多种方法在水文关键要素一致性检验中的比较

Comparison of Consistency Testing for Key Hydrological Elements by Using Multiple Statistical Methods

以武隆站的水文资料作为研究基础,分析了不同的一致性检验方法的特点、适用性以及不足等。结果表明:就趋势性检验方法而言,其过程线法、滑动平均法以及线性拟合法可直接显示序列过程走势,也可通过目估法判断序列走势,但是该方法依赖经验性,不能给出趋势的显著性程度;而作为非参数检验的MannKendall趋势检验法和Spearman秩次相关检验法能确定一定置信水平下的趋势显著性水平;就突变点检验方法而言,双累积曲线法、有序聚类法和M-K突变检验法的结果直观,但双累积曲线一致性关系较为模糊,需配合其他方法一起诊断,有序聚类法计算简单但是检验结果不易评价显著性,均值差异T检验法和M-K突变点检验法可以确定多个突变点;总体而言,M-K检验方法最优,原理清晰,结果直观,便于识别,能检验趋势性和突变性,并可以给出一定置信水平下的检验结果的显著性评价。

Based on the hydrological data gauged at Wulong Station on Wujiang River,the characteristics,applicability and shortcomings of different statistical methods for testing consistency of hydrological series were analyzed and compared.The results show that among the trend testing methods,both the process-based linear regression method and the moving average method can directly display variation characteristics of a sequence process with a visual inspection trend.However,these methods rely on expert experience and can′t test significance level of a sequence trend.Nonparametric tests(i.e.Mann-Kendall trend test,Spearman rank correlation test)can detect significance of a sequence trend at a certain confidence level.Methods for mutation point test including double cumulative curve method,ordered clustering method and Mann-Kendall mutation test method can provide direct testing results of abrupt change in a sequence.But the double cumulative curve can only provide a vague testing result,and it is therefore usually applied together with other methods.The ordered clustering method has advantage of easily understanding and calculating,while it cannot evaluate the significance of testing results.The mean difference T-test and the M-K mutation test can determine multiple mutation points.In general,the Mann-Kendall test method is optimal due to clear principle,intuitive testing result,easy to identify,and the fact that it can not only detect trend and mutation but also give a significant evaluation of test results at a certain confidence level.