Cannonsville Reservoir Watershed SWAT2000 model development, calibration and validation

Abstract

The Soil and Water Assessment Tool version 2000 (SWAT2000) watershed model was utilized to simulate the transport of flow, sediments and phosphorus to the Cannonsville Reservoir in Upstate, New York. The available datasets for model development, particularly the phosphorus input and water quality calibration data, in this case study are unique because of the large amount of watershed specific, spatially and temporally varying data that are available for model development. Relative to the default SWAT inputs, alternative model input generation methodologies were tested and shown to produce more representative inputs that generate substantially different simulation results. The successful application of SWAT2000 in this case study required two critical model modifications regarding excess soil water movement in frozen soils and soil erosion predictions under snow cover. The Nash–Suttcliffe coefficient of efficiency ( E NS ) for daily flows at the main flow station in the watershed was at least 0.80 in both the seven-year calibration period and the one year and four year validation periods. Average monthly total phosphorus loads were predicted within 15% of the corresponding measured data and the monthly E NS coefficients for total phosphorus were at least 0.63 in the calibration and validation periods. The results of this study are important for future SWAT modelling studies in gauged and ungauged watersheds, especially those in regions like the Northeast US that are subject to freezing temperatures in winter.