Comparison of Precision Orbit Derived Density Estimates for CHAMP and GRACE Satellites

Current atmospheric density models cannot adequately represent the density variations observed by satellites in Low Earth Orbit (LEO). Using an optimal orbit determination process, precision orbit ephemerides (POE) are used as measurement data to generate corrections to density values obtained from existing atmospheric models. Densities obtained using these corrections are then compared to density data derived from the onboard accelerometers of satellites, specifically the CHAMP and GRACE satellites. This comparison takes two forms, cross correlation analysis and root mean square analysis. The densities obtained from the POE method are nearly always superior to the empirical models, both in matching the trends observed by the accelerometer (cross correlation), and the magnitudes of the accelerometer derived density (root mean square). In addition, this method consistently produces better results than those achieved by the High Accuracy Satellite Drag Model (HASDM). For satellites orbiting Earth that pass through Earth's upper atmosphere, drag is the primary source of uncertainty in orbit determination and prediction. Variations in density, which are often not modeled or are inaccurately modeled, cause difficulty in properly calculating the drag acting on a satellite. These density variations are the result of many factors; however, the Sun is the main driver in upper atmospheric density changes. The Sun influences the densities in Earth's atmosphere through solar heating of the atmosphere, as well as through geomagnetic heating resulting from the solar wind. Data are examined for fourteen hour time spans between November 2004 and July 2009 for both the CHAMP and GRACE satellites. This data spans all available levels of solar and geomagnetic activity, which does not include data in the elevated and high solar activity bins due to the nature of the solar cycle. Density solutions are generated from corrections to five different baseline atmospheric models, as well as nine combinations of density and ballistic coefficient correlated half-lives. These half-lives are varied among values of 1.8, 18, and 180 minutes. A total of forty-five sets of results emerge from the orbit determination process for all combinations of baseline density model and half-lives. Each time period is examined for both CHAMP and GRACE-A, and the results are analyzed. Results are averaged from all solutions periods for 2004-2007. In addition, results are averaged after binning according to solar and geomagnetic activity levels. For any given day in this period, a ballistic coefficient correlated half-life of 1.8 minutes yields the best correlation and root mean square values for both CHAMP and GRACE. For CHAMP, a density correlated half-life of 18 minutes is best for higher levels of solar and geomagnetic activity, while for lower levels 180 minutes is usually superior. For GRACE, 180 minutes is nearly always best. The three Jacchia-based atmospheric models yield very similar results. The CIRA 1972 or Jacchia 1971 models as baseline consistently produce the best results for both satellites, though results obtained for Jacchia-Roberts are very similar to the other Jacchia-based models. Data are examined in a similar manner for the extended solar minimum period during 2008 and 2009, albeit with a much smaller sampling of data. With the exception of some atypical results, similar combinations of half-lives and baseline atmospheric model produce the best results. A greater sampling of data will aid in characterizing density in a period of especially low solar activity. In general, cross correlation values for CHAMP and GRACE revealed that the POE method matched trends observed by the accelerometers very well. However, one period of time deviated from this trend for the GRACE-A satellite. Between late October 2005 and January 2006, correlations for GRACE-A were very low. Special examination of the surrounding months revealed the extent of time this period covered. Half-life and baseline model combinations that produced the best results during this time were similar to those during normal periods. Plotting these periods revealed very short period density variations in the accelerometer that could not be reproduced by the empirical models, HASDM, or the POE method. Finally, densities produced using precision orbit data for the GRACE-B satellite were shown to be nearly indistinguishable from those produced by GRACE-A. Plots of the densities produced for both satellites during the same time periods revealed this fact. Multiple days were examined covering all possible ranges of solar and geomagnetic activity. In addition, the period in which GRACE-A correlations were low was studied. No significant differences existed between GRACE-A and GRACE-B for all of the days examined