基于加权最小二乘与AR模型的极移预报

Prediction of polar motion based on the weighted least squares and AR model

目的建立加权最小二乘(Weighted Least Squares,WLS)与自回归(Autoregressive,AR)组合模型(WLS+AR)进行极移预报,以更好地体现钱德勒极移和周年极移的时变特性。方法利用极移长时序求解钱德勒项和周年项的模型参数时,通过引入权函数增强当前观测值对模型参数解算的作用,提出权比概念来解决观测值定权的问题,据此设计幂权函数和分段权函数。结果与结论利用国际地球自转和参考系服务组织(International Earth Rotation and Reference Systems Service,IERS)发布的极移数据进行预报实验,结果表明:基于2种权函数的定权策略均能提高极移预报精度,尤其对极移30~180 d的中长期预报改善更加明显;2种权函数的定权策略均适用于WLS+AR模型,其中,分段权函数的定权策略优于幂权函数。

Purposes—To propose a method for Polar Motion(PM)prediction combing the weighted least squares(WLS)and autoregressive(AR)model,so as to reflect the time-varying characteristics of the Chandler and annual wobbles in PM.Methods—The model parameters of the Chandler term and annual wobbles are solved by using long PM time-series through a WLS algorithm.In this algorithm,the concept of weight ratio between adjacent observable epoch is introduced to solve the problem with the observation weighting.The two functions,namely power and piecewise function,are designed and then employed as the WLS weighed function according to the theoretic analysis.Results and Conclusions—The PM time-series from the International Earth Rotation and Reference Systems Service(IERS)are taken as data basis for both short-term and long-term predictions.The results show that the accuracy of the predictions can be enhanced through the two weight functions introduced into the WLS+AR method.Specially,the accuracy of the medium-and long-term predictions from 30 days to 180 days is obviously improved.It is also found that the piecewise function is more potential for weighting PM observations than the power one.It is therefore concluded that the both functions can be applied to WLS+AR model for PM prediction.