Calibration and validation of the SWAT model for a forested watershed in coastal South Carolina
Modeling the hydrology of low-gradient coastal watersheds on shallow, poorly drained soils is a challenging task due to the complexities in watershed delineation, runoff generation processes and pathways, flooding, and submergence caused by tropical storms. The objective of the study is to calibrate and validate a GIS-based spatially-distributed hydrologic model, SWAT, for a low-gradient, third-order Turkey Creek watershed (7,260 ha) within the Francis Marion National Forest in South Carolina Coastal Plain. The model calibration used GIS spatial data and two years (2005 – wet and 2006 - dry) of stream flow and climate data, and was validated with one very dry year (2007) of data. Based on limited field measurements, results showed that the SWAT model with an improved one-parameter “depletion coefficient” can predict the stream flow processes of this watershed reasonably well and better than the classical CN method. The model performed “Good (E = 0.74; RSR = 0.51)” to “Very Good (E = 0.98; RSR = 0.15)” for the monthly and only “Satisfactory (E = 0.65; RSR = 0.60)” to “Good (E = 0.67; RSR = 0.57)” for the daily calibration and validation periods, respectively. It was concluded that the refined SWAT model was still unable to accurately capture the flow dynamics of this forest ecosystem with high water table shallow soils for very wet saturated and very dry antecedent conditions which warrants further investigations on these forest systems. Finally, the three-year average annual runoff coefficient of 17% and ET of 900 mm predicted by the model were found reasonable compared to other published data for the region.