Constraining SWAT Calibration with Remotely Sensed Evapotranspiration Data

Historically, many watershed studies have been based on using the streamflow flux, typically from a single gauge at the basin\u27s outlet, to support calibration. In this setting, there is great potential for equifinality of parameters during the optimization process, especially for parameters that are not directly related to streamflow. Therefore, some of the optimal parameter values achieved during the autocalibration process may be physically unrealistic. In recent decades a vast array of data from land surface models and remote sensing platforms can help to constrain hydrologic fluxes such as evapotranspiration (ET). While the spatial resolution of these ancillary datasets varies, the continuous spatial coverage of these gridded datasets provides flux measurements across the entire basin, in stark contrast to point-based streamflow data. This study uses Global Land Evaporation: the Amsterdam Model data to constrain Soil and Water Assessment Tool parameter values associated with ET to a more physically realistic range. The study area is the Little Washita River Experimental Watershed, in southern Oklahoma. Traditional objective metrics such as the Nash-Sutcliffe coefficients record no performance improvement after application of this method. However, there is a dramatic increase in the number of days with receding flow where simulations match observed streamflow