Optimizing Radiation Pressure Modeling for Improved Thermospheric Density and Wind Estimation from GRACE Accelerometer Data

Accurate estimation of thermosphere mass density and horizontal winds from satellite accelerometer measurements is crucial for understanding the environment experienced by low-Earth-orbit satellites. A critical step in this process is removing non-aerodynamic forces, such as radiation pressure, from calibrated accelerometer data. However, uncertainties in surface reflection and absorption coefficients, as well as incomplete thermal property information and calibration parameters for the accelerometer, often limit the accuracy of modeling. Therefore, this study presents a method for jointly optimizing radiation pressure parameters and accelerometer scale factors in the cross-track and radial direction and demonstrates their impact on wind observations.

During initial studies, the acceleration residuals (the difference between modeled and measured acceleration) in the cross-track direction exhibited a geographical pattern correlated with the magnetic field for both GRACE-A and GRACE-B. However, the residual is opposite in sign for both satellites in the orbital frame. As the satellites are in nearly the same orbit, with an along-track distance separation of only approximately 220km, this cannot be attributed to a vector-based force. The root cause has not yet been identified but could possibly be attributed to an instrument issue. However, it can be empirically corrected in the cross-track accelerometer measurements using quadratic functions of the magnetic vector components.

To isolate radiation pressure as much as possible during the optimization, we use GRACE data from 2009, a period when radiation pressure dominated over aerodynamic drag due to the low solar activity. Following the optimization, a significant reduction in residuals was observed for both GRACE-A and GRACE-B, despite the coefficients being tuned using only GRACE-A data. Including the magnetic correction increased consistency between GRACE-A and GRACE-B. Overall, the method achieved RMS reductions in unmodeled accelerations of more than 13% in the cross-track direction and 32% in the radial direction, indicating improved accuracy of the radiation pressure model.

Using the proposed radiation pressure model, we demonstrated increased consistency in observed crosswinds between GRACE-A and GRACE-B during periods of higher thermosphere mass density. The proposed approach is generalizable to future missions and improves neutral density and crosswind estimation from precise accelerometer measurements, thereby supporting space weather monitoring and forecasting efforts in the thermosphere.