Simulating nature's wrath

Insurance companies that cover damage from natural disasters need to a way to evaluate the risks they face. In essence, they must ask, "What are the chances that a hurricane, earthquake or flood might damage a particular building in a particular city?"

The question takes on front-page importance as Colorado battles giant forest fires, Texas and parts of the Southwest suffer exceptional drought conditions and Duluth recovers from the worst flooding in more than a century.

EQECAT is a "catastrophe risk modeling" firm based in Oakland, Calif., that uses computer simulations of disasters to help quantify their risks. Its executives visited Minneapolis recently to explain how it can help insurers.

Jose Miranda is the firm's director of client advocacy and a meteorologist by trade. This interview was edited for length and clarity.

QHow do insurance companies use your software to help set premiums for natural disaster insurance?

AUsually our clients have a portfolio of property somewhere in the world. They input the actual location of a site or structure into our computer model, and the model outputs their risk of encountering a natural disaster. Clients make their own decisions about whether to write an insurance policy or how to price it; we don't make any recommendations.

QHow long have insurance companies been relying on computer simulation risk models?

QWhat's your software called?

ACurrently it's called "WorldcatEnterprise," but we have a new version called "Risk Quantification and Engineering Version 13" that we're rolling out this fall. It's part of a three-year upgrade of our software that is the largest undertaking of any company in our industry. We've updated most of our "peril models" for countries around the world.

QHow does your simulation deal with some of the latest natural disaster risks?

AWe update our models. Our North Atlantic hurricane model has been updated eight times over the last 10 years. And after last year, when Japan had its worst earthquake in a long time, we're incorporating the latest research on tsunamis and earthquake aftershocks into our model.

Sometimes the model doesn't need to be changed. In the last year or two, we've seen what appeared to be an increase in the number of tornadoes and, as a result, more financial losses. But, if you look at the last 60 years of tornado activity, there isn't really any increase. Instead, there are more storm spotters; Doppler radar has improved, and in many areas population has increased so that tornadoes affect them more.

AOur computer model is like a pie cut into three slices.

One piece is the "hazard module," which takes into account the entire range of possibilities for a given region in a given amount of time. We now simulate 180 natural hazards in 96 countries.

Another part is the "vulnerability module." If we have a hurricane, what's likely to happen to a building the hurricane hits, based on the building's structure?

Then there's the "loss module," which combines the hazard, the vulnerability and the property information. It gives benchmarks. What size of loss will you see once in 100 years, or once in 500 years?

The biggest issue for our clients is that catastrophe models are only as good as the information you give them. If the location data isn't accurate, or we only have hazard data by state or, in some regions, by country, the amount of uncertainty in the model results becomes a problem.

AWe don't issue predictions such as "There's a 22 percent chance that North Carolina will be hit by a hurricane in 2013." We're not in that business.

Instead, we provide clients with data on the "possible risk of ruin." We might look at a case where there are 100 McDonald's restaurants in Miami. How big would the loss be for those restaurants if a Category 5 hurricane [such as Andrew] hit the city? And, by running our simulation, the client could determine which restaurants would be hurt the most based on their location.

Steve Alexander • 612-673-4553