Influence Of Rainfall Station Network Layout On Runoff Simulation Of Small Watershed And Its Uncertainty Analysis

As an important factor driving the watershed hydrological model,rainfall data are crucial to the simulation of runoff processes in watersheds.The main access to rainfall data comes from surface rainfall stations,whose spatial distribution largely determines the accuracy of runoff simulation.It is important to explore the runoff process and the response of runoff to the deployment of the rainfall station network in the watershed to optimize the deployment of the rainfall station network and reduce the uncertainty of the runoff simulation.At present,most of the related researches focus on the wet areas with abundant precipitation,and there are few studies on the influence of spatial distribution characteristics of rainfall stations in arid and semi-arid areas on the flooding process in the watershed,and there is a lack of analysis of different time scales.This paper takes the typical watershed in the hilly and ravine area of Loess Plateau,the Chabagou watershed,as the research object,and sets 51 different rainfall station network densities and spatial distribution methods based on SWAT model,analyzes the differences of watershed runoff process under different rainfall station network densities and spatial distribution methods,clarifies the influence of rainfall station network deployment on watershed runoff simulation and its uncertainty,and draws the following main conclusions.(1)Based on 21 years of rainfall data from 11 rainfall stations in the Chabagou watershed and the SWAT hydrological model,the runoff processes in the watershed were simulated and analyzed under three different time scales from 1990 to 2010.The results show that the coefficient of determination(R~2),Nash coefficient(NS)and relative error(R_e)in the runoff simulation results of SWAT model for the calibration period are 0.72,0.65,10.4;0.62,0.61,9.4;0.68,0.67,-8.6 at annual,monthly and daily scales,respectively,and R~2,NS and R_e for the validation period are 0.57,0.54,-0.9;0.60,0.54,24.3;0.50,0.50,-10.1,respectively.The calibration results are satisfactory.(2)The rainfall data from 11 rainfall stations in the Chabagou watershed are used as the driving force,corresponding to the simulated runoff process as the benchmark,and the simulation of the watershed runoff process is carried out by gradually reducing the number of rainfall stations and changing the spatial distribution of rainfall stations.The results show that the influence of rainfall station network density on the accuracy of watershed runoff process simulation varies with the time scale.On the annual time scale and monthly time scale,when the rainfall station network density is less than 30km~2/each,a better accuracy of annual and monthly time scale runoff simulation can be obtained;on the daily time scale,when the rainfall station network density is less than20 km~2/each,a better daily time scale runoff simulation result can be obtained.The spatial distribution of rainfall stations also has an important influence on the simulation accuracy of watershed runoff process.Under the same rainfall station network density,whether to choose a rainfall station at the outlet of the watershed and the uniform distribution of rainfall stations in the watershed have a great influence on the simulation accuracy of watershed runoff.Therefore,we should try to ensure that the rainfall stations are evenly distributed in the watershed,and at least one rainfall station should be placed at the outlet of the watershed,which can significantly reduce the error of the watershed runoff simulation process.(3)Uncertainty analysis was performed by 95%confidence intervals of runoff simulations at different time scales.The results show that the density of rainfall station network has a great influence on the uncertainty of runoff process simulation in the watershed,and increasing the density of rainfall stations can significantly reduce the uncertainty of runoff simulation.The spatial distribution of rainfall stations also has a great influence on the uncertainty of runoff simulation,and the influence is greater when the number of rainfall stations is small,and the more uniform the distribution of rainfall stations,the less uncertainty it brings to the runoff simulation;the causes of errors in the simulation include not only the hydrological model itself,but also the effect of silt dams in the watershed on the sediment storage of runoff,and the model input data and other kinds of errors.