Forewarned is Forearmed

Do giant earthquakes, like the Magnitude-9.0 quake that struck off the coast of Japan just a year ago, occur randomly over time or in cycles? As scientists increasingly support the latter theory, concerns are mounting about the possibility of another mega-quake — measuring Magnitude 8.5 or higher — striking in the foreseeable future. This apparent global clustering of devastating temblors has far-reaching implications for how the re/insurance industry manages earthquake risk, especially in Canada.

Clustering Defined

To put the potential impact of this phenomenon into perspective, it is useful first to understand the concepts of clustering. At the most basic level, spatial clustering of earthquakes — those tending to occur in some locations rather than others, without respect to time — is a reflection of plate tectonics. Most seismically active geographies are located near tectonic boundaries, and so the causality of spatial clustering is thereby implied. These geographic concentrations form the underpinnings of both modeled hazards and building code definitions of seismic zones.

Spatial and temporal clustering in the short-term refers to the fact that after a significant earthquake, other, usually smaller, quakes in the same area — aftershocks — are more likely to occur in the ensuing days, weeks and months.

Spatial and temporal clustering over the longer-term is still in the early stages of discovery. According to this concept, a major earthquake that relieves stress on the ruptured segment of a fault can then concentrate and therefore increase the stresses on adjacent segments of the same fault. This phenomenon of an earthquake being triggered by an adjacent one is related to calculation of time-dependent probabilities. Therefore, it can currently be quantified only for select faults, such as in California, Turkey and Japan, for which there is significant data.

Four earthquakes exceeding Magnitude 8.5 have struck since 2004. Two occurred in Sumatra (Magnitude 9.1 in 2004, Magnitude 8.6 in 2005), one in Chile (Magnitude 8.8) in 2010, and last year’s Great Tohoku (Magnitude 9.0) in Japan. The previous cluster, from 1950 to 1965, saw six events ranging between Magnitude 8.6 and 9.5. Although we do not know the cause of these quakes, we do know that between the two clusters, no earthquake exceeding Magnitude 8.5 occurred anywhere. The random chance of such clustering is less than 1%. If the current cycle follows the one that occurred in the 1950-65 timeframe, it stands to reason we may only be about midway through it, with the largest quake yet to happen.

A logical next question: What other locations across the globe are most susceptible to that future giant earthquake? And what consequences might be reasonable to expect?

Historically, giant earthquakes have occurred on mega-thrust fault zones, where subducting tectonic plates converge with overriding plates. Such zones encircle the Pacific Ocean and lie beneath the Himalaya — including, not surprisingly, Indonesia, Japan, the Philippines, the west coast of both Central and South America, and also the Pacific Northwest of North America.

The Cascadia subduction zone off the west coast of North America converges only half as fast as tectonic plates in Chile or Japan, but it does have a history of generating Magnitude-9.0 earthquakes, albeit not since 1700. A quake of this magnitude in the Cascadia zone would affect the cities of Vancouver, Seattle, Washington and Portland — representing a total population exceeding 10 million. Estimated insured losses could approach $100 billion, with Vancouver and Seattle each sustaining about 20% of losses. Total economic damage could exceed $400 billion.

Impact on Risk Management

The potential threat of this current cluster calls for innovative assessment of the extraordinary perils associated with giant earthquakes. However, although it is always critical to conduct a full

accounting of uncertainty, current catastrophe models characterize the loss consequences of such effects only minimally — if at all. Too little data exists for the science to have matured or even developed. Put simply, we cannot calculate what cannot be measured. How then to quantify the effect of surprises from giant earthquakes?

As demonstrated in the past, shaking-induced surprise perils can be as significant as the shaking itself. To establish rational expectations, risk owners must account for such consequences even if they are typically excluded from catastrophe models. Case in point: The largest losses from the events of 2011 stemmed from surprise perils of tsunami and flood. The sea walls in coastal Japan, built for a tsunami estimated to be a maximum of 10 metres high, suffered a deluge nearly four times as high.

It should be no surprise to the industry that such “known unknowns” would be agents of major damage and loss in any future large earthquakes. The industry must find ways to anticipate these surprises, even if they are not well captured by existing modeling techniques. The prudent first step is to consider worst-case scenarios. Comprehensive planning requires going beyond the limitations of current catastrophe models and thinking more broadly, more holistically. You may not be able to quantify what you do not know, but it is important to imagine it.

As with Japan’s Tohoku quake, the most significant losses from a Magnitude-9.0 quake in the Cascadia zone would include widespread destruction of seaside communities caused by unprecedented tsunami heights. Damage would include disrupted utility and transportation networks, causing delayed recovery; extensive ground failures including landslides and liquefaction; hazardous materials contamination; and severe shaking amplification in tall buildings and long-span bridges due to a long period of shaking.

Some concerns would be locally focused. For example, Vancouver has started to relocate underground its outdated, wood pole-mounted transformers, which have exploded and sparked fires in past earthquakes. Other concerns, related to long-term business disruption, could affect the entire country. For example, what if the Alaskan pipeline was severed for four months? How would it affect your business? What if the main trucking lanes were shut down because multiple bridges were out?

From an internal perspective, paint a picture of how the disaster would affect your portfolio, earnings and reserves. In most instances, a model can provide the dollar value of insured losses, but for this exercise it is important to go deeper. In particular, you should reconsider strategies and decisions for underwriting and pricing that are driven by tail risk. When assessing the tail losses most affected by giant quakes, applying a standard treatment of uncertainty that uses mean values in intermediate steps and synthetic uncertainty at the end can lead to underestimating the loss, often severely. The multi-decadal time span of clusters makes it possible to optimize capital accordingly. Presuming that capital adequacy reserves are maintained to a specific recurrence probability, the chosen strategy would depend on the current position within the cluster. A catastrophe modeling firm can help re/insurers think through such scenarios.

The word “clustering,” as it is most often used in the scientific and academic communities, implies a known mechanism. In contrast, when considering global clustering, we acknowledge there is no known mechanism. If we ultimately accept the evidence of a current global cluster, and our likely position midway through it, then the probability of another giant earthquake striking at some point in the future can no longer be dismissed. If industry leaders put into practice new insights on clustering, then one can expect these new insights will be incorporated into leading-edge models as well. If we acknowledge awareness of the possibility of this phenomenon, then we can pla
n for it. As the saying goes, forewarned is forearmed.