Local and remote sources of water are estimated by removing the surface source of water vapor in a numerical simulation without significantly influencing the large-scale dynamical circulation. This is done by reducing the surface moisture flux in the model's hydrologic equations but maintaining latent heat in the surface energy balance and evaporation in the soil water budget. Simulations of the 1993 midwestern flood are used as test cases. Without the local surface source of water vapor, the flood region's atmospheric hydrology budget reduces to an approximate balance between precipitation and moisture flux convergence. Comparison with the control simulation hydrology approximated 12% and 20% of the precipitation with a local source for June and July 1993 respectively. The difference between the two months are reasonable considering that June is dominated by a transient synoptic-scale wave pattern, but July precipitation is largely related to mesoscale convective complexes. The southern Great Plains source influences the flood region precipitation slightly less than the local anomaly for each month. The differences between June and July are striking. A great deal of variability exists in the modeled interactions between the surface and atmosphere. This indicates that simply using any one month to generalize the interactions or feedback of water does not address the complete problem. ¿ 1999 American Geophysical Union