Mapping wheat water use, yield gap and productivity based on climate and satellite data in the Berg River Catchment

Thesis (MSc)--Stellenbosch University, 2020.ENGLISH ABSTRACT: The world population is forecast to increase substantially in the coming decades and doubling of current food production to meet expected demand is envisaged. The Western Cape (WC) of South Africa (SA) is a dryland winter wheat production area with average yield levels below the national average. It was important to ascertain whether the current wheat production levels are optimal or potential exists to increase yields. The study intended to establish wheat yields, estimate seasonal water use and subsequently compute study area wheat yield gaps and water productivity. The study was implemented in part of the Berg river catchment area for the 2016 season winter wheat crop. The Penman Monteith (PM) and Hargreaves Samani (HS) reference ET0 were compiled and computed respectively using Python scripts and compiled into a Julian day (JD) indexed databases. The daily distributed reference crop evapotranspiration ET0PM and ET0HS surfaces were interpolated using the spline with tension interpolator implemented in ArcMap. Seasonal distributed wheat crop water use (ETc) was determined from Sentinel 2 (S2) and Landsat 8 (L8) NDVI modelled crop coefficients and ET0. Wheat crop yield surfaces were modelled from L8 and S2 scenes of the 4th and 5th of August 2016 respectively. From the statistical analysis of validated yield data, the within study area, regional and globally benchmarked wheat yield gaps were estimated. The at-station generated ET0HS were validated based on ET0P Musing the absolute and relative difference statistical measures. The results indicated that climate station ET0HS deviated from the monthly ET0PM values in the year 2016 with the magnitude of the differences higher in some stations and certain months. Although the ET0HS and ET0PM were generally in close range of each other, the ET0HS values were consistently higher in nearly all the months, except in January, March and December. The climate station ET0HS and ET0PM differences although large in absolute terms had a low annual RMSE of 0.426 mm.day-1. The averaged percent relative difference (RD %) indicated that ET0HS estimates were within 20% of the ET0PM.The results indicate that the ET0HS adequately estimates reference crop evapotranspiration in the study area. Both the S2 and L8 based crop coefficients (KcS2 and KcL8) were low at the start of the season at 0.236 and 0.185 level, and increased to a maximum of 0.954 and 0.66 mm.day-1 by mid-season on JD 217, respectively. Thereafter the Kc values decreased. The computed seasonal KcS2 and KcL8 values and plotted curves conform to those indicated in the literature, validating the NDVI based Kc in crop water use estimation studies. However, the KcS2 had consistently higher values than those estimated using the Landsat 8 NDVI (KcL8).The seasonal wheat crop water use patterns indicated that the ET0HS and the KcS2 based ETc consistently overestimated crop water use when compared to the ET0PM and KcL8 based crop ETc. Study area wide crop water use data based on PM and HS methodologies, and S2 and L8 modelled Kc were extracted using the zonal statistic as table tool in ArcMap. The median statistic values indicated that the season wheat crop water use was generally low and less than the expected values cited in literature. This was likely due to the drought conditions during the 2016 winter wheat growing season. The range in ETc values obtained were due to differences observed at the level of ET0 and Kc computation. The L8 modelled crop coefficients had lower values than the S2coefficients. In addition, the ET0HS had consistently higher vales than the ET0PM.The best wheat yield metrics were obtained from the L8 based wheat yield maximum statistic with RMSE less than one t.ha-1 while the corresponding S2 based wheat yield estimate showed a RMSE of 3 t.ha-1. The mean wheat yield for S2 was relatively more accurate than the L8 sensor based wheat yield mean statistic with lower RMSE of 1.408 compared to 1.865 t.ha-1. For both S2 and L8 the median maximum yield statistic results were higher than the corresponding mean of the maximum statistic values.The mean-based within study area wheat yield gaps gave comparable results using S2 and L8 imagery of 1.51 and 1.78 t.ha-1 respectively. Similarly the median-based statistic was computed giving yield gaps of 1.36 and 1.47 t.ha-1 respectively. These were compared with yield gaps obtained using mean SA irrigated, SA dryland and Northern Cape irrigated wheat study area median of maximum zonal statistics of 2.39, -1.51 and 3.65 t.ha-1 respectively. These results indicated that the study area had yields that were higher than the national dryland average and that an additional 3.65 t.ha-1 to current yield level could be attained in the WC with irrigation. The observed study area water productivity metrics of 0.39 to 0.43 kg.m-3 established using the L8 yields to ETcPML8 and ETcHSL8 mean statistics respectively, indicated the existence of wheat production augmentation potential considering the possible 2kg.m-3 attained in high wheat production regions. The corresponding S2 yield to ETcPMS2 and ETcHSS2 WP were 1.89 and 1.79 kg.m-3 respectively, which was more than 4 times the L8 estimates. The study results showcased the utility of geographic information systems (GIS), remote sensing (RS) and tabular climate data in determining wheat water use and yield characteristics. The crop performance characterisation has been specified in United Nations sustainable development goals (SDGs). Studies of this nature are important decision support tools in the planning and management for sustainable irrigation projects to meet the escalating world food demand.AFRIKAANSE OPSOMMING: Daar word voorspel dat die wêreldbevolking in die komende dekades aansienlik sal toeneem en dat die huidige voedselproduksie sal moet verdubbel om aan die verwagte aanvraag te voldoen. Die Wes-Kaap (WK) van Suid-Afrika (SA) is 'n droëland-winterkoringproduksiegebied met 'n gemiddelde opbrengsvlak onder die nasionale gemiddelde. Dit is dus belangrik om vas te stel of die huidige koringproduksievlakke optimaal is en of daar potensiaal is om opbrengste te verhoog. Die studie het ten doel gehad om koringopbrengste te bepaal, seisoenale watergebruik te skat en vervolgens koringopbrengskoers en waterproduktiwiteit in die studiegebied te bereken. Die studie is uitgevoer in 'n gedeelte van die Bergrivier-opvanggebied vir die winterkoringoes van 2016. Die verwysingsevapotranspirasie ET0 van Penman Monteith (PM) en Hargreaves Samani (HS) is met behulp van Python-kode bereken en saamgestel in 'n databasis op Juliaanse dag (JD) geïndekseer. Daaglikse verspreide verwysingsevapotranspirasie oppervlaktes (ET0PM-en ET0HS) is geïnterpoleer met behulp van die spline met spanningsinterpolator in ArcMap sagteware. Seisoenale verspreide koringgewas- waterverbruik (ETc) is bereken vanaf Sentinel 2 (S2) en Landsat 8 (L8) NDVI-gemodelleerde gewaskoëffisiënte en ET0. Koringoes-opbrengs-oppervlaktes is gemodelleer vanaf L8-en S2-beelde van 4 en 5 Augustus 2016 onderskeidelik. Uit die statistiese ontleding van gevalideerde opbrengsdata is die gapings binne die studiegebied, streeks-en wêreldwye maatstaf vir koringopbrengskoerse geskat.Die per-stasie gegenereerde ET0HS is bekragtig deur middel van ET0PM absolute en relatiewe verskil statistieke. Die resultate het aangedui dat die klimaatstasie ET0HS afwyk van die maandelikse ET0PM-waardes in 2016 met die omvang van die verskille hoër in sommige stasies en sekere maande. Alhoewel die ET0HS en ET0PM oor die algemeen naby mekaar was, was die ET0HS-waardes in alle maande hoër, behalwe in Januarie, Maart en Desember. Die klimaatstasie ET0HS-en ET0PM-verskille, hoewel groot in absolute terme, was die jaarlikse RMSE van 0.426 mm.dag-1 laag. Die gemiddelde persentasie relatiewe verskil (RD%) het aangedui dat ET0HS-ramings binne 20% van die ET0PM was. Die resultate dui aan dat die ET0HS die evapotranspirasie van gewasse in die studiegebied voldoende skat. Beide die S2-en L8-gebaseerde gewas-koëffisiënte (KcS2 en KcL8) was laag aan die begin van die seisoen, en het op JD217 onderskeidelik tot 'n maksimum van 0.954 en 0.66 mm.dag-1 in die midseisoen toegeneem. Daarna het die Kc-waardes afgeneem. Die berekende seisoenale KcS2-en KcL8-waardes en kurwes stem ooreen met die aangedui in die literatuur, wat bevestig dat NDVI-gebaseerde Kc gebruik kan word in beramingsstudies vir gewas watergebruik. Die KcS2 het egter deurgaans hoër waardes gehad as wat geskat is met behulp van die Landsat 8 NDVI (KcL8) gelewer met behulp van S2-en L8-beelde van onderskeidelik 1.51 en 1.78 t.ha-1. Net so is die mediaan-gebaseerde statistiek bereken, wat opbrengingskoers van onderskeidelik 1.36 en 1.47 t.ha-1 gegee het. Hierdie waardes is vergelyk met opbrengskoers wat verkry is met gemiddelde SA besproeiings-, SA droëland-en Noord-Kaapse besproeiings-koringstudiegebied, gemiddeld van maksimum sone statistieke van onderskeidelik 2.39, 1.51 en 3.65 t.ha-1. Hierdie resultate het aangedui dat die opbrengste vir die studie area hoër was as die nasionale droëland gemiddelde en dat 'n addisionele 3.65 t.ha-1 tot die huidige opbrengsvlak vir die studiegebied bygevoeg sou kon word onder besproeiing. Die waargenome water produktiwiteit statistiek vir die studiegebied van 0.39 tot 0.43 kg.m-3, bepaal vanaf die L8-opbrengste en onderskeidelik ETcPML8 en ETcHSL8-gemiddelde statistieke, het die potensiaal van addisionele koringproduksie aangedui, met inagname van die moontlike 2 kg.m-3 wat in hoë koringproduksie streke verkry word. Die ooreenstemmende waterproduktiwiteit vanaf S2-opbrengs en onderskeidelik ETcPMS2 en ETcHSS2 was 1.89 en 1.79 kg.m-3, meer as vier keer die L8-ramings. Alhoewel die resultate beïnvloed is deur sensor eienskappe en opbrengsmodel, het sommige van die resultate 'n akkurate uitbeelding van die studiegebied se koring watergebruik, waterproduktiwiteit en opbrengsvlakke aangedui. Die studie het dus die gebruik van geografiese inligtingstelsels (GIS), afstandwaarneming (RS) en getabuleerde klimaatdata bevestig in die bepaling van die koringoes prestasie-eienskappe in die studiegebied, soos gespesifiseer in die Verenigde Nasies se doelwitte vir volhoubare ontwikkeling (SDG's). Studies van hierdie aard is belangrike hulpmiddels vir besluitneming in die beplanning en bestuur van volhoubare besproeiingsprojekte om in die toenemende wêreldaanvraag na voedsel te voorsien.Master