| DEM and land use data accuracy significantly affects the simulation effectiveness of hydrological models.To improve work efficiency,reduce model uncertainty,and enhance simulation accuracy,this study selected Maojiahe,Yangjiaping,and Zhangjiashan as control stations in the Jinghe River Basin.SWAT models were built using ten different resolutions of DEM(30-3000 m)and land use data(30～3000 m)to explore the impact of input data accuracy on water and sediment simulation.The TOPSIS theory method was used to comprehensively evaluate the models.In addition,the study also investigated the combined effects of watershed area threshold and different calibration algorithms on the simulation of water and sediment in the SWAT model.The main research findings are as follows:(1)Different resolutions of DEM and land use data can affect the extraction of hydrological parameters.Decreasing the resolution of DEM leads to a more blurred representation of terrain details,increased errors in basin elevation information,and a decrease in elevation range and standard deviation.Low-resolution DEM data also leads to the loss of slope and gradient details within the basin.DEM resolutions ranging from 30 to 1000 meters are suitable for modeling needs.When the DEM resolution is below 1000 meters,it can lead to false river channels and discontinuous river networks,and the river network boundaries may also become concave.When the DEM resolution is below 2000 meters,only macro-scale slope information can be expressed,and almost all slopes are below 7 degrees.As the DEM resolution decreases,the calculated area of the Jing River basin decreases significantly,and the number of sub-basins also shows a decreasing trend overall.The number of HRUs decreases significantly with the decrease of DEM resolution,but changes in land use resolution have a smaller impact on the number of HRUs.(2)The effect of DEM resolution on water-sediment simulation is greater than that of land use.There is a strong correlation between DEM resolution and water-sediment simulation values at Maojiahe and Zhangjiashan stations,but the correlation is weak at Yangjiaping station.With the increase of watershed area,both runoff and sediment yield will increase.Most calibrated SWAT models show good applicability at the three hydrological stations,with Maojiahe and Yangjiaping stations performing better than Zhangjiashan station.The study shows that low-resolution models may also have good applicability,and the best model combinations of DEM and land use resolution are 750 m and 300 m,30 m and 90 m,and 150 m and 3000 m,respectively.(3)The impact of DEM on parameter sensitivity is greater than land use,and the uncertainty of sediment simulation is higher than that of runoff.Among all the models,USLE＿C,EPCO,ALPHA＿BNK,and CN2 are parameters that are sensitive,and when the DEM resolution is reduced to 3000 m,BIOMIX,LAT＿SED,USLE＿K,and CH＿N1 replace OV＿N,SMTMP,SURLAG,and USLE＿P as the parameters with strong sensitivity.Most models have small deviations in runoff simulation and have small uncertainties.The deviation of sediment simulation is higher than that of runoff,with 92 models having a deviation percentage of runoff simulation values within 25%,while only 69 models have a deviation percentage of sediment simulation values within 25%,and some models have a large deviation in sediment simulation.In addition,higher accuracy DEM data typically reduces model uncertainty.(4)The accuracy and uncertainty of hydrological model simulation are significantly influenced by the threshold of watershed area,number of sub-basins,and number of hydrological response units(HRUs).The threshold of watershed area shows a power-law correlation with the number of sub-basins and HRUs,while the number of sub-basins and HRUs exhibit a linear positive correlation.Both too small and too large values of the threshold of watershed area can affect the model’s accuracy and uncertainty.In the Jinghe River Basin,the SUFI-2 and GLUE algorithms demonstrate good applicability,with NSE values of 0.63 and 0.59 for runoff and sediment,respectively,and 0.57 and 0.62 for PSO algorithm,which has a larger simulation deviation.When the threshold of watershed area is set to 2000 km2,the model shows better simulation results for both runoff and sediment,with NSE values of 0.7and 0.65 for runoff and sediment. |
