| Rainstorm and flood both are very typical extreme events, which have low frequ ency, but they exert a huge influence whenever they happen. Comparing with relativel y normal samples, observation data of them are comparatively insufficient, which resu lt in that larger uncertainties are existed in the estimation of distribution parameters an d quantiles in frequency analysis. Extreme value statistics is specialize in the modelin g and statistical analysis method for the extreme variability of random variables, such extremum which rarely happens but its influence is tremendous whenever it occurs, it could assess the risk of extreme events, and has an extensive application prospect in n umerous areas such as hydrology, meteorology, seismology, insurance, finance and etc. In contrast with traditional statistics, the development history of extreme value statist ics are relatively short, and up to now, it keeps evolving. Dadu River is one of the pri mary tributaries in the upper reaches of Yangtze River, flow through heavy rain area i n western Sichuan, rainstorms are quite frequent. This paper choose Dadu River as ty pical river basin, About the modeling method for extreme value statistics, studying its application effect in the analysis for rainstorm frequency. The main research contents and conclusions in this paper are as following:(1) Depend on the precipitation data acquired from Dadu meterological stations, choosing the catchment area where between the middle and lower reaches of Dadu River and Pu Bugou cascade reservoir, utilizing the interpolation method for spatial rainfall to do the interpolation in basin. By various interpolation method, considering the situation that under different altitude and basin, in the region with a higher change elevation, the Kriging method considered as the larger; but in the region with a relatively lower change, the results come from different methods are presented from big to small: IDW>LP>GP>Kriging>RBF. Depend on a comprehensive comparison of diverse methods, it is showed that the result from IDW is relatively accord with observed rainfall in the middle and lower reaches of Dadu River.(2) Extremum statistic method of GEV and GPD distribution model had been introduced, and analyze the frequency of extreme rainfall by separately applying the GEV and GPD distribution model. In the middle reaches of Dadu River, preferable simulations had shown in both models, obtaining the estimated values and intervals in diverse return levels. Among the diverse return levels, the estimated values of GEV and GPD are invariably within the 95% confidence interval, the estimated values of each hydrologic stations will increase with the return level; comparing the return level which between 10 to 50 years with the return level between 50 to 100 years, the former’s estimated value increased largely than the latter, and its variation trend is relatively stable. The samples of GPD are more comprehensive, due to this reason and by analysis, the estimated values in different return level of GPD are greater than GEV in concentrated heavy rainfall stations, while in stations with a lighter rainfall, the difference between two models are limited.(3) The utilization of Copula functions is used to analyze the joint probability distribution of rainstorm in catchment areas, depend on the joint distribution of constructed variables of Copula function, and based on test of goodness of fit, Gumbel Copula works better in fitting variables. On the basis of Copula function to calculate the return period, the result shows that the joint distribution created by Copula could get the conditional return period under a certain rainfall in single region, providing the references for regional engineering planning design and risk assessment. |
PDF Full Text Request