Development regional ionospheric total electron content prediction model based on deep learning approach

The Global Navigation Satellite System (GNSS) provides total electron content (TEC), an essential parameter for understanding the ionosphere phenomenon. Given the impact of the ionosphere on communications and navigation systems, accurate and timely TEC prediction is crucial. This study aims to develop a deep learning model for predicting the regional ionospheric TEC map in the vicinity of the Korean Peninsula (26°-40.5°N, 120.5°-135°E) for the next twenty-four hours using convolution long short-term memory (ConvLSTM) and LSTM techniques. We use TEC maps generated by the Deep Convolutional Generative Adversarial Network – Poisson Blending model (Jeong et al. 2022). To do this, we first generate synthetic TEC maps by training the DCGAN with the International Reference Ionosphere-based TEC maps and then optimize the synthetic TEC maps using TEC observations. Finally, we produce complete TEC maps by implementing TEC observation onto the optimized synthetic TEC maps using Poisson Blending. The training data range from 2002 to 2018; the test period is set to 2019. We compare the model results with observations to evaluate the prediction model’s performance.