Small basin flash floods occur at higher spatial and temporal resolutions than the hydrologic simulation models typically run at River Forecast Centers (RFCs). The Flash Flood Guidance (FFG) method currently used does not explicitly account for this scale mismatch and cannot fully utilize the benefits of modern data collection systems like NEXRAD radar. The study described here has been initiated to test the hypothesis that a statistical-distributed modeling approach applied to predict flash floods in small basins will provide more appropriate guidance than the current FFG system. The statistical-distributed modeling approach involves calculating estimates of flood frequency for each cell in a distributed model using archived radar data. The same model is then run in forecast mode and the flood frequency estimates available for each pixel are used as an indicator of relative flood risk. Initial testing of this approach will use a priori rainfall-runoff and routing parameter estimates that have proven reasonable in other studies. The approach provides a framework for defining relative flood risk at ungaged stream locations in a way that accounts for model uncertainty. Preliminary results from a proof-of-concept study will be presented and future research and development plans will be discussed.