论文部分内容阅读
论文基于鄱阳湖流域降水数据,采用平稳和非平稳GEV模型进行极值降水的模拟和分析。检测各站年最大1 d降水量序列(AMS1)的非平稳特征,将时间作为位置参数的协变量进行非平稳AMS1序列的GEV模拟。结果表明:1)鄱阳湖流域AMS1序列的形状参数基本均大于0,服从Fréchet分布;位置和尺度参数的空间分布较一致,形状参数则有差异。2)在较高重现期下由轮廓似然方法估计的置信区间比Delta方法更准确;重现水平的轮廓似然函数曲线在较高重现期之下呈较显著不对称性。3)不同重现期下的鄱阳湖流域极值降水等值线图的空间分布特征,与位置和尺度参数的分布图更为接近,与形状参数的差别则较大。4)基于非平稳GEV模型得到赣县站随时间变化的极值降水设计值,其在1951年的100 a一遇设计值到2010年下降为接近50 a一遇,预示着未来发生极值降水和洪灾的风险加大。
Based on the Poyang Lake Basin precipitation data, the paper uses the stationary and non-stationary GEV models to simulate and analyze extreme precipitation. The non-stationary characteristics of the maximum 1-year precipitation series (AMS1) at each station were measured and the GEV simulation of the non-stationary AMS1 sequence was performed using the time as a covariate of the position parameters. The results show that: 1) The shape parameters of AMS1 sequences in Poyang Lake Basin are basically larger than 0, which are subject to the Fréchet distribution; the spatial distribution of location and scale parameters are consistent and the shape parameters are different. 2) The confidence interval estimated by the contour likelihood method under the higher recurrence period is more accurate than the Delta method; the contour likelihood function curve at the recurrence level shows more significant asymmetry under the higher recurrence period. 3) The spatial distribution characteristics of the Poyang Lake Basin contour precipitation map at different recurrence periods are more similar to the distribution of location and scale parameters, but different from the shape parameters. 4) Based on the non-stationary GEV model, the design value of extreme precipitation over time at Gan County Station was obtained. The design value of once-in-100-a-year precipitation in 1951 dropped to nearly 50 a in 2010, indicating the occurrence of extreme precipitation in the future And the risk of flooding increases.