In this paper, we present a novel approach to improve the accuracy of TEC prediction through data augmentation. Prior works that adopt various deep-learning-based approaches suffer from two major problems. First, from a deep model perspective: LSTM models exhibit low performance on long-term data dependency, while self-attention-based methods ignore the temporal nature of time series, which results in an information utilization bottleneck. Second, the existing TEC actual data is limited and existing generative models fail to generate sufficient high-quality datasets. Our work leverages a two-stage deep learning framework for TEC prediction, stage 1: a time series generative model synthesis of sufficient data close to real data distribution, and stage 2: an Anto-correlation-based transformer to model temporal dependencies by presenting series-wise connections. Experiment on the 2018 TEC testing benchmark demonstrates that our method improves the accuracy by a large margin. The models trained on synthetic data had a notably lower RMSE of 1.17 TECU, while the RMSE for the IRI2016 model was 2.88 TECU. Our results show that the model significantly reduces monthly RMSE, displaying higher reliability in mid, high, low latitudes. Our model shows higher reliability and significantly reduces monthly RMSE and latitude RMSE. However, although our model performs better than IRI2016, low latitudes RMSE needs improvement, as values are generally above 2.5 TECU. This finding has important implications for the development of advanced TEC prediction models and highlights the potential of transformer models trained on synthetic data for a range of applications in ionospheric research and satellite communication systems.

Total Electron Content (TEC) of the ionosphere under fluctuating geomagnetic conditions plays a pivotal role in space-based communication and navigation systems. Periods of geomagnetic storms have a significant effect on the ionospheric TEC, thereby impacting the precision of Global Navigation Satellite Systems (GNSS) as well as other communication systems. This study evaluates the performance of three predictive models: Diffusion, Transformer, and Seasonal Autoregressive Integrated Moving Average (SARIMA) by forecasting TEC during different geomagnetic storm conditions. Diffusion model predicts by simulating the random processes that govern ionospheric variability, and SARIMA model relies on statistical principles to emulate seasonality, trend, and autocorrelation structure within data. Our study uses TEC and geomagnetic data to differentiate between storm and quiet periods, evaluating each model’s predictive accuracy in these distinct scenarios. Performance was benchmarked against the Center for Orbit Determination in Europe’s (C1PG) 1-day ahead ionospheric forecasts. This study specifically concentrated on the performance of the prediction models within a singular grid cell located at the exact geographical coordinates of 114.3° E longitude and 30.5° N latitude. The study indicates that all three models exhibit exceptional forecasting abilities during both geomagnetic storm and quiet periods. Significantly, Diffusion Model surpasses the others, achieving an outstanding 85.64\% of its predictions within the high-correlation interval ranging from 0.95 to 1.00 during quiet geomagnetic periods, while the C1PG model records 51.66%. Otherwise, during strong geomagnetic storm periods, Diffusion model operates at a 75.41% accuracy rate, while the C1PG achieves a 71.77% accuracy rate.