Parameter Sensitivity Analysis of the Liuxihe Model Based on E-FAST Algorithm

Liuxihe Model is a physically based distributed hydrological model proposed for river basin flood forecasting,which has 12 different model parameters in every cell.Due to the huge model parameters,it is difficult to optimize all the model parameters,and the parameter sensitivity analysis is needed for simplifying the parameter numbers.This paper,based on the E-FAST algorithm that has been widely used in quantitative sensitivity analysis,presented a parameter sensitivity analysis framework for Liuxihe model,which has 4 steps including parameter simplification and ranging,parameter sampling,sensitivity analysis evaluation function determination and sensitivity index calculation.Four river basins in Southern China including Huanglongdai River Basin,Liuxihe River Basin,Changhu River Basin and Xinanjiang River basin with basin areas ranging from several hundred to several thousand square kilometers were chosen as the study cases to validate this method,and 4 different evaluation functions were tested.The results showed that the water content at saturation condition and water content at field condition are highly sensitive model parameters,soil hydraulic conductivity,soil thickness,soil porosity characteristics,hill slope Manning’s coefficient and river channel Manning’s coefficient are sensitive parameters,while the other parameters including potential evaporation,evaporation coefficient,underground water recession coefficient and water content at wilting condition are less sensitive parameters.