The foundations of a scientific approach to flood warning systems had been laid down by numerous researchers during the 1970s and early 1980s. They culminated in a decision-theoretic methodology for modeling and evaluating forecast-response systems for floods on main-stem rivers [ Krzysztofowicz and Davis, 1983], a comprehensive set of design requirements for flash flood warning systems [ Georgakakos, 1986], and development plans for improving hydrologic services to the Nation [ NWS, 1982; Hudlow, 1988].
The revived interest in warning systems as a means of reducing negative consequences of floods and a viable alternative or an addition to traditional structural flood controls such as dams, levees and diversion channels, stimulated new research activities in the last quadrennium. This research had three prongs: (1) a multiobjective decision technique has been offered as a tool for planning and operating flood warning systems, (2) a Bayesian theory has been formulated for comprehensive studies of local flood warning systems, and (3) several applications of the Bayesian theory have been pursued.