Diagnosing thermospheric density uncertainty from LEO satellites using data assimilation

The growing constellation of low-Earth-orbit satellites allows us to characterize the thermosphere-ionosphere system (TI). One of the most valuable LEO measurements are accelerometer derived neutral density estimates, which play a central role in satellite drag estimations, TI modeling, and space weather operations. Despite their importance, the measurement uncertainty of satellite-derived neutral density for most LEO missions remains unknown. In this study, we use a data assimilation (DA) based framework to diagnose the observation uncertainty directly from neutral density measurements.

Using the Coupled Thermosphere Ionosphere Plasmasphere electrodynamics model (CTIPe) and TIDA, the TI Ensemble Kalman filter data assimilation scheme, we perform controlled experiments with varied assumed uncertainties. Two complementary diagnostics are applied: the Desroziers method, which estimates the effective observation uncertainty required for a self-consistent DA system, and an ensemble-spread method, which isolates the true measurement error by removing model-projected variability from the innovation variance.

We apply both diagnostics to CHAMP, Swarm A/B/C, and GRACE-A/B across low and high solar-activity periods. Results confirm the expected 10–15% uncertainty for CHAMP during quiet conditions, while GRACE (15–35%) and Swarm (25–50%) exhibit larger values, reflecting differences in altitude, solar activity, instrument characteristics, and thermospheric variability. The two methods provide complementary perspectives and the limit of the estimated uncertainty range: Desroziers quantifies the upper bound, and the ensemble-spread method provides the lower bound uncertainty. The framework provides a pathway to systematically quantify uncertainty in current and upcoming LEO missions, supporting improved density models, drag prediction, and space weather services.