A Comparison of the Curve Number and Green-Ampt Models in an Agricultural Watershed

Abstract

The curve number and Green-Ampt rainfall-runoff models were compared in the highly agricultural San Joaquin River watershed in California using the Soil and Water Assessment Tool (SWAT). The rainfall-runoff models were left uncalibrated to objectively assess model performances; however, streamflow simulations showed high accuracy compared to observed data caused by the large impact of reservoir releases on streamflow. For daily simulations, the Nash-Sutcliffe model efficiency coefficients were 0.81 for the curve number model and 0.78 for the Green-Ampt model. A Nash-Sutcliffe coefficient of 0.93 was found for both models for the monthly simulations. The Green-Ampt model more accurately predicted large streamflow events than the curve number model, while the curve number model better predicted normal flow events. Both models tended to overpredict streamflow. The average monthly hydrologic components of surface runoff, groundwater flow, lateral soil flow, and the amount of water in the soil column were also compared to quantify the underlying differences between the two rainfall-runoff models. These comparisons yielded equal or comparable average monthly surface runoff values between the two rainfall-runoff models, but higher subsurface flows (lateral soil and groundwater inflows) and soil water volumes for the Green-Ampt model. These results are largely due to the difference in model assumptions, where the curve number model assumes an initial abstraction before surface runoff and the Green-Ampt model assumes surface runoff only when the precipitation rates is greater than the infiltration rate. The selection of the most appropriate rainfall-runoff model should be based on the watershed physical characteristics and the overall goal of the watershed modeling.