By KATE STILLWELL, an earthquake products manager at Oakland, Calif.-based catastrophe risk modeler EQECAT Inc.
Earthquake models recently incorporated the 2008 update to the U.S. Geological Survey (USGS) hazard model, which is the basis for the national seismic hazard maps used in building codes. The new science underlying the USGS update has dramatically reduced our perception of earthquake risk.
The most significant aspect of this update was the incorporation of next-generation attenuation (NGA) equations, which describe how seismic shaking intensity varies with distance and magnitude. For most earthquakes, the newer equations estimate lower shaking intensity than previously thought--up to 40 percent lower, but on average around 15 percent to 25 percent lower. This means, for instance, we now think earthquake hazard in California is less than we thought prior.
This new science alters how we model earthquakes by affecting how the models estimate the geographic distribution of shaking intensity caused by a given earthquake.
But, in the next update of the scientific standards on which catastrophe
models are based, this downward trend could be reversed.
Factoring soil-based attenuation (SBA) into the models, however, could pre-empt this potential flip-flop in results, making it a critical solution for insurers, risk managers and other model users looking for a consistent outlook on earthquake risk.
THE BASICS: HOW MODELS WORK
An earthquake model is used to estimate the financial consequences of potential future earthquakes. It can inform insurance and reinsurance premiums, capital reserves and other transactions that transfer risk.
Users input information on assets exposed to risk: their location, physical characteristics and insurance structures. The model then simulates tens of thousands of possible earthquakes and the resulting damage and losses. The resulting output includes a probabilistic representation of monetary losses. For example, the model output can answer the question: What is the probability of losing $100 million from earthquakes in any given year? Or, considering the full suite of all possible future events, what is the average annual cost of risk?
The first step in modeling earthquake risk is defining the set of possible earthquakes, their size, location and probability of occurrence. Then, for each earthquake, the model estimates the geographic distribution of ground shaking. Using shaking intensity and information on exposures, the model next estimates the resulting damage and, given the policy structures, calculates insured losses.
THE CASE FOR SBA
One of the most sensitive components of earthquake modeling is the calculation of the shaking footprint, also known as the attenuation relations.
The multiyear NGA research project, coordinated through the Pacific Earthquake Engineering Research Center (PEER), improved the confidence of this calculation by incorporating data from three times as many earthquake records than previously used.
Embedded in the formulation of the NGA is an assumption about the medium through which the seismic waves propagate, also known as the reference site class--i.e., whether the earthquake's waves are traveling through soil, rock, etc. NGAs can be formulated (in other words, the variables in the equation arranged) to accommodate any number of assumptions about the underlying site class.
The conventional formulation of NGAs, and the one used by the USGS and engineers, is to assume that rock is the reference site class. Yet the vast majority of insured exposure is on soil, where buildings are founded. For these exposures located on soil, the shaking intensity must be altered by so-called site adjustment factors, which can be as high as a factor of two.
The SBA approach uses soil as the reference site and requires far less correction to the NGA results (only for those exposures located on rock, rather than vice versa.) For buildings on soil, SBA eliminates an important source of uncertainty and more faithfully reflects physical conditions.
The site adjustment factors are due for an update. They were last revised before the most recent USGS update, the one that caused dramatic reductions in modeled earthquake loss estimates. This time lag creates inconsistency and bias. The next update to site adjustment factors is likely to result in increases: Even though shaking has decreased, NGA data shows that soil amplifies shaking more than we previously thought. Therefore, the current bias in a rock-based approach underestimates the hazard, particularly for exposure located on soil sites. If the hazard is underestimated, insured losses could be too.
What is the magnitude of this potential underestimation of earthquake risk? In comparing results for a SBA versus a rock-based approach, SBA produces shaking intensities up to 20 percent higher.
Using SBA in modeling circumvents the potential rock-based bias because it minimizes the reliance on outdated factors, thereby reducing the magnitude of changes to modeled results.
Until the standard site adjustment factors catch up to the new ground-motion prediction equations, SBA can help avoid the roller coaster ride of results. It enables the industry to realistically and rationally analyze earthquake risk for sites, policies and portfolios.
September 1, 2010
Copyright 2010© LRP Publications