Research On The Delay Correction Model And Forecasting Model Of Regional Ionosphere

Monitoring and forecasting the regional ionosphere is a significant work and research orientation. This paper focuses on the forecasting model of regional ionosphere to provide users with real-time ionospheric VTEC (Vetical Total Electron Content) information. It ultimately can improve the GPS positioning accuracy, especially with single-frequency GPS machines; and also be able to provide an effective way for ionospheric research activities.195 IGS grid VTEC data and data from 12 JSCORS stations are used in this paper. The main content of the work and conclusions are as follows.1) With the 195 IGS grid VTEC data, the spatial distribution and spatial correlation of the reginal VTEC have been analyzed, the results show that the spatial distribution of ionospheric VTEC have great correlation with latitude. The VTEC value will be higher at low latitudes. And the ionosphere VTEC has a very distinct spatial correlation. The VTEC of the adjacent points will be influenced each other. Thus when we forecast the VTEC, a single point forecasting model is not enough, its spatial distribution characteristics should be considered.2) With the 195 IGS grid VTEC data, the time-varying characteristics of reginal VTEC have been analyzed, the results show that the ionospheric VTEC has cyclical variation with a period of 24 hours. From 6:00 to 20:00 of local time, the ionosphere VTEC has dramatic changes. VTEC typically reaches a maximum at a certain time between 12:00 to 14:00. However, from 20:00 to 6:00 the next day, the VTEC fluctuat in a small range or stay steady (around 6:00 reached the minimum value).3) With the 195 IGS grid VTEC data, ARIMA model and BP neural network model of the ionosphere VTEC were conducted respectively with forecast lenfth of 24 hours. The results show that the ARIMA forecasting model has an average relative error of 6.43%. Moreover, the BP neural network model has and an average relative error of 8.30%. Results demonstrate that the ARIMA model is better.4) Based on the decomposition and reconstruction procedures with reginal ionospheric VTEC, the reginal VTEC prediction model is proposed. Therefore, EOF-ARIMA model, EOF-BP model and EOF-Fusion model were established. The results show that the EOF-ARIMA model has an average relative error of 4.05%; the EOF-BP model has an average relative error of 4.64%; the EOF-Fusion model has an average relative error of 3.68%. Compared to the AMIMA single-point prediction model, the accuracies of EOF-ARIMA model could improve 40.3%; Compared to the BP neural network model single-point prediction model, the accuracies of EOF-BP model could improve 42.8%; the accuracy of EOF-Fusion model is the highest, compared to the single-point prediction model, its accuracy could improve 46.1% and 54.5% respectively.5) After the prediction of the regional ionosphere VTEC, the ionosphere interpolation model has been conducted with data from 12 JSCORS stations. The polynomial model, spherical harmonic function model and BP-polynomial fusion model were introduced. The results show that the polynomial model has an average absolute error of 0.24TECU, the spherical harmonic function model has an average absolute error of 0.35TECU, BP-polynomial fusion model has an average absolute error of 0.20TECU. Compared to the polynomial model, the accuracies the accuracy of BP-polynomial fusion model can improve about 15%.