The effect of climate change on quality and quantity of runoff from a northern, agricultural watershed was simulated using the Soil and Water Assessment Tool, 1996 Version (SWAT96). SWAT's snow evaporation submodel was modified. SWAT was calibrated using water quality and quantity data measured in the Cottonwood River near New Ulm, MN (drainage area: 3400 km2). The standard errors after calibration were 3.31 mm, 157 kg/d, 752 kg/d, 3744 kg/d, and 85 t/d for mean monthly streamfiow, P yield, ammonia (NH3)/organic N yield, nitrate (NO3) yield, and sediment yield, respectively. The standard error for monthly streamflow was 9.62 mm. SWAT96 was then used to simulate the effect on the Cottonwood River watershed of a 2xCO2 climate scenario, obtained from the Canadian Climate Center's global circulation model. Assuming land cover and land management remained constant, SWAT96 projected a decrease in mean annual streamflow, P yield, NH3/organic N yield, NO3/nitrite (NO2) yield, and sediment yield. Mean monthly values changed significantly for many months of the year under the 2xCO2 climate scenario. The standard errors in SWAT's baseline simulations, however, were too high for the simulated climate change effects to be measurable for NO3/NO2 and sediment yields. The model assumptions and calibration methods used to obtain the accuracy required for simulating the effects of climate change lead to the conclusions that land use/land cover and land management practices are likely to have a greater impact on water quality than climate change and that SWAT must be calibrated to be used for climate change analysis.