中文摘要：

利用自回归模型进行日长变化中长期预报时，预报精度逐渐降低．跳步自回归模型在中长期的预报中具有良好的预报精度，且具有较好的预测稳定性．因此，尝试采用跳步自回归模型替代自回归模型进行日长变化预报．最后，利用国际地球自转参数与参考系服务（International Earth Rotation and Reference Systems Service，IERS）提供的EOP 08 C04日长变化序列进行实验，并分析比较两种模型的预报结果．实验结果表明，跳步自回归模型用于改善自回归模型中长期预报精度是可行有效的．

英文摘要：

It is known that the accuracies of medium- and long-term prediction of changes of length of day （LOD） based on the combined least-square and autoregressive （LS＋AR） decrease gradually. The leap-step autoregressive （LSAR） model is more accu- rate and stable in medium- and long-term prediction, therefore it is used to forecast the LOD changes in this work. Then the LOD series from EOP 08 C04 provided by IERS （International Earth Rotation and Reference Systems Service） is used to compare the effectiveness of the LSAR and traditional AR methods. The predicted series resulted from the two models show that the prediction accuracy with the LSAR model is better than that from AR model in medium- and long-term prediction.