We have developed multiple modeling approaches towards investigating aspects of change within the Minnesota River Basin. These include constructing fully distributed and parameterized models such as SWAT as well as developing reduced complexity models that capture the most critical variables and relationships. By taking multiple approaches in parallel we learn more about each individual model’s ability to capture the system response. This is especially critical when drivers and responses are not occurring at the same time or space scale. For example, 82% of the landscape is row-crop agriculture but most sediment is derived from less than 1% of the land area. Is it necessary to model the entire landscape? Similarly, most of the sediment is generated in a few storm events each year. Is it necessary to model the entire hydrograph?