Modeling and Predicting Total Electron Content by Semi-Parametric Autoregressive (AR) Model

Abstract

Based on dynamic data system(DDS) modeling methodology, after transformed a seasonal time series for total electron content(TEC) of the ionosphere into a stationary time series by differencing technique, stationary TEC values are modeled by the autoregressive(AR) model. In order to correct model’s systematic errors, authors proposed that AR model is improved by non-parameters introduced to AR model and the ionospheric TEC is predicted using the improved AR model which is called semi-parametric AR model. Preliminary results show that the semi-parametric AR model has a good performance than one of the AR model for short-term TEC prediction while, for relatively long-term TEC prediction, the performance of the semi-parametric AR model is no less than one of AR model.