Investigation of SCS-CN and its inspired modified models for runoff estimation in South Korean watersheds

• We investigated the SCS-CN and its inspired modified models in Korean watersheds.
• Runoff estimated using the calibrated CN is a better alternative to tabulated CN.
• For same CN values, small λ (< 0.2) exhibit better results for SCS-CN model.
• To estimate runoff, the initial abstraction (Ia=0.02P) performs better than Ia=0.2S.

Employing a dataset of 658 large storm-events from 15 South Korean watersheds (48.6–249.63 km2), this study established the initial abstraction (Ia) as 2% of the rainfall amount (obtained based on three different scenarios), instead of the originally assumed 20% of the maximum potential retention (S) in the Soil Conservation Service Curve Number (SCS–CN) model. After investigating 8 different models, including the original SCS-CN and its inspired modified models, it was found that lower values of the initial abstraction coefficient (λ) exhibited better runoff estimation than the fixed λ (=0.2) as lower λ < 0.2 was recommended by researchers and supported by this work for CN values calculated from observed storm-events. To reduce errors in runoff estimation, CNs should be calibrated using observed rainfall-runoff data from regional watersheds. The proposed model, which incorporates the newly suggested initial abstraction based on rainfall-runoff rank-order data, outperformed in 14 out of 15 watersheds. Using the optimized/calculated CN values, the proposed model ranked first (as best) based on the evaluation of three different performance indices, followed by the model of Hawkins et al. (2002), one of the models of the Mishra and Singh, and the original SCS-CN model respectively. Owing to a significant degree of agreement between the observed and calculated runoff, the proposed model is recommended for field applications in this study area.