SWAT Literature Database for Peer-Reviewed Journal Articles

Title:Evaluation of the SWAT model on a coastal plain agricultural watershed 
Authors:Bosch, D.D., J.M. Sheridan, H. L. Batten, J. G. Arnold 
Journal:Transactions of the ASAE 
Article ID: 
URL (non-DOI journals):http://ddr.nal.usda.gov/handle/10113/9717 
Broad Application Category:hydrologic only 
Primary Application Category:input effects 
Secondary Application Category:hydrologic assessment 
Watershed Description:22.1 km^2 subwatershed of the Little River (south central Georgia) 
Calibration Summary: 
Validation Summary:Daily/monthly (1997-2002) E values: LRD = -.24/.64 HRD = -.14/.55 HRM = -.04/.80 HRO = -.03/.80 
General Comments:SWAT2000 was used within BASINS 3.0 to perform comparisons between low & high resolution input data, and between default, modified, or mixed input parameters for existing groundwater conditions. Low resolution included BASINS DEM (90 m), USDA GIRAS land use data, and STATSGO soil data. High resolution data included 30 m DEM, field surveys of land use, and SSURGO soil data. The results improved with better resolution. Daily results could potentially improve with more calibration & model modification (LRD=low resolution/default; HRD = high ressolution/default; HRM = high resolution/modified; HRO = high resolution/mixed) 
Abstract:The Better Assessment Science Integrating point and Nonpoint Sources (BASINS) system was developed by the U.S. Environmental Protection Agency to facilitate developing total maximum daily loads (TMDLs). The Soil Water Assessment Tool (SWAT) is one of the watershed-scale simulation models within BASINS. Because of the critical nature of the TMDL process, it is imperative that BASINS and SWAT be adequately validated for regions on which they are being applied. BASINS and SWAT were tested using six years of hydrologic data from a 22 km^2 subwatershed of the Little River in Georgia. Comparisons were made between water balance results obtained using high and low spatial resolution data as well as those obtained using default initial parameters versus those modified for existing groundwater conditions. In general, all scenarios simulated general trends in the observed flow data. However, for the years with lower precipitation, the total water yields simulated with the low spatial resolution data and the default initial conditions were overpredicted by up to 27% of the annual precipitation input. Total water yields simulated using the high spatial resolution input data were within 20% of the observed yields for each year of the assessment. Nash-Sutcliffe model efficiencies (E) for monthly total water yields were 0.80 using the high spatial resolution data with the modified initial conditions and 0.64 using the low spatial resolution data with the default initial conditions. While the model simulated general streamflow trends, discrepancies were observed between observed and simulated hydrograph peaks, time to peak, and hydrograph durations. A one-day time lag between the simulated and observed time to peak was the primary cause of large errors in daily flow simulations. Model modification and more extensive calibration may be necessary to increase the accuracy of the daily flow estimates for TMDL development.