Initial Abstraction Ratio in the SCS-CN Method in the Loess Plateau of China

The Soil Conservation Service curve number (SCS-CN) method is widely used to predict runoff because of its simplicity. The initial abstraction ratio (λ), which is assumed to have a default value of 0.2 in the SCS-CN method, is an important influence factor for runoff estimation. A more reasonable value may improve the performance of this method. The objective of this study was to determine what value for λ would give the most reasonable estimation of runoff in the Loess Plateau of China. Rainfall-runoff data from farmland plots in Shaanxi (205 rainfall events) and Gansu (552 rainfall events) provinces were used. The results indicated that λ varied from 0.01 to 0.46, with an average of 0.08 and a median of 0.05 for rainfall events. The difference in the event runoff predicted using λ = 0.05 versus λ = 0.2 was significant at a significance level of 0.01. The prediction accuracy for λ = 0.05 was greater than that for λ = 0.2 in accordance with the assessed events, especially for events with a rainfall amount less than 25 mm. Nevertheless, the SCS-CN method with λ = 0.05 still did not significantly improve the prediction accuracy under the event runoff conditions. The Nash-Sutcliffe model efficiencies (ME) at Zizhou and Xifeng were 0.25 and 0.51, respectively. However, the SCS-CN method using both λ values performed well in simulating annual runoff. The ME of the annual runoff for both λ values was approximately 0.75, which was obviously greater than that of the event runoff. Compared with the annual runoff predicted using λ = 0.2, those using λ = 0.05 did not obviously improve the prediction accuracy. Further studies are needed to focus on the effect of variation of λ values with antecedent soil moisture on model performance of the SCS-CN method.