Thermospheric density model biases at the 23rd sunspot maximum

Abstract

Uncertainties in the neutral density estimation are the major source of aerodynamic drag errors and one of the main limiting factors in the accuracy of the orbit prediction and determination process at low altitudes. Massive efforts have been made over the years to constantly improve the existing operational density models, or to create even more precise and sophisticated tools. Special attention has also been paid to research more appropriate solar and geomagnetic indices. However, the operational models still suffer from weakness. Even if a number of studies have been carried out in the last few years to define the performance improvements, further critical assessments are necessary to evaluate and compare the models at different altitudes and solar activity conditions.Taking advantage of the results of a previous study, an investigation of thermospheric density model biases during the last sunspot maximum (October 1999—December 2002) was carried out by analyzing the semi-major axis decay of four satellites: Cosmos 2265, Cosmos 2332, SNOE and Clementine. Six thermospheric density models, widely used in spacecraft operations, were analyzed: JR-71, MSISE-90, NRLMSISE-00, GOST-2004, JB2006 and JB2008. During the time span considered, for each satellite and atmospheric density model, a fitted drag coefficient was solved for and then compared with the calculated physical drag coefficient. It was therefore possible to derive the average density biases of the thermospheric models during the maximum of the 23rd solar cycle.Below 500km, all the models overestimated the average atmospheric density by amounts varying between +7% and +20%. This was an inevitable consequence of constructing thermospheric models from density data obtained by assuming a fixed drag coefficient, independent of altitude. Because the uncertainty affecting the drag coefficient measurements was about 3% at both 200km and 480km of altitude, the calculated air density biases below 500km were statistically significant. The minimum average biases were obtained with JB2008, NRLMSISE-00 and GOST-2004.Above 500km, where only one satellite was analyzed (at 630km), and errors tend to increase with altitude, it cannot be asserted that the calculated biases are significant. Nevertheless, they are presented to show how the various models diverge at higher altitudes. Around 630km, NRLMSISE-00 had a negligible average bias, while the other models underestimated (GOST-2004) or overestimated the average density, by amounts varying between 6% and 16%. However, in terms of semi-major axis root mean square residuals, JB2006 and JB2008 were the best in any case.Below 500km, the short-term behavior of the models was also investigated by fitting the semi-major axis decay over 30-day arcs. The resulting fitted drag coefficients displayed a significant variability, probably associated with mismodeled density variations, but JB2008, followed by JB2006, provided the smallest semi-major axis residuals and a reduced short-term variability of the density bias at just a few frequencies, having been probably successful in removing a significant fraction of the mismodeling sources.