THE DEVELOPMENT OF SWAT MODELLING SYSTEMS FOR LARGE CORN BELT RIVER BASINS PART 2: MODEL PERFORMANCE AND EVALUATION

Agricultural nonpoint source pollution is the main source of Nitrogen and Phosphorus in the intensely row-cropped Upper Mississippi River Basin (UMRB) and Ohio-Tennessee River Basin (OTRB) stream systems, and is considered the primary cause of the Northern Gulf of Mexico hypoxic zone according to the US Environmental Protection Agency. Thus, this region mostly located in the Midwestern USA, provides a natural location for research on how intensive agriculture for food, feed and biofuel production can coexist with a healthy environment. The use of a process-based hydrologic and water quality model is considered indispensable to address this objective at such a large spatial scale. An integrated modeling system has been constructed with the hydrologic Soil and Water Assessment Tool (SWAT) model (see Part 1 paper describing overall modeling system), capable of estimating river basin responses to alternative cropping and/or management strategies. To facilitate the identification of optimum locations for a cost-effective river basin management, this SWAT large-scale application incorporates a greatly refined subwatershed structure based on the 12-digit hydrologic units (subwatersheds), while the land phase of the hydrologic and nutrient cycles are calculated at thousands of Hydrologic Response Units, which represent unique combinations of land, soil and topographical features. However, given the very large scale and the need to ensure the reliability of flow and pollutant load predictions at various locations within such hydrologic systems, a model's calibration becomes a time-demanding and challenging task. The purpose of the current article is to demonstrate a semi-automatic calibration approach for large-scale and spatially detailed modeling studies, with the use of the Sequential Uncertainty Fitting algorithm (SUFI-2) and the SWAT-CUP interface developed exclusively for SWAT hydrologic projects. The calibration framework provides estimates of the uncertainty of predictions at various locations and can be finalized within a reasonable timeframe with a powerful personal computer (PC). This study presents the whole calibration approach providing guidance on similar efforts at the regional scale.