Ionosphere and Plasmasphere Simultaneous Tomography Constrained by A Deep Learning Topside Model

The plasmasphere is an important part of the solar-terrestrial system and contributes prominently to the Global Navigation Satellite System (GNSS) Total Electron Content (TEC). Due to the large discrepancy in electron density (Ne) magnitudes between the ionosphere and plasmasphere, as well as the asynchronous variations between the two, traditional tomography methods face significant challenges in simultaneously reconstructing the three-dimensional structures of both the ionosphere and plasmasphere. This paper introduces a novel Ionosphere and Plasmasphere Simultaneous Tomography (IPST) method constrained by a deep learning topside model. The model, based on Long Short-Term Memory (LSTM) neural networks and trained using electron density observations from Alouette and ISIS satellites, was combined with Thermosphere-Ionosphere-Electrodynamics General Circulation Model (TIEGCM) simulations to infer the topside constraints for the tomography. The proposed method reconstructs ionospheric and plasmaspheric electron density profiles with higher accuracy compared to traditional methods such as Improved Constrained Simultaneous Iterative Reconstruction Technique (ICSIRT) and ICSIRT for ionosphere and plasmasphere (ICSIRT-IP). Simulation results show that IPST achieves lower errors, particularly above 1000 km altitude. Real data experiments during both quiet and storm days confirm that IPST produces more accurate Ne profiles and F2 peak density (NmF2) values, as well as smaller absolute errors in Slant TEC (STEC), compared to ICSIRT-IP. The reconstructed electron density from IPST showed greater consistency with ionosonde and DMSP satellite observations, with Root Mean Square Error (RMSE) reductions of 31% in NmF2 and 36% in STEC. These results highlight the potential of the IPST method for improving ionosphere and plasmasphere electron density reconstructions.