Recommendation: Evaluation of flash flood warnings should be based on their contributions to improved decision making and should employ metrics that take account of the magnitude and scale of the events and the increasing specificity of the warnings. The NWS should improve the database of flash flood events and warnings to include more complete and accurate listings of both warnings and events.
The traditional national measures for evaluating flash flood warnings should be augmented to better take into account the myriad needs of the local population at risk. For example, the value attributed to the lead time of a warning should take into account factors such as the response time of a basin. In addition, warnings indicating the magnitude of the threat and designating specific regions within county warning areas are more useful to both the intended responders and others within the same county who are not affected by the warning. Evaluation statistics such as the probability of detection (POD) and the false alarm ratio (FAR) could be stratified by the scale of the events. The warning and event databases should be adapted to capture the increased specificity of the warning capability, in contrast to the current practice of recording warnings and events at the county level, with little indication of magnitude and impact.
Establishing a reliable database of events for verification purposes involves well-recognized difficulties, but recording the warnings issued is (or should be) a straightforward process. Nevertheless, situations involving a series of overlapping warnings, or extensions of warnings, and possibly multiple flash flood events within a county appear not to be well represented in the national database. This results, at least partly, from the rules employed in establishing that database; such rules may be required to ensure uniformity that permits comparisons across the nation. However, a more comprehensive listing would be useful for detailed analyses.