CNN-LSTM-Attention-Based Electron Density Modeling Using CSES-01 Observations

The F2-layer peak electron density (NmF2), the F2-layer peak height (hmF2), and the α-Chapman scale height (H) are key parameters for characterizing the vertical structure of ionospheric electron density, and of great significance for space environment monitoring and space weather applications. Using Global Navigation Satellite System (GNSS) radio occultation (GRO) observations from the CSES-01 satellite, this study proposes the CSES-NET model. A reliable sample data set is first constructed through sample screening, α-Chapman fitting, and physical constraints. On this basis, a multitask model integrating convolutional neural network (CNN), long short-term memory (LSTM) network, and an attention mechanism is developed to simulate the three parameters. The results show that the proposed model reproduces the spatiotemporal characteristics of NmF2, hmF2, and H well and effectively improves the reconstruction accuracy of electron density profiles (EDP). Baseline and ablation experiments are also carried out to ascertain the optimal modeling scheme. This study may provide a reference for reconstructing ionospheric profiles from the key parameters of the Chapman profile under quiet geomagnetic conditions.