Meteorologist Mark Bove describes his work and current research on severe thunderstorms.

Over the past two decades, insured losses due to natural catastrophes have increased drastically. Besides the record-breaking losses incurred by Hurricane Andrew in 1992 and the Northridge Earthquake in 1994, losses from smaller events have also risen dramatically. And, with the American population shifting toward major cities and coastal regions, the resulting dense concentrations of exposure will help create the potential for even higher losses in the years to come.

In an effort to address these and other issues regarding natural hazards, American Re-Insurance Co established a Geoscience Department in July 2000. Consisting of a meteorologist - myself - and an earthquake engineer, the department is charged with improving the understanding of North American natural catastrophes and the losses they produce, as well as protecting the company and its clients from debilitating losses due to natural disasters.

To accomplish these goals, we spend a lot of time providing technical and scientific underwriting support, as well as conducting specialised research on risk from natural perils. What follows are some of the key ways in which we help the company manage its exposure to natural catastrophes and some highlights of the department's research over the past two years.

Model issues
One of the principle areas of underwriting support comes from the assessment of commercially available catastrophe models. These models use proprietary probabilistic techniques to simulate a natural disaster (earthquake, tropical cyclone, flood, etc.) and estimate the potential losses it incurs. They have become increasingly valuable tools over the past decade, and are now established as the primary method by which insurers and reinsurers quantify natural catastrophe risk. However, they should not be used as a black box. All models have idiosyncrasies and nuances within their methodologies that need to be understood to ensure proper usage.

With our experience and background in computer modelling, the Geoscience Department is able to investigate the scientific and statistical foundations of catastrophe models, using information provided by modelling companies and available scientific literature. We can then make recommendations to underwriters and the catastrophe management team on making better business decisions based on the model output.

We can help evaluate the effects of perils not included within a given model. For example, a hurricane model may only estimate losses due to high winds. However, insured losses in a hurricane can also stem from freshwater and coastal (surge) flooding. So, if a property portfolio is being modelled for hurricane loss without considering potential impacts from flood, the results give an incomplete loss picture. By assessing regions where unmodelled perils can have an impact on portfolio losses, we get a more accurate picture of potential loss, protecting the company from unexpectedly high losses.

Many business opportunities require extra specialised/customised research to fully assess the natural hazard risk associated with them. When these opportunities arise, the Geoscience Department works on a case-by-case basis, providing expert advice on the issue at hand. Some issues we have addressed include the exposure and vulnerability of fibre optic cables to freezing rain, workers' compensation catastrophe covers, tornado and hail risk in the central US, and industry loss warranties in Hawaii.

In addition to working with underwriters to improve the underwriting of risk from natural catastrophes, we spend a significant amount of time researching the natural perils themselves. Ultimately, this research can lead to better understanding of natural hazards, how they produce insured loss and how the reinsuring of these perils can be improved.

Severe thunderstorms
One area of ongoing research covers losses due to severe thunderstorms in the US. Over the past five years, annualised insured losses from severe thunderstorms in the US have averaged $4.3bn, the largest average loss from any natural peril over that same period. The annual average loss due to thunderstorms has nearly tripled since the early 1980s, when they averaged $1.5bn per year. Correspondingly, losses from individual thunderstorm outbreaks have increased as well. Since 1995, there have been five thunderstorm outbreaks that produced economic losses over $1bn. Before 1995, only one thunderstorm outbreak produced a loss of that magnitude. There have now even been instances of a single tornado and hailstorm alone causing $1bn in insured loss.

These increases are due to rapid urbanisation over the past 20 years in regions prone to thunderstorm activity, resulting in more dense concentrations of exposure to these risks. The trends associated with thunderstorm risk show no sign of stopping, and research is now being conducted to examine the behaviour of hail losses in the US.

Tropical storm Allison
No matter how unfortunate they are, natural catastrophes do provide unique opportunities to investigate how insured property responds to water, wind or earth movement. So when a significant disaster occurs, we conduct detailed investigations into the event, examining the underlying science and the losses incurred. Depending on the event, the investigations can be either on-site surveys of loss or detailed studies via literature and research. Whatever the method, the lessons learned from these investigations can then be applied to future underwriting practices.

Since the department was started, we have had the opportunity to investigate a few loss events. One such event was the extensive flooding caused by the remnants of tropical storm Allison in Houston, Texas in early June 2001. Allison was a relatively weak, disorganised and short-lived tropical system. Classified as a tropical storm on the afternoon of 5 June, it only produced sustained winds of 95 km/h (60 mph) at its peak, and moved ashore near Galveston, Texas, the next morning.

While Allison's winds were too weak to cause any widespread damage, the system's rains were torrential and weak steering currents in the upper levels of the atmosphere let the storm remain over the region for five days. By the time its remnants exited Texas four days later, parts of the city of Houston had been inundated by over 1000mm (36 inches) of rain, far too much for the low-lying area to handle. Downtown, Houston was among the areas hardest hit, but flooding was widespread over the region. By the time the water receded, there was over $5bn in economic loss across Texas, of which $2.7bn was insured. Allison also affected five other states for a total economic loss of $6bn of which $3.5 billion of which was insured, making it the fourth largest insured loss from a natural disaster in US history.

The losses from Allison's flooding came as a surprise to the insurance and reinsurance industry, since weak storms of this type were perceived as unable to produce insured flood losses totalling billions of dollars. Further, as residential flood insurance is federalised, it was assumed that floods in commercial and industrial centres were sufficiently rare and limited in size to not pose a significant loss threat. Allison, however, showed that catastrophic flooding needs to be considered when addressing the hazards from tropical cyclones. Further, loss patterns indicate that mitigation efforts, such as waterproof doors and walls, can help protect critical equipment in buildings, reducing the amount of property loss and business interruption.

One of our responsibilities is the education of the company's actuaries and underwriters, and those of our clients, about natural hazard and risk. We do this through several different forms of outreach, including papers and reports posted on the company's internal and external websites and seminars to colleagues and clients.

The department has just released its first print publication, Topics, which is a review of North American natural catastrophes from the previous year and the underwriting lessons that can be learned from the events. This inaugural issue of Topics also contains statistics of `significant' US natural catastrophes, 1950-2001

In our research, we can draw on the extensive experience of our counterparts at our parent company, Munich Re, who have been conducting insurance-focused geoscience research since 1974, and in turn we contribute to the group knowledge base. In summary, by leveraging our expertise, as geoscientists we help our company manage risk from natural catastrophes. The end result is a better understanding of natural perils and improved underwriting of the risks associated with them. In a world where losses - both potential and realised - due to natural catastrophes continue to increase, the union of reinsurance and science becomes increasingly valuable, a partnership that can hold great value for the entire industry.

By Mark Bove

Mark Bove is Senior Research Meteorologist at American Re-Insurance Co. (Topics is available online from