TY - JOUR T1 - Assessment of the Soil Conservation Service–Curve Number method performance in a tropical Oxisol watershed JF - Journal of Soil and Water Conservation SP - 500 LP - 512 DO - 10.2489/jswc.74.5.500 VL - 74 IS - 5 AU - G.J. Alves AU - C. Rogério de Mello AU - S. Beskow AU - J.A. Junqueira, Jr. AU - M.A. Nearing Y1 - 2019/09/01 UR - http://www.jswconline.org/content/74/5/500.abstract N2 - The Soil Conservation Service–Curve Number (SCS-CN) method is a rainfall-runoff model intended to estimate direct surface runoff (DSR) from rainfall. This method is based on an empirical approach of the relationship between rainfall (P) and ground conditions (soils, management, and antecedent moisture content). SCS-CN method can be applied using published tables containing a CN value for each combination of soil hydrology, land use, and management, and the total five-day antecedent rainfall (P5). However, its accuracy increases as observed rainfall-runoff events are used for calibration. The standard SCS-CN method assigns 0.2 to the initial abstraction coefficient (λ), which refers to the ratio between initial rainfall abstraction (Ia) and the potential maximum retention (S). In this study, λ was evaluated for a tropical watershed with predominance of Oxisols using 15 methodologies for CN characterization, based on CN values published by National Engineering Handbook, Section 4 (NEH-4) and also derived from 166 monitored rainfall-runoff events. Such methodologies include the use of λ equal to 0.2—standard approach, 0.02—average value obtained from the 166 events, and 0.05—a value that has been suggested in some studies. The results demonstrated that a fixed area-weighted CN representing the entire watershed gave poor accuracy even when antecedent runoff condition (ARC) was considered. In contrast, the methodology based on calibration to the observed hydrological events provided good results. Improvements in DSR estimates were found for most of the methodologies applied when λ was set equal to 0.05 and 0.02 instead of 0.2, except for the spatially distributed CN model (Heterogeneity Model), in which the use of λ = 0.2 implied improvement of DSR estimates. ER -