The U.S. Geological Survey's Distributed Routing Rainfall-Runoff Model [DR3M] was applied to a 2.27-square-mile portion of Bear Branch watershed at Murfreesboro, Tennessee, to demonstrate the application of this model to small urban watersheds in central Tennessee. Kinematic wave theory was used to route excess rainfall overland and through a branched system of stream channels. The model was calibrated with hyetographs from two raingages, hydrographs from two streamflow gages, and peak-stage elevations from two crest-stage gages that were operated in the watershed from March 1989 through July 1992. Standard errors of estimate for peak discharge at Northfield Boulevard and Compton Road are 41.1 and 92.2 percent, respectively. Standard errors of estimate for runoff volumes at Northfield Boulevard and Compton Road are 53.5 and 97.6 percent, respectively. The calibrated model was used to simulate flood hydrographs for 73 large storms occurring during the period 1901-1990 and the simulated flood peaks were used to develop flood-frequency relations for present (1992) conditions in the watershed. Flood discharges for the 100-year recurrence-interval storm were estimated as 350 cubic feet per second at Northfield Boulevard, 1,000 cubic feet per second upstream of DeJarnett Lane, 610 cubic feet per second downstream of DeJarnett Lane, 800 cubic feet per second upstream of Osborne Lane, 790 cubic feet per second downstream of Osborne Lane, and 1,000 cubic feet per second at Compton Road. The effect of detention storage on flood hydrographs was simulated at several locations in the watershed. Detention storage upstream of DeJarnett Lane significantly reduces downstream flood peaks, whereas detention storage upstream of Osborne Lane has almost no effect. The results of this study indicate that the Distributed Routing Rainfall-Runoff Model could be an important tool for testing the effects of future development and flood storage alternatives on flooding in small urban watersheds throughout the area.