Radiation force modeling for ICESat precision orbit determination
Precision orbit determination (POD) for the Ice, Cloud and land Elevation Satellite (ICESat) relies on an epoch-state batch filter, in which the dynamic models play a central role. Its implementation in the Multi-Satellite Orbit Determination Program (MSODP) originally included a box-and-wing model, representing the TOPEX/Poseidon satellite, to compute solar radiation forces. This “macro-model” has been adapted to the ICESat geometry, and additionally, extended to the calculation of forces induced by radiation reflected and emitted from the Earth. To determine the area and reflectivity parameters of the ICESat macromodel surfaces, a high-fidelity simulation of the radiation forces in low-Earth orbit was first developed, using a detailed model of the satellite, called the “micro-model”. In this effort, new algorithms to compute such forces were adapted from a Monte Carlo Ray Tracing (MCRT) method originally designed to determine incident heating rates. After working with the vendor of the Thermal Synthesizer System (TSS) to implement these algorithms, a modified version of this software was employed to generate solar and Earth radiation forces for all ICESat orbit and attitude geometries. Estimates of the macro-model parameters were then obtained from a least-squares fit to these micro-model forces, applying an algorithm that also incorporated linear equality and inequality constraints to ensure feasible solutions. Three of these fitted solutions were selected for post-launch evaluation. Two represented conditions at the start and at the end of the mission, while the third comprised four separate solutions, one for each of the nominal satellite attitudes. In addition, three other sets of macro-model parameters were derived from area-weighted averaging of the micro-model reflectivities. They included solar-only and infrared-only spectral parameters, as well as a set combining these parameters. Daily POD solutions were generated with each of these macro-model sets, for eight-day intervals in four different ICESat mapping campaigns. As a group, the fitted parameters slightly outperformed the averaged parameters, based on a variety of metrics. Their impact on POD accuracy, however, was limited to the sub-millimeter level, as measured by independent satellite laser ranging (SLR) residuals. As a result, no change to the nominal macro-model parameters is recommended.