How to Read the Mind of a Wildfire

From studying tree rings to creating intricate computer models, scientists are trying to understand why flames behave the way they do.

Ecologist Don Falk points out a fire scar on a fallen tree stump. (Brian L. Frank)

In a stand of ponderosa pine trees high in the Santa Catalina Mountains overlooking Tucson, Arizona, forest-and-fire ecologist Don Falk squatted with me next to a 100-foot-tall tree born a decade or two before American independence. At the base of the trunk, the tree's thick cinnamon-colored bark gave way to a shallow opening a foot wide and two feet high that looked like a series of successively smaller triangles. Falk ran his hand along the charred edges of the opening and explained what we were looking at: a window into the forest's past, and fire's role in shaping it.

Falk studies fire-scarred trees to understand how frequent, severe, and widespread fires have been in an area, and how those patterns have shifted over the centurieswhich is also a key to understanding why some fires are bigger, more unpredictable, and more destructive these days, “How do you know anything on Earth has changed?” he asks. “You have to be able to compare it to how things were in the past. This is how we know the history.”

Long before the Mexican-American War, when this land still belonged to Mexico, a fire swept up this mountain slope. Short flames wrapped around the tree and curled like an eddy in a stream, lingering on the back side, where accumulated leaves and pine needles caught fire. The flames stayed long enough to penetrate the bark and killed a portion of the cambium, which produces new cells. The tree slowly healed itself, pushing edges of new growth onto the dead area, year after year. But the scar remained. The next fire that came through left another scar, and the next fire another. If we examined a cross-section of the tree, we could use the rings to figure out the exact year of each fire.

Falk works down in the valley at the University of Arizona's Laboratory of Tree-Ring Research, which occupies a gleaming new four-story glass-and-metal cube and holds 2 million wood specimens from around the world, the largest archive of its kind. The lab's founder, an astronomer named Andrew Ellicott Douglass, created a new discipline called dendrochronology: the analysis of tree rings to interpret and date past events. He used rings to date ancient Aztec and Pueblo ruins in the southwest by studying trees used in their construction, and he found that trees in the region grow more in wet years than in dry years, a first step in our understanding of climate change.

Falk, his face tanned by long days in the field, walked with me through the pines. He stopped at a large ponderosa-pine stump, two feet across, cut smooth by a chainsaw. To understand wildfire today, everything we've done to try to control it, and the problems those efforts have wrought, this was a good place to start. He brushed fallen pine needles from the stump and offered a quick reading of the tree's fire history: Born in the mid-1700s, it shows scarring from fires every decade or two, the rings curled like breaking waves around the wound. But something curious happens after the marks from an early-1900s fire: the scars stop. The tree rings continue out toward the edge, for decades, slowly healing that last fire wound, until the tree died several years ago.

Where did the fires go? Grazing animals consumed some of the fuels that would have carried fire. Then, a century ago, we embarked on a campaign to banish fires from forests, with a goal of extinguishing them soon after they started. But that wasn't such a good thing for the forest. When fires don't come through regularly, fuels accumulate. A couple of centuries ago, forests like this one in the southwest might have had a few dozen trees per acre, widely spaced, with an open, savannah-like floor. Today an acre might be crowded with thousands of mostly smaller trees. When fires do burn, they're more destructive, often killing the big trees along with the small.

“What's being released in a fire is the accumulated capital stored up through years of photosynthesis,” Falk says. “You're not destroying the carbon, hydrogen, or oxygen molecules. They're just being liberated.” And on a tremendous scale: even a relatively small fire of a couple hundred acres can pump out energy equivalent to the atomic bomb dropped on Hiroshima, and can push a mushroom cloud of hot air, ash, and soot miles into the sky.

Falk and I stopped for lunch in a meadow and sat on a scorched and toppled pine tree. Evidence of the 2003 Aspen Fire, which burned 84,000 acres across these mountains, surrounded us: dead trees that stand like giant toothpicks; huge patches of pine forests that have been replaced with shrub. Falk regarded this with a scientist's remove. “There's no good or bad fire,” he said. “That's a human construct. We can say what's normal, or historical, or expected, but how can you say it's bad?”


While Falk studies fires to better understand how they have changed over the years and altered the landscape, wildland firefighters study fire as soldiers might analyze enemy capabilities. They catalog mental snapshots of fire behavior they have encountered: how the flames ripped through a grassy canyon or hopped off the ground and leaped into the treetops; the strange calm before a sudden wind shift; the fire tornados that can spin flame in new directions. Early on, a rookie matches these images with what he's learned in training or heard from the veterans.

“The mind pulls up the most similar slide, and it says, ‘I recognize this situation as similar to this other one that I experienced, so probably a similar approach is a good way to go about it,’” Larry Sutton says. He heads up risk management for the Forest Service at the National Interagency Fire Center, in Boise, Idaho. “The danger is, what if this situation doesn't match any of your slides?”

For nearly a century, scientists have been trying to create a bigger reference library to explain why fires behave the way they do. Starting in the 1920s, Harry Gisborne gathered data on weather, terrain, and fuel conditions to predict the likelihood and potential severity of a fire in a given area. The legacy of that work can be see today at the entrance to many parks and forests—an arrow pointing to a color-coded scale for fire danger. When the Mann Gulch Fire flashed up a grassy Montana hillside and killed 13 firefighters in 1949, Gisborne went there to study the fire's progression, and died from a heart attack while hiking the steep terrain.

An aeronautical engineer named Richard Rothermel continued Gisborne's quest. At the Forest Service's Fire Sciences Laboratory in Missoula, Montana, he used wind tunnels and combustion chambers to test how fast fires ignite and spread in different wind speeds, slopes of terrains, temperatures, and types of vegetation. He translated his observations into mathematical equations that could be used on the fire line by sketching a graph or punching a few numbers into a special calculator.

This was a reductionist model that treated fire as a single wall of flame moving through a homogenous fuel bed. It didn't allow for much nuance, but it was easy to use and gave fire crews an idea of what they could expect: Will the flames be small enough for a hand crew to knock down, should a bulldozer or airplane be used, or should everyone just stand back and watch the fire burn? In the 1970s, Rothermel began training the first generation of fire-behavior analysts to sketch the bigger picture for fire crews: where the fire will move and when, based on the fuel, terrain, and weather.

Today, fire-behavior analysts often run computer simulations for fire spread. Since the 1990s, one of the most widely used has been FARSITE, a Windows-based fire-progression model based on Rothermel's calculations from the 1960s. FARSITE conceives of a fire as a long chain of individual fires, side by side, each influenced by the specific wind speed, terrain, and vegetation in that area. The program calculates the spread of each small fire and incorporates the weather forecast to create an animation of how quickly and how far the larger fire will spread.

Programs like this are good at predicting where a fire will spread over a few days, which is crucial information for fire managers divvying up limited assets. But they can falter when confronted with fast-changing variables and complex terrain or weather conditions.

“One of the big things about fire behavior is looking at the exceptional events, the 1 in 100,” says Richard Bahr, the head fire-behavior analyst at the National Interagency Fire Center, who was trained by Rothermel in the 1980s. Those outliers have always existed, but are exacerbated today by climate change, bug infestations, invasive grasses, and fires so intense that they create their own weather. “We are living in a time that is unprecedented, with the extremes we're seeing in temperatures, precipitation, and winds, and with that, the effects are unpredictable,” Bahr says. “If you don't have that in your slide tray, you aren't going to believe it.”


In Los Alamos, New Mexico, Rodman Linn showed me the next generation of fire models, which can map out fire not just in the middle of the bell curve, but out on the tails as well, where firefighters more often meet the unexpected. Linn is a researcher at Los Alamos National Laboratory, which primarily develops and studies weapons and their effects for the military. Models of weapon effects and the spread of fire share the same scientific underpinnings. For his doctorate dissertation in theoretical fluid dynamics, Linn built FIRETEC, which uses physics-based models for combustion, heat transfer, and turbulence to simulate the spread of wildfire. Unlike FARSITE, which can run on laptops, FIRETEC requires the supercomputers at Los Alamos National Laboratory.

FIRETEC doesn't treat fire as a wall moving through a uniform fuel bed, but divides a landscape into cells as small as a cubic meter and approximates the type and condition of fuel, the terrain, and the weather within each small space. One cell might have a tree, while the surrounding cells are grass, or the wind might be blowing south in a cell, but west in an adjacent space and north in another. Similarly, fires create their own turbulence and wind patterns. As warm air from a fire rises, for instance, cool air rushes in to replace it and fans the flames, and different parts of a fire can compete with each other for air, causing winds and fire to grow and shift in unexpected ways.

A FIRETEC animation of a 1996 blaze in Malibu

Linn and his colleagues coupled FIRETEC with a fluid-dynamics model called HIGRAD that represents airflow across different types of terrain and vegetation. The combined program can simulate the constantly changing interaction of wildfire with the surrounding atmosphere, which is a huge missing piece in most fire models. Without it, there's no way to predict extreme fire spread driven by the complex mess of winds that slam into one another and mix with a fire’s own convective forces.

Linn popped open his laptop and played a simulation that he and his colleagues built for Los Angeles County to re-create a fire that had raced up a steep hillside and injured several firefighters who were trying to protect houses along a ridge near Malibu in 1996. The FIRETEC simulation behaves like the real fire did, covering one kilometer from the canyon to the houses in the same time, about 24 minutes. Next Linn showed me the FARSITE simulation for the same fire. The flames spread up the hillside and wash into the houses, much the same as Linn's model, but FARSITE predicts the fire would need three hours to reach the firefighters—six times slower than the fire actually moved.

Linn showed me another simulation, an elegant purple-and-pink mushroom cloud rising over Bandelier National Monument, just south of Los Alamos. As the simulated column in the model sheers and collapses, purple and pink tentacles cascade down, spread across the landscape, and pour through the canyons, like dry-ice vapor snaking along the ground. This was the Las Conchas Fire, which started at 1 p.m. on June 26, 2011, after a gust of wind blew an aspen tree across a power line about 10 miles southwest of town. An easterly wind pushed the flames hard, and by 7 p.m. the fire had swept over 8,000 acres, mostly in a narrow east-west band, with little lateral spread.

Firefighters had several things working in their favor as night came on: the winds were easing; temperatures were cooling; the fire was headed downhill, which slows progress; and it was moving from mixed conifer forest into piñon and juniper trees, which don't carry fire nearly as efficiently. The incident-management team predicted that the fire would grow to a total of 9,000 to 12,000 acres by morning. Instead, by 3 a.m. the fire had exploded across 43,000 acres near Los Alamos, where some 18,000 people were sleeping, spreading at more than an acre a second.

“It's a really scary scenario,” Linn says. “No one anticipated this growth at all.”

And no one knew why it had happened. Linn and his colleagues suspected the smoke plume that rose to 36,000 feet over the hills and canyons as the fire burned hot and fast that day was to blame. Plumes like that are inherently unstable. Soot cools as it rises, creating water vapor and more energy and turbulence. If a fire loses intensity or the wind shifts, chunks of the plume can sheer off like a calving glacier, or the whole column can collapse and whoosh down, driving the fire like an enormous bellows.

Linn’s few seconds of animation took a week to build and run on the lab's supercomputers, and they only simulate the air currents that drove the fire. Next, he and his colleagues will incorporate the fire behavior for a full re-creation of the blaze. But funding for wildfire science is scarce, so this is slow going. “This is a drop in the bucket of one year's [fire] suppression budget,” Linn says. “It's really incredible how little money is put into this problem, in the context of how much damage it causes and how much money we spend to fight it.”


These new models can be particularly useful for putting healthy fire back in the landscape, which can, in time, limit the number of disaster fires. How will the way a forest is thinned—the size and pattern of tree removal—influence fire behavior? And what's the best way to start a fire under chosen conditions instead of waiting for a lightning strike or a tossed cigarette, when the forest might be much drier or the winds higher?

Getting that kind of intentional fire right can be tricky, and if it's near houses, there is little tolerance for error. In 2000, a prescribed fire caused $1 billion in damage when it escaped the intended burn area, torched 235 Los Alamos homes, and damaged several buildings at the national laboratory.

The presence of human structures means that forests and shrublands aren't allowed to burn the way they once did. It’s a problem that keeps getting worse. Because the federal government pays most of the tab for firefighting, local governments don't have as much incentive to regulate development in the most fire-prone areas. Some insurers charge higher premiums if homeowners don't mitigate fire dangers, and more communities are adopting building codes that require landscaping and construction materials that can better withstand wildfire and not carry flames through a neighborhood.

But most community-protection programs are voluntary, with progress outpaced by influx into the wildland-urban interface, and more homes at risk means more firefighters at risk.

Last year, commissioners in Montana's Lewis and Clark County told their local firefighters that the location of homes should not dictate fire-suppression tactics or the placement of fire lines—just because someone builds a house in the woods doesn't mean firefighters are obligated to protect it. That urge is understandable in structural firefighters, trained to save houses when someone dials 911, but it can be just as strong among hotshots and smokejumpers, a humanistic impulse. “You know there are houses being lost, and it makes you sick,” said Travis Dotson, an analyst at the Wildland Fire Lessons Learned Center, in Tucson, Arizona. “I'm never going to be on a fire and see a house and say, ‘Well, they shouldn't have built there, it's their own damn fault.’ That still represents a human being's life. And it's a test of my ability as a firefighter.

“The safer thing would be to just go down the road and get in the safety zone,” he added. “But if there's any chance, I don't know very many firefighters who are going to just not give it a shot.”

The Lessons Learned Center itself was born of disaster, created after a blowup at the 1994 South Canyon Fire, in Colorado, killed 14 firefighters. Modeled on Army and Marine Corps programs, the center analyzes best practices, picks apart accidents, and suggests new tactics to boost safety and effectiveness. Wildland firefighters have also instituted military-style leadership-development courses, mission rehearsals, and after-action reviews.

Dotson has spent his adult life in wildland fire, and would be more comfortable sleeping on the ground in soot-covered Nomex clothing than sitting in a cubicle, where baseball caps from all the fire crews he's worked with line a shelf. He still works several fires a year, and he has had some close calls. He's lost sight of the flaming front while moving, listened to the growing roar, and wondered if he'd just put himself in a fatal situation. “Just about anybody who has done full-time fire has those instances,” he says.

The challenge, then, has been differentiating between inherent and preventable risk, and creating a culture in which firefighters can learn from others’ miscalculations, rather than live through—or die from—their own.

Along with lessons learned, wildland firefighters have borrowed another military tool: the staff ride. For more than a century, soldiers have toured American and overseas battlefields to see the terrain as the combatants did, and to better understand how decisions were made in the fog of war, with incomplete or conflicting information. “By its very nature, war is a highly complex affair with a virtually infinite number of variables,” William G. Robertson writes in the Army's handbook for staff rides. “Conducted in a dynamic environment by human beings, themselves infinitely variable in personality and intellect, war is played out on the three-dimensional chessboard of terrain.”

So, too, for fire. Robertson helped design and lead the first fire staff ride, for the 1990 Dude Fire in Arizona, which killed six firefighters. Dotson now regularly guides groups through the Dude Fire site.

“People often come in with all of that hindsight bias,” Dotson says. “What were they thinking? Why would they be down in this canyon at that time of day? But when you walk them through the process, and what people knew at the time, by the time they get to that spot they say, 'Oh, wow, now I understand how it is they ended up here. At that moment, they realize, 'This could have been me.'”