Storm Surge and Flooding

Current Understanding

Although the dominant characteristic of hurricanes is their strong winds, most fatalities and a major fraction of economic losses are caused by water. In a hurricane, coastal regions may experience flooding both from high river flows induced by the storm’s torrential rains and the storm surge propelled by the hurricane’s winds. The storm’s strong onshore winds produce a powerful stress that forces water onto the shore. A secondary effect is a rise in sea level induced by the storm’s low pressure. These effects add to the normal variation in tides and are further complicated by water level setup due to waves (most significant over relatively steep continental shelves) and runoff due to precipitation. Storm surge is determined primarily by meteorological forcing (storm intensity, path, and spatial and temporal scales) and topographic parameters (width and slope of the continental shelf, geometry and character of local coastal and shelf features such as barrier islands, headlands, bays, sounds, inlets, marshes, channels, levees, and barriers). In a low-lying region such as the Gulf Coast the storm surge can extend inland for miles. During Katrina, storm surges approaching 30 feet were experienced in some areas.

The basic physics of storm surge is relatively well understood, although for complex coastal areas the governing equations are not amenable to simple solutions. Storm surge estimates in the 1950s and 1960s were obtained assuming a steady force balance between the onshore wind stress, the water surface slope, and bottom friction. These calculations were intended for use along straight coastlines with only slightly varying bathymetry. They were the basis for much of the storm surge protection infrastructure designed until the 1970s and that in many areas is still in use today.

Beginning in the 1960s, the availability of high-speed computers permitted development of numerical approximations to the governing equations using structured computational grids (often composed of squares), first in two spatial dimensions (vertically integrated) and more recently in three dimensions. This capability eliminated the need to assume a steady state force balance (which may not be appropriate for the rapidly changing winds that occur in hurricanes), allowed inclusion of the effect of mass redistribution (the balance between the net amount of water flowing into an area and the change in water level), and the representation of more realistic shorelines and bathymetries. A primary example of models of this type developed in the 1980s is NOAA’s SLOSH (Sea, Lake and Overland Surges from Hurricanes) model that has been used to compute FEMA Flood Insurance Rate Maps. It still is used by the National Weather Service to forecast hurricane storm surge and is the basis for many emergency management decisions. Unfortunately, the SLOSH code is proprietary, so the lack of access to the code has tended to stifle progress, thus making the model inflexible to changing local conditions or new knowledge. Other countries, for example, Britain, Netherlands, and Denmark, also have developed national surge models that have overcome many of the deficiencies of the U.S. models.

Recently, advances have been made in a number of areas of direct relevance to storm surge modeling. Unstructured, finite element numerical techniques have been developed for storm surge models that allow the use of computational grids composed of triangles. These grids are readily configured to represent complex topographic or bathymetric features such as irregular coastlines, rivers, channels, inlets, barrier islands, causeways, levees, etc., and can therefore provide very high resolution in localized areas of interest. With accurate wind fields, these models have demonstrated accuracy of a few feet when run using data from past hurricanes. Combined storm surge and tidal models are now able to account fully for the effect of the astronomical tide on the total water level and therefore the potential flooding during a storm. Such models should be used to plan coastal protection systems. Finally, interactions between storm surge and wind wave models are being pursued to allow inclusion of the water surface setup due to gradients in the radiation stress associated with wave breaking, the influence of wave conditions on the air-water drag, and the influence of waves on bottom friction. Research is also under way to include the influence of different land covers on the air-water drag and bottom friction. The latter effects are particularly important in areas of inundation or coastal wetlands.

Near-Term Needs

Improve geographic, bathymetric, and geodetic data, and develop and apply high-resolution unstructured grid models for storm surge prediction. Accurate representation of local bathymetric and topographic features is critically important for computing storm surge elevations and flooding. These features reflect both natural processes and human actions and are complicated by subsidence and uncertainty in datums. Recently developed very high resolution geographic/geodetic/land use data products offer the potential for significant advances in this area, if they can be transferred effectively into the computation grids required by storm surge models. The representation of conveyances and obstructions to surges must be improved, particularly in developed areas. This suggests the continued development of unstructured grid models and tools to construct the required computational grids from the continuously growing catalogue of very high resolution data products.

Develop a well-defined reference equipotential surface that is needed to connect water level variations accurately to land. The vertical datum or knowledge of the geoid and its changes (accounting for natural processes and subsidence) are usually not adequately known. There are some areas along the Gulf Coast that are particularly problematic. Gravity surveys in addition to LIDAR/InSAR/GPS mapping should provide more accurate storm surge calculations.

Improve meteorological data needed for storm surge analysis and prediction. Accurate storm surge predictions necessarily begin with accurate meteorological fields. These are required in three different contexts: historical, to provide storm surge model validation for past events; probabilistic, to provide forcing fields for use in the design of storm surge protection systems and emergency planning; and forecast, to provide forcing fields for storm surge forecasts to be used for emergency response.

Enhance the historical database needed for research to advance prediction and as the basis for storm protection design. Our knowledge of the meteorological fields associated with historical tropical cyclones has steadily improved due to the development of techniques for interpolating increasingly available observational data from satellites, aircraft, buoys, and land stations. The NOAA Hurricane Research Division’s H*Wind product is composed of gridded wind fields at snapshots in time for hurricanes in the North Atlantic basin for at least the past 5 years. Storm protection system design has most often been based on idealized storms such as the “standard project hurricane” or the “maximum probable hurricane.” These are based on storm parameters such as the central pressure, radius to maximum winds, and maximum wind speed that are independently determined based on the historical probability distributions of corresponding storm parameters. Resulting storm parameters are used with a parametric vortex wind model to create a corresponding wind field that is then propagated on selected tracks. Nevertheless, any probabilistic representation of tropical cyclones is quite difficult to elicit due to the relatively small numbers of storms that have occurred since the beginning of observational data collection, the significant geographical variability associated with those storms, and the indication that the storm statistics (e.g., frequency, intensity, duration) may be changing as a function of global warming. As a result, considerable uncertainty exists in current probabilistic representations of tropical cyclone meteorological fields.

Improve the detailed representation of tropical cyclones in forecast models. Tropical cyclones are not well characterized in current forecast models. Predictions are limited mostly to basic storm parameters such as central pressure, eye location, and maximum wind speed. While considerable improvement has been achieved over the past decade in track forecast skill, errors in the forecast track can still lead to substantial errors in surge forecasts due to the localized nature of storm surge. The same improvement has not been seen in the forecast skill for other tropical cyclone parameters.

Improve the coupling of wave and storm surge models and the parameterization of land types, to increase the accuracy of storm surge predictions. Even with an accurate wind field, considerable uncertainty will still remain in the calculation of surface wind stress and frictional forces, particularly at high wind, in fetch-limited conditions, and in areas of varying land types and covers.

Incorporate into forecast models the interactions between storm surge and hydrologic runoff. Significant flooding can result from the combination of both coastal storm surge and hydrologic runoff from extreme precipitation (e.g., Hurricane Fran in 1996 and Floyd in 1999 in eastern North Carolina). While hydrological runoff often lags storm surge by hours to days, the processes may interact and these interactions should be modeled when appropriate. To date, this has been attempted only rarely.

Longer-Term Directions

Given the considerable uncertainty in tropical cyclone characteristics, probabilistic and forecast assessments of the associated storm surge are best done through ensemble modeling. While different ensembles can be constructed, the most obvious method is to make multiple model runs using the same initial conditions and surge model, but with different meteorological forcing to simulate errors in the tropical cyclone parameters or different surge model parameterizations to simulate uncertainty in these relationships. A consensus projection of the resulting storm surge would then need to be developed. This requires the development of techniques to define ensembles of tropical cyclone meteorological fields as they approach landfall to allow corresponding ensembles of storm surge model runs to be assembled.

Information currently available for assessing either the general susceptibility of an area to storm surge and flooding or the likelihood that an area will flood during a specific event is not being communicated to the public in the most effective way to affect human behavior. Better integration of the new generation of high-resolution geographic data products (to street and building scales), current 3-D computer animation technology, and storm surge model output could make such information more relevant, effective, and more readily available to the public.

High-resolution storm surge calculations, whether part of an ensemble forecasting system or part of a statistically based design process, potentially require the interfacing of multiple models (e.g., meteorological model, wave model, hydrologic model, storm surge model) and the execution of multiple model runs. These combinations will require access to significant computational resources and advanced model integration techniques.

While water levels and inundation are the primary concerns, storm surges may have additional impacts. For example, flood waters may entrain contaminants whose dispersal following the storm needs to be assessed to aid with mitigation efforts. Waves and surge can cause significant sediment erosion, transport, and deposition, causing failure of critical infrastructure or significant disturbances to ecologically significant landscapes (e.g., barrier islands, marshes, or wetlands). Storm surge also may introduce high-salinity water into ecologically sensitive areas that are not accustomed to such conditions. In each case, models will be called upon to help address issues that are far beyond the realm of traditional storm surge computations. These newer applications may require further development within the framework of current models, as well as incorporation of processes not included in traditional storm surge models, for example, three-dimensional, baroclinic, coupled wave/current/sediment bed change, etc. These needs represent active areas of research in coastal modeling.