The lessons heeded from the events of 1954 when hurricane Hazel ripped across the US can be applied to the latest catastrophe modeling techniques, explains Dr Atul Khanduri

Fifty years ago, history was made when the first Boeing 707 tore through the skies on its maiden flight, heralding the launch of the jet age. Many will also remember 1954 for the roar of Hazel - the monster hurricane that left a path of destruction stretching from the Caribbean to Canada. Half a century later, as we move into the active months of what is forecast to be an above-average hurricane season, it is an opportune time to look back and examine one of the most destructive storms in North American history. It is also an opportunity to look ahead and turn our attention to the catastrophe models used today by insurers and reinsurers to tame the big cats yet to come.

Hurricanes develop and intensify over warm ocean water. Once over land, they are deprived of this source of energy as the 'eye' begins to fill with cold air. Additionally, at ground level, frictional forces induced by surface roughness impede wind flow, slowing down the winds near the ground. But while all storms weaken after landfall, some storms - the fastest and most powerful - move far inland, bringing damaging winds into areas that many would not think at risk from hurricanes. Fifty years ago, on the morning of 15 October 1954, such a storm ripped through the Mid-Atlantic states, across Pennsylvania and New York and on into Canada.

Hurricane Hazel killed 95 people in the US, 80 more in Canada and caused economic losses of about $281m, a huge amount five decades ago.

Hazel wreaks havoc

Hurricane Hazel embarked on its destructive course well before it hit the US mainland. It began life as a low-pressure trough over the warm waters of the tropical Atlantic, near the Windward Islands. On 5 October, hurricane hunter planes located a small storm about 50 miles east of Grenada.

Though the eye was not yet clearly defined, winds had already reached an estimated 95mph. Days later, having been caught up in an upper level low, Hurricane Hazel swept over Haiti, destroying several towns, wiping out nearly half of the island's coffee and cocoa plantations, and leaving hundreds dead. Hazel then headed toward the Bahamas, where six more died.

On 14 October, as Hazel neared the Carolina coast, hurricane hunters measured Hazel's winds at 150 mph and her forward speed at a lively 30 mph.

Hazel made landfall as a category 4 hurricane near the border between South Carolina and North Carolina. Maximum sustained winds were 140mph.

Coinciding with an autumnal high tide, Hazel was preceded by a storm surge that reached nearly 18 feet in some areas, breaking records for the region.

The town of Garden City, SC was almost completely destroyed. Only two of the 275 homes remained habitable.

By the time Hazel crossed North Carolina, its forward speed had increased to over 50 miles per hour. The storm tracked through Virginia and Pennsylvania by late evening. Although by this time Hazel had taken on extratropical characteristics, the storm had lost little of its punch. On 16 October, Hazel moved across upstate New York and, a mere 14 hours after landfall, it plowed into Ontario, Canada. Peak winds reached between 70mph and 80mph in many areas of central New York and northeast Pennsylvania leading to significant damage. Gusts of up to 100mph were reported in locations throughout Virginia, Maryland, Pennsylvania, Delaware, New Jersey and New York. In Washington DC, 100mph gusts tore trees from the White House lawn.

Tree damage was widespread, as is indicated by the following passage from Jay Barnes' North Carolina's Hurricane History : "In the aftermath of the storm, some sections of highway were littered with 'hundreds of trees per mile.' Some were uprooted and tossed about, and others were snapped off ten to twenty feet above the ground. In the city of Raleigh, it was reported that an average of two or three trees per block fell."

Conventional wisdom, prior to Hazel, was that hurricanes dissipated before reaching as far inland and as far north as Ontario, Canada. And in fact the Allegheny mountain range, which extends from northern Pennsylvania to southwestern Virginia, did break up Hazel's intensity, though only slowly. Hazel, however, did not die. Instead, it combined with a low-pressure system moving in from the east across the plains. The storm, with winds by then of 70mph (just below hurricane classification), picked up extra moisture at the edge of the low-pressure system. It proceeded across the Great Lakes and dumped torrential rain on the Toronto area, submerging several highways and washing away bridges. Ontario saw its worst flooding in more than 200 years.

Forward speed

The forward, or translational, speed of Atlantic hurricanes averages about 15-20mph. Typically, storms speed up as they move to more northerly latitudes. The Great New England Hurricane of 1938, for example, reached forward speeds of near 60 mph. Other storms, like 1998's Hurricane Georges, stall and batter coastal communities with hurricane force winds for hours on end. Without doubt, Hazel's forward speed was unusually high given the latitude of landfall and it was a significant factor in the extent of damage the storm inflicted, both geographically and monetarily.

Fast forward speeds can increase damage by increasing wind speeds where the wind's direction coincides with the storm's direction, i.e., on the storm's right side, and by quickly producing a large field of damage.

But slow moving storms also have the potential for inflicting greater-than-expected damage, if for a different reason. In this case, damage has more to do with the inability of building components to withstand constant pounding for long durations than with the intensity of the wind itself.

When design loads are exceeded, a fastener or connector that has been pulled out as a result of uplift can compromise the integrity of the building envelope. Wind damage manifests itself at the weak link in a structural system. As each connector surrenders to fatigue failure, loads are transferred to the next point of vulnerability. The longer the duration of high winds, the longer this process will continue and the greater the likelihood that the envelope of the building will be breached. Once a breach occurs, internal pressure can build up uncontrollably, blowing out doors and windows, pulling roofs away from their framing and even, in the case of wood frame buildings, resulting in complete structural failure.

So will a slower storm result in higher losses? For a given exposed property, damages resulting from the storm of longer duration will clearly be higher (all else equal) because of the cumulative effects of wind. A faster moving storm, however, will bring potentially damaging winds further inland, affecting a larger geographical area. In the end, total losses depend on the geographical distribution of exposures relative to the storm's footprint. In either case, when estimating damage and loss, a model should capture the effects not only of peak wind speed, but also of wind duration.

A look ahead

Much has changed since 1954. Populations in coastal communities have increased, as have property values and the costs of repair and replacement.

Building materials and designs have changed, as have building codes, and new structures may be more or less vulnerable than were the old ones.

Clearly, however, Hurricane Hazel would cause losses in the many billions of dollars were it to recur today. Adjusting for inflation, population growth, increased wealth and today's take-up rates (the percentage of properties actually insured), AIR estimates that insured losses from a recurrence of Hurricane Hazel would be close to $11bn.

Moreover, things could be worse. An analysis of Hazel's windfield shows that it largely spared Washington DC. With a shift of Hazel's track just ten miles to the east, the storm's radius of maximum winds would fall directly over Washington DC and as much as double losses in the area.

Of course, the real purpose of catastrophe models is not to estimate losses from historical storms. Rather, the purpose of modeling is to anticipate the likelihood and severity of future events so companies can appropriately prepare for their financial impact. Standard actuarial techniques rely on past losses to project future losses. But the scarcity of historical loss data resulting from the infrequency of these events, as well as the changed landscape of insured properties, makes standard actuarial techniques for loss estimation inappropriate for catastrophe losses. That is why modelers, using sophisticated stochastic simulation techniques, produce catalogues of hundreds of thousands of potential future events capable of causing loss.

The re/insurance industry has experienced a period of relative calm in recent years in terms of major hurricanes landfalling on the US mainland.

But 2003's Hurricane Isabel taught that large losses ($1.7bn) can result even from much weaker storms than those of the intensity of Hazel. As population density continues to increase in areas of high risk, so does the potential for large losses. Catastrophe modeling enables the proactive decision-making and strategic planning necessary for financial stability in the face of such extreme events.