中文摘要：

针对LS＋AR模型在日长变化预报过程中存在的端部效应现象,采用时间序列分析方法对日长变化的序列进行外推,形成一个新的序列,用这个新序列求得LS模型的系数,然后再用LS＋AR模型对日长变化原始序列进行预报。实验结果表明,利用端部效应改正的LS＋AR模型与LS＋AR模型相比,在日长变化的预报精度上有一定的改善,尤其在跨度为中长期时改善更为明显。

英文摘要：

Aiming to resolve the edge effect in the process of predicting length of day （LOD） by the least squares and autoregressive （LS＋AR） model, we employed a time series analysis model to extrapolate LOD series and produce a new series. Then, we used the new series to solve the coefficients for the LS model. At last, we used the LS~AR model to predict the LOD series again. By comparing the accuracy of LOD prediction by edge-effect corrected LS ＋AR and that by LS＋AR, we conclude that edge-effect corrected LS＋AR can improve the prediction accuracy, especially for medium-term and long-term predictions.