5.3. The effect of soil moisture
The antecedent moisture condition (AMC) plays an important role in the progress of runoff. The conventional SCS-CN method uses the 5-day antecedent rainfall amount prior to a rainfall event to select three levels of AMC. Measured values of CN are not limited to the three discrete AMC level defined by the 5-day AMC which causes a sudden jump in runoff prediction (Shi et al., 2017). Moreover, Koelliker (1987) found that the probability of an AMC2 condition occurred prior to a storm event was only 7%, rather than 50% (Hjelmfelt et al., 1982), which indicated the AMC value was underestimated and thereby resulting in the underestimation of runoff by the standard SCS method; Huang et al.(2007) also indicated there was no evident relationship between CN and the 5-day AMC, and that tabulated CN values are almost always less than measured values, which confirmed the underestimation of runoff. Therefore, the 5-day AMC used by the standard CN method is not a reasonable.
Soil moisture is an alternative solution to characterize the AMC would be to use direct measurements, or estimated values, of soil moisture prior to a rainfall event, rather than using antecedent rainfall, which is the most important factor defining the initial abstraction of the SCS-CN method. Improved knowledge of antecedent moisture would greatly enhance rainfall runoff prediction (Wood, 1976; Michele and Salvadori, 2002). The results of Huang et al.(2007) method preformed better than Huang et al.(2006) and standard SCS-CN method (Fig. 2) which indicated the equation of soil moisture (Eq.7) is more suitable for AMC than the antecedent 5-day rainfall.
In the new CN value, three factors of slope, soil moisture and storm event were introduced, which represented the characteristic of the land condition, antecedent moisture condition and characteristic of the storm event respectively, which all have great impact on the progress of runoff. However, as compared with Huang et al.(2006) and Huang et al.(2007) method, the performance of the Method 1 was significantly improved after accounting for the storm duration factor (Table 3). From this standpoint, it may be indicated the characteristic of the storm event (storm duration) has greater impact on runoff prediction of the SCS-CN method which has been frequently ignored in other modified SCS-CN methods and followed by soil moisture and slope. Moreover, the λ = 0.001 is more suitable for the initial abstraction coefficient in the study area according to comparison between Method 1 and 2, the same conclusion was also obtained by Huang et al.(2007) for four plots in the Loess Plateau. The slope parameters in Method 2 which is obtained from Huang et al.(2006) with data from slopes ranging from 14 % to 140 % are also more suitable for application as compared with Method 3. This inference can be further confirmed by sensitivity analysis which also suggested that the initial abstraction ratio λ and the slope parameters (a1 and a2 ) are less sensitive, thereby the Method 2 can be applied to similar semi-humid, semiarid, and arid regions with the optimized parameters, but may need adjustment for humid regions because the soil moisture condition and rainfall characteristic may be different from those tested in this study.