Thermospheric density predictions during quiet time and geomagnetic storm using a deep evidential model-based framework

Abstract

Knowledge of the thermospheric density is essential for calculating the drag in low Earth orbit satellites. Existing models struggle to predict density accurately. In this paper, we propose thermospheric density prediction using a deep evidential model-based framework that incorporates empirical models, accelerometer-inferred density from the CHAMP satellite, and geomagnetic and solar indices. The framework is investigated on both quiet and storm conditions. Our results demonstrate that the proposed model can predict the thermospheric density with high accuracy and reliable uncertainty in both quiet and storm times. The predicted results from the evidential model are advantageous over the Gaussian Processes (GPs) model in our previous studies. Furthermore, the proposed model can also provide insightful aleatoric and epistemic uncertainties.