Break the Ice

Canadians are certainly no strangers to harsh winter weather.

The City of Winnipeg, for example, on average experiences just 15 days from December through February with temperatures above freezing and 50-plus days a year when the temperature drops below a frigid -20°C. Quebec City has a 50% chance of seeing some form of frozen precipitation on any given day during the same time frame, averaging a formidable three metres of snow annually.

From an insurance perspective, however, the harsh winter weather story is a bit different. Information on catastrophic loss as a result of winter weather in Canada is somewhat sparse: Over the last 35 years, industry sources indicate there have been slightly more than 30 catastrophic events attributed to winter storms, with insured loss ranging from only a few million dollars to over $2.5 billion (trended to 2012 dollars).

Without a doubt, the scant data available represent an incomplete picture of winter storm risk in Canada, and for two significant reasons:

• the short time frame and relatively lean claims history make it difficult to assess the risk of loss on a catastrophic scale; and

• exposure growth in Canada is increasing exponentially.

To the first point, by their nature, catastrophic storms are low-probability, high-risk events. Even the best statistical or actuarial models would find it difficult to predict the details of a 1-in-250-year event given only 35 years of data.

To the second point – and perhaps more important – past claims history may have little to nothing to do with a company’s current book of business if it has experienced significant growth over the last 30 years. Also, even the small amount of catastrophe loss data the industry has reflects exposure growth, with four catastrophic events in the 1980s, 10 in the 1990s, and 16 in 2000s.

While one might be tempted to blame this increasing trend on climate change or variability, the more likely culprit is the growth of Canadian exposure.

What is the best way to manage and explore this ill-defined, yet ever-growing, source of risk? How can the industry take what little data it has and combine it with the best science has to offer to quantify how winter weather might affect the bottom line? Enter catastrophe models.

CATASTROPHE MODELS CAPTURE WINTER STORM RISK

Modelling catastrophic events became accepted as a key tool for managing risk in 1992. This is when catastrophe modelling firm AIR Worldwide used modelling to accurately project the insured losses from ruinous Hurricane Andrew, which levelled parts of Florida and bankrupted 11 insurance companies.

Since that early catastrophe model (and Andrew’s legendary destruction) shocked insurers into understanding why such models are important, the catastrophe modelling landscape has evolved to include multiple regional and worldwide modelling vendors and a range of modelled perils – from tropical cyclones, severe thunderstorms and winter storms to earthquakes, inland flood, pandemics, terrorism, and most recently, cyber risk.

Canada had its own “Hurricane Andrew” of winter storms in the late 1990s with the arrival of the Great North American Ice Storm of 1998. Although the aftermath of the storm was not as devastating as Andrew, the $2 billion-plus (2012 dollars) insured loss total and the weeks of ensuing power outages caught many insurance companies off guard. Before 1998, the most costly winter storm event in Canada had been the 1996 Victoria Blizzard, which totaled over $200 million (2012 dollars) in insured losses.

Similarly, events since the 1998 Ice Storm have come in significantly smaller punches, the most expensive being the 2013 Toronto ice storm at around $225 million (2013 dollars).

While it is tempting to call the 1998 storm an outlier, catastrophe models show that, in reality, nature is capable of creating even bigger and costlier winter wallops.

THE NUMERICAL WEATHER PREDICTION MODEL

Catastrophe models seek to better quantify the view of risk provided by actuarial data by accounting for the scientific principles that underlie each modelled phenomenon. For weather-related models, this often involves incorporating observational data from automated weather stations and eyewitness reports of severe weather, and even from a type of atmospheric data called reanalysis, which contains basic-state variables, such as temperature, wind speed and humidity at multiple layers of the atmosphere over a historical time frame.

Since the atmosphere is highly non-linear and chaotic in nature, it can be unclear what information these sometimes-disparate data sources are giving us.

Numerical weather prediction (NWP) models – the same models that local meteorologists use to predict the temperature – can help with understanding the effects of these variables. Specifically, when provided with past weather data, NWP models can re-simulate significant historical meteorological events.

Importantly, NWP models can serve as tools to understand how slight perturbations in atmospheric conditions might result in drastically different historical outcomes. Coupled with information about the vulnerability of certain exposures to various hazard impacts, such as large ice accumulations, and to a financial module capable of handling complex insurance terms and uncertainty, how small changes in atmospheric conditions can translate to large changes in insured loss can be studied.

EXTREME DISASTER SCENARIOS CLARIFY RISK POTENTIAL

Perhaps the best way to understand the value of catastrophe models is to see them in action – for instance, to model the 1998 Great North American Ice Storm in a slightly different location.

First, review the hazard. Freezing rain occurs when precipitation falls through a layer of relatively warm air aloft into a layer of very cold air at the ground. If the cold layer is shallow enough, precipitation will melt in the warm layer and then fall to the earth’s surface as super-cooled water, freezing on contact with objects on the ground.

The resulting ice build-up can be particularly damaging to high-power transmission lines and other forms of infrastructure. Power loss resulting from damaged lines can contribute to a high volume of business interruption losses, as well as to building damage as a result of the freezing of pipes no longer being kept warm by heating systems that rely on electricity.

Similarly, the ice load on branches and trees increases the amount of damage from falling limbs, and ice damming contributes to building losses.

Given the sensitivity of precipitation to the vertical structure of temperature in the atmosphere, the difference between 10 inches of snow or an inch of ice accumulation is often only a couple of degrees.

Now, imagine what would happen if the 1998 Ice Storm happened today, but this time situated over Toronto. With catastrophe models, the loss from such an unlikely (but possible) event can be calculated.

The map on this page shows the ice accumulation from an event with the characteristics of the 1998 storm, but modelled to occur farther to the west, over downtown Toronto and the surrounding area. In terms of uniform radial ice accumulation, this event would produce a layer of ice nearly 60 millimetres in diameter (the width of about two toonies, side by side). Given today’s building stock (and typical insurance policies), insured losses could approach $25 billion (2012 dollars).

This extreme case demonstrates how small changes in a previously observed event can have a lasting impact on a company’s bottom line. While this event was created with the intention of having the most impact possible, the point remains that, given the short observational period mentioned above, explicit modelling of these phenomena may be the only way to fully develop a well-rounded view of risk.

Ongoing exposure and population growth make winter storm risk in Canada an attractive market for many insurers
; however, the relatively sparse historical record can challenge quantification of risk. Accurately quantifying risk can be even more difficult for other, even less well-observed perils facing Canada, such as severe thunderstorm.

Supplementing claims experience with catastrophe models, coupled with responsible risk management and underwriting practices, can make the Canadian marketplace significantly more manageable.