Development and testing of a modified SWAT model based on slope condition and precipitation intensity

Abstract

The SWAT model is a semi-distributed hydrological watershed model with an explicit physical mechanism, which is widely applied in areas characterized by a wide variety of precipitation distributions and slope conditions. However, the Soil Conservation Service-Curve Number (SCS-CN) value and precipitation initial abstraction ratio were not related to slope, soil moisture content, precipitation intensity, and land cover interception within the original SWAT model. Thus, a modified SCS-CN model is proposed by considering the slope, rainfall intensity, soil moisture, and interception effects of the landscape in this study, specifically revising the critical sensitive parameters including CN2 and initial abstraction ratio (λ). The modified SCS-CN model was added to the original SWAT model and named the SWAT-CH model, which is applied to Chao River watershed which experiences obvious differences in precipitation and slope. For this application, the daily streamflow was simulated for the calibration and validation periods of 1993–2005 and 2006–2013, respectively. The simulated variable is monthly runoff at watershed outlet, namely Xiahui hydrological station. The parameter sensitivity and uncertainty and the model applicability were analyzed using as Chao River watershed as a case study. The main results are as follows. ① Compared with the original SWAT model, the parameters of SOL-K, HRU-SLP, SLSUBBSN, and CANMX were identified as the most sensitive parameters for the runoff simulation in the new SWAT-CH model. ② The prediction efficiency of the SWAT-CH model surpasses that of the original SWAT model. The Nash-Sutcliffe efficiency, KGE index, and the PBIAS were 0.94, 0.83 and 4.55%, respectively for SWAT-CH during the calibration stage; while, those corresponding indexes were 0.91, 0.84, and 7.00% for the validation period. The modified SWAT-CH model can better reproduce the hydrological process than the SWAT model. In detail, the Nash-Sutcliffe efficiency, KGE index, and the PBIAS were improved by 0.02, 0.10, −4.01 percentage points for the calibration period, as well as by 0.18, 0.25, and −28.55 percentage points for the validation period in the modified SWAT-CH model compared with the original SWAT model. ③ Influenced by soil moisture, there is a substantial difference between the SWAT-CH and SWAT models in wet periods at various temporal scales. ④ Parameter uncertainty is the main source of the uncertainty in hydrological estimation for the Chao River watershed. The modified SWAT-CH model has, to some extent, reduced the effect of parameter uncertainty with increasing structure uncertainty from 6.00% for the SWAT model to 9.00% for the SWAT-CH model. The model adaptability has been enhanced by introducing terrace slope, precipitation intensity, and soil moisture for the modification of CN2 and the abstraction ratio, which makes the revised SWAT-CH model have better simulation performance in areas with the obvious characteristics of terrain & rainfall heterogeneity and drastic dry-wet cycles in the soil profile.