Pol Mesalles-Ripoll, SpaceNav; Zachary Waldron, SpaceNav; Alex Sanchez, SpaceNav; Roman Rositani, SpaceNav; Matt Duncan, SpaceNav
Keywords: drag, space weather, space situational awareness, atmosphere, density, model, orbit determination, orbit prediction
Abstract:
As low-Earth orbit (LEO) becomes increasingly crowded, the risk of collision continues to grow. Effective mitigation of this risk requires more reliable and accurate predicted trajectories of resident space objects, which in turn requires improved specification and forecasting capabilities of the Earth space environment via upper atmospheric models. The U.S. Space Force uses its own dynamically-calibrated High-Accuracy Satellite Drag Model (HASDM) for orbit predictions and conjunction assessment, but the real-time and predicted densities from HASDM are not publicly available. For satellite owner/operators generating orbit predictions, there are several semi-empirical models such as the MSIS, DTM, and Jacchia-based series of models. These are often computationally fast and accurate for climatological uses, but their ability to accurately project into the future is closely tied to the fidelity of their drivers (i.e., space weather indices such as the F10.7 solar flux and the Kp and Ap geomagnetic activity) and similarly limited by their lower temporal and spatial resolutions. Physics-based models offer greater potential for forecasting but lack the accuracy of semi-empirical models in near real-time scenarios. With no clear consensus on which atmosphere model performs best for which scenario, in this paper we assess the accuracy of each model’s predictive capabilities using their respective drivers for a variety of spacecraft operating in different altitude regimes and inclinations: NASA’s GPM (∼430km, 65°), PlanetiQ’s GNOMES-4 and GNOMES-5 (∼550km and ∼580km, 97°), and JPL’s OCO-2 (∼700km, 98°). We focus our analysis on the following models: NRLMSISE-00, NRLMSIS 2.1, DMT2020, JB2008, and WAM-IPE.
For this study, we build on top of SpaceNav’s previously presented framework for predictive orbit accuracy evaluations. By processing GNSS tracking data for each satellite through our orbit determination (OD) pipeline over the span of multiple months—where both calm and storm conditions were observed—we generate a large dataset of definitive ephemeris data. Using the latest estimated state every 12 hours, we propagate predictive ephemerides using the solved-for drag coefficient (CD) and the most up-to-date space weather predictions available at the time (using data from the Canadian Penticton observatory, NOAA SWPC, the U.S. Air Force, and Space Environment Technologies). Differencing the predictive orbits against the reference definitive trajectory, we generate a set of position differences (ephemeris overlaps) for the analysis period. These differences are presented in relative time from the epoch of each OD and can be aggregated together to obtain cumulative distribution functions of the position error at each propagation time (e.g., 6, 12, 24, 36, and 48h, representing the time range where operators typically commit to risk mitigation maneuvers). Statistics on the prediction errors are generated and analyzed in terms of the norm of the position differences.
Finally, we review the contribution to the errors caused by space weather forecasts alone by propagating the orbits with both predictive and definitive space weather drivers. Previous work has shown that errors in the F10.7 are one of the largest contributors to orbit prediction accuracy; comparing the position difference statistics for each model using the two space weather datasets, we want to revisit these results with what is now an expanded set of atmosphere models.
Date of Conference: September 16-19, 2025
Track: Atmospherics/Space Weather