Thermosphere density from accelerometer and GNSS data

Thermosphere density is one of the largest sources of uncertainty in the orbit estimation and prediction for satellites in low Earth orbit. With the rapidly growing number of satellites in this region, accurate knowledge of thermosphere density is becoming increasingly important for e.g. collision risk assessment and avoidance, mission design and lifetime predictions. Accurate in situ density observations are valuable to improve our knowledge of the thermosphere. Such observations can be obtained from a limited number of space-borne accelerometers and can also be derived from precise GNSS tracking data of satellites in low Earth orbit, which are more commonly available.

TU Delft maintains a database of precise thermosphere density and crosswind observations derived from the CHAMP, GRACE, GOCE, Swarm, and GRACE-FO satellites. We continually strive to improve the accuracy of these observations by enhancing our density retrieval strategy. This presentation provides an overview of the most notable improvements. They consist of accurate accelerometer data calibration via precise orbit determination, using high-fidelity satellite geometry models for simulating the aerodynamic and radiation pressure forces, and accounting for the satellite thermal emissions. Recently, an improved retrieval of aerodynamic accelerations from GNSS data was implemented, which leads to a substantially larger signal-to-noise ratio of the GNSS-derived density data, triggering a full reprocessing of the Swarm density observations derived from those data. In a next step, we propose a new gas-surface interaction model that accounts for surface roughness. Since the surface roughness is unknown for the satellites in our database, the model’s roughness parameter must be determined from in-flight data collected during attitude maneuvers and orbital conjunctions. Finally, we have developed a method to comprehensively quantify the uncertainty of density observations, enabling us to augment the observations with uncertainty estimates in the future.