Runoff from natural and managed landscapes has been modeled using the rainfall excess over infiltration. The Precision Agricultural-Landscape Modeling System (PALMS) uses this method by first estimating infiltration on each 5-m by 5-m grid cell in a field and then routing the excess (runoff) over the landscape, which is composed of many grid cells. Infiltration and routing of runoff are calculated simultaneously on each time step, which varies from a few seconds to minutes depending on various landscape characteristics. In the past mean values of soil hydraulic properties were used to simulate infiltration over a field. In this study we imposed a range of hydraulic conductivity values over a 0.5 ha field by using a Monte Carlo method for generating a log normal distribution. Thus each 5-m by 5-m grid cell was assigned a different hydraulic conductivity that was chosen from the log-normal distribution randomly. Several simulation runs were conducted with the same log-normal distribution but with different randomizations of hydraulic conductivity values over the landscape. Using a single averaged hydraulic conductivity over the landscape causes predicted runoff to occur only for the largest rainfall events, and no runoff for small or moderate events. Simulations that use a distribution of hydraulic-conductivity values result in runoff from a wide range of rainfall events and agree much better with measurements than model runs with averaged hydraulic conductivity. No scheme for averaging hydraulic conductivity values appears to accommodate the interaction between runoff and infiltration that occurs on heterogeneous landscapes.