Messing with models: executives from a disaster-modeling firm travel to the Gulf Coast looking for clues to fine-tune their computer models

Risk & Insurance, Oct 1, 2005 by Matthew Brodsky

Frank Fischer and Mohit Pande cruised the neighborhood in their big black sport-utility vehicle, nodding and waving at folks out on the street. They counted the houses and checked out their build, looked closely at the property behind the homes, and estimated their value.

The structures, however, were not tidy ranchers surrounded by neatly manicured lawns. In this case the properties consisted of front walls ripped apart by a neighbor's roof and glass shattered by flying tree limbs. In the yards, massive trees lay uprooted. The damage came courtesy of one nasty killer hurricane by the name of Katrina.

The property owners in this neighborhood of Gulfport, one of the communities that suffered most, were outside because their houses are tipping, sinking, smelling, simmering. As for the two men in the SUV, they were engineers from one of AIR Worldwide's Hurricane Katrina post-disaster survey teams.

Fischer and Pande were one of three pairs that the modeling agency dispatched to the Gulf Coast to measure Katrina's footprint--an apropos engineer's term for the sweep and depth of the storm's destruction.

AIR has been doing such post-disaster surveys since Hurricane Hugo in 1989. Now, with nearly two decades of experience and a systematic and in-depth methodology to follow, Pande and Fischer provided a ground-floor estimation of Katrina's power.

"The analysis is at a property level," Fischer explains.

The purpose of their detailed investigation was twofold. AIR wanted to validate its existing hurricane model.

"We come here with the best scenario, the median scenario wind-speed footprint," Pande says. "So then the damage survey is to validate the wind-speed footprint that we have, to see that the damage that the model is predicting is in line with what we see here, with the actual observations."

AIR's clients can then better serve their customers, by wielding the refined footprint to triage their claims--and redeploy and fill in the cracks of their adjuster coverage to cover the hardest-hit areas.

In the technical terms of modeling, it means the insurance carrier can shift its forces to cover areas with the highest "average claims severity level," says Fischer. "If it's in Biloxi instead of Paseagoula, they'll send (more resources) into Biloxi."

The second objective, Pande explains, was to record details of the actual damage footprint and return to Boston to use them for future research to enhance the model. For instance, after the four storms in Florida last year, AIR modified the 2005 model to account for what's known as "aggregate demand surge," or the price increase in the order of magnitude for products and services needed to rebuild--concrete and lumber, for example. After one storm, prices for construction materials tend to inflate. After four storms in a row in the same region, they rise exponentially, and this has to be factored into the model.

For the Katrina post-disaster survey, the process began when Katrina first was christened, thousands of miles out in the Atlantic, and AIR meteorologists and engineers put a bead on it. As Katrina took aim at the U.S. mainland, Fischer says, the AIR team cranked up their modeling program, spinning out thousands of simulations.

Then Katrina hit--Monday at approximately 6:10 a.m. local time in southern Plaquemines Parish, La., near the town of Buras.

AIR immediately set out to narrow down its thousands of simulations to a few dozen potential scenarios, Pande explains, based on meteorological parameters such as central pressure and wind speed, and more earthly assumptions such as property distribution and construction. From these scenarios, the modeling company formulated those estimated insured damage ranges reported in the news. For Katrina, AIR's initial insured loss estimate was between $17 billion and $25 billion.

OUT OF MANY, ONE

The next step--and the reason Pande and Fischer traveled into Katrina's destruction--was to narrow down these scenarios further to one footprint, the scenario.

Armed with potential scenarios and footprint printouts, which look like a rainbow depiction of where Katrina stomped inland, each color representing a different wind-speed range, AIR dispatched three survey teams on Friday, Sept. 2, four days after Katrina made landfall. The teams dispersed though the stricken area, south, east, and west, to conduct their neighborhood-level studies and to find the edges of the footprint, where wind speeds dropped below 40 mph.

Their teammates surveyed inland for the footprint fringes, which extended about 200 miles, while Pande and Fischer surveyed along the coast, making squares on their maps. In each given town or region--Biloxi, Mobile, Gulfport--the survey team selected random three by three, or four by four, sections of town, usually four to each area of the makeshift grid. Then they counted the houses and business, and estimated the percentage of structures damaged, then the percentage of damage to these structures. They completed a survey of about 10 blocks to 15 blocks a day.