Catastrophe risk expert Karen Clark explains different types of floods and contrasts the challenges of modeling storm surge and inland flooding. Catastrophe modeling for storm surge associated with hurricanes is straightforward, although detailed data is required. By contrast, a lack of date and the absence of a predictable pattern for the occurrence of future inland flooding suggest that fully transparent, scenario-based approaches will provide more meaningful results for inland flood risk than opaque catastrophe models, she says.
One is the flooding associated with storm surge from hurricanes.
Hurricanes can cause the coastal waters to rise several meters leading to the inundation of low-lying areas along the coast. Storm surge is measured as the difference between the height of the storm tide and the astronomical tide, and is produced when the force of hurricane winds pushes water toward the shore
The most vivid recent example of flooding from storm surge was caused by Superstorm Sandy, which inundated many coastal neighborhoods in New Jersey and the boroughs of New York City. In 2005, Hurricane Katrina caused record storm surge heights of 28 feet near Pass Christian, Mississippi. Other notable surge-producing U.S. hurricanes are the Galveston Hurricane of 1900 and the Great New England Hurricane of 1938.
Another type is inland flooding caused by heavy rain. Tropical storms typically cause this type of flooding, and in some events the damage can be severe.
In 2001, Tropical Storm Allison dropped over 40 inches of rain in Texas over a period of a few days. Inundation in Houston caused severe damage to businesses, hospitals, homes, and other structures. More recently, Tropical Storm Irene (2011) caused extensive inland flooding in New England.
Inland flooding can also result from heavy rains over an extended period of time. The great Mississippi River flood lasted from June to August of 1993, damaging 50,000 homes and displacing over 70,000 people.
Most areas of the world are susceptible to this type of flooding and major events occur almost every year.
Flash flooding is caused by heavy or excessive rainfall in a very short
period of time, generally less than six hours. Flash floods often occur in normally dry areas that suddenly experience heavy precipitation, and they also occur in areas where the soil is already saturated. Flash floods can occur along streams and riverbeds, in mountain canyons, and urban areas.
Tsunami waves are generated when the ocean floor abruptly shifts and displaces vertically the overlying water. These waves can travel long distances at speeds approaching 500 mph. Large magnitude earthquakes that occur below or near an ocean, such as the recent Tohoku earthquake in Japan, are the most common causes of tsunamis.
As opposed to the storm surge from a hurricane, which will continuously crash on shore then roll out, a tsunami wave slows down as it approaches the coastline thereby allowing the water to pile up to heights of tens of meters and to come ashore with even greater force.
All ocean regions of the world can experience tsunamis, but destructive tsunamis are most frequent in the Pacific Ocean because of the nature of the tectonics (structure of the earth’s crust) and plate boundaries in this region.
Modeling flood damage
Modeling the damage from storm surge caused by hurricanes is relatively straightforward. The meteorological parameters driving the peak surge are generally the same as those causing high wind speeds in a hurricane. And the higher the wind speeds, the higher the potential surge heights. The location of peak surge is associated with the storm’s radius of maximum winds.
The storm direction relative to the coastline and the forward speed also influence the generation of storm surge. Storms traveling perpendicular to the coast will generate higher storm surge as will larger hurricanes.
Other important factors are coastal bathymetry (ocean depth), astronomical tide, and the presence of inlets, bays and rivers.
For sections of the coast where the continental shelf is relatively wide and shallow, the storm surge can build up much more as the hurricane moves ashore. In areas where it falls off more sharply, such as southeast Florida, storm surge is less of a threat.
In coastal areas where the astronomical tide heights vary significantly between low and high tide, the combined water level can be elevated if the storm occurs at high tide.
Finally, in areas where there are significant bays, rivers and inlets, such as Galveston, Mobile and Tampa, the storm surge potential is greater because as a volume of water is funneled into a narrower area, the water heights will increase.
Inundation occurs when the water heights exceed the land elevation. Detailed, high resolution elevation data is available for most parts of the world and can be used to identify areas and properties likely to be impacted by storm surge.
In summary, the following factors will influence the damage caused by storm surge from a hurricane or typhoon:
- Storm intensity and peak winds at landfall
- The size of the storm
- The angle of the track relative to the coastline
- The forward speed
- The coastal bathymetry
- Presence of inlets, bays or rivers
- Local topography
All of these factors can be credibly modeled, but very detailed data is required. The images below show the flooding that would be caused by a repeat of the 1938 Great New England Hurricane at high tide.
Estimated inundation area around Old Saybrook, Conn. from 1938 Great New England Hurricane occurring at high tide.
Estimated inundation area around New London, Conn. from 1938 Great New England Hurricane occurring at high tide.
Estimated inundation area around Mystic, Conn. from 1938 Great New England Hurricane occurring at high tide.
Estimated inundation area along eastern Connecticut coastline from 1938 Great New England Hurricane occurring at high tide.
The vendor models include storm surge for certain regions. For example, all of the major catastrophe modeling firms include storm surge for U.S. hurricane, but not all include it for the Caribbean and Pacific regions.
Modeling inland flooding is much more challenging for a number of reasons, and the catastrophe model vendors have recently begun to tackle this peril. Heavy rains can occur just about everywhere, but the exact location and amount of rainfall is hard to predict. It’s a further extrapolation to pinpoint where flooding will occur.
Even the definition of an “event” can be problematic when the flooding occurs from heavy rains over a number of days, weeks, or even months. For perils such as hurricanes and earthquakes, government organizations typically maintain databases of the important parameters for historical events. For inland floods, the record keeping consists primarily of tracking flood heights at different locations at different recording times. Flood maps, for example, provide the estimated return period floods for specific locations; the maps do not provide information on the aggregations of exposures likely to experience flooding from the same event.
While sophisticated hydrologic computer models are used by the National Weather Service and other organizations to forecast floods in real time, there’s no predictable pattern to how future inland flooding will occur. A catastrophe model must project future scenarios and assign credible loss probabilities, and it is the probabilistic aspect that is so difficult for this peril. While the modeling companies are attempting to develop probabilistic inland flooding models, insurance companies should not expect these models to give reliable “answers.”
Developing a highly credible exceedence probability (EP) curve from a probabilistic model for inland flooding is not realistic given the data limitations. While the science can indicate locations that are likely to flood and how the flooding will occur if a significant event happens, many less reliable assumptions are required to generate an EP curve, specifying for example, the 1-in-100 year probable maximum loss (PML). Opaque catastrophe models may not be the best tool for the inland flooding peril.
While sophisticated hydrologic computer models are used by the National Weather Service and other organizations to forecast floods in real time, there’s no predictable pattern to how future inland flooding will occur. A catastrophe model must project future scenarios and assign credible loss probabilities, and it is the probabilistic aspect that is so difficult for this peril.
A more robust method for assessing inland flood risk might be a fully transparent, scenario-based approach. This type of approach is being utilized by some of the large reinsurance brokers and other organizations.
In this method, hypothetical scenarios are developed for different geographical regions based on topography and past flood events. The flood levels for the scenario “footprints” are adjusted using detailed high resolution elevation data and then superimposed on the property exposures.
A more sophisticated version of this method “floats” the scenario events to provide a range of damage and loss estimates for each scenario. This approach can also be utilized for tsunamis—another peril for which credible probabilistic models do not exist.
In summary, there are several types of flooding that can cause significant property damage, and the availability and credibility of probabilistic models varies among the different types.
Flooding from storm surge is the most straightforward to model because the data required is available and consistent with the information used to generate hurricane wind intensities. Because catastrophe models for inland flooding are much less credible and are not likely to improve significantly in the near future due to the nature of this type of flooding and the data limitations, insurance companies would be well served by utilizing alternative approaches.