对长江宜昌站月平均径流量突变的分析

本文简要叙述了一阶矩突变扫描式t检验和二阶矩突变扫描式F检验算法,并应用于长江宜昌水文站1882-2002年间历年月平均径流量的实测序列,划分出24个年际至十年际时间尺度的相对5级旱、涝时段及其变率特征,与有关长江上游洪水、干旱灾害的历史记载相吻合.

An Analysis of Abrupt Changes in Monthly Streamflow at Yichang Section of the Changjiang River

In this article, two algorithms of the scanning t test and of the scanning F test were briefly described separately. The scanning t test detects multiscale abrupt changes in the first moment (subsample mean), i.e. the differences of subsample means between two adjoined sub-samples in a long time series. The scanning F test detects multiscale abrupt changes in the second moment (subsample variance), i.e. the differences of subsample variances (or standard deviations) between two adjoined sub-samples in a long time series. An application of these two algorithms to monthly Streamflow observations at Yichang section of the Changjiang River (Yangtze River) was carried out for the period from January 1882 to December 2002. Based on the scanning t test, 12 change points at which the Streamflow abruptly increases, and 11 change points with decrease of the Streamflow were detected. Then, twenty-four episodes of wetness were partitioned in 5 grades on interannual to decadal time-scales, according to the average level of streamflow over each episode. Based on the scanning F test, 10 change points, at which the sub-sample variance reduces (becomes steady), and 7 change points toward larger sub-sample variances were detected. By combining with the results from the scanning t test, each of the twenty four wetness episodes was characterized as steady (with small variance) or unsteady (with large variance), and listed in Tab. 1. It is confirmed that each of the twenty-four wetness episodes coincides well with historical records of floods or droughts in the upper reaches of Changjiang River. Tab. 1 summarizes the main results in this work.