Comparative Analysis of HRU and Grid-Based SWAT Models

A grid-based version of the Soil and Water Assessment Tool (SWAT) model, SWATgrid, was developed to perform simulations on a regularized grid with a modified routing algorithm to allow interaction between grid cells. However, SWATgrid remains largely untested with little understanding of the impact of user-defined grid cell size. Moreover, increases in computation time effectively preclude direct calibration of SWATgrid. To gain insight into defining appropriate strategies for future development and application of SWATgrid, this research considers the simulated differences between commonly-employed hydrologic response unit (HRU)-based and grid-based SWAT models and the implications of resolution on their simulation and calibrated parameter values for a Midwestern, agricultural watershed. Results indicate that: (1) the gridded approach underpredicted simulated streamflow between 5% and 50% relative to the baseline model, depending upon the input spatial resolution and routing algorithm implemented; (2) gridded models generally underpredicted total phosphorous and sediment loads while overpredicting nitrate load; and (3) ranges of values of optimized model parameters remained similar up to 90 m. Results from this analysis should help in defining future applications of the SWATgrid model and the effects of differing spatial resolution of the model input data.