Modeling Water Quality in the Sandusky Watershed, Ohio:

The objective of the research was to model water quality using the Soil and Water Assessment Tool (SWAT) interfaced with ArcView GIS. SWAT requires the integration of a range of GIS data layers; digital elevation model, soil types, land use, stream networks, water quality monitoring sites, point pollution sources and climatic observations. Prior to modeling water quality with SWAT, ArcView was used to delineate the subbasin boundaries and create soil-land use complexes. The interface of SWAT with ArcView enables the seamless integration of model input pre-processing, simulation of water quality and visualization of the spatiallydistributed results. The sensitivity of SWAT outputs to threshold area values for subbasin delineation was examined. Introduction, Rationale & Goalss SWAT (and many other environmental models) is a complex water quality model relying on numerous parameters to represent hydrological, climatic, water quality, plant and soil processes within a watershed. This poses problems when attempting to calibrate the model to a specific study area due to the number of parameters and the possible correlations between each other (Vandenberghe et al., 2002). As a result many researchers have proposed that a sensitivity analysis should be performed before model calibration to identify the sensitive parameters. For example, Vandenberghe et al. (2001) presented a calibration approach for ESWAT, a modified version of SWAT, which involved two steps, first, sensitivity analysis was performed, secondly, an automatic calibration method was used with the parameters that were identified as having a significant impact on the model output. To date, most of the sensitivity analysis studies for SWAT have focused on parameter uncertainty, e.g. curve number (CN), snowmelt parameters etc. This is performed after a level of spatial discretization has been selected. Fewer studies have considered the impact that different levels of spatial resolution have on the simulated outputs. Spatial sensitivity studies have focused on the input digital elevation model (DEM) resolution (Cotter et al., 2002) or the number of subbasins used to subdivide the study area (Bingner et al., 1996). The number of subbasins determines the number of different land use-soil combinations that are represented in the model. This in turn has impacts on the model output. Therefore, the goals of this study was to examine the spatial sensitivity of the SWAT model.