Polar Motion Prediction Based on the Combination of Weighted Least Squares and Vector Autoregressive Models

This paper presents the application of weighted least squares (WLS) extrapolation and vector autoregressive (VAR) modelling in polar motion prediction. The simultaneous predictions of pole coordinates \(x_{{\text{p}}}\), \(y_{{\text{p}}}\) are generated by the combination of (1) WLS extrapolation of harmonic models for the linear trend, Chandler and annual wobbles, and (2) VAR stochastic prediction of the WLS residuals (WLS+VAR). For WLS fit the weights are computed by a power function to accurately extract the Chandler and annual wobbles. Moreover, the VAR technology is used to model and predict the WLS residuals of pole coordinates \(x_{{\text{p}}}\), \(y_{{\text{p}}}\) together considering the correlation between \(x_{{\text{p}}}\) and \(y_{{\text{p}}}\). . The results show that the accuracy of the ultra short-term predictions up to 10 days into the future obtained by WLS+VAR is equal to or slightly better than that by LS+AR, whilst the predictions out 10 days are substantially more accurate than those by LS+AR. Furthermore, the medium-term predictions beyond 90 days can be improved by WLS+VAR in comparison with the prediction values provided by the IERS Bulletin A. The improvement in the prediction accuracy can reach up to 20%.