Study Of The Short-term Forecasting Of The Ionospheric TEC In China Region

The ionosphere is an important part of our survival space environment, and has important significance in the research of modern space weather. To understand the variation of ionospheric, seeking to overcome the disasters caused by the ionosphere and ionospheric search method for human better service, it is necessary to carry out a more accurate prediction of the ionosphere. The main contents are as follows:1. Based on auto-correlation method,improved it into the subsection(equal interval and unequal interval) prediction for single station forecasting, and combined with Kriging interpolation method for regional reconstruction, Forming a complete short-term prediction method of ionospheric TEC in China. The study shows: the prediction error of single station plays an key role in the forecasting error of the whole scheme. For 1 hour ahead forecasting, when using the unequal interval auto-correlation method to predict single station, compared with the existing literature, it has decreased 1.3TECu. And the improved single station forecast value is used to reconstruct the region, the accuracy was decreased 0.2TECu. The whole scheme error is 1.8TECu, which is lower than the existing error. The improvement of forecast accuracy of single station is beneficial to the improvement of the whole forecast precision.2. To forecast the TEC, the Distance Correlation method was introduced, which is a sort study on the physical quantities of ionospheric TEC, then combined with the nonlinear model to predict TEC. The previous results show that the selection of input parameters has great influence on the prediction accuracy. However, the existing research lack of theoretical basis, because they only show the importance analysis of variables from the experimental point of view. This paper focuses on how to select the important variables, by using the Distance Correlation method, which can be used to describe the correlation between TEC and input parameters. This provides the basis for the selection of the input parameters and the important parameter,furthermore, this paper established the models of DC-SVM and DC-BP, and the experimental results show that the accuracy is improved by 0.4 percentage points.