Aug. 3 (Bloomberg) -- The computer screen shows a toxic cloud rolling slowly over buildings, a visualization that permits scientists to record every lethal swirl and eddy and to calculate the toll for a city's residents.

It's a far cry from ``The Sims,'' an addictive computer game that lets users create a virtual universe. As displayed on computers nicknamed ``Coyote'' and ``Thunderbird,'' disaster simulations conducted under a $25 million program run by Sandia and Los Alamos national laboratories are helping the U.S. government predict the impact of chemical or biological attacks, killer hurricanes, or accidents such as the collapse of the bridge on the main highway into Minneapolis.

For years, simulations have helped manage the nuclear- weapons stockpile and conduct war games. Now disaster planners are using them to create a ``virtual U.S.'' in which scientists gird for worst-case scenarios to test the vulnerability of the country's infrastructure, former White House counterterrorism chief Richard Clarke said.

Simulation programs permit scientists to ``imagine a whole series of events and one by one run the tests,'' said Clarke, an early booster of the technique. ``It's as close to reality as you can get.''

Spurred by the twin disasters of Sept. 11 and Hurricane Katrina, officials at the Department of Homeland Security have gotten the message. They're using graphic modeling to predict a disaster's human and economic toll, expose weak spots in defenses and train policy makers in improving their crisis responses.

Flu Pandemic

As early as this month, the Homeland Security Department will release results from the second part of a simulation study of a flu pandemic. The initial phase of the test concluded that a merely adequate vaccine given immediately would be more effective than delayed inoculations with a better drug.

By the end of this year, the lab will complete a study on the Midwest's New Madrid Fault, a break in the earth's crust that some scientists fear could produce an earthquake more devastating than one caused by the San Andreas Fault in California.

Lab staffers' enthusiasm about the insights they've gained from simulations sometimes gets the better of them. After discussing how the U.S. financial system has ``the most robust'' security of any electronic network in the world, Los Alamos economist Sam Flaim said ``there are two ways'' to disrupt it.

As he began to elaborate, he was shushed by colleagues who reminded him that the information shouldn't be made public.

Limited Impact

The scientists who create them say that the computerized simulations can be useful tools for state and local disaster officials, with limitations. For instance, the labs came close to predicting that 2.35 million Gulf Coast residents would lose power in 2005 because of Hurricane Katrina, yet could do little to shore up the region's defenses, scientists said.

After Sept. 11, Congress included a provision in the USA Patriot Act that focused the Los Alamos-Sandia simulation partnership on protecting the buildings, shipping, telecommunications and financial networks, bridges and power plants that are essential for the country to function.

With a shifting staff of 80 researchers, the National Infrastructure Simulation and Analysis Center, or NISAC, creates models that resemble a multi-dimensional chessboard.

Layer upon layer of graphics represent different industries interacting with each other as they respond to a major disruption. A tangle of jumping colored dots and lines representing infection rates shows the effect of a raging flu pandemic.

Spidery Legs

Some of the visualizations look like spiders, their arcing legs the supply routes that could be cut by terrorists or a natural disaster.

Most simulations are based on reams of information, such as weather, census or medical data, that would make a researcher bleary-eyed in raw form. Scientists plug the data into special formulas and rules they've constructed to govern their virtual worlds.

For the pandemic study, they created ``synthetic people'' based on census information. Each was given an age, race and socioeconomic level and was assigned behaviors that determine whether they would stay home, go to work or flee in the face of an unfolding epidemic.

While scientists can simulate some behavior, they haven't been able to create a computerized version of the human thought process. ``What's really tripped us up is the cognitive,'' said Bill Wimbish, a retired Army colonel who uses simulations in teaching crisis management at the National Intelligence University in Langley, Virginia. ``It's hard to model irrational behavior.''

Holding Back Data

Simulations sometimes suffer from a lack of critical data provided by businesses, which own 85 percent of the country's infrastructure and are concerned about revealing trade secrets, scientists said.

Aware of simulations' shortcomings, scientists are trying to improve their models before the next big disaster hits.

Since Katrina, the labs have created programs that can show within hours a disaster's potential impact on a region's economy, power and phone systems -- conclusions that used to take days to reach.

Still, even the best simulation isn't much good if policy makers don't put the results into action, said Senator Pete Domenici, a New Mexico Republican who backs the use of graphic modeling. ``It's doing a great job for government if the government would just use it,'' he said.

To contact the reporter on this story: Jeff Bliss in Washington at jbliss@bloomberg.net.

Last Updated: August 3, 2007 00:02 EDT