Comparative Application Of Xinanjiang,SWAT And BTOPMC Models

This paper has used the lumped hydrological model--double parabolic cruve Xinanjiang model, response unit semi-distributed hydrological model--SWAT model and grid-distributed hydrological model--BTOPMC model to simulate the daily runoff in Hanjiang river basin and Shuangqiao Basin respectively.This study ran hydrological model with the input data of SRTM3 DEM, IGBP land-use and land-cover, FAO soil data, NOAA-AVHRR NDVI data and CRU TS2.0 meteorological data. For comparison, all of three hydrological model used Shuttleworth-Wallace model to calculate the evapotranspiration and model parameters were calibrated using SCE-UA algorithm with Nash-Sutcliffe efficiency as their object functions.The application in Hanjiang basin, choose the data of 1981 to 1984 as calibration period, 1985 to 1988 as verification period. Simulated the daily runoff of Guanyinqiao station, Baokeng station, Yangjiafang station, Hekou station, Shanghang station, Shuikou station, Xikou station, Hengshan station and Chaoan station respectively. Compared model parameters, Nash-Sutcliffe efficiency, water volume balance, pertinence of Nash-Sutcliffe efficiency and hydrologic features, analysis of runoff process in each sub-basin. The simulation indicates that:(1) All parameters in three models were calibrated in their reasonable ranges and can objectively reflect the underlying surface regime, precipitation distribution and nested relationship of each sub-basin.(2) For the Nash-Sutcliffe efficiency of each sub-basin: Xinanjiang model performed best, BTOPMC model second and SWAT model worst. For the mean square deviation of Nash-Sutcliffe efficiency in calibration and validation periods: BTOPMC model performed consistently of simulation results at all stations, Xinanjiang model secondly, and SWAT model worst. For the simulation water volume compared to the observation of each sub-basin: BTOPMC model simulated best in water volume compared to the observation, SWAT model second and Xinanjiang model worst.(3) BTOPMC model's Nash-Sutcliffe efficiency is positive correlation to catchment area in verification period, and this indicates BTOPMC model is more appropriately for large watershed. Xinanjiang model and SWAT model's Nash-Sutcliffe efficiency is uncorrelation to catchment area, average elevation, elevation's Cv, elevation factor, mean annual precipitation, rainfall distribution uniformity, vegetation distribution uniformity and soil distribution uniformity both in calibration period and verification period. BTOPMC model is the same in verification period.(4) In lumped Xinanjiang model, the time arrving at flood peak is accurate in large and small basin. However, the big peak discharge is too high in large basin and the small peak discharge is too low in large and small basin. Response unit semi-distributed SWAT model and grid-distributed BTOPMC model had a lower peak discharge in large and small basin, also ahead of the time arrving at the flood peak. BTOPMC model performed best in low flow simulation, SWAT model second and Xinanjiang model worst.The application in Shuangqiao basin, choose the data of 1991 to 1994 as calibration period, 1995 to 1998 as verification period. Compared the performance of different hydrological model from total simulating effect and each flow class' s simulating effect. The simulation indicates that:(1) Xinanjiang, SWAT and BTOPMC models could represent daily rainfall-runoff processes in Shuangqiao basin very well. From the overall, Xianjiang model got a better simulation than SWAT and BTOPMC models; all of three models were good at water volume simulation.(2) According to different discharges, SWAT model had a better simulation of water volume than Xinanjiang and BTOPMC models; the accuracy of Xinanjiang model was highest in simulation of high discharge and BTOPMC model performed best in simulation of middle and low flow.