That a flood forecast based solely on rainfall observed up to the forecast time carries with it an implicit prediction of no more rain, and that such a prediction is the worst possible in the course of a severe storm has been recognized by many. Yet a wider acceptance of the Quantitative Precipitation Forecast (QPF) as an input into a hydrologic model has been slow. Vis-a-vis the history of developments, recounted by Zevin [1994], the progress in the last quadrennium appears phenomenal. To wit, a number of NWS offices now routinely prepare QPFs for the purpose of river forecasting. New methodologies for producing QPFs are being developed. Techniques for preparing QPFs are being taught to hydrometeorologic forecasters as part of the NWS modernization program [ Stewart et al., 1993]. And case studies of recent floodings have documented the enormous utility of QPF-based river forecasts for flood warning decisions [ Hughes, 1993], reservoir control [ Eiben and Philips, 1993], and commercial navigation [ Eiben and Yess, 1993]. This review concentrates on methodological aspects of producing QPFs operationally for the purpose of forecasting floods on main-stem rivers.