Abrupt change is an important manifestation of hydrological process with dramatic variation in the context of global climate change, the accurate recognition of which has great significance to understand hydrological process changes and carry out the actual hydrological and water resources works. The traditional method is not reliable at both ends of the samples. The results of the methods are often inconsistent. In order to solve the problem, we proposed a comprehensive weighted recognition method for hydrological abrupt change based on weighting by comparing of 12 commonly used methods for testing change points. The reliability of the method was verified by Monte Carlo statistical test. The results showed that the efficiency of the 12 methods was influenced by the factors including coefficient of variation (Cv), deviation coefficient (Cs) before the change point, mean value difference coefficient, Cv difference coefficient and Cs difference coefficient, but with no significant relationship with the mean value of the sequence. Based on the performance of each method, the weight of each test method was given following the results from statistical test. The sliding rank sum test method and the sliding run test method had the highest weight, whereas the RS test method had the lowest weight. By this means, the change points with the largest comprehensive weight could be selected as the final result when the results of the different methods were inconsistent. This method was used to analyze the daily maximum sequence of Jiajiu station in the lower reaches of the Lancang River (1-day, 3-day, 5-day, 7-day and 1-month). The results showed that each sequence had obvious jump variation in 2004, which was in agreement with the physical causes of hydrological process change and water conservancy construction. The rationality and reliability of the proposed method was verified.
跳跃变异是全球气候变化背景下水文过程发生剧烈变异的重要表现形式,其准确识别对认识水文过程发生的变化及开展实际水文水资源工作具有重要意义.针对目前传统方法在样本两端有时检验结果不可靠、多种方法检验结果常常不一致等问题,本文通过对常用的12种变异点检验方法对比研究,基于赋权方式提出了水文跳跃变异综合加权识别方法,并利用Monte-Carlo统计试验验证了该方法的可靠性.结果表明: 12种跳跃变异检验方法的效率受变异点前变差系数(Cv)、变异点前偏态系数(Cs)、均值变幅系数、Cv变幅系数、Cs变幅系数等因素的影响较大,但与序列均值无明显关系.以各方法的性能差异为基础,基于统计试验结果对各检验方法赋予权重.其中,滑动秩和检验法和滑动游程检验法的权重最高,RS检验法的权重最低.通过赋权方式,可以在各方法检测结果不一致时选取综合权重最大的变异点作为最终结果.将该方法应用于澜沧江下游戛旧年径流极大值(1日、3日、5日、7日、1月)序列进行分析,结果显示各个序列均在2004年发生了明显跳跃变异,且与该流域水文过程变化、水利工程建设等物理成因十分吻合,验证了该方法的合理性和可靠性.
Keywords: Lancang River; abrupt change; climate change; hydrological time series; statistical experiment.