Hydrological models are used to assess many water resource problems, from agricultural use and water quality to engineering issues. This study shows how the combination of rainfall sources, rain gauges (accurate at a point) and NEXRAD (cover a wide area), can provide accurate, timely rainfall amounts to drive the Soil and Water Assessment Tool (SWAT) model, which provides a more reliable tool for producers and policy makers alike. Treatment of rainfall data was tested at South Fork of the Iowa River, which covers about 780 square-kilometers (193,000 ac) and is one of 15 benchmark watersheds of the USDA Conservation Effects Assessment Project (CEAP). Results show significant improvement in model predictions when correctly modifying the rainfall data. Gauge records that were given a value for trace events performed better than when trace events were given zero value even though the annual rainfall only differed by 25mm yearly. The combination of rain gauge data and NEXRAD data provided superior results even when the watershed has good spatial distribution of rain gauges. Gauge and NEXRAD data adjusted with inverse distance weighting or kriging with external drift predicted stream flow at four gauges the best among tested. In absence of rain gauge data, simple adjustment of the radar estimates to match total rainfall amounts in the region resulted in adequate stream flow estimates. Furthermore, uncorrected NEXRAD precipitation, though only 15% below annual precipitation, resulted in over a 50% difference in stream flow. The choice of rainfall treatment should be considered carefully to provide accurate model predictions.