水文时间序列长期相关性的识别

IDENTIFYING LONG TERM DEPENDENCE OF HYDROLOGICAL TIME SERIES

本文根据Ｈｕｒｓｔ系数估计值ｋ的特征，用蒙特卡洛方法建立了ｋ的经验分布表，通过观测序列的ｋ值与表中相应的值进行比较以判别序列的长期相关性．此外，本文还借助于ｂｏｏｔｓｔｒａｐ法提出了长期相关性识别的非参数法．使用这两种方法所得出的结论是一致的，其有效性通过实例得到了验证．

This paper studies the estimator of Hurst coefficient k and then develops the empirical distribution of the Hurst k via the Monte Carlo method.The long term dependence of the hydrological time series can be determined by comparing the Hurst k of an observed series with the corresponding empirical percentage point for the k .In addition,a non parametric method based on the bootstrap method is proposed to identify this dependence.Both methods result in the same conclusion,as is verified by two practical cases.