Submitted to: Hydrological Processes
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/10/2012
Publication Date: 5/11/2012
Citation: Cho, J., Bosch, D.D., Vellidis, G., Lowrance, R.R., Strickland, T.C. 2012. Multi-site evaluation of hydrology component of SWAT in the Coastal Plain of Southwest Georgia. Hydrological Processes. DOI: 10.1002/hyp.9341.
Interpretive Summary: Watershed scale simulation models can be useful for examining the impacts of climatic variation and changes in land-use. However, the utility of these models is dependent upon proven reliability or accuracy of the simulations. While hydrologic measurements collected within watersheds can be used to assess the accuracy of model simulations, the utility of these results for extrapolation into watersheds where no measurements are available is largely unknown. This study examined the accuracy that could be obtained in hydrologic simulation within watersheds where no hydrologic data were available when using parameter sets obtained through calibration using similar watersheds where hydrologic data were available. Model simulation results yielded hydrologic simulations within 2% for the calibration watershed. For validation results where no hydrologic data were used to establish modeling parameters the errors ranged from 6% to 23%. These results indicate that that assessments of water quality responses to conservation practices based upon SWAT simulations will accurately predict the direction of the change in water quality (up or down), but uncertainty in absolute magnitude of the predicted change will be at least as large as the percent error in predicted flow…in this case +/- 23%.
Technical Abstract: Land use changes and development within mixed use watersheds can result in adverse or positive environmental impacts. In this study, the feasibility of using the Soil Water Assessment Tool (SWAT) watershed model to predict hydrologic responses in watersheds within the Coastal Plain of southwest Georgia was evaluated. SWAT was calibrated on the 16.7 km2 subwatershed K (LRK) within the Little River Experimental Watershed (LREW) by varying six model parameters. The optimized parameter set was then applied to a watershed of similar physical conditions (LRJ), a smaller watershed with different land use and soils (LRO), and three larger watersheds within the same drainage system (LRI, LRF, and LRB) without further calibration . In addition to graphical comparisons between simulated and observed streamflow, three quantitative measures were also calculated, percent error (PE), RMSE-observations standard deviation ratio (RSR), and Nash-Sutcliffe efficiency (NSE). Simulation results with PE < ±25%, RSR = 0.70, and NSE > 0.50 were considered to be satisfactory. Following calibration for LRK, a PE of 1.2% and daily and monthly NSE value of 0.76 and 0.94, respectively, were obtained. Validation results on LRJ (22.1 km2) were considered to be good, PE was -5.6%, daily NSE was 0.71, and monthly NSE was 0.89. The poorest simulation occurred on LRO (15.9 km2) with corresponding measures of 22.7%, 0.61, and 0.83. For the nested watersheds, PE increased with increasing watershed size, equal to -8.7% for LRI (49.9 km2), 8.4% for LRF (114.9 km2), and 15.6% for LRB (334.3 km2). SWAT performed well in simulating temporal trends of discharge within all three larger nested watersheds with daily and monthly NSE values greater than 0.7 in all cases. The variations in total runoff among subwatersheds can be attributed to increases in storages and transpiration from shallow aquifer within the wetland areas, which are greater near the watershed outlet. The modeling results indicate that SWAT is capable of simulating temporal trends of streamflow in ungauged watersheds with similar geophysical characteristics to the LREW. Furthermore, the ability to extrapolate model parameterization from gauged watersheds to similar ungauged watersheds lends significant credibility to watershed modeling studies being conducted throughout the U.S. Results indicate that assessments of water quality responses to conservation practices based upon SWAT simulations will accurately predict the direction of the change in water quality (up or down), but uncertainty in absolute magnitude of the predicted change will be at least as large as the percent error in predicted flow…in this case +/- 23%.