A Method for Polar Motion Prediction Based on LS Model of Error Compensation

Abstract

Polar motion is an important component of the Earth Rotation Parameter (ERP). Not only is it one of the necessary transformation parameters between International Celestial Reference Frame (ICRF) and International Terrestrial Reference Frame (ITRF), but it is also essential for deep space exploration mission and satellite ultra-fast precise orbit determination. Polar motion is usually available with a delay of hours to days, thus polar motion prediction is needed to meet the growing demands for spacecraft navigation and physical geography science research. The least square model (LS) is a kind of mature polar motion forecasting model, but it has the problem that, though, the inner residual value of LS fitting is reasonable, the residual value of LS extrapolation, however, is poor, which will cause prediction errors to accumulate greatly. This paper proposes a LS model of error compensation (ECLS model) to solve this problem with the LS model, and hence improve the accuracy of predictions. Restrictions are first added to the two endpoints of LS fitting data to place them on the fitting curve of LS. If the LS interpolation residual sequence and the extrapolation of the residual sequence have a good correlation, then the LS interpolation errors can be used to correct the LS extrapolation value and to make the adjustment to the prediction errors of LS model. Feasibility and effectiveness of the ECLS model can be proved for predicting polar motion by comparing CELS with EOP_PCC, RLS + AR, RLS + ARIMA and LS + ANN. In addition, for the short term (30 days) prediction, examples show that the ECLS model can effectively improve prediction accuracy of polar motion, and the results show prediction accuracy equal to that achieved/observed at International level.