Parameters optimization based on the combination of localization and auto-calibration of SWAT model in a small watershed in Chinese Loess Plateau

This study simulated the watershed flow and sediment responses based on calibration of the SWAT model in the semi-arid Chinese Loess Plateau (LP) where soil erosion intensively occurs. After the model’s initiation and manual modification, a 7-year inconsecutively observed flow and sediment data from 1984 to 1990 was used to analyze the model’s application in the selected watershed called AJW in the Chinese LP region. The model procedure included sensitivity analysis, parameter calibration and model validation. The best parameter set was finally determined based on the combination of parameter localization and auto-calibration. Then the model was assessed for its accuracy based on the NSE estimation, resulting in 0.77 and 0.67 for calibration and 0.46 and 0.32 for validation on simulations for flow and sediment, respectively, which is a moderately satisfactory accuracy among the applications of the SWAT model. Annual watershed assessment on flow and sediment with the calibrated SWAT model resulted in a multiyear averaged annual runoff coefficient of about 2.7% and an erosion modulus of 797 t/(km²·a⁻¹) in the AJW, indicating a beneficial consequence from the implementation of the historical soil and water conservations.