Untitled Document

The Increasing Costs of U.S. Natural Disasters
The Princeton University Geoscience 499 Class

Forecasting hurricane damages
Costliest hurricanes Print exclusive

The U.S. government, as the insurer of last resort, is becoming increasingly vulnerable to the costs of natural disasters through disaster declarations and spending by the Federal Emergency Management Agency (FEMA). The number of presidential disaster declarations has generally increased over the last half century, since the federal government has assumed continuous responsibility for disaster aid. The federal government’s costs for natural disasters are increasing both in terms of the federal budget and the gross domestic product (GDP).

In July 1993, flooding on the Mississippi River wreaked havoc in Missouri. A total of 534 counties in nine states were declared federal disaster areas, and 168,340 people registered for federal assistance. Federal disaster aid has been on the rise over the past 50 years for hazards, including floods, hurricanes and earthquakes, as disaster response and mitigation efforts inadvertently have encouraged development in high-risk areas. Image by Andrea Booher/FEMA Photo.

Even when accounting for the exponential rise in GDP over the last four decades, costs of natural disasters as a percentage of GDP have more than tripled. This figure does not include the recent costs from Hurricane Katrina, which will most likely be the most expensive disaster in U.S. history, and has raised fundamental questions about high-risk land use.

The costs of natural disasters are driven by relatively few events — fewer than 1 percent of disaster declarations are responsible for the majority of costs. To reduce our nation’s exposure to natural disasters, we need to determine what factors cause these few events to be so expensive.

In general, the increase in cost correlates strongly with the large increases in population and wealth in disaster-prone areas — in particular, East Coast regions vulnerable to hurricanes and West Coast regions vulnerable to earthquakes. By directly comparing disaster costs with local infrastructure costs, event size and event frequency, however, more subtle relationships emerge. Earthquakes, hurricanes, tornadoes and floods vary in frequency and impact, but all have the capability of inflicting great damage and incurring high costs to the federal government. U.S. hazard mitigation efforts and disaster relief policies may inadvertently be contributing to these increased costs by making us more vulnerable to expensive low-probability, high-cost events. The political and social forces that support these counterproductive policies will require a national change in how we perceive these disasters — a change that hurricanes Katrina and Rita may help initiate.


In the last century, more than 170 hurricanes have hit the United States. Each year in the Atlantic, approximately six hurricanes form, with one or two making landfall on the United States. Hurricane frequency, strength and location are affected by wind shear and sea-surface temperatures, both of which are part of the greater El Niño-Southern Oscillation and multi-decadal cycles and patterns.

Despite this year’s highly active Atlantic hurricane season, there does not appear to have been a significant increase in either the occurrence or the severity of hurricanes over the last century; El Niño and La Niña provide an explanation for which years have more or fewer. During an El Niño phase, the sea-surface temperature anomaly in the Pacific Ocean is high, and the number of storms in the Atlantic is low (in 1983 for example). During La Niña (the opposite of El Niño), the anomaly is low and the number of storms is high (in 1988 for example).

Hurricanes are complex events, having a diverse set of factors that drives damages in the areas they strike. Wind speeds, intense rainfall, coastal storm surges, unpredictable paths and varying travel speeds are among the characteristics that, together, define each hurricane event. In general, the physical and temporal uniqueness of each hurricane event mean that a large portion of the cost is the result of coincidental damages. Katrina, for example, would not have generated so much flooding if it had not moved so slowly over the coastal areas it affected.

The quantifiable storm characteristic that is most strongly correlated with FEMA spending is population density in the county where landfall occurred for large events, of Category 3 or higher on the Safir-Simpson scale.

Hurricanes typically strike coastal areas with strength, but weaken quickly after making landfall. Coastal areas, already at risk of storm-surge flooding, are therefore subject to more forceful winds. Increases in sea level, the inevitable landward migration of East Coast barrier islands and the continued population shift to coastal areas on the East Coast — which includes building high-value beachfront property — result in higher damage costs from hurricanes that make landfall, though the number of storms has not been increasing.

The more typical and less severe Category-1 or 2 events, which are not hugely different from a harsh rainstorm, show little or no correlation between FEMA spending and population density at landfall. The more forceful (and rare) events, as with earthquakes above magnitude 6.0, show a strong correlation between FEMA spending and population density.


Between 1989 and 2004, more than 180 earthquakes of magnitude 5 or greater have struck in the continental United States. U.S. presidents have declared 11 of these events disasters.

As with hurricanes, earthquake costs are determined by the population impacted. The relationship between cost and population becomes apparent by comparing earthquake size and population density between 1989 and 2004. When FEMA awarded money to an area with a moderate or high population density (200 to 1,800 people per square mile) affected by an earthquake of a magnitude greater than 6.5, the funding increased with the population density irrespective of the magnitude of the event (for example, the Northridge, Loma Prieta and Olympia quakes). When FEMA awarded money to an earthquake-affected area with low population density (20 to 200 people per square mile), the funding scaled with the magnitude of the event (for example, the Landers, Petrolia, San Simeon, Clackamas, Napa and Plattsburgh quakes). When FEMA awarded money to an earthquake-affected area with a very low population density (between 0 and 20 people per square mile), the funding was small — approximately $5 million — irrespective of the magnitude of the event (such as with the Denali and Klamath quakes).


Between 1989 and 2004, 155 disaster declarations have included the word “tornado” in their description. The size of tornadoes is described by the Fujita scale F0 to F5 (similar to the magnitude of an earthquake or the category of a hurricane). A dramatic increase in reported F0 events occurred in the late 1980s, due to the introduction of Doppler 88-D surveillance radar (see story, this issue); more events were detected and therefore more reported. After removing the F0 events, the frequency of tornadoes over the last 20 years shows no increase.

Most tornadoes occur in the central United States, where population densities are low (between 20 and 200 people per square mile) and relatively uniform. Tornadoes are also short-lived and affect only small areas. For a tornado to incur a high cost, it must occur along with significant flooding, or be a unique event in terms of what it destroys. On March 14, 1997, for example, an F5 tornado destroyed an airplane hangar in Kentucky containing expensive aircraft, and in another case, on May 4, 1999, an F5 event included a cluster of 94 tornadoes occurring over two days in the Oklahoma City region.

In contrast to earthquakes and hurricanes, which are low-frequency, high-cost events, tornadoes exemplify high-frequency, low-cost events. Tornado events are similar in cost — typically less than $25 million — because they occur frequently in areas of moderate to low population density.


Because of the complex nature of rivers and their drainage basins, defining one standard that successfully measures a flood is difficult. No scale exists for floods that is similar to the magnitude scale for earthquakes, the Fujita scale for tornadoes or the category system for hurricanes. Recurrence intervals are used to measure the size of a flood, but the measurements are unique to a drainage basin and do not allow comparisons among different drainage basins.

More than 85 percent of U.S. counties have been declared federal disaster areas due to floods in the past 50 years. With increasing development in floodplains, less soil is available to soak up water, and flooding occurs more easily than ever. Floods are such high-frequency, ubiquitous events that much has already been done to ameliorate effects of common flooding disasters. On a local scale, the cost-effectiveness of further flood control measures is questionable, especially in areas not normally prone to flooding. On a national scale, however, floods constitute a high portion of FEMA’s disaster obligations.

The 20 most expensive floods in the United States show that costs roughly depend on population density. However, five of these 20 floods do not follow this trend. These outlying events are either extremely large, such as the Mississippi River flood in 1993, or occurred in areas unprepared for a flood, such as Detroit, Mich., in 2000. As with tornadoes, the relationship suggests that the driving factor of the cost of expensive floods is the unusual nature of the event with regard to its magnitude or its location.


For extreme events, disasters can also be a test of governments, as the world has witnessed with Hurricane Katrina. Disasters can bond a community and provide opportunities for politicians to demonstrate leadership. In fact, political support for incumbent politicians commonly increases following a disaster. Conversely, if a government fails to respond properly, disasters can also foment political unrest. On a small scale, they can change public opinion of a leader, as illustrated by polls that showed lower approval ratings of President Bush following Katrina. And on the larger scale, they can even result in the overthrow of governments, as in the case of the 1972 Managua earthquake, when the vast destruction contributed to the unrest that eventually led the Nicaraguan people to oust General Somoza.

This seawall was built to protect homes from migrating coastlines along the New Jersey shore. Such mitigation practices against moderate hazards (supported with federal funds) can result in making regions more vulnerable to large hazards, as it not only leads to the loss of the beach, but also encourages the development of expensive infrastructure in high-risk areas. Image courtesy of Gregory E. van der Vink.

In the United States, usually the governor of the affected state makes a formal request for FEMA assistance. Because a presidential disaster declaration is required before FEMA can provide an area with federal disaster relief funding, political factors can be introduced into the process. Based on records from 1952 to 2002, whether the U.S. president and the state governor shared a party affiliation had no significant effect on whether a disaster declaration would be approved or turned down. Furthermore, Republican and Demo-cratic administrations have similar approval and denial rates.

Although party politics do not seem to play a major role in disaster funding, almost every election year shows a small spike in the number of disaster relief requests approved — particularly when the incumbent is running for reelection. Meeting symbolically with disaster victims and approving requests for disaster funding are attractive opportunities for a candidate to improve his or her public image. The minor year-to-year political fluctuations are small, however, when viewed in the context of the overall increase in disaster declarations with time.

Opportunity for change

The timescale of human experience is short compared to the recurrence interval of many natural phenomena. While we develop infrastructure resilient to common events, such as routine seasonal weather, we remain vulnerable to those events that occur less frequently or over longer timescales. For example, a 6-inch snowfall in Boston, where such storms occur annually, has much less impact than a 6-inch snowfall in Washington, D.C., where such storms occur only once a decade.

When considering events that garner a FEMA disaster declaration, two cost-frequency trends emerge. Earthquake and hurricane disasters are generally high-cost, low-frequency events, whereas tornado and flood disasters are low-cost, high-frequency events. The common factor for high cost is the extent to which the disaster is unusual — either in terms of its recurrence interval or its size. In other words, rare events such as earthquakes and hurricanes just need to happen in populated areas to be costly. More frequent events, such as tornadoes and floods, need to be unusually large or to occur in areas where they usually do not occur.

The government of Japan maintains this house, destroyed by a volcanic eruption, as a reminder to improve awareness of such low-probability high-impact events. Image courtesy of Gregory E. van der Vink.

In the United States, the increase in costs to the federal government is most likely an unforeseen consequence of our own disaster management policies. Mitigation strategies require public support; public support requires awareness; and awareness usually requires the occurrence of an event. As a result, resources are almost always available to respond to the last event, but rarely to mitigate against the next. We tend to view natural disasters as random unfortunate acts, rather than the predictable consequence of high-risk land use. As a result, we rebuild communities in the same high-risk areas — thus inadvertently using taxpayer dollars to put more people in harm’s way.

In many cases, the influx of federal assistance and rebuilding can actually boost the local economy — resulting in more infrastructure and increased population in these high-risk areas. Such efforts may, themselves, then become responsible for increased costs. Whether it is cost-effective for communities to maintain a state of readiness for low-probability events, such as a “storm of the century” or even a “storm of the decade,” is questionable.

Mitigation against moderate events can make us more vulnerable to large events. For example, a levee may be enlarged to handle a 100-year storm rather than a 50-year storm. If the net result of the rebuilding is four times as many people move into the area, the risk ends up increasing. Paradoxically, our mitigation efforts, like our response efforts, subsidize such high-risk land use as living on migrating barrier islands, in floodplains or on active fault zones.

Consequently, we are becoming more vulnerable to low-probability events. For any specific individual city, this short-term solution is politically and, with federal disaster relief, economically attractive. When such an approach is adopted across a nation, however, the result increases already large costs to the federal government.

Population trends, mitigation efforts and federal disaster relief policies all contribute to encouraging high-risk land use and ultimately to making our society more vulnerable to the costs of natural disasters. Hurricane Katrina has created an opportunity to change this trend at the national scale. If New Orleans is built to accommodate the inevitable next extreme hurricane, it will set an example for future land-use management and urban planning. Absent such change, the costs of natural disasters, and the government’s liability, will simply continue to increase as we place more people and infrastructure in harm’s way.


Forecasting hurricane damages

On Sept. 20, five tropical cyclones — three typhoons in the Pacific and two hurricanes in the Atlantic — were brewing. The three typhoons were way out in the Pacific, not immediately threatening any land, and Hurricane Philippe looked like it was staying out in the Atlantic. Hurricane Rita, however, was posing an imminent threat to land. It was strengthening over the Bahamas and beginning to affect the Florida Keys and southern Florida. Rita was headed west-northwest, aiming directly at storm-weary Florida and the Gulf Coast; New Orleans was evacuated again.

In the meantime, scientists at the Kinetic Analysis Corporation facility in Savannah, Ga., were busy calculating an important facet of any storm: just what damages could be expected as Rita and the others continued along their tracks. Before Rita even became a hurricane, these researchers predicted damages around $130.8 million in Louisiana, $407.52 million in Florida and $15.95 billion in Texas. Worse perhaps for the embattled U.S. energy sector were predictions that 40 percent of Gulf Coast oil rigs and platforms (and 55 percent of natural gas infrastructure) could be affected. The storm’s impending hit sent oil prices up $4 per barrel in one day.

Over the next four days, as the storm stretched to Category 5 and then receded to Category 3 by the time it reached Texas and Louisiana shores, using computer models, the forecasters adjusted their predicted damage assessments to match Rita’s projected track. The day before the storm blew ashore, they had downgraded damage estimates to $1.95 billion in total, but their estimates of affected oil and gas infrastructure had increased. The researchers’ goal was to alleviate loss of life or serious harm by forecasting what damages would likely be seen in a particular area — affecting everything from energy infrastructure to emergency responses — says Chuck Watson, president and director of research and development at Kinetic.

Over the past decade, Watson and colleague Mark Johnson of the University of Central Florida in Orlando have been developing computer models that calculate projected damages to structures in the path of a storm based on wind speeds, sea-surface temperatures, pressure systems, humidity levels and the severity of waves. Every six hours or less, the models, which run on two supercomputers, update the damage forecast based on new atmospheric data from weather stations around the world, Watson says. The second major component of the damage forecast is the nature of the structures in the storm’s path.

“Just what effect a storm will have on an area depends on what’s in the way,” Watson says. Wind speeds will dictate whether or not a roof will blow off a house, but whether a tree falls down or just loses branches is dependent on how healthy the tree is and what kind of storm hits. For example, for a weak storm, healthy trees with lots of moisture will be more resistant, but for a strong storm, it is much better to have weak, dry trees that will just break, Watson says, as strong trees will cause more damage in a strong storm. Wetlands are another example of a component in the model. Wetlands can help block storm surges and protect coastal property, so where there are no wetlands, waves can cause more damage (see story, this issue).

Johnson and Watson start tracking storms as soon as they form at sea. And while their program, the results of which are available on a public Web site, provides damage assessments for up to five days out, Johnson says the most accurate forecast will be within 48 hours of a storm’s projected hit on land; he suggests checking back frequently. Logging onto the site will suggest what damages people may expect to see in their neighborhoods, Johnson says, and may help them know what steps to take to avoid catastrophic loss.

Johnson and Watson designed the models for use by emergency managers and the insurance industry in Florida. “We were working on local mitigation strategies to identify various hazard zones, primarily from winds and floods,” says Hank Erikson, manager of mitigation and planning at the emergency management office in Tallahassee, Fla. Johnson and Watson mapped specific hazard zones and put their models and results online, “making it accessible to a lot more people,” Erikson says. Then they expanded their program. The models and Web site “are very user friendly — you don’t have to be a GIS specialist to know how to use it,” he says. “If you can use the Internet, you can use this site.”

In addition to emergency managers and insurance companies, “I would think the site could be very useful to energy companies,” Johnson says. For example, if they know certain fuel storage tanks withstand winds better full, they could be sure to fill those tanks, or vice versa, Watson says. They could also prepare by evacuating sites or putting resources elsewhere, if the model suggests that certain rigs will sustain enough damage to be down 30 days or more, he says.

Although Watson and Johnson’s damage assessment forecasts are still experimental, “we’ve been pretty close” on the effects of hurricanes Ivan and Katrina, Watson says. Their model predicted that 11.5 percent of rigs and platforms in the Gulf would be out of commission for 90 days or more after Ivan: In reality, 10 percent were.

Megan Sever


"Louisiana's marshland mess," Geotimes, November 2005.
Forecasting Tropical Cylone Damages Web Site

Back to top

The authors are the teacher and students of Geoscience 499 at Princeton University, a senior seminar on natural hazards. The team members are Gregory E. van der Vink, faculty, and his students Steven Q. Andrews, L. Kamran Bilir, Richard O. Lease, Lisa A. Newman-Wise, Margaret Prat and Ashley M. Prescott. The work is the result of two years of research and synthesis.

"Storm of progress for tornado forecasts," Geotimes, November 2005 Print exclusive

Back to top

Untitled Document

Geotimes Home | AGI Home | Information Services | Geoscience Education | Public Policy | Programs | Publications | Careers

© 2014 American Geological Institute. All rights reserved. Any copying, redistribution or retransmission of any of the contents of this service without the express written consent of the American Geological Institute is expressly prohibited. For all electronic copyright requests, visit: