| Since the 20th century,under the influence of global warming and large-scale human activities,the hydrological series is no longer stationary,and the traditional hydrological frequency analysis method based on the stationarity assumption is no longer applicable.In practice,if the traditional method continues to be used,it will undoubtedly increase the risk or the cost of water conservancy projects.Therefore,it is of great scientific and practical significance to carry out nonstationary hydrological frequency analysis and establish a robust nonstationary frequency analysis theory for the calculation of hydrological design values under the background of environmental change.In this thesis,a new mechanism-based reconstruction(Me-RS)method for nonstationary hydrological frequency analysis is proposed and applied to Jialu River basin(JRB),Chabagou River basin(CRB)and Mahuyu River basin(MRB)in Yulin region in northern Shaanxi.Then,we compared the Me-RS method with the "reliability"-based methods such as "Design life level(DLL)","Equivalent reliability(ER)","Average design life level(ADLL)" and the decomposition and synthese method commonly used in China,to verify the rationality and applicability of the Me-RS method from multiple perspectives.The main constents and results of this thesis are summarized as follows:(1)To deal with the problem of nonstationary hydrological frequency analysis,this thesis proposes the theory of the Me-RS method based on physical mechanism,including the establishment of the Mechanism function and Me-RS function,calculation of design values of Me-RS series and the final realization of the design value of the nonstationary hydrological variables.The Me-RS method is applied to JRB,CRB and MRB,and the obtained design flood peak discharge is more consistent with the decreasing characteristics of the flood peak discharge series in recent years.The Me-RS method uses the time-invariant Mechanism function reflecting the physical mechanism of nonstationary changes to describe the influence law of factor X on Y.And through the theoretical analysis and derivation of the nonstationary variance expression ability,the dimensional harmony and independence requirements of different model structions,the multiplication model is adopted as the basic model of the Me-RS method.Based on the Mechanism function and the basic model,the RS function is constructed,according to which the nonstationarity of the original hydrological series is removed,and the nonstationary series is transfored into the stationary RS series.Once the RS series is stationary,the design value of the RS series can be calculated by the traditional frequency analysis method easily.Then,according to the value of the nonstationary influence factors at the design stage,the corresponding design values of the original nonstationary series can be obtained by the RS function again.(2)Using the reliability-based DLL method,ER method and ADLL method to deduce the nonstationary design flood peak discharge of JRB,CRB and MRB.The time-varying parameters of the flood peak discharge series are estimated according to GAMLSS model,with time and underlying surfactor factor as covariables,respectively.The design life level of projects was selected as 100 years after the initial year of the observation(1957～2056 of JRB,1961～2060 of CRB,1962～2061 of MRB)and 100 year after the series has been significantly reduced(20112110),respectively.It is found that the calculation results of the three reliability-based methods are greatly affected by the design life period of projects.In the three methos,the results of the DLL method are significantly larger than that of ER and ADLL methods,and even far larger than the design flood peak discharge derived by the traditional method.The design values obtained by ER method and ADLL method are relatively close,and the ADLL method are slightly smaller.Different covariates will also lead to different results.Compared with the time covariate,taking underlying surface factor as the covariate has more physical significance.In addition,in the estimation,the time-varying characteristics is usually fitted with the linear or some simple nonlinear curves,but the actual conditions must be extremely complicated.Therefore,there may be great uncertainty in the estimation of time-varying parameters.(3)This thesis also uses the decomposition and synthesis method which is popular in China to analyze the flood peak discharge of the three basins.In the decomposition of the flood peak discharge series of each basin,it was found that the removal of the periodic components and the trend components can effectively eliminated the significant linear trend of the first-order moment,and can significantly improved the nonlinear trend,but it cannot improve the nonstationarity of the second order moment.The reason may be that the definition of deterministic components and stochastic components is not clear,and the physical mechanism that causes the nonstationarity is not considered during the decomposition and synthesis process.The continuous decrease of trend component leads to different values at different moments,and the values in 2010 is used for synthesis analysis in this thesis.However,it is also lack of physical significance to assume that the flood peak discharge series decreases indefinitely with time.(4)This thesis makes a comprehensive comparison of the above three kinds of nonstationary frequency analysis methods from the perspectives of the calculation results(design values)and uncertainty of the calculated design values to analyzes the rationality and reliability of the calculation results of each method.In the application of the reliability-based methods,the design life level of projects was selected as 100 years covering the entire record period(1957～2056 of JRB,1961～2060 of CRB,1962～2061 of MRB).And the underlying surface factors are selected as the covariates to make a more intuitive comparison with the Me-RS method and the traditional method.The deterministic components in the decomposition and synthesis method determined according to the values of the current stage(2010).Both the qualitative analysis and the comparison with the measured data of the future stage("7.26" flood event of CRB in 2017)and the past stage(flood peak discharge of JRB in 1958)prove that the Me-RS method is more reasonable.The design values obtained by the reliability-based methods and the decomposition and synthesis method are generally larger.The Bootstrap method was used to analyze the uncertainty of nonstationary design values of flood peak discharge series calculated by different methods.The analysis results show that the uncertainty of the design flood peak discharge calculated by the reliability-based methods and the decomposition and synthesis method is larger than that of the Me-RS method.The Me-RS method is a nonstationary hydrological frequency analysis method with more physical significance and less analytical uncertainty,and has the prospect of practical application. |
PDF Full Text Request