SWAT application to estimate design runoff curve number for South Korean conditions

Abstract

AbstractRunoff estimations based on the standard USDA–NRCS curve number (CN) table without calibration have a tendency to give inaccurate results when the CN values are applied in South Korea which has many high slope watersheds and that has a continental monsoon climate. Particularly for the design flood estimation, accurately calibrated CN values are required because the estimated peak flow is very sensitive to the selection of CN. However, the lack of flood data makes it difficult to calibrate and assign runoff CNs to Korean watersheds. Even if sufficient data are available to estimate CN values, it is also difficult to obtain the direct flows by separating base flows from total runoff hydrographs due to the temporal irregularity of rainfall events and the resulting complex pattern of runoff. Therefore, an alternative method for estimating CNs needs to be developed to overcome these issues.The purpose of this study is to present a method for estimating runoff CNs using the soil and water assessment tool (SWAT) model which can take into account watershed heterogeneities such as climate conditions, land use and soil types. The proposed CN estimation method uses the simulated flow data by SWAT instead of using measured flow data. This method has advantages in estimating CN values spatially for each subbasin division considering watershed characteristics. The use of daily data can reduce the sensitivity to the abnormality that is commonly involved in flow data with a small time scale. The SWAT‐based CN estimation method, combined with the asymptotic CN method, was applied to the Chungju dam watershed in South Korea. A regression equation was then developed from this approach, which was used to estimate CN values that decrease exponentially as rainfall amounts increase and that converge to 60·6 and 79·4 without and with considering subsurface lateral flow, respectively. Furthermore, the CN values for the antecedent moisture conditions were determined using the probabilistic approach. The CN associated with the 50% probability for the Chungju dam watershed is 87·8 which can be taken to be representative of antecedent moisture condition (AMC) II. The CNI and CNIII associated with 90% and 10% probabilities are 78·9 and 94·1, respectively. The estimated CNII = 87·8 differs markedly from the geographic information system (GIS)‐based CN 65·0, which implies that the standard USDA–NRCS CN method should be calibrated to the studied area of interest. Copyright © 2010 John Wiley & Sons, Ltd.