Study On The Relationship Between Runoff And Sediment In Ganjiang River Basin Under Changing Environment

The relationship between runoff and sediment can reflect their matching degree,which has important impact on river channel evolution,water resource security and ecological environment.Due to the highly nonlinear of relationship and their scale difference,the statistical method provides a powerful tool for the study of relationship between runoff and sediment.Therefore,it is necessary to make the research on their relationship under the changing environment considering the influence of external factors,which is of great significance for the comprehensive channel management,the protection and utilization of water resources and security of regional water ecological environment.In this study,the Ganjiang River basin is chosen as the study region and studied by both of the statistical method and hydrological model.Based on the identification of the non-stationarity of hydrological variables,the quantitative attribution analysis of annual runoff and annual sediment are analyzed under the changing environment,and the changing relationships of runoff and sediment are researched using the time-varying Copula function with the physical covariates.The main findings and conclusions of this thesis are summarized as follows:(1)The annual runoff of three stations and sediment of four stations display nonstationary in the Ganjiang Riverbasinduring 1958-2014,respectively.As for the time-varying characteristics of hydrometeorological variables,the trend and changepoint of the variables are analyzed by non-parametric test method.Then,a time-varying moment model which can be used to diagnose nonstationarity of multi-order moment is introduced and used to constructing local nonstationarity.The results show that the optimal distribution function of annual runoff is Gamma for the three hydrological stations.Except Waizhou station,Xiajiang and Xiashan show abrupt changes while Julongtan shows local trend for annual runoff.The optimal distribution function of annual sediment is Weibull,and the four stations all show nonstationary characteristics,among which the Waizhou and Julongtan stations show local trend,while the Xiajiang and Xiashan stations show abrupt changes.The time-varying moment model constructed in this paper shows a good goodness of fit,and can identify the timevarying characteristics of the first-order moment,the second-order moment and even the higher-order moment.In addition,the results of local nonstationarity are more scientific and reasonable as well as meet the actual situation.(2)The annual runoff variation of three stations in Ganjiang River basin is mainfly attributed to the climate change.Based on the recognition of the time-varying characteristics of the annual runoff,the monthly water balance model is selected from the simulated result of accuracy and uncertainty.Then,the GLUE method is introduced to analyze the parameter distribution and their correlation,and the reduction method and reoccurrence method are proposed to quantitatively analyze the attribution of runoff.The results show that the max NSEs of DWBM,ABCD and VWBM monthly water balance models are above 0.85,and the uncertainty analysis index CRs are also above75%.ABCD,selected as the attribution analysis model,is most superior and stable among them.The annual runoff of the three hydrological stations(Xiajiang,Xiashan and Julongtan)has increased during the changing period due to climate change.The maximum and minimum monthly values are concentrated in April and June,respectively.The threecandidate monthly water balance models have good simulation results while parameters of the optimal model are robust.The two attribution analysis results of runoff are similar,which are more comprehensive and reasonable.(3)The annual sediment variation of four stations in Ganjiang River basin is mainfly attributed to the human activities.Based on the recognition of time-varying characteristics of annual sediment,a multiple linear regression model is proposed considering different rainfall intensity and slope gradient,and linear regression attribution analysis and elastic coefficient attribution analysis are constructed.Introducing three kinds of physical factors,the GLAMSS model is used to separate the contributions of each factor on annual sediment.The research results show that the human activities play an important role in the decrease of sediment on the Ganjiang Riverbasin.Further,the separation contribution results of multi-factor indicate that rainfall erosion have less effect on the sediment,while the forest cover and large reservoirs are two key factors,which indicate thepackageprojectsof soil and water conservation play a very important role in Ganjiang River basin.The cumulative annual rainfall of different rainfall intensity can represent the effect of climate change on annual sediment.Meanwhile,considering the impact of different slopes,the binary regression linear model is more scientific,reasonable and convenient.GLAMSS model based on physical covariates could simulate the influence of different factors on annual sediment because of clear mechanism.(4)The relationshipbetweenannual runoff and sedimentat the four stations in Ganjiang River basin has changed significantly,with the weakeningcorrelation.As for the relationship between the runoff and sediment affected by the changing environment,the multi-variable joint probability distributionis used to display the change of relationship through the joint probability and conditional probability.In detail,the timevarying Copula function is introduced and constructed through the time-varying construction of correlation structure.The results show that the optimal Copula of Waizhou station is Clayton Copula with constant parameter,and the annual sediment is decreasing and the reduction amplitude is decreasing under the same joint probabilityfrom 1958 to 2014.The optimal Copula of Xiajiang,Xiashan and Julongtan is Frank Copula.The Copula parameter of Julongtan remain constant while the parameters of Xiajiang and Xiashan are time-varying with the weakening correlation.The optimal three-dimensional Copula function of flood peak,water level and sediment in Waizhou is time-varying Frank Copula,and the correlation among the three variables is also weakening.Time-varying copulas can well representthe relation of the three variables based on the covariates of time and physical factors,while copulas with physical covariant could more fine depict the changeof relationship.Meanwhile,the result shows that under the small sample data,different types and structures of copulas have a little difference in data mining,which means they have a good applicability in practice.