Impact Of Revised Seismic Hazard Maps On Earthquake Models And Risk Outlined For Casualty Actuaries

Coronado, CA – The U.S. Geological Survey’s (USGS) 2008 revisions to seismic hazard maps for the lower 48 states incorporated new data on active faults, updates to regional earthquake catalogs, and changes in general methodology, attendees at the Casualty Actuarial Society’s Spring Meeting were told.

The new maps revealed a number of changes from the 2002 versions, particularly for California and the West, which will impact commercial earthquake models used to estimate insured losses.
“Even though there hasn’t been a recent major earthquake in the United States, all three of the major modeling firms have released, or are releasing, significant updates to earthquake models for the U.S.,” said Don Windeler, director, Americas Earthquake Models, Natural Catastrophe and Portfolio Solutions for Risk Management Solutions (RMS).

He described how the construction of an earthquake model entails looking at where earthquakes occur, how often they occur, their size, and their probabilities, and said the USGS national hazard maps are used as a baseline for all current earthquake models for the United States.  The models develop a future set of earthquakes, representing a range from large and infrequent to smaller and more frequent.

 “For each of those events, we have to have some form of severity measure as we move away from the center of the quake, including how the soil conditions might affect the shaking at an individual building or a portfolio of buildings,” he said.  “Given the shaking at each site, there’s a vulnerability module that looks at the potential damage and the variability for different coverages – from building coverage, contents coverage, workers compensation, to the potential for cost escalation after an event, that is, an economic demand surge that may require an insurance company to pay more over time simply because the repair costs escalate when everyone is trying to find the same scarce resources.”

“Over the last two years, major studies from the U.S. Geological Survey’s National Seismic Hazard Mapping Project (NSHMP) and the Uniform California Earthquake Rupture Forecast (UCERF) provided input for each of the models as to where the earthquakes are, how big they are, how often they occur, and what type of ground motions are likely to be seen,” he noted.

“In 2008, one of the single biggest changes that affected the USGS hazard maps and consequently the models,” the RMS official emphasized, “was the release of a study called NGA, the Next Generation Attenuations, which are the equations used to predict ground motion as a function of distance from an earthquake, the magnitude of the quake, and the kind of quake.  It is one of the key drivers of uncertainty in the model because it affects not only the impact at a single site, but the correlation between multiple locations.”

Windeler said the new NGA relationships are the largest single driver of changes to modeled losses for California and most of the western U.S.  “Once you get into the Pacific Northwest, there’s also the contribution of offshore subduction-type earthquakes, like what was seen in Chile earlier this year,” he said.

While the results of the NGA project resulted in reductions in hazard estimates throughout much of the western U.S., the California-specific outputs from the UCERF study helped drive complexity in the pattern of change throughout the state.  These two studies, the NSHMP and UCERF, represent the state of the science – a benchmark against which we are going to be compared in regulatory settings, he concluded.

Outlining his company’s earthquake model, David Lalonde, senior vice president, AIR Worldwide Corporation, told his fellow actuaries that earthquake models are now based on improved science; the underlying risk has not changed, but our understanding and measurement of the risk has advanced.

“AIR scientists participated in the USGS process to develop the updated hazard maps,” he said, “In addition, AIR has gone beyond the USGS approach in several ways: by employing state or local maps that have more detailed geologic information at high spatial resolution, by including the spatial correlation of ground motion parameters generated by the same earthquake, and by explicitly modeling ground motion effects for major basins in California, Washington, and Nevada.”

Lalonde said that ground motion generated by an earthquake can produce spatial patterns in which pockets of ground motion are either higher or lower than predicted by attenuation equations. According to Lalonde, “spatial correlation can have a significant impact on modeled losses in cases where the larger/smaller-than-expected ground motion pocket occurs in a densely populated area.”

Addressing the vulnerability component of the earthquake model, Lalonde said the damage functions in the model were updated based on engineering analysis, a thorough understanding of building codes, the results of published studies, and analysis of detailed claims and damage data.

 “As catastrophe modelers, it is our job to incorporate the most advanced science and engineering in our models,” Lalonde said.

Kate Stillwell, earthquake product manager for EQECAT, Inc., told the group that “As scientists and engineers, EQECAT follows scientific guiding principles when developing our model.  Along with those principles, we try to foster a sense of humility in understanding that a model is just a representation of reality and not necessarily reality itself,” she said. “But in order to best capture that reality we want to focus our scientific rigor on minimizing modeling uncertainty.”

Stillwell said the Chilean earthquake earlier this year was the type of quake that is very appropriate for use in studying the probability of what could happen with a quake in California because of similar construction methods and similar adoption of stringent building codes.  She said the Chilean quake showed that the use of strong building codes reduces risk, that uninsured damage also affects insured losses, and that the duration of shaking affects the level of damage.

“Still on the horizon,” Stillwell said, “is resolving the questions of how to capture earthquake duration, alternate formulations of vulnerability, and how to deal with non-modeled risks, such as tsunamis that might occur after quakes off the coast.”

Shawna Ackerman, principal and consulting actuary, Pinnacle Actuarial Resources, Inc., moderated the session, which was held at the CAS Spring Meeting in Coronado, California on May 25, 2010.

The Casualty Actuarial Society fulfills its mission to advance actuarial science through a focus on research and education.  Among its 5,100 members are experts in property-casualty insurance, reinsurance, finance, risk management, and enterprise risk management.