线型选择对西江流域水文频率计算不确定性的影响

Influence of Model Selection on Uncertainties of Hydrological Frequency Analysis in Xijiang River Basin

针对水文频率计算中线型选择的不确定性问题。研究以我国南方典型洪涝灾害易发区和多发区西江流域为例,利用AIC、AICC和BIC信息准则,对水文频率分析中常用的7种线型在西江和柳江的适用性进行了分析。结果表明:NORM、GEV和P-Ⅲ线型在西江梧州站具有较好地适用性,GUMBEL、GEV和P-Ⅲ为柳江柳州站比较理想的频率分布线型。基于线型优选的研究结果,研究进一步应用Bayesian MCMC模型,对优选线型计算结果的不确定性进行了分析,结果表明:在优选的3种优选线型中,西江梧州站以NORM和P-Ⅲ线型的不确定性最低,柳江柳州站以GUMBEL和P-Ⅲ线型的不确定性最低。综合考虑线型的不确定性及其水文频率的拟合效果,P-Ⅲ线型更适合于西江和柳江水文频率计算。

Accounting for uncertainty in model selection of hydrological frequency analysis, a case study is carried out in Xijiang river basin where the flood occurs more frequently than in other hydrologic areas of China. The applicability of the seven typical models is analyzed on the basis of several model selection criteria such as AIC, AICc, BICin Xijiang river basin and Liujiang River basin. The results show that the three distribution models（ NORM、GEV and P-Ⅲ） have better applicability at Wuzhou Station of Xijiang River,and the other three distribution models（ GUMBEL, GEV and P-Ⅲ） have better applicability at Liuzhou Station of Liujiang River.The Bayesian MCMC method is employed to analyze the uncertainties of three selected distribution models on the basis of the model selection results. The results prove that the NORM and P-Ⅲ have a low uncertainty in hydrological frequency analysis in Xijiang River,the GUMBEL and P3 show a low uncertainties in Liujiang River. Considering model uncertainty and fitting results, the P-Ⅲ shows a preferable applicability for the hydrological frequency analysis in Xijiang River and Liujiang River.