Floods remain one of the most frequent and devastating natural hazards worldwide. In the United States alone, there are 20,000 flood-prone communities; 3,000 of them receive site-specific flood forecasts from the National Weather Service (NWS), and 1,000 have local warning systems; the remaining communities receive county-wide warnings. Between 1965 and 1985, floods accounted for 63% of the federally declared disasters (337 out of 531), took 1,767 lives, and caused $5 billion worth of damage annually, on the average.
The Great Flood of 1993, documented by the National Oceanic and Atmospheric Administration [ NOAA, 1994], provided a vivid demonstration of Cromwell's rule: one should never assign the exceedance probability of zero to any prior observation. The duration (from March to November), the extent (over nine states) and the magnitude (exceeding previous floods of record at 95 forecast points) made this the most catastrophic flooding in modern U.S. history: 54,000 persons were evacuated, 50,000 homes were damaged, and economic losses of $15-20 billion have been estimated. The event also tested the limits of the Nation's forecast and warning services as flood stages were exceeded at about 500 forecast points.
While forecast and warning systems in existence at the time of past events significantly contributed to the reduction of losses, the potential for further prevention of avoidable losses through system improvements remains considerable. This potential was assessed a decade ago from both theoretical [ Krzysztofowicz and Davis, 1984] and operational [ NWS, 1982] perspectives. Recently, a comprehensive review of past efforts, in both research and operations, has reinforced the need for an integrated approach to hydro-meteorological services [ AMS, 1993; Zevin, 1994]. And the Great Flood of 1993 has resulted in an updated list of desirable improvements to forecast and warning systems [ NOAA, 1994]. These documents provide us with an opportunity to compare recent theoretical advances with future operational needs, to identify emerging problems, and to discuss research opportunities.