Woody plant encroachment is a world wide phenomenon with implications on the hydrologic cycle at the catchment scale that are not well understood. In this study, we use observations from two small semiarid watersheds in southern Arizona that have been encroached by the velvet mesquite tree and apply a distributed hydrologic model to explore runoff threshold processes experienced during the North American monsoon. The paired watersheds have similar soil and meteorological conditions, but vary considerably in terms of vegetation cover (mesquite, grass, bare soil) and their proportions with one basin having undergone mesquite removal in 1974. Long-term observations from the watersheds exhibit changes in runoff production over time, such that the watershed with more woody plants currently has less runoff for small rainfall events, more runoff for larger events, and a larger runoff ratio during the study periods (summers 2011 and 2012). To explain this observation, we first test the distributed model, parameterized with highresolution (1 m) terrain and vegetation distributions, against continuous data from an environmental sensor network, including an eddy covariance tower, soil moisture, and temperature profiles in different vegetation types, and runoff observations. We find good agreement between the model and observations for simultaneous water and energy states and fluxes over a range of measurement scales. We then identify that the areal fraction of grass (bare soil) cover determines the runoff response for small (large) rainfall events due to the dominant controls of antecedent wetness (hydraulic conductivity). These model-derived mechanisms explain how woody plants have differential effects on runoff in semiarid basins depending on precipitation event sizes.