Nicholas Dietrich, Tomoko Matsuo
Abstract Forecasting orbit positions in the low Earth orbit (LEO) environment is necessary for avoiding collisions, which is particularly difficult during geomagnetic storms due to highly variable neutral mass density and the associated uncertainty on atmospheric drag. Orbit position errors are commonly assumed to follow a Gaussian distribution, being represented with a covariance ellipsoid, but nonlinear neutral density variability during storms is likely to disrupt this assumption. This study attempts to quantify storm-time non-Gaussian neutral density uncertainties by using a new physics-based particle filter framework. The framework is implemented in a reduced state space of forcing parameters, while utilizing a high-dimensional first-principles physics-based model of the ionosphere-thermosphere (I-T) system as part of the likelihood calculation. The National Center for Atmospheric Research Thermosphere Ionosphere Electrodynamics General Circulation Model (TIEGCM) is used as the physics-based model. The framework is applied to an isolated storm event in July 2022, using neutral density observations retrieved from the Gravity Recovery and Climate Experiment Follow-On (GRACE-FO) mission's onboard accelerometer. The outcome is a time-varying uncertainty characterization of non-Gaussian forcing parameters and three-dimensional neutral density fields arising from nonlinear storm-time I-T dynamics. Filter results show bimodal distributions for the day and night sides and reveal how regions with high non-Gaussian forcing parameter distributions vary through the storm period. The development of this physics-based particle filter framework represents the first steps towards future efforts in quantifying the impact of nonlinear LEO neutral density dynamics on orbit position errors.