Use of the SWAT Model to Quantify Water Quality Effects of Agricultural BMPs at the Farm-Scale Level

摘要

Best management practices (BMPs) have been implemented on a farm-by-farm basis within the Cannonsville Reservoir watershed (CRW) as part of a New York City watershed-wide BMP implementation effort to reduce phosphorus (P) loads to the water supply reservoirs. Monitoring studies have been conducted at selected locations and at the watershed outlet on one of the farms, which spans an entire subwatershed within the CRW, with the aim of quantifying effectiveness of the BMPs installed on the farm. This study applied the Soil and Water Assessment Tool (SWAT) and a recently developed BMP characterization tool to the farm over pre- and post-BMP installation periods with the object of determining the extent to which model results incorporating all installed BMPs match observed data, and the individual impact of each of the BMPs installed on the farm. The SWAT model generally performed well at the watershed level for flow, sediment, and phosphorus simulations. Annual Nash-Sutcliffe (NS) coefficients for the components ranged between 0.56 and 0.80, while monthly NS coefficients ranged between 0.45 and 0.78. The model also performed well at the field level, with simulated in-field P loads closely matching observed data. Because the fields had various combinations of BMPs installed on them, it was difficult to separate out individual BMP impacts based on SWAT simulations. It was, however, possible to determine the effects of BMP combinations such as nutrient management plans and rotations (31% dissolved P; 25% total P). For dissolved P, integration of BMP tool efficiencies allowed individual BMP impacts to be incorporated while still maintaining the same level of representation as was obtained using model simulations. As the SWAT model is often used with little or no post-BMP data to verify simulation results, this study served to validate SWAT model suitability for evaluating BMP impacts. The BMP tool was found to suitably complement the model by providing insights into individual BMP impacts, and providing BMP efficiency data where the model was lacking.