中文摘要：

分别在正态分布和 t 分布假设下，通过改变样本容量、自由度、变异值、变点位置等参数用模拟试验的方法比较分析了3种常用的水文序列变异点检验方法的检验性能。结果表明，Pettitt 检验法对于端点位置的均值变点具有最好的检验效果，Brown-Forsythe 检验法适合小样本数据，秩和检验无明显优势。最后用这3种方法分析了西宁市1951年至2000年3月份的月均降水量序列，发现在1977年存在一个变异点。

英文摘要：

By varying sample size ,free degree ,change size ,change point location ,the size and power performances of three most popular used hydrological time series change point testing methods are studied via simulation both under the normal and t distribution hypothesis. Simulations indicate ,Pettitt test has the best performance for those mean change point which nears the edge of series ,Brown-Forsythe test per-forms well for small sample size series ,and rank sum test has not significant performance among these three tests. Finally ,these tests are used to analyze the average monthly precipitation series of Xining City in March from 1951 to 2000 ,and found a change point in 1977.