水文序列跳跃变异点的滑动相关系数识别方法

跳跃变异点识别与检验是处理非一致性水文序列的首要工作。针对目前方法未能直观反映序列跳跃变异实质的缺陷,本文通过推求原始序列与跳跃成分序列的相关系数表达式,构建其与跳跃变异前后均值差异程度的关系,在此基础上提出水文序列跳跃变异点的滑动相关系数识别与检验方法。统计实验结果显示:所提方法易于操作且容易求解,在受到序列长度、变异点位置、均值变化系数、Cv变化系数与Cs变化系数等多种因素影响时,其变异点识别效率与常用的Pettitt检验法、B-F检验法基本相同,但性能更加稳定。将所提方法应用于澜沧江下游允景洪水文站不同时间尺度的径流序列进行跳跃变异点识别与检验,并结合物理成因分析进行验证,结果显示了该水文站径流过程受其上游水利工程施工与运行影响很大,云南省境内澜沧江干流梯级水电站的调蓄作用(特别是年内的"削丰补枯"作用)十分明显,导致不同时间尺度上的径流过程多在2004年前后呈现出显著的跳跃变异。

Moving correlation coefficient-based method for the detection of change-points in hydrological time series

Detection of change-point is a primary task for the analysis of nonstationary hydrological time series. Those methods used presently for the practice cannot directly reflect the essence of changes in a time series. To overcome the problem,in this paper we first derived the expression of correlation coefficient between the original series and the jump-component series, and then quantified the relationship between the correlation coefficient and the difference degree of mean value before and after the change-point. Based on it,we proposed a new method for the detection of change-point in a time series,called moving-correlation coefficient-based detection method. Results of statistical experiments indicate that the proposed method is easy to be conducted,and the correlation coefficient is easily calculated,too. When encountering the influence of series length, position of change-point, and difference of mean value, variation coefficient and skewness coefficient before and after change-point, the proposed method has similar performance as those of the Pettitt test and Brown-Forsythe test, both of which have been widely used for detecting change-point. But the former has more stable efficiency. The proposed method was then applied to detect the change-points in the runoff processes at multi-time scales at the Yunjinghong hydrological station in the lower reaches of the Lancang River basin. The results indicate that the construction and operation of the water conservancy projects have great influence on the streamflow regimes of the basin, causing more water in dry season and less in the wet season in a year, and the obvious variability and abrupt changes in the runoff process in around 2004.