| The SCS-CN method is one of the most widely used hydrological models to predict surface runoff from watershed for a given rainfall event. However, runoff generation is governed by multiple factors including topography, landform, soil, climate, vegetation and land use with spatial and temporal heterogeneity in nature, the standard SCS-CN method could produce large errors in predicting surface runoff. Therefore, it is an effective way to modify the original model for particular region, specific watershed for improving the accuracy. In this study, the measured event rainfall-runoff datasets from Qingshui River watershed and three watersheds located in Caijiachuan watershed on the Loess Plateau of China were splinted for calibrating and validating the original model and its revised models to estimate surface runoff from. Main results as follows:(1)This paper employed a dataset of rainfall and runoff events from 1980 to 2003 in Qingshui River watershed (436km2), those datasets were splinted for calibrating (1980 to 1996) and validating (1997 to 2003). Due to the division of antecedent moisture condition is not clear in the standard SCS-CN model, antecedent precipitation index has often been used to define soil moisture conditions prior to the storm events, estimating M based on the API, which established the relationship between M and API (M= αebAP1), named MMS-API(M8). After investigating the nine different models, including the original SCS-CN and its eight inspired modified models, it was found that: ①Qingshui River lacked soil moisture data, model M8 estimated M on the base of API, therefore, compared with the other eight models, the proposed model M8, performed the best. Therefore, the M8 is recommended for surface runoff prediction in this study area, ②for the Qingshui River watershed with daily average storm rainfall 10-25mm, the M8 model performed the best, λ=0.01, CN=22.52; When daily average storm rainfall≥25mm, the M8 model performed the best, λ=0.02, CN=20.33.(2) The measured event rainfall-runoff datasets from Caijiachuan watershed on the Loess Plateau of China during 2005 and 2009 were splinted for calibrating (2005 to 2006) and validating (2007 to 2009). Rainfall intensity was incorporated into the standard SCS-CN model in Caijiachuan Watershed, while optimizing rainfall intensity adjusting factor β, the initial abstraction coefficient (λ) and the potential maximum retention S, it was found that, in farmland and grassland dominated watershed: λ=0.001, β=-0.2, S= 1849mm; in forest plantation dominated watershed:λ=0.002, β=-0.49, S=321.5mm; in secondary forest plantation dominated watershed:λ=0.008, β=-0.55, S=928mm.(3)During the validation period, compared with the standard SCS-CN model the fitting precision was significantly improved. ①in farmland and grassland dominated watershed:the model efficiency increased to 0.7 from -0.17, the value of Root mean squared error decreased to 0.11 from 0.23;②in forest plantation dominated watershed:the model efficiency increased to 0.60 from-0.66, the value of Root mean squared error decreased to 0.45 from 0.92; ③in secondary forest plantation dominated watershed:the model efficiency increased to 0.81 from 0.78, the value of Root mean squared error remained unchanged. The results indicated that the SCS-CN model incorporating rainfall intensity could improve runoff estimation accuracy.(4) In Caijiachuan Watershed, absolute value of rainfall intensity adjusting factor β was the smallest in farmland and grassland dominated watershed, which indicated that effective fainfall was the highest. The initial abstraction coefficient (λ) and initial abstraction (Ia) were the highest in secondary forest plantation dominated watershed, which indicated that the capacity to store rainwater was superior to the other two watersheds.(5) The initial abstraction coefficient (λ) of the optimized models was much smaller than 0.2. |
PDF Full Text Request