Afforestation as a climate change mitigation option has been the subject of intense debate and study over the last few decades, particularly in the tropics where agricultural activity is expanding. However, the impact of such landcover changes on the surface energy budget, temperature, and precipitation remains unclear as feedbacks between various components are difficult to resolve and interpret. Contributing to this scientific debate, regional climate models of varying complexity can be used to test how regional climate reacts to afforestation. In this study, the focus is on the gauged Nzoia basin (12,700 km2) located in a heavily farmed region of tropical Africa. A reanalysis product is dynamically downscaled with a coupled atmospheric-hydrological model (WRF-Hydro) to finely resolve the land-atmosphere system in the Nzoia region. To overcome the problem of Nzoia river flooding over its banks we enhance WRF-Hydro with an overbank flow routing option, which improves the representation of daily discharge based on the Nash-Sutcliffe efficiency and Kling-Gupta efficiency (from −2.69 to 0.30, and −0.36 to 0.63, respectively). Changing grassland and cropland areas to savannas, woody savannas, and evergreen broadleaf forest in three synthetic numerical experiments allows the assessment of potential regional climate impacts of three afforestation strategies. In all three cases, the afforestation-induced decrease in soil evaporation is larger than the afforestation-induced increase in plant transpiration, thus increasing sensible heat flux and triggering a localized negative feedback process leading to more precipitation and more runoff. This effect is more pronounced with the woody savannas experiment, with 7% less evapotranspiration, but 13% more precipitation, 8% more surface runoff, and 12% more underground runoff predicted in the Nzoia basin. This study demonstrates a potentially large impact of afforestation on regional water resources, which should be investigated in more detail for policy making purposes.