The New Crystal Ball for Firefighters

By Matthew Frank, 7-16-08

	Caption: Fire modeler Mark Finney of the Forest Service demonstrates the FSPro (Fire Spread Probability) program from his office in Missoula, Montana. Photo by Anne Medley

Mark Finney mouses around a computer screen and up pops an elaborate map of where the Carey Fire in Northern California is burning. Overlaid in splotches of color, he sees, more importantly, where it's likely to grow.

"What we can do here and what it'll show you is mind-blowing," he says.

The map is the product of the Forest Service's innovative computer-modeling program called Fire Spread Probability (or FSPro), a program Finney began developing in 2005 "to take a more strategic look at fire management." Today, its predictions are informing fire managers on how -- if at all -- to suppress hundreds of blazes around the country.

Tools like Finney's are becoming increasingly important for fire managers dealing with two big shifts in firefighting that can seem contradictory: The first being the recognition that some fires are good for the forest ecosystem and therefore, not every fire should be actively controlled. The second being the push to aggressively suppress fires to protect the ever-increasing number of homes sprouting up in the Wildland-Urban Interface.

"It's the paradox of fire: the more you suppress them, the worse they get," Finney says. By fighting every fire, "we end up destroying the very thing we're trying to protect."

	A RAVAR (Risk Assessment of Values-at-Risk) map, overlaid with an FSPro map, shows the Gap Fire burning in Santa Barbara County, California -- and all that's in its way. Click the image for the full PDF.

Modeling, too, helps top brass know where scarce firefighting resources are needed the most -- crucial as huge firefighting bills hamstring the Forest Service and fires burn hotter and longer. About half of the agency's budget is spent on wildfires.

The colors on the map on Finney's screen are in concentric circles and represent probabilities. The red in the center is the area where the fire, if not actively suppressed, has a greater than 80 percent chance of spreading to in the next 14 days. There's a 60 to 80 percent chance the surrounding orange splotch will burn, and on down the spectrum to the all-encompassing light purple, where there's a less than 1 percent probability.

Finney is a former firefighter who works in the Fire Sciences Lab in Missoula and he is one of the preeminent fire modelers in the country. He explains, with puzzling formulas and jargon, how FSPro works. The gist is this: Computers in Kansas City and Sioux Falls run 1,000 weather scenarios, based on a bank of historical records, and mixes them with data on the other two ingredients that determine the magnitude of a fire, topography and fuels, to determine how many different ways the fire can grow. In an hour or two it spits out a map of probabilities -- for any fire manager in the country that's requested it -- often to be pinned up at fire camp.

Finney and fellow forester Rob Seli use the example of a fire that had been burning in wilderness near Oregon's Cascades. The question facing fire managers was whether to actively fight the fire and deploy some of the few remaining crews in the country, or to let it go and risk it escaping the wilderness. FSPro suggested a 20 percent chance of escape. So they held off, and the crews were sent where they were more urgently needed. Decisions like these can save millions of dollars.

	An FSPro analysis from late June displaying all the wildfires burning in Northern California. Click the image for the full PDF.

FSPro becomes even more powerful when overlaid with RAVAR maps, or Rapid Assessment of Values-at-Risk. These values include endangered species habitat, houses, power lines, HAZMAT sites, watersheds and municipal water sources. The maps allow managers to make better-informed decisions about the benefits and losses associated with a fire, and the probability of it all happening. Finney likens it to the way a car insurance company calculates your premium.

Most often, FSPro confirms fire managers' instincts. "It's pretty obvious which way the wind is blowing; you don't need a computer to tell you that," Finney says. On last summer's Black Cat Fire west of Missoula, the models showed that if the fire jumped Highway 93 into very dry and thick fuels it might run rampant, possibly into the Rattlesnake Recreation Area and Wilderness. The maps validated the plan, and crews held strong along the highway. (The fire actually blew back to the West into grass, brush and homes there, surprising managers and FSPro both.)

The program isn't perfect. Its accuracy is a function of the data it runs on. One hundred years of weather data is of course better than 20. And fuels data doesn't always account for recent fire activity.

Nor does it always necessarily lead to saving money. Finney says letting a fire burn through wilderness, for example, can end up costing more in long-term monitoring than it would in up-front suppression.

They're in the process of evaluating the program's accuracy, but it takes a lot of fires to know with any confidence if, say, 80 percent of them burn in the red splotch on the map. Plus, since the success of suppression is a constant variable, Finney and his team can never know where a fire might have otherwise burned.

"I'm not going to say it's the final approach to doing this," Finney says. "It's one more piece of the puzzle to bring to bear on solving a problem."

So far this season, FSPro has been has been able to keep up with the rash of fires burning throughout California, Finney says. Before long, it'll surely be put to the test throughout the rest of the West.

"It runs well enough it doesn't need me anymore," Finney says. "That's a sign of success."
[End of article]