Skip to content.


Marylee Murphy, University of Minnesota, St. Croix Watershed Research Station
James E. Almendinger, St. Croix Watershed Research Station

The Soil and Water Assessment Tool (SWAT) is a spatially-referenced, daily-step watershed modeling program developed by the Agricultural Research Service. We are applying SWAT to assess the effects of land use and management changes in the Willow River watershed in western Wisconsin. This river drains 717 km2 of predominately agricultural land in St. Croix County and enters the St. Croix River at city of Hudson. Principal spatial data used to construct the model included 10-m digital elevation models, soil-survey geographic (SSURGO) data generalized by soil hydrologic group, and land-use satellite imagery updated to 1999 and 2006 conditions, the two years for which river monitoring data were available. Despite some significant problems with the model code, the model was calibrated to daily flows and monthly loads of suspended sediment and total phosphorus as measured during water year 1999. As a validation test, the model was updated to 2006 conditions and output was compared to monitoring data from that water year. The model fit the data well except for one large (> 75 mm d-1) rainstorm event.

The model required careful configuration to avoid improper parameterization during the calibration process. First, knowledge of closed-depression drainage areas was critical for proper model function. About 29% of the Willow River watershed drains to closed depressions which trap all sediment and virtually all phosphorus from runoff. If these areas been allowed to contribute to the main channel in the model, loads of these constituents would have been greatly overestimated, or the model would have been inappropriately parameterized to control these loads. Likewise, about 13% of the watershed area drains through large riparian wetlands which were presumed to remove virtually all sediment, though phosphorus was allowed to pass through. Second, residue production, decomposition rates, and implement mixing depths were carefully adjusted so that different tillage practices (conventional, mulch, and no-till) resulted in appropriate amounts of residue between the time of harvest in the fall and planting the next crop in the spring. Model defaults were inadequate for these processes. In addition, mulch till and no-till practices required increased infiltration (reduced curve numbers), increased surface roughness (Manning’s N), and increased bio-mixing parameterization for the model to mechanistically simulate the effects of reduced tillage on sediment and nutrient loads. Third, soil chemistry parameters were adjusted so that soils had appropriate labile and total phosphorus contents, which allowed phosphorus loads to correspond with sediment loads. Labile phosphorus was set to 41 ppm, which resulted in a total soil phosphorus content of about 500 ppm. Because coarse grains are preferentially trapped during overland transport, this concentration was enriched to about 1000 ppm by the time the sediment reached the stream channel. Fourth, rural residential lands were modeled as grassland, but with soil phosphorus content initialized as agricultural land and with infiltration reduced by one soil hydrologic group.

The model is being used to simulate the effect of selected agricultural management practices, including the conversion to conservation tillage (mulch or no-till), conversion of all cropland to a simple corn-soybean rotation at the expense of alfalfa (dairy) acreage, reduction of soil labile phosphorus to optimal levels for corn (about 20 ppm), reduction of dietary phosphorus in dairy feed, and conversion of cropland to rural residential. Results from model runs of these scenarios will be discussed as available.