Suitability of the Natural Resources Conservation Service (NRCS) curve number (CN) model of run-off prediction was evaluated on three humid tropical forested catchments in Kimakia, Kenya. The catchments were dominated by Pinus patula (catchment A), Arundinaria alpina (catchment C) and Pennisetum clandestinum (catchment M). The study used discharge and rainfall data collected between 1958 and 1986. Seventy-three (73) isolated storms were graphically separated into baseflow, interflow and surface run-off. Forest cover types significantly improved catchments characteristics that influence baseflow and interflow generation in catchment C but not those that influence surface run-off production. In its original form, the NRCS CN model resulted in direct run-off estimates that deviated from observed ones by between 43.8% and 55.3%. These discrepancies were minimized through modification of the β and CN parameters. CN generated empirically using storm rainfall predicted the direct run-off satisfactorily. Therefore, the modified NRCS CN model adequately estimates direct run-off from humid tropical forested catchments.