A modified curve number method for runoff prediction of different vegetation types at the slope scale in China

The Soil Conservation Service Curve Number (SCS-CN) method is widely used to estimate surface runoff, relying on the Curve Number (CN) from the SCS handbook. However, its reliance on three discrete hydrologic condition (HC) categories (e.g., good, fair, poor) reduces sensitivity to land surface variability, leading to abrupt CN changes and inconsistent runoff estimates. To enhance CN estimation accuracy, this study assessed CN values for each HC using the median (CN _C) and least-squares fit (CN _F) methods based on rainfall-runoff observations from 65 monitoring sites across China. Although CN _F slightly improved CN estimation, it still resulted in unsatisfactory performance, with Nash–Sutcliffe efficiency (NSE) values below 60 % under many HC conditions, highlighting the limitations of categorical HC classification and internal CN variability. To address this, the study developed three equations integrating CN values with tabulated CN values for bare soil and vegetation coverage, tailored to grassland, shrubland, and woodland ecosystems. This method was calibrated and validated using data from 58 sites and tested at 7 independent sites. Results showed marked improvements in runoff prediction accuracy: for calibration, NSE values increased from 60.55 % (original method) to 78.57 % for grassland, from 58.51 % to 82.90 % for shrubland, and from –21.20–64.39 % for woodland. Similar improvements were observed in validation, with NSE increasing from 61.78 % to 79.11 % for grassland, 57.46–81.28 % for shrubland, and –43.68–62.71 % for woodland. These findings demonstrate the superior performance and broader applicability of the proposed method for runoff prediction in China’s diverse vegetated landscapes.