PCA of the forecast magnitude (2.5 Db) did occur at the forecast time; however the 7 May 1978 event was the largest rela- tivistic solar proton event since 1960. 423 t ] 9800 ] 0 ] 88-434 - ,l SMS/GOES _. _,, ,o-_oo ,,._,,_ sl'l._ 18.c - 3_l.c, +2', 2'L,o..J '.jr- _ _l.O,_ 4_ . _00'_ CT 4D| F" _M_ 11_ t 50 ,!r I LO I 0:_ o GI% 50 . _00 Ir22_ : =. I I _ ....................... I|r f _ z: '' -__:-_" "::ii,, (-3 C_ III ¢t ",,2\ 121 Fig 7. Predictedtime intensityprofilefor the 30 April 197b solar proton event. The energychannelsshownare thosemonitoredby the SMS/GOESsatellite. N , ! ii i _+ _ + !% ' _i__ -"_ O " l|l I_ I|I I|O l_,il llg Fig.8 The 30 April 1976 solarprotoneventas observedby the JohnsHopkins ; UniversityAppliedPhysicsLaboratoryexperimentson I_ 8. "+ 424 ,+.. L..--- ,-,?-:-+_ + ,,.. +.++ 1980010188-435 REFERENCES Axford,W. I., (1965): Anisotropicdiffusionof solarcosmicrays. Planet. SpaceSci., 13:1301. : Bakshi,P., and W. Barron,(1974): SpectralCorrelationsBetweenSolar Flare RadioBurstsand AssociatedProtonFluxesI. AFCRL-TR-74-0508.Air Force : CambridgeReesearchLaboratories,HanscomAir ForceBase,Massachusetts. Bakshi,P., and W. R. Barron,(1975): SpectralCorrelationBetweenSolar Flare RadioBurstsand AssociatedProtonFluxes,II. AFCRL-TR-75-0579.Air ForceCambridgeResearchLaboratories,HanscomAir ForceBase, Massachusetts. Bakshi,P. and W. R. Barron,(1979): Predictionof SolarFlareProtonSpectral Slope FromRadioBurstData.J. Geophys.Res., 84:131. i Barouch,E., M. Gros, and P. Masse,(1971): The Solar LongitudeDependence of ProtonEventDelay. Sol. Phys., 19:483. Burlaga,L. F., (1967): Anisotropicdiffusionof solar cosmicrays,J. Geophys. Res___.:.,72:4449. Castelli,J. P., J. Aarons,and G. A. Michael(1967):FluxDensityMeasurements of RadioBurstsof Proton-ProducingFlaresand NonprotonFlares, J. of _eoph_s.Res. 72:5491. Castel]i,J. P., W. R. Barron,and J. Aarons(]973): SolarRadioActivityin August]972. AFCRL-TR-73-O086.Air ForceCambridgeResearchLaboratories, HanscomAir ForceBase,Massachusetts. Caste]l,J. P., and D. A, Guidice(1976): Impactof CurrentSolarRadio PatrolObservations.Vistasin Astronomy, 19:355. Castelli,J. P., and G. L. Tarnstrom(1978): A Catalogof ProtonEvents1966- 1976HavingNon-ClassicalSolarRadio Spectra. AFGL-TR-78-O]2].Air ForceGeophysicsLaboratory,HanscomAir ForceBase,Massachusetts. . Cliver,E. W., (1976): ParentFlareEmissionat 2.8 GHz As A Pr_uictorof , the PeakAbsorptionof Polar-CapEvents. NELC-TR-2015.NavelElec- tronicsLaboratory,San Diego,C_lifornia. C1iver,E. W., J. A. Secan,E. D. Beardand J. A. Man]ey (1978): Prediction of Solar ProtonEventsat the Air For_eGlobalWeatherCentra]'sSpace EnvironmentForecastingFacility. In: Proceedingsof the 1978Symposium on the Effectof the Ionosphereon Spaceand TerrestrialSystems. 425 1980010188-436 iDodson, H. W., E. R. Hedeman, R. W. Kreplln, M. J. Martres, V. N. Obridko, M. A. Shea, D. F. Smart and H. Tanaka (1975): ,.Catalo_of Solar Particle Events 1955-1969. Vo1. 49 of the Astrophysics and Space Science Library, l'B'(t'e'd"by Z. Svestka and P. Simon, D. Reidel Publishing Co., Dordrecht, HolIand. Forman, M. A., (1970): The Equilibrium Anisotropy in the Flux of lO MeV Solar Flare Particles and Their Convection in the Solar Wine. J. Geophys. Res., 75:3147. Gold, R. E., E. C. Roelof, J. T. Nolte and A. S. Krieger (1975): Relation of large-scale coronal x-ray structure and cosmic rays: 5. Solar wind and coronal influence on a Forbrush decrease lasting one solar rotation. Proc. 14th InternationalCosmic Ra_'Conference (Munich), 3:1095. Krimigis, S. M., (1965): Interplanetarydiffusion model for the time behavior of intensity in a solar cosmic ray event. J. Geophys. Res., 70:2943. Kuck, G. A., S. R. Davis and G. J. Krause (1971): Prediction of Polar Cap Absorption Events. AFWL-TR-71-1. Air Force Weapons Laboratory, Kirtland AFB, New Mexico Lanzerotti, L. J., (1973): Coronal Propagation of Low-Energy Solar Protons. J. Geoph.ys.Res., 78:3942. McCracken, K. G., and U. R. Rao (1970): Solar Cosmic Ray Phenomena. Science Reviews, 11:155. McCracken, K. G., U. R. Rao, R. P. Bukata, and E. P. Keath (1971): The Decay Phase of Solar Flare Events. Sol. Phys., 18:100. Newe11, D. T., (1972): Forecasting Peak Proton Flux and PCA Event Magni- Cudes Using "Flash-Phase' Integrated Radio-Burst Flux Density. AFCRL- TR-72-0543. Air Force Cambridge Research Laboratories, Hanscom Air Force Base, Massachusetts Palmer, I. D., R. D. Zwickl, W. R. Webber and F. B. McDonald (1978): Spectrum of Mean Free Paths Near the Earth in the Solar Cosmic Ray Event of April 29, 1973. J. Geophys. Rest, 83:2461. Reid, G. C., (1964): A diffusive model for the initial phase of a solar proton event. J. Geophys. Res., 69:2659. Reinhard, R., and G. Wibberenz (1974): Propagation of Flare Protons i the Solar Atmosphere. !:vl.Phys., 36:473. Roelof, E. C., (1973): ,_Jewaspects of interplanetary propagation revealed by 0.3 MeV solar proton evo.ntsin 1967, _n: Solar-TerrestrialRelations 411. University of Calgary, Cdnada. 426 i 1980010188-437 Roelof,E. C., (1975): Scatter-freecollimatedconvectionand cosmic-ray transportat l AU. Proc. 14th InternationalCosmicRay Conference (Munich), 5:1716. Roelof,E. C., (1976):SolarParticleEmission.In: Physicsof Solar PlanetaryEnvironments,1:214. Publishedby the _nericanGeophysical Union,Washington,D.C.,USA. Roelof,E. C., R. E. Gold,S. M. Krimigis,A. S. Krieger,J. T. Nolte, P. S. Mclntosh,A. J Lazarusand J. D. Suillvan(1975): Relation of large-scalecoronalx-raystructureand cosmicrays: 2. Coronal controlof interplanetaryinjectionof 300 keV solar protons.Proc. 14th InternationalCosmicRay Conference(Munich), 5:1704. Roelof,E. C. and S. M. Krimgis,(1973): Analysisand synthesisof coronal and interplaneteryenergeticparticle,p!asmaand mdgneticfielaobserva- tionsover three solarrotations. J. Geopn_s.Res._._.,78:5375. 4 Selle,'_,B., F. A. Hanser,M. A. Stroscioand G. K. ,ates(1977): The night and day relationshipsbetweenPolarCap RiometerAbsorptionand Solar Protons. RadioScience, 12:779. Straka,R. M., (;970): The Use of SolarRadioBurstsas Pre ictorsof P_'oton EventMagnitude. AFCRL SpaceForecastingResearchNote 2. Straka,R. M. and W. R. Barron,(1969): MultifrequencySolarRadioBurstsas Predictorsfor ProtonEvents. AGARDConferenceProceedings,Noo 49. Svestka,Z. (1976): SolarFlares,Volume8, Geophysicsand Astrophysics Monographs,D. ReidelPublishingCo.,_ordrecht,Holland. Van Hollebeke,M. A. I., L. S. Ma Sung,and F. B. McDonald,(1975): The Variationof SolarProtonEnergySpectraand Siz,_Distirbutionwith Heliolongitude.Sol.Phys., 41:189. Wibberenz,G., (1974): InterplanetaryMagneticFieldsand the Propagation of CosmicRays. J._.Geophys., 40:667. 427 t Jl 1980010188-438 • jr- • _ Summary of Forecasters' Meeting on April 25, 1979, .I During the International Solar-Terrestrial Predictions Workshop J. B. Smith i. INTRODUCTION The presence at the workshop of many who are involved in Solar- Terrestrial predictions and of others who are the users of the monitoring and prediction services offered a unique opportunity to discuss currently available data and forecast services and to suggest improvements in those products. T,,erefore, a meeting of solar-environmental forecasters and customers was held on April 25, 1979, (following two days of Working Group meetings) to discuss the early progress of each of the working groups. To introduce some order to the discussion of several subjects by the large group of individuals, the working groups were first discussed by represun- tatives of those working groups _ followed by open group discussion. Notes of the discussion° and recommendations are presented below. A summary of the discussion of solar _ctivity forecast verification appears after the working grc'JD remarks and is followed by a list of the participants in the forecasters' meeting. Although recommendations generally represent a m'jority _iew, some comments may be neither a consensus statement uor a m.Jjority opinion. 2. WORKING "ROUP DISCUSSIONS 2.1 Long-Term Solar Activity Predictions 1. Histor;,ally, the sunspot count, particularly as recorded by the Zurich OLser• =tory, has served as a valuable index of solar activity and offe-s the primary long term data base for analysis and prediction. (A more objective and much used activity index is the solar radio flux at I0.7 cm, observe," and reported by Ottawa since 1974.) The continuation of these observations is considered of primary importance to long-term solar activity predictions and to the understanding of the cyclic variations in solar activity. 2. For physical reasons, forecasters and users emphasize that observa- tions at._predictions from other nun-traditional data, particularly EUV, are very desLrable. Improvement in more traditional data bases, such as solar rasio flux and geomagnetic indices, is needed. 3. The validity of the short-term, mid-term, and long-term periodicities in the data should be verified. Are the periodicities real and predictable? Statistical analysis should be applied to determine the validity of the data fluctL,_tions. 428 - I F ] 9800 ] 0 ] 88-439 4. It was suggested that the long-term predictions working group could have benefited by the addition of a representative from the area of dynamic modeling. This was certainly a popular opinion and should be considered for any future meeting. 5. Group discussion indicated that intermediate time-scale forecasting (on the order of a solar rotation to a few months) has not been given deserved attention. Techniques for predicting the evolution of active regions and of active longitudes should be evaluated for current application. A 27 day forecast'of daily 10.7 cm flux and Ap values is a recently added service at SESC in Boulder. 2.2 Short-Term Solar Activity Predictions I. Users and potential users of solar-terrestrial forecast services are not always fully aware of the services and data available and of how to interpret the data provided. Similarly, the forecasters may not always understand the requirements of the users. Increased interaction between the forecasters and their customers is clearly desirable. Z. Forecast centers, although closely related by data and product exchange, do not presently share all available data. Different data sets are availaEie to the different forecasters, although an internationally exchanged subset is available to all. A real-time exchange of more of the important infor_=tion is strongly suggested. The discussions and submitted workshop papers ,,ave also made apparent some difference in the forecast techniques applied by the centers. The various forecasting techniques should be shared among centers and perhaps similarly evaluated. 3. Although many current solar prediction techniques are quite qualita- tive, or input data are used quite qualitatively, objective solar predic- tion techniques have received a good deal of attention and are discussed at length in separate sections of these proceedings. It seems necessary to quantize the forecasts where possible. New observational parameters may be necessary for further progress. The opinion was expressed that prediction accuracy has progressed as far as is possible with the data currently available. 4. F_ asts for periods of less than 24 hours are usually made on an as-required basis. Forecasters should determine whether there is a recur- ring need for forecasts of a few hours, as was implied here. 5. Predictions of an active regions' flare potential should include a prediction of ;*s capability for producing proton events. 6. RECOMMENDATION. There is a strong recommendation (with general agree- ments) for forecaster exchange between centers. " 429 1980010188-440 2.3 Geomagnetic Disturbances There were no members of these study groups present in the forecasters meeting. Discussion of the forecasting techniques and application was therefore limited. 2.4 Magnetosphere - Ionosphere Interaction Only questions were posed: Why does a geomagnetic disturbance produce, or not produce, ionospheric effects? What are the cause and effect relation- ships, the coupling mechanisms, and the geomagnetic distribution? 2.5 Ionosphe,-ic Predictions i. Improved measurements of more directly involved wavelengths, e.g. EUV, may be beneficial in ionospheric monitoring and prediction. 2. There is a need to know what ionospheric instrumegtation is currently operational. 3. An expressed opinion that ionospheric predictions are not always based on "real physics" was perhaps not fully subscribed to. 2.6 Solar - Weather Predictions I. Further identification should be made for solar parameters that should be evaluated for possible correlation with long-term (drought, climate, planetary wave studies) and short-term (pressure changes, electrical storms, etc.) weather phenomena. Greater interaction with solar_.activity prediction groups is desirable. 2. The coupling of the troposphere to solar activity is certainly not ur,derstood. Some things to be looked at are the energy input, the solar constant, coronal holes, sector boundaries, chemistry, triggering, geomag- : netic storms, etc. 3. At this point, it appears that the need for data and research is greater than for solar predictions. We are clearly not now in a position to : use solar forecasting to predict the Earth's weather. 2.7 Propagation Predictions Those who use the propagation forecast were well represented and the discussion and comments centered on communication between the forecaster and the user. Those who use the predictions are not always familiar with the standard terminology and may not understand the implications of the forecast. A partial solution is certainly user education and some forecaster education about tailoring their product for the customer. Dissemination of the short- 0 term predictions is still often too slow. 430 I 1980010188-441 &2.8 Geomagnetic Applications 1. This area is of interest to many commercial users. The predictions are important and are used because that can be translated into financial savings. The emphasis should be strongly toward the geomagnetic effect, not the cause of the activity. Very specific daily forecasts are of value, and forecasts that can be simp]y interpreted are desirable. 2. Not only the predictions of an active geomagnetic field are of value, the prediction of quiet conditions is also very important. 3. Division of geomagnetic predictions according to geographic zones would be beneficial. 4. Problem Areas: Again the too slow dissemination of short-term fore- casts was mentioned. 3. FORECAST VERIFICATION Many of the principles and problems of forecast verification schemes mentioned in the meeting are discussed in NOAA-ERL-TechnIcal Memorandum (in press) by Gary R. Heckman, entitled "Verification of Solar Flare Fore- casts at the Space Environmental Services Center for tbe Years 1969-1974". This paper has an excellent discussion of several scoring methods and con- siders them in light of a "proper" scoring method A "proper" verification scoring method is one where the test forecast for scoring purposes represents the forecaster's true belief about what will happen. This paper also con- siders how to weigh probability forecasts according to the "cost" to a particular user of making one of two types of error. One type of error is predicting an event which does occur. A method by which this can be done is explained fully in Heckman's paper. Some comment5 made in the general discussion of forecast verification are summarized below: 1. There is a need for more utilization of statistical analysis to develop and verify forecasting techniques. 2. Verification may indi,_ate which variables are the good predictors as well as indi'cate the quality of the forecast. Verificatiop is valuable as feedback to the user _nd the forecaster. 3. A summary of all currently used verification techniques and their advan- tages is needed. : 4. It is imperative that all published forecast accuracy statements include: _< (a) a clear statement of exactly what is b_ing forecast, (b) the climatology :_ during the forecast period, and (c) a clear statement of exactly how the _" verification was performed. • 431 I 1980010188-442

Geophysical predications: Sun-weather

Abstract

Extracted data

Geophysical predications: Sun-weather

Abstract

Extracted data

Related items

Related items