Understanding how precipitation is partitioned into evapotranspiration and streamflow is important for assessing water availability. In the Budyko framework, this partitioning is quantified through the ω parameter. Previous studies have modeled the physical representation of ω; however, the spatial heterogeneity of the relationship between ω and the variables that it represents has not been investigated. This study uses a geographically weighted regression model to identify spatial variations in the factors that control the water balance in 126 reference watersheds with minimal human disturbance and 765 non-reference watersheds in the continental United States. Results show that snowfall and forest coverage are important predictors of ω in the reference watersheds. Relative cumulative moisture surplus, dam storage, and developed land in riparian areas are important predictors in non-reference watersheds. Climate is a primary control of the relative importance of forest coverage. The importance of forest coverage is greater in arid watersheds than in humid watersheds. Snowfall is more important than forest coverage in the Northeast and Midwest. This study demonstrates that dam construction and urban sprawl have a significant impact in non-reference watersheds. Dam storage is the most important predictor in 21% of the non-reference watersheds, and riparian developed land is more important in 13% of the non-reference watersheds. Overall, there are statistically significant relationships between climatic, physiographic, and human-related factors and the ω parameter. The spatial variations in the relationship quantified in this study can help to improve regional watershed management.