Study On Hydrologic Model Uncertainty Based On Watershed Delineation And Spatial Resolution Of Meteorological Data

Hydrological model is an important tool to explore and understand the principle of water cycle and hydrological process.With the rise of geographic information technology and remote sensing technology,hydrological model has gradually become the foundation and core of digital watershed management and modern water resources management.The sources of hydrological model uncertainties include parameters,structure and inputs.For distributed hydrological models,the simulation results vary because of the inaccuracy of model generalization and the mismatch between data and model scale,which is a matter of model input uncertainty.In this paper,SWAT was used to build the distributed hydrological model of the Upper Mississippi River Basin(UMRB)in order to study the uncertainty from watershed delineation and spatial resolution of meteorological data.The main content of research and results are as follows:(1)5 scenarios were set up by different subbasins and Hydrologic Response Units(HRUs)delineation schemes and spatial characteristic of the basin under different scenarios were calculated.Differences on the output of hydrological elements were also analyzed.Results showed the uncertainty of the simulated hydrological elements were related to meteorological factor in terms of time scales.The uncertainty of potential evapotranspiration was strongly related to temperature factors and decreased with the increase of temperature.The correlation between runoff uncertainty and temperature factors was less than that of evapotranspiration,and it increased with the increase of air temperature.From the view of space scales,the difference in lateral flow output was greater in steep terrain areas.The difference between surface runoff and groundwater runoff output was mainly related to land use and soil type transformation caused by watershed delineation.(2)Simulated runoff for 13 hydrological stations in UMRB under different delineation scenarios were compared with annual observed runoff.Results show that the accuracy of runoff simulation increased with finer watershed delineation.Then,the range of variation of annual runoff with nested catchments was analyzed.Uncertainties of simulated runoff under different watershed delineations had scale effect.For the same HRU division threshold,the uncertainty of simulated runoff between different subbasin divisions increased with larger catchments.However,the scale effects of runoff uncertainties between HRU divisions were different,relating to the value of drainage density.The runoff uncertainties at different scales increased with the drainage density.Based on the results of accuracy and scale effect,strategies for delineation is proposed for large-scale watershed modelling.(3)18 parameters were selected to calibrate based on the comparison between observed and simulated flows of 3 calibration stations in the study area.Those parameters involved the processes of snowfall,surface runoff,groundwater,soil water,evapotranspiration,and lateral flow.The delination of subbasins had greater influence on the parameter estimations than the Hydrological Response Units(HRUs),and the optimal parameters may be the same with the same number of subbasins.Some parameters were correlated with spatial characteristic factors,among which the correlation between groundwater runoff and soil hydrological grouping was stronger.Also,the oriders of parameter sensitivity were different under different delineations of subbasins.(4)Based on the validation results of 10 stations at tributaries after parameter transplantation,the correlation between evaluation indicators and spatial characteristic of corresponding catchments were quantified,and model suitability under different delineations were evaluated by the global performance indicator.Results showed that R~2,NSE,and KGE have a significant correlation with land use and soil type;PBIAS had a significant positive correlation with the total area of the catchments.For large-scale watersheds,although the accuracy of watershed outlet can meet the requirements,accuracies of different regions across the basin varied a lot.The more the subbasins,the greater the difference was between different regions in the basin.When the number of subbasins are equal,the average accuracy of simulated runoff at stations increaseed with HRUs.(5)A site data and a gridded data were selected to evaluate the reliability of meteorological data,and the grid meteorological data with higher average simulation accuracy was assessed to be an optimal data source.Then,the grid data was further matched with the number of sub-basin divisions to generate different densities of climate data and to be compared and analyzed.Results showed that with the increase of the meteorological data density,the runoff capacity of the model increased,the annual average runoff grew and the evaluation criteria based on monthly flow also improved.The finer the watershed delineations are,the lager the uncertainty caused by spatial resolution of climate data are.In conclusion,based on research tool of the distributed hydrological model SWAT,building model for the Upper Mississippi River basin,this research applies the theoretical methods of uncertainty analysis to reveal the influences of watershed delineation and spatial resolution of meteorological data on the hydrological model uncertainty.The influence law of hydrological model uncertainty provides a sight for the application and improvement of distributed models.