How Densities Derived from Satellite Drag are Affected by Composition and Drag Coefficient Modeling

The ability to model atmospheric drag, or the main perturbing force of the atmosphere on the orbital trajectories of Low Earth Orbit (LEO) objects, is critical for precise satellite orbit determination. Uncertainty in atmospheric drag comes primarily from estimates of the atmospheric mass density of the space environment and the spacecraft drag coefficient, C_D. Deriving density from atmospheric drag measurements requires accurate modeling of C_D . Physical gas-surface interaction models, like the Cercignani-Lampis-Lord (CLL) framework, can model C_D based on a set of assumptions about the energy and momentum exchange between the atmospheric gas and the satellite surface which depend on atmospheric composition and temperature. Measured density offsets for satellites of different shapes can arise from non-uniform sensitivity to the CLL C_D parameters attributable to different geometries. Thus we can test the consistency of the CLL assumptions by comparing densities derived from satellite drag measurements for satellites of different geometries. We compute C_D , derived observational densities, and atmospheric model-dependent densities for the GRACE satellite (of complex shape) and a set of 22 compact satellites located within a scale height of GRACE for 4 one-week time periods representing different geomagnetic conditions. Measurements come from accelerometer-derived densities that have been revised with CLL C_D values for GRACE and Two-Line-Element (TLE) tracking data for compact satellites. Modeled densities, composition, and temperature come from the Thermosphere-Ionosphere-Electrodynamics General Circulation Model (TIE-GCM). Comparing the measured-to-modeled density ratios for GRACE and the spherical satellites allows us to quantify the density offsets attributable to C_D assumptions and to modify C_D model parameters to achieve greater consistency. Results indicate that the C_D model assumptions produce significant differences between dayside and nightside density comparisons. The unmodified CLL model tends to perform well on the dayside but can be improved on the nightside by adjusting the assumed surface scattering process.