A combined forecasting method of ionospheric tomography TEC based on LSTM and IRI model

Total electron content (TEC) is one of major parameters that evaluates the errors caused by radio waves passing through the ionosphere, and its accurate estimation and prediction have important research significance. Based on the reconstructed TEC data with the ionospheric tomography algorithm, a forecasting model combined the long short-term memory(LSTM）and international reference ionosphere(IRI) model is proposed to realize the prediction of the quiet-time ionospheric TEC over Europe. Besides, the prediction results are compared with those of the LSTM model and the IRI gradient method. The experimental results show that the one-hour prediction result of the IRI gradient method is the best, but the two-hour and three-hour prediction accuracy is gradually reduced. The forecasting results of the LSTM model is well within two days in advance, but with the increase of iterative prediction time, there are more outliers in the prediction results. The statistical error shows that the combined forecasting model is more stable than the IRI gradient method, and has a significant effect on the correction of the LSTM model so as to eliminate most of the outliers in the prediction results, which effectively improves the prediction accuracy of the single model. The predicted root mean square error (RMSE) between the combined forecasting model and the actual tomography TEC is 1.1 TECU, and the predicted RMSE between the forecasting prediction model and center for orbit determination in Europe- TEC(CODE-TEC) is 1.7 TECU.