The prediction accuracy of a distributed hydrological model depends on how well the model input spatial data describe the characteristics of the watershed. Especially in a large catchment, could a higher resolution of input data contribute to improving the accuracy of model simulations? In this study, surveyed soil data with two different spatial resolutions were used as input data for a SWAT model simulation in a large catchment of Xinjiang River basin (15535km(2)) in China. The purpose of this study is to (1) evaluate the effect of different spatial resolutions of soil type on the simulation of hydrologic components of the SWAT model and (2) examine the adaptability of applying high precision of soil resolution in large catchments. The first soil data set has a coarse resolution of 1: 3,000,000 and five major types of soils are classified. The second one has a finer resolution of 1: 1,000,000, in which 36 major soil types were classified. Evaluation of the distribution of SWAT hydrologic response units (HRUs) is conducted by examining the changed percentages of soil class and land use of the two soil types. Result show that when the threshold of soil area definitions in the watershed is increased, the number of HRUs decreases when the higher resolution soil data was used, but there is little change when the lower resolution soil data was used. In order to evaluate the differences in model predictions of the two SWAT setups with differing soil data, modelled streamflows were compared before and after calibration. Before calibration, the coarse resolution soil data performed marginally better than the fine resolution soil data. After calibration, streamflow predictions of both SWAT setups improved. In the study watershed, using a higher resolution soil data didn't yield improvements in monthly streamflow modelling with the SWAT model. This lack of difference in streamflow predictions between the two data sets is attributed to theapplication of the SCS curve number method in SWAT. Monthly soil water storage and evapotranspiration outputs of the SWAT model were also compared. Results show that the finer resolution data produced higher monthly soil water storage estimates than the coarse resolution data across the entire watershed during the simulation period. However, there was little difference in evapotranspiration output. This insensitivity of evapotranspiration to soil properties suggests that perhaps a relatively simplified computation method for the evapotranspiration module in SWAT could be considered. The implications of this study are that improvement of the resolution of soil data does not necessarily contribute to a more accurate prediction of streamflow in large catchments.
