Prediction and Mitigation of Flash Floods
(Adopted by AMS Council on 14 February 2000)
Bull. Amer. Met.Soc., 81, 13381340
In spite of decades of effort by government and the private sector to improve observations and warnings, flash floods continue to be one of nature’s worst killers. For example, during an approximately 1-h period on 5 May 1995, 16 lives were lost and over $1 billion in property damage occurred as a result of a flash flood in Dallas, Texas. Similarly, on the evening of 28 July 1997, a flash flood in Fort Collins, Colorado, took five lives and caused over $100 million in property damage. Events such as these ensure that flash floods, along with lightning, remain the leading weather-related causes of deaths in the United States.
Flash floods are distinguished from other types of flooding by the short timescales over which flood-producing rainfall occurs (generally less than 6 h) and the small spatial scales (generally less than 1000 km2) of drainage basins in which flooding occurs. Most flash floods occur at night and, as would be expected, produce the most damage in urban areas. Aside from intense rainfall and small net storm motion, factors that contribute strongly to flash flooding are low permeability or saturated soils, impervious ground surfaces, and steep slopes. Failure of small to medium-sized dams, including debris dams, contributes significantly to the fatalities and damage associated with flash floods. A majority of flash-flood-related deaths occur in motor vehicles as people seek shelter and/or try to escape from rising waters.
Recreational activities and economic development continue to place pressures on floodplain utilization. As a result, the number of flood-prone communities grows larger and therefore the need for more timely and accurate warnings continues to increase. Although structural measures of protection (e.g., flood control reservoirs) can provide some protection from flooding from larger streams with longer flood lead times, it is too costly to use flood control structures on the large number of small streams flowing through populated areas. Moreover, it is difficult and costly to construct and then continuously monitor the safety of all structures that could mitigate flash flood disasters. Therefore, it is vitally important to reduce the population living in floodplains by providing disincentives to encroachment. This is one of the goals of the national flood insurance program of the Federal Emergency Management Agency (FEMA). Nevertheless, since too many people still dwell along small streams that can easily flood, community warning systems and self-help programs provide the only practical safeguard for many small flood plain communities.
2. Warning systems and new technologies
Essential components of flash flood warning systems include the following.
• preparedness programs that include local citizen involvement.
• Rainfall-observing systems.
• Electronic data communication systems.
• Diagnostic/predictive models.
• Model calibration procedures.
• Warning dissemination systems.
• Action plans for local civil authorities.
In fact, many of the accomplishments in forecasting and improved warnings for flash flooding have come from the application and utilization of a number of these program components. For example, the National Weather Service (NWS) lead time for flash flood forecasts has increased to over 50 minutes. Much of this improvement is directly tied to new technology and also training along with the introduction of hydrometeorologists in the Weather Forecast Offices. The deployment of the new weather surveillance radars nationwide has significantly improved our capability to continuously monitor intense localized rainfall.
Furthermore, upgrading the new radar network to include polarimetric capabilities may offer substantial potential for significant additional improvements in rainfall measurements. The new radar technology, when combined with satellite and advanced automated surface rainfall and streamflow technologies, enhances the prospects for detecting and quantifying intense rainfall and rapid rises in streamflow. Still further, the databases created by the new observing capabilities facilitate better studies of the physical character of such rainfall events. This is especially true for understanding flash flood dynamics and microphysical processes, knowledge that is vital for the development of improved radar rainfall estimates. For example, improved understanding of radar underestimation of "warm-process" rain will ultimately lead to further improvements in diagnostic and forecast models. Because both remotely sensed and on-site recorded data are necessary for the production of high-resolution (to a few square kilometers) optimal estimates of rainfall, deployment and maintenance of rain gauge networks will still be important components of hydrologic research and development. A major challenge for remote sensing systems is enhancement of rainfall estimation capabilities in complex terrain.
3. Integrated hydrometeorological approaches
The coupled meteorological–hydrological nature of flash floods is becoming more and more evident. Prediction of flooding events will require interactive meteorological and hydrological models that introduce the new weather radar data and include feedbacks from the near-surface soil water to the atmosphere. Coupled hydrological–meteorological models should be constructed in a manner that permits prediction of the time and space distribution of both the rainfall and the resultant flooding. Efforts to predict the temporal and spatial distribution of flooding are expected to dominate research and development of diagnostic/predictive models. Considerable efforts have also been expended to develop probabilistic approaches to heavy rainfall and flood forecasting. Early efforts during the late 1990s have shown dramatic improvements in warnings when forecasters are trained and fully engaged in understanding the strengths and drawbacks of utilizing probabilities to express their confidence in their rainfall forecasts. Additional work also must be done to improve automated stream gauging systems and to improve integrated rainfall–streamflow prediction systems that integrate models with observations.
Important research challenges include
• understanding the processes that govern the production of extreme rainfall rates in convective weather systems,
• determining why some convective systems become quasi-stationary,
• defining the effects of the initial spatial distribution of soil moisture on the development of surface runoff,
• developing reliable distributed hydrologic models for simulating the hydrologic response of urban areas,
• linking local distributed hydrologic models to the larger-scale hydrologic models used operationally by the river forecast centers of the NWS,
• establishing the fundamentals of small-catchment hydrology/hydraulics, and
• quantifying forecast uncertainty by providing probabilistic forecast guidance.
The influence of the spatial scales of soil moisture on flash flood modeling is central to supporting these basic research efforts. Data requirements for conducting the research and developing predictive models necessitates the utilization of geographic information systems for specifying catchment geometrical properties pertinent to surface runoff development.
Because the occurrence of flash floods depends strongly upon the local nature and dynamics of rainfall and upon the wide spectrum of spatial and temporal scales that such dynamics span, fundamental uncertainties are inherent in any attempt to forecast these events. These uncertainties make it necessary to use statistical–dynamical predictive rainfall models. They also encourage exploration of probabilistic approaches, including ensemble forecasting techniques.
4. Preparedness and self-help programs
Dissemination of flood warnings has improved in recent years, largely in response to the enhanced coverage and attention of the local commercial media. Public response to warning would improve if individual streams were identified in the warning messages. Also, enhanced National Oceanic and Atmospheric Administration Weather Radio coverage of flash flood-prone areas over the United States would undoubtedly contribute to smaller disaster impact. Dissemination methods involving the Internet should be explored and tested.
Proactive preparedness programs remain indispensable for loss-of-life and flood-damage reduction. Continued efforts by the NWS, FEMA, the media, and state and local emergency management agencies to educate the public regarding the occurrences and destructive force of flash floods are essential. Improved community monitoring, detection, and warning programs with emphasis on individual warning responses are a must. Real-time feedback from local designated persons or authorities to the National Weather Service Forecast Offices as to the hydrological and meteorological aspects of flood development will make for a more effective warning system.
Present-day flash floods are calamities with the potential for a very high death toll and huge losses of property. Although forecasting such events remains a tremendous challenge, the availability of the new weather radar data is expected to enhance forecast
[This statement is considered in force until September 2013 unless superseded by a new statement issued by the AMS Council before this date.]