Updated Weather Forecasting

When it comes to natural catastrophes, especially weather-related events, history will not only repeat itself, researchers warn, but it will do so in a much more frequent, severe manner. Scientific research points to the likes of the 1998 Ice Storm and the August 2005 Toronto storm becoming regular occurrences.

Has the technology employed by catastrophe modellers been able to keep pace with Mother Nature? This summer will bring to the Canadian market the release and enhancements of winter storm, tornado, hail and thunderstorm models. Earthquake models are also becoming more robust, modellers say. But the question of creating a model for overland flooding still has modellers and insurers scratching their heads, wondering how they will plug that particular gap in the Canadian marketplace.

CHANGING LANDSCAPE

Property values in Canada have increased significantly since the last major Canada industry exposure update in 2004, says Jayanta Guin, senior vice president of research and modelling at AIR Worldwide. Using indexed census data, AIR updated the number of residential risks. Replacement costs were updated based on data from Statistics Canada to reflect large increases in material and labour costs in the country over the last several years. “The largest increases were seen in Alberta and Saskatchewan,” Guin says. “For example, according to Statistics Canada, residential replacement costs in Alberta surged 26% from May 2006 to May 2007 alone. Replacement costs in Saskatchewan increased 21% during the same period.”

Significant increases were also seen in B. C. Guin attributed most of these changes to a re-evaluation following the Kelowna wild fires, which revealed substantial undervaluation. “Changes to commercial property values have by comparison been modest,” he continues. “However, coverage splits have been updated, resulting in a significant shift in value from building to contents value.” This has had the effect of reducing modelled losses for the earthquake peril, he says. The effect on severe thunderstorm losses is much smaller.

STORM WATCH

AIR has used the updated Canadian property value database to update its AIR Canada Severe Thunderstorm model, due to be released in July. Risk Management Solutions (RMS) plans to launch major upgrades to its tornado hail model in August 2008 (the enhanced model will be a convective storm model and will include straight-line wind and lightening in addition to tornado and hail), as well as release a winter storm model. This is potentially good news for insurers, who last year suffered losses due to an F5 tornado that pummeled Manitoba; in addition, a series of hail storms across the prairies caused a spike in the number and severity of claims -particularly in Saskatchewan, where the number of claims reported, 13,700, was 10% higher than the five-year provincial average.

“We [at RMS] are using more and more advanced techniques and we’re bringing in techniques that are currently employed in weather forecasting — namely, numerical modelling — in order to get a much more complete event set,” says Claire Souch, the senior director of model management at RMS.

Numerical models use equations to determine the behaviour of the atmosphere at any given point in time. Essentially, this form of weather prediction samples the state of fluid in the atmosphere at ay given time; equations common to the fields of fluid dynamics and thermodynamics are then used to estimate the state of the atmospheric fluid at some time in the future. This allows modellers to predict how the weather or the atmosphere is going to evolve.

“We can also force [our] systems to create new storms,” Souch says. “So it’s a way of creating storms in a simulation environment, which can then be put into our models and allow us to get to a very, very high, detailed resolution.”

She says the use of numerical modelling techniques, coupled with more robust information on building stock and claims data, allows modellers to build an event set that contains all of the possible permutations of these types of storms that might strike Canada. As well, it allows for the modelling of what the damage might be from these storms on the ground.

Looking at Insurance Bureau of Canada data from 1984 to the present for cat losses due to winter storm, “you were probably only modelling for one-third of the losses incurred,” suggests Keith Fillmore, senior vice president of Aon Re Global. Among the losses not modelled for was the 1998 Ice Storm -which RMS estimates say could potentially incur US$3 billion in losses if it were to happen today (more than double the 1998 insured losses).

SHAKING THINGS UP

Models targeting atmospheric events are not the only ones becoming more robust. Canadian earthquake models have existed for years, but new developments in the United States may trickle their way onto the Canadian market, says Julie Serakos, senior vice president of Willis Re Inc.

The biggest update in earthquake research stems from work led by the U. S. Geological Survey, Serakos says. “Essentially, it’s a better representation of a risk at a local level,” she says. “So when an earthquake occurs, there is more understanding about how the ground motion will propagate away from the fault line and specifically affect buildings and the areas affected by the event.”Also, there is now a better understanding of the “interplay of the ground motion with the soil and the structures of the buildings,” she says.

It’s quite possible there will be some changes to the Canadian models in the future as a result of this research, modellers say.

PLUGGING THE FLOOD GAP

Although earthquake and tornado models have been available for years, comprehensive overland flood models for the Canadian market continue to elude modellers. Some suggest it’s a combination of the degree of difficulty in modelling this particular peril, a lack of sufficient data and a perceived lack of demand from (re)insurers.

Unlike earthquake and tornado perils, overland flood is not a localized event, Souch says. “With flood modelling, you have to really be able to estimate the flood depth, which changes very rapidly over the surface of the ground, so you’re very dependent on having good information available on the elevations [and] terrain of the ground,” she says. “It’s very data input-intensive. It’s also very modelling intensive, because you have to model the entire hydrological cycle in order to get it right. That includes modelling rainfall, rainfall run-off, evaporation, snow melt — all sorts of things.”

Souch says RMS is in the process of developing a robust overland flood modelling technology for the United Kingdom market, which will hopefully find its way into the United States market. Whether or not that will include Canada, though, is still up in the air. One major complication is that, in contrast to the United States, Canada has a lack of detailed government flood maps and information, Souch says.

Steven Jakubowski, senior vice president and group manager of Impact Forecasting, says his firm is almost ready to release an overland flood model for the U. S. market within the coming months. The riverine model uses the U. S. Geological Survey’s storm gauge data for river networks across the country to build a forecast for future events.

“We have to have actual street address position to determine the exact position of a structure,” he explains. “We can marry that information with a digital elevation model. Then we can have a forecast flood based on historical storm gauge data. We can say at your particular residence or commercial building that we expect there to be three, five or 10 feet of flooding. And then we’re going to model the losses based on that.”

Fillmore is not convinced such a model will make it to the Canadian market. He says the earthquake peril is, and will likely remain, the big issue here. “The issue with adding perils for Canada is that there is no problem sel
ling an earthquake model for Canada, but as you get further down the list of perils — like brush fire or flooding — you have fewer and fewer people who are prepared to pay for models.” Fillmore says he doesn’t mean to suggest that non-earthquake perils are insignificant in Canada, just that from an economic point of view “the real driver that everyone looks at is earthquake.”

If the industry wants third-party cat modellers to add more models to the Canadian table, “there’s going to have to be a push from reinsurance brokers, reinsurance underwriters and the insurers,” he says.