This dissertation describes the modeling efforts on the Upper Mississippi River Basin (UMRB) using the Soil and Water Assessment Tool (SWAT) model. The UMRB extends from the source of the river at Lake Itasca in Minnesota to a point just north of Cairo, Illinois, and covers a drainage area over 490,000 km2. SWAT is a long term, continuous watershed scale hydrologic model. The main goal of this study is to apply the SWAT model to the UMRB and selected subwatersheds to evaluate the model as a tool for agricultural policy analysis and climate change impact analysis.

The SWAT model was first applied to the Maquoketa River Watershed, which covers approximately 5,000 km2 area in Northeast Iowa. A sensitivity analysis using influence coefficient method was conducted for eight selected hydrologic input parameters to identify the most to the least sensitive parameters. Calibration and validation of SWAT were performed for streamflow on annual and monthly basis. Model performance was evaluated by two statistical measures: the coefficient of determination (R2) and the Nash-Sutcliffe simulation efficiency (E). These values computed for the monthly comparisons were 0.86 and 0.85 for the calibration period and 0.69 and 0.61 for the validation period. The model was then validated for the entire UMRB streamflow at Grafton, IL and evaluated for a climate change impact analysis. Calibration and validation were preformed for 1968-87 and 1988-97, respectively; R2 and E values computed for the monthly comparisons were 0.74 and 0.65 for the calibration period and 0.81 and 0.75 for the validation period. The impacts of eight climate change scenarios (changes in temperature, precipitation, and/or CO2 levels) including a simplified replication of a previously reported future climate projection were then analyzed, relative to a baseline scenario. The results indicate that the UMRB hydrology is very sensitive to potential future climate changes.

The impact of future climate change was then explored for the streamflow by using two 10-year scenario periods (1990s and 2040s) generated by introducing a regional climate model (RegCM2) to dynamically downscale global model (HadCM2) results. The combined GCM-RCM-SWAT model system produced an increase in future scenario climate precipitation of 21% with a resulting 18% increase in snowfall, 51% increase in surface runoff, 43% increase in groundwater recharge and 50% increase in total water yield in the UMRB. Furthermore, evaluation of model-introduced uncertainties due to use of SWAT, GCM, and RCM models yielded the highest percentage bias (18%) for the GCM downscaling error. Change in stream flow (50%) due to climate change exceeds both the individual model biases and also the combined - model bias, thereby providing a relatively high confidence in the prediction.

Building upon the above SWAT validation for the entire UMRB with less detailed input data available in the Better Assessment Science Integrating Point and Nonpoint Sources (BASINS) package, a SWAT modeling framework was constructed for the entire UMRB. The framework incorporates more detailed input data and is designed to assess the effects of land use, climate, and soil conditions on streamflow and water quality. An application of SWAT is presented for the Iowa and Des Moines River watersheds within the modeling framework constructed for the UMRB. The next step of the research will focus on validation of the model for nitrogen and phosphorus, and simulation of the agricultural policy scenarios for the entire UMRB.