Study On The Influence Of Input Data Uncertainty On Hydrological Model Simulation Results

Hydrological model is not only an effective approach to understand complex hydrological cycle processes and mechanisms,but also a useful tool to solve many practical hydrological problems.With the continuous progress of the science and technology,hydrological models are more widely used,which leads to more prominent uncertainty problems in the fields of the hydrological process simulation and prediction.The uncertainty of hydrological simulation mainly comes from the model input data,model structure and model parameters,and the uncertainty of the input data is one of the important factors that affects the uncertainty of the model simulation results.In this paper,the upper reaches of the Heihe River basin is selected as the study area,and the Soil and Water Assessment Tool(SWAT)model is used to simulate the runoff in the study area.On this basis,the influences of the input data uncertainty on the model simulations are focused on,in which the digital elevation model(DEM)and precipitation data are selected as the two main points in the study of model input data.The effects of DEM uncertainty on the runoff simulation results are investigated by inputting different resolution DEM data,and at the same time,keeping the other input data and the optimal parameter estimates unchanged.We select four different spatial resolution of DEM data with 100 m × 100 m,300 m × 300 m,500 m × 500 m and 1000 m × 1000 m,respectively The monthly,seasonal,and annual runoff are discussed with the changes of DEM resolutions.The uncertainty of precipitation data is studied by introducing a precipitation error model to the original obseravtions,and the observed precipitation data of five stations(Yeniugou Station,Zhamushike Station,Yingluoxia Station,Qilian Station and Ebo Station)in the upper reaches of Heihe River Basin are then reconstructed.The influences of precipitation data uncertainty on model simulations are studied by analyzing the monthly,seasonal,annual and multi-year average runoff simulations which are obtained from the reconstructed precipitation data,instead of the original observed data.The results show that,with the improvement of DEM data precision,the accuracies of model runoff simulation results are improved.When the resolution of DEM is 100 × 100 m,the error between the simulated runoff and observed runoff is least and the flow curve has the best goodness of fit.The greater uncertainties of the precipitation data,the greater impacts on the simulation results,and the wider uncertainty intervals of the runoff simulation results.Furthermore,larger precipitation observations may result in greater uncertainties in runoff simulations since the greater errors are added to the precipitation observations.Comparing with the influences of model parameter and model structural uncertainties on the model simulation results,it can be found that the uncertainty of precipitation data has the greatest influence on the runoff simulation results.Therefore,it is an important way to reduce the uncertainty of hydrological simulation by improving the accuracy of input data.