Comparison between Curve Number Method and Green-Ampt Method to Estimate Sediment Yield in a Small Watershed Using SWAT
Daiki Kamei1, Kayoko Ogura1, Wang Chunying1, Sadao Eguchi2, Kazunori Kouyama2, Noriko Yamaguchi2, Kanta Kuramochi1, Ryusuke Hatano1
1Laboratory of Soil Science, Graduate School of Agriculture, Hokkaido University
2National Institute for Agro-Environmental Sciences
Key Words: curve number method, green-ampt method, sediment yield, small watershed, SWAT (Soil and Water Assessment Tool)
Cesium-137 (137Cs), one of the radioactive elements, was scattered on wide area around Fukushima Prefecture due to Fukushima Daiichi nuclear power station accident caused on March, 2011. Half-life of 137Cs is relatively long (30.1 years) (Brown et al., 1955) and 137Cs is strongly absorbed by soil particles (Kirk and Staunto, 1989). Therefore, to predict dynamics of 137Cs, it is needed to predict long term dynamics of sediment and stream flow which transports sediment. For prediction of dynamics of sediment in watershed, SWAT (Soil and Water Assessment Tool), which is one of hydrological models, is widely used in the world.
There are three inflow routes of sediment into streams; surface runoff, lateral flow and groundwater flow. On prediction of 137Cs dynamics, estimation of surface runoff is especially important. SWAT has two surface runoff calculation methods, curve number (CN) method and green-ampt (GA) method. However, there are few studies estimating sediment yields with GA method. The objectives of this study is to estimate sediment yield in a small watershed with CN method and GA method and compare and evaluate the performances of both methods.
The study site is the upper stream of the Hiso River, which is located in Iitate, Fukushima Prefecture and the watershed area is 445 ha. Daily stream flow and sediment load data from 2013 to 2014 were used in this study; data from 2013 were used for the calibration of parameters in SWAT and 2014 for the validation using the parameters. Model performance was evaluated by using the coefficient of determination (R2) and the Nash and Sutcliffe (1970) efficiency (ENS). Empirically, a good model performance would have R2 and ENS values higher than 0.5.
The statistical evaluation for the daily stream flow showed the simulation results using both methods were acceptable. However, both methods might considerably underestimate surface runoff. GA method estimated higher lateral water flow than CN method. The model simulation using both methods underestimated the peaks of sediment load in summer, but GA
method overestimated sediment load during the periods of base flow. As a result, statistical evaluation for daily sediment load showed that CN method performed better than GA method.
This may be because overestimation of saturated hydraulic conductivity (SOL_K) in GA method caused underestimation of surface runoff and sediment load at the peaks, and overestimation of lateral water flow and sediment load during the periods of base flow.