关键词: 地球自转参数, 预报, 最小二乘, 自回归

Abstract: Earth rotation parameters (ERP) are integral parameters for transformation between the celestial coordinates and the terrestrial coordinates in satellite precise orbit determination, and are also important products for Intenational GNSS Service (IGS) and International GNSS Monitoring and Assessment System (IGMAS). To improve the prediction precision of the linear prediction model used by Chinese Academy of Surveying & Mapping (CGS), the best method of ultra short-term forecasting based on LS+AR is researched. The optimal data length is determined with the CGS data, and then the LS+AR is used to predict the ultrshort term ERP. The results are compared to that of the IGS and iGMAS and show that the optimal data length is eight days (sessions). The prediction precision of LS+AR is much better than that of LS, and also better than the linear model used by CGS. The results also show that the x and y direction prediction of ERP is better with the session's data than the day's data, while the LOD is better with the day's data than the session's data. The LS+AR ultrashort term prediction is a good method to predict the ERP for IGS and iGMAS analysis centers.

Key words: earth rotation parameters, forecasting, the least squares, autoregression model