Aggregate Evaluation Of Water And Sediment Statistical Model In Typical Watershed Of Loess Plateau

The process of runoff and sediment production in the middle reaches of the Yellow River has always been concerned by hydrologists.As the main runoff and sediment production area in the middle reaches of the Yellow River,the Loess Plateau has attracted many scholars' interest.Hydrological model method is an important simulation tool to study the hydrological process of river basin.There are many kinds of hydrological models.There are differences in the basic data sources,model principles,technical means,research period,factors of concern,etc.of each model,resulting in differences in the simulation results and simulation accuracy of each model in a specific region.On the other hand,the evaluation of simulation accuracy of different models.There are many quantitative indexes in price,and the model accuracy evaluation is lack of systematicness.In this paper,Wuding River and its nested small watershed are selected as the research object,the basic differences of existing hydrological models are compared and analyzed,the variables of the models are extracted,and PCIS and PMIS are used to construct hydrological models suitable for typical watershed of the Loess Plateau,and the applicability of each model is analyzed;principal component analysis,AHP and comprehensive score method are used,based on the existing single evaluation index.A multi-level and multi-index model accuracy evaluation system is constructed to comprehensively evaluate the accuracy of the existing model and the model proposed in this study,in order to verify the validity of the accuracy evaluation system.The contribution rate of water and sediment and the accuracy of prediction model are calculated to veri fy the accuracy and applicability of the accuracy evaluation system.The main results are as follows:(1)The existing models can be roughly divided into four types according to the composition:multiple ratio model,linear model,power function model and hybrid model.Different types have different composition factors and data requirements on different scales,and show different fitting accuracy under different model accuracy evaluation indexes.The research shows that the existing models have diversity.The multiple ratio model is composed of large-scale rainfall factors,so the applicability of the model is good;the data and component factors of the linear model are required to be higher;the power function model and the mixed model are also required to be higher,and some data need to be processed before use,but the simulation accuracy is higher.Eight factors,namely,NSE,R,Re,D,RMSE,CR,B and D,are selected to analyze the reliability of the simulation results.Under the analysis of error reliability of simulation results,evaluation refers to the fitting sensitivity of NSE and R2 is higher than other indexes in the range of standard difference.According to the rate of abrupt change years of water and sediment in each basin,the periodic and validation periods are divided,and the existing model series are simulated.Single index shows that NSE and R2 are more sensitive to model accuracy,while RMSE is less sensitive.(2)Calculation of PCIS and PMIS under the river basin water and sediment model construction.Except that the number of runoff optimization variables of Wuding River and Dali River basins is 3,the optimization parameters of other conditions are 2 and the variables are basically the same;the simulation parameters of runoff in each basin are mainly large-scale variables,and the simulation parameters of sediment transport are mainly monthly scale variables;with the decrease of research scale,the variability and difference of the optimal variables increase gradually.It is worth noting that PMIS can exclude the redundant information included in PCIS,which means that PMIS excluding P9 and Pmx produces a smaller set of prediction variables than PCIS,which is conducive to the model development of data scarce areas.Under single precision index,NSE and R2 values of water and sediment model precision determined by PCIS and PMIS model are higher than those of the existing model,while RMSE is slightly lower,that is to say,the model precision is higher than the existing model.(3)A model accuracy evaluation index system consisting of 1 objective layer,2 criteria layers and 11 variables layers is established,and the accuracy evaluation of all 7 models is calculated.The comprehensive score shows that the average comprehensive score of each basin under different models is above 0.7,showing a trend of decreasing simulation accuracy from large-scale to small-scale;PMIS and PCIS models are in the top two of the scores respectively;the optimal simulation formula of rate regular runoff is PMIS and PCIS models,and the scores of each model in the verification period are decreased,but PMIS and PCIS models still show good accuracy.The fitting results of sediment discharge are similar to runoff.The above shows that the precision of PCIs and PMIS models can reach a high level.By calculating the contribution rate of water and sediment change of the simulation data and the measured data of the optimized model,the contribution rate is similar to the measured sequence;the accuracy evaluation of the three prediction models shows that the comprehensive scores of SMA and BMA are higher than that of most models,which further verifies that all the selected models of the accuracy evaluation system have the applicability of hydrological calculation.