|SHARIFI, AMIR - University Of Maryland|
|LIANG, SUN - US Department Of Agriculture (USDA)|
Submitted to: American Geophysical Union
Publication Type: Abstract Only
Publication Acceptance Date: 9/1/2016
Publication Date: N/A
Technical Abstract: Process-based watershed models typically require a large number of parameters to describe complex hydrologic and biogeochemical processes in highly variable environments. Most of such parameters are not directly measured in field and require calibration, in most cases through matching modeled fluxes (such as river discharge) with field observations from gauging stations. Problems emerge when field measurements are not available to properly calibrate the model (e.g. ungauged basins), or are only available for a short period of time, thereby limiting model validation. In this study, we present a novel approach that uses another source of independent observations, i.e. remotely sensed evapotranspiration (ET) data, to calibrate the Soil and Water Assessment Tool (SWAT) for flow predictions at daily scale. The ET retrievals come from the disaggregated Atmosphere Land Exchange Inverse model (DisALEXI), which provides ET estimates daily at 30m resolution by fusing satellite information from several platforms. In this method, an efficient optimization algorithm (Dynamically Dimensioned Search Algorithm) is implemented to find an optimal combination of SWAT parameter values that leads to convergence between SWAT and DisALEXI ET estimates. The proposed method was applied to a 290 km2 watershed within the upper region of the Choptank River basin, on the Delmarva Peninsula. Results show that when SWAT was calibrated under the proposed method, daily flow predictions improved significantly in comparison to the uncalibrated model – particularly the baseflow component. This study has many implications for applications in ungauged/poorly gauged watersheds, and in watersheds under irrigation, specifically where irrigation information is not available. Irrigation information is usually provided to watershed models, and the proposed method can be applied to rectify the uncertainty associated with lack of irrigation information.