| With the global climate change, frequent human activities and the rapiddevelopment of socio-economic, extreme flood events are increasingly occuring theworld, the flood security situation is becoming more serious, and better estimation ofannual maximum flow is very important for flood hazard mitigation.Due to the fact that the observed data often exhibit two or more segments whichcould not all fit into a single curve but each fit into one smooth curve, this thesissummaied the literatures of hydrologic frequency analysis, using higher probabilityweighted moments which was proposed by Dr. Wang Q.J. in the1990s. The authorused higher probability weighted moments, ordinary probability weighted moment,method of moments and maximum likelihood to calculate the flood frequency ofannual maximum flow series in Northern Shannxi, we expect to provide hydrologicaldesign values for these regions. The main contents and conclusions are as follows:(1) Derive the expressions of higher probability weighted moments andparameters for Generalized Extreme value (GEV) distribution and P-III distribution.According to the principles of higher probability weighted moments of GEVdistribution, this paper gives the S function of probability weighted moments of P-IIIdistribution, and uses the principles of Gauss-Legendre numerical integration tocomputer the expressions of higher probability weighted moments for P-IIIdistribution.(2) Analysis the flood frequencies of study area based on the method of higherprobability weighted moments are given. This paper used higher probability weightedmoments to calculate the parameters of GEV distribution and P-III distribution and fitthe theoretical curve to annual maximum flows for different higher PWMs using the12hydrological stations’ annual maximum flow series in Northern Shannxi. Theresults show that with the increasing of PWM order, besides a few stations, GEVdistribution theoretical frequency curve fit better to the higher flow tail part ofexperience frequency, and the error values smaller. The P-III distribution does notgive a stratifying result owing to its high dimensional integration errors. (3) Application of method of moments and maximum likelihood in floodfrequency analysis. This paper used method of moments and maximum likelihood tocalculate the parameters of GEV distribution and P-III distribution and fit thetheoretical curve to annual maximum flows based on the each station of annualmaximum flow series in study area. The results show that there is a higher error andpoor flood frequency curve fitting to the higher flow tail part when use the method ofmoments. On the other hand, the method of maximum likelihood fitting better resultsthen the error is small, but it need to solve the likelihood function which is complexespecially to the P-III distribution.(4) Determine the distribution model and the method of estimating parametersfor the annual maximum flow series in the study area. Analyze the parameterestimates and the effect of fitting into large floods values of each hydrological stationin northern Shaanxi, of which there are seven stations selected GEV distribution as itstheoretical distribution model, the other five stations selected P-III distribution isbetter. The better method of parameter estimation for GEV distribution is higherPWM.(5) Using Monte Carlo simulation to analyze the statistical performance ofhigher PWM method. The author uses Monte Carlo method to simulate the randomsample which is independent and multiple repeate, and comparative analysis of theparameter estimates and flood design value based on the bias, standard error and rootmean square error, in order to further study the application of the parameterestimation method of higher PWM in flood frequency distribution. The results showthat higher probability weighted moments of GEV distribution gives better fitting. |
PDF Full Text Request