A Validation Study of SWAT in an Urbanizing Multiuse Watershed in the Central U.S

Watershed-scale distributed hydrologic/water quality (H/WQ) models such as the Soil and Water Assessment Tool (SWAT) are increasingly important tools for land managers. They are powerful prediction tools, and are also useful to assess and develop Best Management Practices (BMP) and estimates of Total Maximum Daily Loads (TMDL). A major impediment to successful calibration and validation of H/WQ models is a general lack of distributed monitoring data including discharge, sediment yield and climate data. The current study evaluates SWAT model simulations of water and sediment yield using publically available meteorological and USGS flow data versus simulations developed using comprehensive data sets from a nested network of five hydroclimate stations collecting continuous discharge, sediment and climate data. This work is taking place in the Hinkson Creek Watershed (HCW, 230.8 km2) located in central Missouri. Elevation ranges from 170 m at the outlet to 287 m at the headwaters. Landuse in the upper HCW consists of rural pasture and wooded areas, while the lower HCW contains the city of Columbia. Soils in the upper HCW are characterized as loamy till with a well developed clay pan, and in the lower HCW as thin cherty clay and silty to sandy clay. The transitional climate of Missouri includes influences from winter dominant continental polar air masses, and summer prevalent maritime and continental tropical air masses. Average annual temperature and precipitation are 12.8 °C and 1016 mm, respectively. A USGS streamflow gauge (contributing area 179.5 km2) has been operating from 1966-1981, 1986-1991, and 03/2007 to the present. SWAT model inputs were configured for the HCW using ArcSWAT with six subcatchments corresponding to the contributing areas of each of the five gauge sites and the watershed outlet. Using 2001 National Land Cover Database (NLCD) landuse and State Soil Geographic Database (STATSGO) soils data, the model was parameterized for 26 hydrologic response units. The model was forced with measured climate data from the publically available University of Missouri Sanborn Field meteorological station for the period from January 1, 2000 to December 21, 2008. Simulated discharge was compared to measured discharge at the USGS gauge from March 8, 2007 to December 21, 2008. The simulation yielded R2 values of 0.87 and 0.31 and Nash-Sutcliffe values of 0.81 and 0.29 for monthly and daily flow, respectively. Peak daily flow was 221.2 m3/s and 131.8 m3/s for measured and simulated discharge, respectively. Visual analysis of simulated hydrographs show consistent overestimation of baseflow and leading edge hydrograph discontinuities. Future work will compare current modeling results with model runs developed using data from the distributed hydroclimate station network. Model performance across the continuum of land use, from the forested to urban portion of HCW, will be quantified. This work will supply Central U.S. land managers with improved understanding and tools (i.e. models) for future landuse and development scenarios.