Objective This study aims to evaluate the applicability of the SCS-CN model in small-scale watersheds and improve the prediction accuracy of the SCS-CN model.

Methods This study utilized regression analysis of precipitation and runoff data from 2015 to 2016 to determine initial abstraction. The statistical parameters, including root mean square error (RMSE) and Nash-Sutcliffe efficiency (NSE), were employed to optimize initial abstraction ratios and potential maximum retention values of the SCS-CN model. Validation of the SCS-CN model with optimized parameters was performed using rainfall-runoff data from 2017 to 2018.

Results The initial abstraction ratio was more sensitive to the SCS-CN model than the runoff curve number. The optimized initial abstraction ratios for flat planting, open-ridging, and tied-ridging were 0.033−0.039, 0.030−0.035, and 0.029−0.030, respectively. The corresponding potential maximum retention values for flat planting, open-ridging, and tied-ridging were 104.9−124.7 mm, 186.8−227.4 mm, and 231.2−243.6 mm, respectively, while the curve numbers (CN) were 67.1−70.8, 52.8−57.6, and 51.0−52.3, respectively. For slope gradients of 5° and 10°, the optimized initial abstraction ratios were 0.033 and 0.032, respectively, with potential maximum retention values of 185.2 mm and 178.8 mm, respectively. The CN values for slope gradients of 5° and 10° were 57.8 and 58.7, respectively. Ridge-furrow rainwater harvesting, particularly tied-ridging rainwater harvesting, demonstrated lower optimized initial abstraction ratios and CN values, coupled with higher potential maximum retention values, compared to flat planting. Both optimized initial abstraction ratio and CN value increased as slope gradient increased, while potential maximum retention decreased as slope gradient increased.

Conclusions The optimized SCS-CN model can accurately predict runoff in ridge-furrow rainwater harvesting.