Rainfall-Runoff Simulation In Typical Sub-Basins Of Beiluohe Watershed Based On Swat Model

The Beiluohe watershed on the Loess Plateau was chosen as the research area.The accuracy of PERSIANN satellite rainfall product over the Beiluohe Watershed was verified.The rationality of SWAT model in two sub-basins of Beiluohe River was evaluated.With different DEM data sources,resampling methods,parameter calibration methods,time scale as well as DEM resolutions,different SWAT models were built to evaluate the uncertainty in rainfall-runoff simulation over sub-basin of Huluhe watershed.The main results were as follows:(1)The PERSIANN dataset had the tendency to underestimate precipitation,and could not estimate extreme precipitation very well.For the multi-year average precipitation and the average precipitation during flood season,the extraction method showed better performance than the bilinear interpolation method.Totally,precipitation in summer and fall was underestimated,while in winter and spring was overestimated.For the seasonal time scale,accuracy in summer was highest and was followed by spring.There were errors in the two extraction methods.It showed that the extraction method was better than the bilinear interpolation method.For the PERSIANN dataset,with the increase of elevation,the multi-year average precipitation decreased.(2)The SWAT performed reasonable simulation results over the two sub-basins of the Beiluohe watershed.Two monthly time scale SWAT models were established over Huluhe sub-basin and sub-basin above Wuqi hydrological station seperately based on ASTER GDEM 30 m data.It showed that results of the two models were satisfying and the total water balance was met.However,result of Huluhe watershed was in higher accuracy compared with that of Wuqi hydrological station sub-basin.(3)Many factors could result in uncertainties in SWAT rainfall-runoff simulation.Based on ASTER GDEM and STRM DEM,as well as three resampling methods(nearest neighbor interpolation method,bilinear interpolation method,cubic interpolation method),different SWAT models were built in daily and monthly time scale.SUFI-2 and PSO in SWAT-CUP were used for model parameter optimization.The fuzzy optimum model was introduced to evaluate the simulation results.The 30 m ASTER GDEM data extracted parameters in high precision.For the resampled DEM data,the nearest neighbor interpolation method obtained the most accurate topographic information.The model established by 30 m ASTER GDEM showed superior performance than models established by 90 m STRM DEM.For the models built by resampled DEMs,the nearest neighbor interpolation model got optimal result.Considering different calibration methods,performance of the model established by DEM resampled with the nearest neighbor interpolation,coupled with the SUFI-2 method,turned out to be the best.Monthly-scale simulations were better than daily-scale simulations.(4)To investigate the influence of DEM resolution on extraction of hydrological,topographic characteristics and runoff simulation,the 30 m ASTER GDEM was resampled and SWAT models with ten different resolutions were built seperately in Huluhe sub-basin.All hydrologiical models were built on monthly basis.The DEM resolution could greatly influence slope,area extraction and watershed subdivision,whereas had less effect on elevation.For DEM resolution lower than 150 m,the simulations were reasonable.It took more time for model established by the 30 m DEM to obtain optimal parameters.The 120-150 m DEM performed better simulations.The results of models built by DEMs,whose resolutions were less than 600,showed that resolution had effect on the division of HRU,as well as the model accuracy.