Seismic Changes in Earthquake Prediction

The number and cost of natural disasters that hit the global market in 2005 was remarkable. Driven largely by the active Atlantic hurricane season, insurers paid out more than US$80 billion worldwide. Canada was largely spared direct impact of these catastrophes however Canadian insurers may still feel the pinch of the resulting global capital shortage when reinsurance renewals come up.

Continuing its evolution, cat modeling both answers questions raised and leverages lessons learnt as a result of cat events. In May 2006, Risk Management Solutions (RMS) released version 6.0 of its RiskLink catastrophe modeling system, with major updates to the U.S. Hurricane, U.S. and Canada Earthquake, and Europe Windstorm models. The most relevant of these to the Canadian market are the earthquake modeling updates.

EARTHQUAKE MODELING: SPECTRAL RESPONSE

Major earthquake model changes have been implemented for the eastern half of the U.S. and Canada. These include updates to seismic source models for consistency with:

* the Geological Survey of Canada 2003 and U.S. Geological Survey 2002 national hazard maps;

* the transition to a spectral response-based ground motion and vulnerability approach; and,

* other data and event sampling enhancements to improve the underwriting strengths of the product.

Similar updates were rolled out for the western U.S. and Canada in 2003. The 2006 release includes new methodologies for modeling post-event loss amplification and fire following earthquakes in the western and eastern halves of Canada.

The new eastern Canada Earthquake model includes a transition from loss estimation techniques using Modified Mercalli Intensity (MMI) to a spectral response-based approach. While commonly used in catastrophe models, MMI is a qualitative, subjective metric that assumes all buildings at a site will experience ground motions the same way. However, buildings of different heights and materials will actually experience the same input ground motion differently, depending on each building’s composition. The RMS spectral response-based approach estimates building damage using an objective measure of ground motion intensity, called spectral acceleration, to directly link ground motions to building performance based upon each building’s height and construction material. This technique presents some interesting cases for eastern Canada.

The eastern half of Canada is considered to be “stable” crust, an interior continental setting far away from the active plate tectonics of the west coast. However, earthquakes in the east tend to be felt over a much larger area than an event of the same size in the west. The difference in area is a consequence of both the source and the path; the old, cold eastern crust transmits ground shaking more efficiently, and the earthquakes also tend to be more energetic considering the size of the fault rupturing in an event. That difference in energy release results in a greater proportion of high-frequency ground motions in most eastern earthquakes.

The spectral response approach to loss estimation captures those differences in ground motion and applies them to specific structure types. In the example of eastern rCanada, that translates to a relatively greater impact on short, stiff structures such as low-rise residences. High-rise buildings however, are less vulnerable in the new methodology, as smaller earthquakes do not generate sufficient ground motion with frequencies that would have the greatest effect on these taller buildings.

BEYOND “SIMPLE” DAMAGE

The hurricanes of 2004 and 2005 provided new insights into the amplification of insured losses after severe catastrophes. Following a major catastrophic event, and most notably seen after Hurricane Katrina, claims costs can exceed the normal cost of settlement as a result of economic, social, and operational factors – i.e. factors that go beyond mere “simple” damage. For example, post-event loss amplification encompasses such issues as the escalation of the cost of labor and materials, inflation of insurance claims, and “Super Cat” scenarios. In the last of these, a disaster striking major metropolitan areas such as Vancouver, Toronto, or Montreal could cause evacuations and/or a systematic economic downturn.

The potential impacts of loss amplification in Canada reflect aspects of the country’s seismic risk. In the seismically-active west, insurers and reinsurers using the RMS model will see increased loss estimates begin at very short return periods and exceed 10% by the 100-year return period loss. In the less-active east, 10% increases are not seen until beyond the 250-year return period, but the impacts increase rapidly and outstrip the west at the longest return periods.

Fire following earthquake (FFEQ) has always been of concern in Canada, given the much greater proportion of insureds covered for fire losses relative to earthquake. The update to the RMS FFEQ model focused on localizing the results to a more detailed level for individual cities. The update:

* incorporates major roads and freeways as fire breaks;

* simulates fire spread through a grid of the building density and materials, and

* varies local wind conditions.

In addition, the simulation model reflects new global data on fire ignitions as a function of ground motion, and takes into account reduced capacity for fire suppression in the most heavily damaged areas.

In the aggregate, the updated fire modeling suggests a lower risk from this peril in Canada than previously estimated. Even with the lower penetration of earthquake insurance – much lower in the eastern half of the country – the estimated average annual loss from fire following earthquake is approximately 10% of the gross shake value in the west and 20% in the east. A major part of this phenomenon relates to nature of the peril: it requires high ground shaking in a dense urban area to ignite many fires and overwhelm the local suppression capacity. Since these large urban events are uncommon in Canada in the first place, major fire losses lurk in the far tail of the loss distribution rather than in the 250- to 500-year return period window of greatest concern to most insurers.

The shift of losses further out in the tail is a theme throughout the model updates.

EARTHQUAKE LOSSES

Losses estimated through spectral response tend to decrease for small magnitudes and short return periods relative to MMI, but increase for larger earthquakes. Loss amplification figures most prominently in catastrophic events, as do the fires following such events. If one looks at the ratio of the 1,000-year loss to the 250-year loss as an indicator of how skewed the distribution may be, the ratio for industry-wide losses in eastern Canada has increased from about 4x to between 5x and 6x because of the overall revisions. More concentrated portfolios are likely to see even higher ratios.

This model update also integrates the latest CRESTA zones for Canada into the geo-coding, analysis, and reporting functionality of the RiskLink catastrophe modeling system. The CRESTA organization has officially designated these zones, which are included in RMS products for companies required to report aggregates by CRESTA zones. Some users have already noted a shortcoming of the new zones for western Canada. The main Vancouver metropolitan area has five different zones, rated in terms of relative earthquake hazard. The rest of B.C. is grouped in Zone 11 as “Very Low Risk.” This zone, however, includes both Victoria and the rest of Vancouver Island, which are exposed to both a Cascadia subduction zone and intermediate-depth earthquakes, and therefore represents the highest hazard in western Canada.

Previous Canadian CRESTA zones differentiated both Victoria and the other coastal areas exposed to large earthquakes from the inland regions closer to Alberta. The lack of this distinction in the current zoning has generated interest in alternative approaches to monitoring aggregates.

RMS provides functionality for analyzing “footprints” of accumulation, allowing insurers to determine quickly how much exposure falls within a predefined area. These areas can be created to match political boundaries, the maximum damage extent of a scenario event, or other hazard metrics such as a threshold ground motion on national seismic hazard maps. An accumulation summary does not replace risk analysis; however, it does provide a benchmark for monthly reporting.

Given its exposure to active seismic sources such as the Cascadia subduction zone in the west and the Charlevoix seismic zone in the east, the Canadian insurance market must always consider the threat of earthquakes. Catastrophe models continue to be a critical tool for insurers and reinsurers to understand the risk of experiencing these erratic events. These models continue to evolve, even in the absence of damaging earthquakes within Canada. Lessons learned from other global events inform modeling parameters and hopefully, as a result, better prepare the Canadian market for when that disaster finally does hit close to home.