Watershed Configuration and Simulation of Landscape Processes with the SWAT Model

Recent and future river basin management requires a more spatially distributed description of basin hydrology and nutrient transport processes to enable land use management as a process controlling factor to realize sound river basin management. The spatial description of these processes in the Soil and Water Assessment Tool (SWAT) watershed model is presently realized by aggregating the flows from overlaid soil and land use patches in subbasins with averaged slope angles. Many concepts with different degrees of complexity have been developed in river basin modelling to aggregate units with similar hydrologic behavior (Hydrological Response Units). Watershed configuration for SWAT currently consists of: 1) subbasins defined by surface topography and 2) hydrologic response units in each subbasin to account for heterogeneity in soils and land use. The hydrologic response units do not account for landscape position within the subbasin. Until recently, many existing watershed models did not implicitly account for landscape processes within a subbasin. Other smaller scale models do account for hillslope transfer (e.g. WEPP, REMM, APEX, HYDRUS-2D). In an attempt to account for landscape position and processes, SWAT was modified to simulate landscape units within subbasins. Surface, lateral vadose zone, and groundwater flows are routed between landscape units (while allowing for hydrologic response units within each landscape unit). Surface runoff can be overland or channelized when routed from one landscape unit to the next. The model is being tested on the USDA-ARS experimental Y-watershed at Riesel, Texas, USA, using soil moisture and groundwater data. Using GIS techniques, the watershed was divided into three landscape units - valley bottom, hillslope, and upland. Further development will include landscape unit routing of sediment and nutrients and stream interaction with the valley bottom (i.e.; riparian/flood plain landscape unit). Simulated daily stream flow at the watershed outlet after routing across the landscape units, compared well to measured flow (R-2 = 0.7). Mean annual lateral flows across landscape units were also realistically simulated. Soil moisture (upper 1 m) was compared to measured soil moisture at one monitoring site in each landscape unit with the model predicting drying early in the summer but following general wetting/drying cycles. The revised version of the model is also tested using data collected from a low-gradient watershed near Tifton, Georgia, USA which contains heavily vegetated riparian buffers. The modified model provided reasonable simulations of surface and subsurface flow across the landscape positions without calibration. The application demonstrates the applicability of the model to simulate filtering of surface runoff, enhanced infiltration, and water quality buffering typically associated with riparian buffer systems. Future validation will include comparison with: 1) the Riparian Ecosystem Management Model (REMM) and riparian data sets; 2) with data from larger basins with defined floodplains; and 3) watersheds having well defined variable source contributing areas. The concept assumes the controlling factors for hydrological processes and functions must be adequately described at different spatio-temporal scales to accurately delineate such response units. This requires a sound description of the characteristics by using physically based parameters and indicators, but also simplified solutions at larger scales. Presentation of the new model concept and first results of testing simulations of different aspects of catchment-related control of landscape processes, pattern hydrology, and spatially distributed modelling are discussed.