Application of Empirical Mode Decomposition in the Ultra Short-Term Prediction of Polar Motion

Abstract:

The high-frequency signals discourage short-term prediction of the pole motion (PM). This study applies empirical mode decomposition (EMD) to decomposes PM. Firstly, removing the high-frequency signals in the PM series, then the combined model of least squares extrapolation and general regression neural network (GRNN) is used to predict polar motion without the high-frequency signals from one to ten days in the future. The result shows feasibility of this new model and obvious improvement of the prediction accuracy.