Fu S (Fu, S.)^1,2,* Zhang G (Zhang, G.)^2,1, Wang N (Wang, N.)³, Luo L (Luo, L.)²

Abstract: The Soil Conservation Service curve number (SCS-CN) method is widely used to predict runoff because of its simplicity. The initial abstraction ratio (lambda), 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 lambda 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 lambda 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 lambda = 0.05 versus lambda = 0.2 was significant at a significance level of 0.01. The prediction accuracy for lambda = 0.05 was greater than that for lambda = 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 lambda = 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 lambda values performed well in simulating annual runoff The ME of the annual runoff for both lambda values was approximately 0.75, which was obviously greater than that of the event runoff Compared with the annual runoff predicted using lambda = 0.2, those using lambda = 0.05 did not obviously improve the prediction accuracy. Further studies are needed to focus on the effect of variation of lambda values with antecedent soil moisture on model performance of the SCS-CN method.

Keyword: Initial abstraction ratio; Loess Plateau; Runoff; SCS-CN method