Improvement of the Polar Motion Prediction Model Using Wavelet Decomposition

In order to further improve the prediction accuracy of polar motion, a wavelet decompositionbased method for polar motion prediction is proposed. In this method, the wavelet decomposition is first used to decompose polar motion time-series into approximation and detailed signals. These approximation and detailed signals are then reconstructed to obtain low- and high-frequency components of polar motion. Furthermore, the least-squares (LS) extrapolation of models for the linear trend, Chandler and annual wobbles are used for predicting low-frequency components, while the autoregressive (AR) model is employed to forecast the high-frequency components together with the residuals derived by subtracting the LS extrapolation curve from the low-frequency components. Finally, polar motion predictions are made by the combination of the LS extrapolation and AR model. The results show that the polar motion predictions can be improved using the proposed wavelet decomposition-based method, especially medium- and long-term predictions.