Impacts Of Rain Gauge Network Configuration On Hydrological Modelling Uncertainty

Significant uncertainty is observed in hydrological modelling and forecasting,of which rainfall error is the main uncertainty source.The main causes of rainfall errors are the insufficient rain gauge number and the improper gauge network configuration.Thus,in order to accurately catch the rainfall spatial and temporal characteristics and to reduce the hydrological modelling uncertainty,it’s of vital importance to assess the impacts of rain gauge density and network configuration on hydrological modelling and to study the rain gauge network optimization methods.The densely gauged Xiangjiang Basin is selected as the study area to conduct this study.The results of this study show that:(1)The Monte Carlo sampling method is used to randomly select gauge combinations from all the 252 rain gauges in Xiangjiang Basin.Rainfall data come from these gauge combinations are used as the hydrological model inputs separately.The SCE-UA algorithm and the Bayesian uncertainty estimation framework are used respectively to analyze the impacts of rain gauge density and network configuration on hydrological modelling from the aspects of certainty and uncertainty.The results show that under different rain gauge density and network configuration,the hydrological modelling efficiency varies a lot.With the increase of gauge density,modelling uncertainty induced by the difference of gauge network configurations reduces gradually accompanied by the increase of model efficiency.However when the gauge number is larger than a certain value,no obvious reduction of the modelling uncertainty is observed with further increase of rain gauge number.(2)Through the comparison of modelling uncertainty under different rain gauge density and network configuration,we find two ways to reduce the impacts of rain gauge network configuration on modelling uncertainty.One is to increase the rain gauge number to indirectly reduce the modelling uncertainty caused by gauge network configuration.Another one is to optimize the gauge network configuration to reduce the impacts directly.(3)Based on the Entropy theory,a rain gauge network optimization method was proposed.This method maximizes the transinformation between areal rainfalls derived from selected gauges and all gauges.Then it is compared to other four commonly used methods through the aspects of the network uniformity,the description of rainfall spatial and temporal characteristics and the efficiency in hydrological modelling.The results show that entropy theory is useful in gauge network optimization,while,the optimized gauge network configurations are quite different using different optimization methods with different objective functions based on the theory.When the gauge density is low,the method we proposed or the one which maximizes the uniformity of rain gauge network,gives relatively better gauge network with higher hydrological modelling efficiency.When the gauge density is high,no significant difference is found among these optimized gauge networks derived from different network optimization methods.Through the optimization of rain gauge network configuration,the minimum number of rain gauges is obtained without reducing the rainfall measuring accuracy and hydrological modelling efficiency.