The West’s Infernos Are Melting Our Sense of How Fire Works

42,000-foot plumes of ash. 143-mph firenadoes. 1,500-degree heat. These wildfires are a new kind of hell on earth, and scientists are racing to learn its rules.
Firefighters march into El Capitan Canyon in California
“We’re running 400 to 500 fires a year. In the heat of summer, five or six a day—and most you’ll never see,” says CalFire chief Brian Estes.Photograph: Marcus Yam/Los Angeles Times

On the windy, hot day of July 26, 2018, as record 113-­degree temperatures baked Redding, California, in the northern Sacramento Valley, Eric Knapp toiled in an air-conditioned government office. After work, he planned to meet his wife and 3-year-old daughter, and some family friends, for dinner. Slender and fair-skinned with a gentle smile, Knapp is a research ecologist for the US Forest Service. He was well aware that, three days earlier, in coastal mountains west of town, a wildfire had started when a trailer got a flat tire and the metal wheel rim scraped the asphalt, sending sparks into dry brush.

Like the vast majority of wildfires, this one, called the Carr Fire, burned initially as a wide but shallow band of flames advancing slowly, like a battalion of infantrymen marching shoulder to shoulder, and left behind charred grass and lightly scorched trees. The Carr Fire was also typical in that it moved according to the dictates of wind, ground slope, and flammable fuels—southeast around a lake, then up a hill, in part because heat rises. Early on that particular morning, the fire had crested a rise above Redding and, with a northwesterly breeze at its back, crawled downhill toward town.

November 2020. Subscribe to WIRED.

Photograph: Kevin Cooley

Knapp was finishing up for the day when his friend Talitha Derksen, a wildlife biologist with a daughter close in age to Knapp’s own, sent a text saying that her neighborhood might have to be evacuated. One of the agencies tasked with that judgment call, the California Department of Forestry and Fire Protection—aka CalFire—is one of the world’s largest and most effective wildland firefighting organizations. CalFire bases evacuation recommendations on predictions of where, and how quickly, a flame front will move next. That day, the fire appeared likely to reach the floor of the Sacramento Valley at a subdivision called Land Park, about a mile northwest of Derksen’s house.

Knapp and the others changed plans: They’d meet at Derksen’s, order pizza, and help her get ready to leave in case it came to that. Knapp stopped at his house to grab fireproof Nomex clothing. As he headed to Derksen’s, he considered dropping by the office again to pick up his hard hat and emergency fire shelter—a sort of fire-resistant pup tent—but decided he was unlikely to need them.

As he turned onto Derksen’s street, the flame front was a couple of miles away and hidden by trees, but Knapp could see the smoke rising in a straight and tall plume that turned the sun orange. When he arrived at Derksen’s house, she was already packing bags. Knapp, to be sure he knew what they were dealing with, jogged out the nearby Sacramento River Trail for a view. Upriver, on the far bank, he could see red flames torching gray pines and scrubby oaks.

Knapp was shooting photos when he noticed something odd: The wind where he stood blew out of the south, into the fire, but the flame front still moved the other way, driven by that northwesterly at its back. Then he saw something else: Portions of the smoke plume swirled in different directions, as if beginning to rotate.

Knapp knew this could signal a once rare and dangerous phenomenon known as plume-driven fire, in which a fire’s own convective column of rising heat becomes hot enough and big enough to redirect wind and weather in ways that can make the fire burn much hotter and, with little warning, spread fast enough to trap people as they flee.

As Knapp ran back down the trail, he passed neighbors walking and recommended they turn around. But even he had no idea how much peril they were all in. At the house, as Derksen left, Knapp and others hosed down the roof and rain gutters and cleared the yard of flammable material like cardboard boxes and lawn furniture. Knapp was the last person there, spraying water on the fence and yard.

Even as Knapp cranked the spigot, the swirling smoke he’d seen was fast accelerating, transforming much of the Carr Fire’s enormous lower plume into the biggest fire tornado ever observed, a whirling vortex of flame 17,000 feet tall and rotating at 143 mph with the destructive force of an EF-3 tornado, the kind that erases entire towns in Oklahoma.

While Knapp blithely sprayed water around Derksen’s house, that fire tornado—hidden from him by all the smoke in the air—leaped across the Sacramento River, touched down in Land Park, snapped high-tension power lines, uprooted trees, wrapped steel pipes around utility poles, and obliterated hundreds of homes, igniting and shredding them and hurling their burning debris up to altitudes at which commercial passenger jets fly.

Not far from where Knapp stood, CalFire captain Shawn Raley was evacuating a woman and her daughter in his truck when all the windows imploded, showering them with shattered glass. Close by, a 37-year-old fire inspector named J. J. Stoke radioed Mayday moments before the tornado lifted his 5,000-pound Ford F-150 off the asphalt and flipped it repeatedly down Buena­ventura Boulevard, killing him. Three other CalFire workers were driving bulldozers on that same boulevard when their windows also shattered. One of the 25-ton vehicles got spun around and dropped on top of a truck driven by a retired police officer, who then jumped out and crouched behind the bulldozer’s blade while his truck caught fire.

That’s about when flaming debris that had been sucked into the Carr Fire’s plume of smoke drifted out of the updraft column into what fire meteorologists call the fallout zone, which is exactly what it sounds like. Knapp couldn’t possibly have seen that happening; it was tens of thousands of feet above him. Nor could he see the flaming remnants of homes and trees hurtling downward like firebombs, smashing onto roofs and igniting dozens of houses. While looking up into the black whirling darkness overhead, Knapp, who still thought the Carr Fire was advancing with the slow predictability of a classic shallow flame front, watched embers rain down on the bark chips upon which he stood, lighting them afire. At the same moment, with the very ground at his feet aflame, Knapp felt an even more powerful pulse of heat.

That fire tornado, and the blaze that raged for weeks after, ultimately destroyed more than a thousand homes and buildings, killed eight people, and scorched nearly a quarter-million acres. Yet it was neither the biggest California fire of 2018, nor the most destructive, nor even the only one to behave in frighteningly anomalous ways. The Mendocino Complex fire, about 100 miles south of the Carr, which started the day after Knapp lingered unwittingly below a tornado, was also briefly plume-driven and ultimately burned almost 460,000 acres in what was then the largest California wildfire of all time. In early November, the Woolsey Fire near Malibu destroyed 1,643 structures while ripping trees and power-line posts out of the ground with a force suggestive of yet another fire tornado. The infamous Camp Fire, likewise in November, burned 70,000 acres in 24 hours—about a football field a second, for a while—and created an urban firestorm that destroyed more than 18,000 structures and killed 85 people, mostly in the town of Paradise, generating billions of dollars in insurance claims and bankrupting the state’s largest utility, PG&E.

By the time California’s 2018 fire season was over, it had burned more than 1.6 million acres to become the most destructive on record—a title it maintained for slightly less than 20 months, when it was overtaken not by the 2020 fire season but by a mere four weeks in late summer 2020, during which an estimated 3 million acres burned. But that’s not the truly worrisome part. In making sense of Western wildfires, total acres burned are far less important than the increasingly capricious violence of our most extreme blazes. It is as if we’ve crossed some threshold of climate and fire fuel into an era of uncontrollable conflagrations.

“Not only is the size and severity increasing, but the nature of fire is changing,” says David Saah, director of Pyregence, a group of fire-science labs and researchers collaborating on the problem. Still more concerning, given the trend toward fires dramatically more catastrophic than anything we’ve yet seen: The physics of large-scale wildfires remain so poorly understood that fire-modeling software is often effectively powerless to predict where they will next occur, much less how they will unfold once they do. If there is any good news, it is that, as Saah puts it, “the science for a lot of this stuff is under way.”

Eric Knapp has worked for the US Forest Service in Northern California for 16 years.

Photograph: Andres Gonzalez

About a year after the Carr Fire, on a bright June day in 2019, Brandon Collins, a big-jawed fire-­science researcher at the University of California, drove a white pickup down a cedar-scented mountain road into the Blodgett Experimental Forest, a 4,000-acre university property near Lake Tahoe where he studies the effect of forest management practices on wildfire risk. All of those practices begin with the inescapable fact that California is flammable. It is hard for us moderns to accept—conditioned, as we are, by Smokey Bear—but fire is every bit as natural and inevitable in the American West as flooding in the Mississippi River Basin and hurricanes in Florida. Fire is not only guaranteed by climate and ecology; it is vital to the health of many ecosystems. The 20th century, in fact, during which large wildfires were far less common in the West than they are today, should properly be seen as the unnatural outlier. Prior to that, and especially before Anglo-American conquest, wildfire burned an estimated 6 million to 13 million acres each year in California, according to one study, far more than even the current record-­setting season.

Most of those frequent fires past were different, though, in a critical way: Burning with a shallow flame front, like the early stages of the Carr Fire, they ripped through grass, pine duff, and fallen branches—so-called surface fuels—on the forest floor instead of torching whole trees and leaping crown-to-crown as our biggest fires do today. Those regular surface fires generally kept overall fuel loads so low that each subsequent fire could only do the same—scorch out the understory without harming mature trees. Over time, this sustained forests of old-growth conifers, oak, and madrone widely spaced on carpets of grass and shrubs, which in turn made terrific forage for deer. Indigenous people lit wildfires all over the American West for millennia to manage land for this outcome—with such success that, in the late 19th century, Anglo-American ranchers and even lumbermen adopted the practice.

Collins, to show me what that looked like, stopped the truck at a section of the Blodgett Forest that had been managed for 16 years in the old way, with regular fire. We have all experienced varied responses to landscape, from apprehension at a bleak desert or dark cave to calm in a tropical cove. I can report that a forest, when allowed to burn the way it evolved to burn, feels wonderful, a sun-dappled gallery of enormous sugar pine, Douglas fir, and black oak shading meadow-like ground at once sheltered from weather but open enough to move freely.

The Forest Service, which currently controls about 20 million acres of California, put a well-meaning end to this kind of land management almost from the founding of the agency in 1905. Seeing forest in near-term dollar signs—lumber, watershed, game—and dismissing the idea that wildfire played any positive ecological role, the Forest Service learned to snuff every blaze in every forest as quickly as possible. The wrongheadedness of this approach became obvious to the agency itself by the 1940s, when its researchers began to catch on to the fact that the longer a forest goes without fire, the more fuel will pile up and the worse the blaze will be.

That insight made it into official Forest Service policy by the 1970s, encouraging regional employees to use deliberate controlled burns as a means of keeping fuel loads low. By that point, unfortunately, lumber and paper companies had come around to the burning-­is-bad position, as had civilians who disliked smoky air, enjoyed recreation in national forests, and thought of fire in purely destructive terms. Combined with issues of legal liability—who pays for damage to private property caused by prescribed burns on public land?—it all made Forest Service officials understandably reluctant to follow through with any particular prescribed burn. Private property owners, who control California’s other 13 million acres of forest, were (and still are) even less motivated to light their own land ablaze, much less to tolerate a neighbor doing so. CalFire, meanwhile, tasked with responding to every fire on 31 million acres of nonfederal land inside state borders, has, compared to the Forest Service, almost no fuel-management authority. CalFire’s straight­forward mandate, for which it spends upward of $2 billion a year and operates more than 700 fire engines and 75 aircraft, is to extinguish every blaze, fast—a job it does extraordinarily well on about 6,400 wildland fires annually.

CalFire chief Brian Estes, who commands firefighting operations for just three of California’s 58 counties, says, “We’re running 400 to 500 fires a year. In the heat of summer, five or six a day—and most you’ll never see. Anytime I have a 911 dispatch to a vegetation fire”—a grass fire, say, on somebody’s lawn—“you’re going to get seven engines, a battalion chief, two bulldozers, two air tankers, an air attack, and two hand crews. They’re going to roll out the barn. But if you do that for a hundred years, and you don’t allow people to do prescribed fire, the fuel just gets more and more dense.”

Collins showed me a graphic example at our next stop, a patch of forest that hadn’t been logged or burned in more than 100 years. Crowded tight with young trees among the big, old ones, it was piled deep not only with surface fuels like pine duff and leaves but so-called ladder fuels, the big fallen branches and shrubs that help surface fire leap up into the crowns and spread more quickly up high. That patch of forest also felt intuitively awful: dark, shadowy, mazelike, and cavernous, the nightmare forest of an old fairy tale.

Flammable as it looked, even forests mismanaged like that patch burned until recently in the historical way, at low severity along the forest floor. As a result, the entire field of wildfire science—including every modeling tool with which firefighters make life-or-death decisions and society structures itself in fire-prone areas—is based on that kind of fire behavior. The core mathematics of this science date to the early 1970s, when a Forest Service researcher named Richard Rothermel used small laboratory fires to produce equations expressing the relationship between wind speed, ground slope, and how fast a fire spreads. Rothermel knew his approach worked properly only for wildfire in light surface fuel like that in his lab—and failed to capture what happened when blazes got into treetops and jumped crown-to-crown. But these so-called Rothermel spread equations were applicable to so many wildfires that the Forest Service quickly developed paper-and-pencil ways for firefighters to plug in numbers for wind and slope angle and make reasonable guesses about how fast and in which direction a fire might spread—in a single heading, on a straight line. Eventually that modeling framework was run on cumbersome supercomputers, then on handheld calculators. In the early 1990s, PC-based software finally allowed firefighters to predict fire spread in two dimensions on a map.

That software, created by a Forest Service scientist named Mark Finney, was severely limited by a lack of mapping and fire-fuel data. It didn’t do much good, in other words, if you couldn’t load it with topographical maps and vegetation data for the fire you needed to fight. Over time, though, other researchers compiled these data sets on their own and shared them with one another until, in 2009, they were available for the entire US. Finney’s software now does such a good job predicting fire spread in light ground fuels that it has become the industry standard, used thousands of times each year by firefighters nationwide. And versions that allow simulation of possible future fires are also used by land managers eager to prevent them.

As early as 1994, though, Finney could see that the contemporary modeling framework had more serious limitations. In central Washington state that year, a large and unusual blaze called the Tyee Creek Fire behaved in ways utterly outside the bounds of Finney’s model. Instead of burning with a shallow flame front that followed wind and terrain, Finney says, “the fire basically spread in three directions, all about the same rate, every day in the afternoon”—as if the wind had somehow blown 360 degrees outward from the center of the fire.

The Tyee Creek Fire also kept its huge central area ablaze for days on end, a somewhat speculative phenomenon known as mass fire. “It would just kind of bulge out and put up a giant plume, and then just expand, expand, expand, every day,” Finney says. “I remember thinking, ‘Wow, this is so much beyond anything that we are able to model now, it’d be silly to even try.’”

Finney realized that no amount of modification to the Rothermel spread equations could ever make them account for a fire like Tyee Creek. Not only had they been developed around small lab fires, but 20 years’ experience using them had focused on shallow flame fronts moving quickly through light fuel, with no accounting for slow-burning heavy fuels ignited along the way, much less feedback between ground fire and the immediate atmosphere. Put another way, as Finney recalls saying to a colleague at the time, “the truth is, we have no idea how this stuff really works.”

To chip away at the problem, starting in the early 2000s, Finney went back to first principles, assuming nothing. He lit new experimental fires at a research station in Missoula, Montana, and revisited basic questions like whether wildfire spreads through simple heat radiation—conventional wisdom at the time—or through direct contact with flames.

“It’s a very hard problem,” Finney says, “because if you’ve ever sat around a campfire and watched it, the thing that keeps you transfixed is that the flames are always dancing around. How do you characterize such a nonsteady phenomenon in order to model it?” Light ground fuels, Finney learned, caught fire strictly through convection, and typically consumed themselves in 30 seconds or less at about 1,500 degrees. Heavy fuel like logs and fallen trees would smolder or glow with embers for hours or days, releasing heat all that time. They tended to burst into flaming combustion, quickly releasing their stored energy, under sustained wind. Like when you blow on a campfire.

While conducting that basic research, Finney happened across a book titled Fire and the Air War, about Allied bombing campaigns during World War II. He learned that British and American commanders, while pressing the war against the Germans and Japanese, had discovered that it was easier to burn cities down than to blow them up. The trick lay in first knocking the buildings over, then lighting them on fire. The Royal Air Force did just that to the German city of Dresden in 1945. Military intelligence officers studied recon photographs to identify older districts built largely of wood, then saturation-bombed them with high explosives. A second wave of aircraft hit those same districts with more than 2 million pounds of magnesium-­thermite incendiary bombs. This had the desired effect of lighting the city afire, but it also triggered something unexpected. Shortly after all those buildings got to burning—30 minutes after, as it happened—a single giant plume of heat and smoke rose over Dresden, and took on a shape similar to a giant thunderstorm.

The Dresden firestorm famously produced hurricane-­force winds powerful enough to uproot giant trees and snap them in half, suck up roof gables and furniture, and send countless humans flying like fallen leaves into the whirling fire tornado. Before it was done, that firestorm wholly incinerated several square miles of city.

Finney also unearthed a stack of obscure research reports, published during the Cold War, that analyzed the Dresden firestorm and a similar one above Hiroshima after the detonation of the atomic bomb (yet again, roughly 30 minutes after). One of these reports, commissioned by the Defense Nuclear Agency, compared bombing-induced firestorms with those generated by natural disasters, like during a 1923 Tokyo earthquake when a flaming cyclone lifted floating boats up off a river—and the river water itself nearly 50 feet into the air—before hitting a military depot in which 40,000 people had taken refuge, killing nearly all of them.

Yet another of these reports, titled Mass Fire and Fire Behavior and published by the Forest Service in 1964, looked at what might happen if a national forest got hit by a nuclear weapon. Detonation of a multi­megaton warhead, the authors calculated, could simultaneously ignite as much as 1,200 square miles and cause a firestorm that ultimately burned out 10,000 square miles. The researchers involved were well aware that naturally occurring wildfires could, at least theoretically, cause the same level of damage. This was particularly frightening in light of the population boom in fire-prone wildlands out West. To better understand the risk, the Forest Service conducted a series of gigantic live tests in which, on federal land in Northern California, they laid out street grids similar to those in both urban and suburban neighborhoods. Each home site in these neighborhoods was piled with wildland fuel—juniper and pinyon trees, in one case—and set ablaze. This not only produced small tornadoes; it also confirmed that mass fires of wildland fuel burn in ways remarkably similar to the firestorms of World War II.

As he read all this stuff, Finney told me, something clicked. “I realized, ‘Oh, my gosh, we’re creating the conditions for mass fires,’” he says. “These fires aren’t just big because of, say, climate change or some accident. They’re big because we have a landscape full of long-burning heavy fuels, just like cities.”

A fire smolders in August in Healdsburg, California, part of a complex of blazes that burned nearly 400,000 acres in the area.

Photograph: Ian Bates

The key ingredient in a firestorm, whether in a wartime bombing campaign, a plume-driven fire like the Carr, or a wind-driven fire like the one that destroyed Paradise, appears to be the simultaneous burning of many small fires in a combination of light and heavy fuels over a large area with light ambient wind. As that broad area continues to burn with glowing and smoldering embers over many hours, the separate convective columns of all those many little fires begin to join into a single, giant plume. As the hot air in that plume rises, something has to replace the air at its base—more air, that is, sucked in from all directions. This can create a 360-degree field of wind howling directly into the blaze with the same effect as vents on a forge, oxygenating the fire and pushing temperatures high enough to flip even heavy fuels (giant construction timbers, mature trees) into full-blown flaming combustion. Those heavy fuels then pump still more heat into the convective column, creating a feedback loop: The column rises ever faster and sucks in more wind, as if the fire has found a way to stoke itself.

That seems to be what happened during the Carr Fire. According to Neil Lareau, an atmospheric physicist at the University of Nevada, a weather balloon released on the morning of July 26 detected a lid of warm air, known as an inversion layer, several thousand feet above the Sacramento Valley.

While Knapp settled into work at his office, this inversion layer trapped the Carr Fire’s heat plume near the ground. But as the day wore on, the heat plume forced its way to higher altitudes, steadily cooling.

About the time Knapp jogged out to the river to get a view of the fire, this plume reached 18,000 feet, high enough for water vapor, carried aloft, to condense into liquid cloud droplets, spawning a pyrocumulonimbus, or fire-generated rotating thundercloud. That process of condensing hot vapor or steam into liquid releases heat; you can think of it as the inverse of the cooling effect caused by evaporation, like we’ve all felt emerging from a swimming pool into wind. In the case of the fire plume, this condensation of water vapor into liquid cloud droplets delivers new heat to the plume itself, causing it to rise even faster and higher.

Back down at ground level, meanwhile, the rising plume pulled in new air by sucking at those two pre­existing wind fields, the ones that Knapp noticed blowing into the fire, out of the south and northwest, respectively. Blowing toward one another at an askew angle and intersecting at the flame front, those two winds wrapped around each other and drew in flame to create a whirling vortex of fire. The higher the plume rose, the faster the vortex spun. Lareau likened it to a figure skater: “The skater starts a slow rotation with their arms out wide, and they draw their arms inward and maybe put them up over their head, and they suddenly start rotating very, very quickly.”

As the fire tornado splintered homes and launched flaming debris into the sky over Knapp, it set up one of the most dangerous of plume-driven phenomena—the raining of firebrands. A classic surface-driven wildfire ignites only the immediate area crossed by the fire’s own shallow flame front; falling firebrands, by contrast, allow plume-driven fires to propagate miles from the core burn, as if launching incendiary bombs to ignite entirely new mass fires like the one that burst up around Knapp.

Fires of this type can be nearly impossible to suppress, because they can move too quickly for firefighters to get out of harm’s way and burn too hot to extinguish, but also because so many people in the West have settled in places where these fires are increasingly occurring—the wildland urban interface, or WUI (pronounced woo-ee), exurban sprawl in California’s many mountain ranges.

“We have crammed millions and millions of people and roads and homes and yards into this highly volatile Mediterranean climate,” says CalFire chief Estes, who grew up in the town of Paradise. Worse, Estes says, enormous numbers of these people have gravitated to quaint old Gold Rush towns that, like Paradise, happen to sit atop river and creek drainages where wildfire fuel accumulates and winds tend to blow especially hard.

“If you roll a map of California out,” Estes says, “I can give you 150 communities that have exactly the same combination of factors as Paradise.”

In every one of those communities, according to Estes, “when we have disastrous fires, we have to get those people out, and that makes it so much more complex, I can’t even tell you.” For at least the first 16 hours of the Camp Fire in his hometown, Estes adds, firefighters were mostly just pulling residents out of homes and using bulldozers to clear roads blocked with cars abandoned by drivers who’d gotten trapped in traffic and fled on foot. During that whole period, Estes says, “there was not a single fire engine fighting that fire. They were all trying to rescue people.”

Knapp took this photo of the Carr Fire in Redding as it began to twist itself into what would become one of the strongest fire tornadoes on record.

Photograph: Eric Knapp

The final elephant in the room, of course, is climate change—and the likelihood that it is already pushing even our current nightmares toward holocausts beyond imagining. Knapp, Finney, Collins, and several other researchers (most of whom are now involved in Pyregence, the fire-science consortium) have already identified an especially frightening way in which that might happen. Current climate-change patterns suggest we are headed for ever-less winter snowfall in the West, with hotter summers, ever-­worsening droughts, and ever-more acute spells of extreme fire weather—long periods of dry heat that bake moisture out of grass and trees, combined with winds ferocious enough to whip even a small spark into a conflagration. The collapse of commercial logging, meanwhile, mostly due to environmental regulation, has combined with our collective intolerance for prescribed burns (nobody likes smoky air) to let forests grow unnaturally dense with young trees. More trees means more roots competing for the same underground water. During the drought of 2011 to 2016 in California, that competition, with help from bark beetles, killed a breathtaking 150 million trees in the largest mass die-off ever recorded in the United States.

Nobody knows how all those dead trees will affect wildfire. Initial research suggested that tree mortality would moderately increase risk of severe fire for several years, as dry needles helped fire spread crown-to-crown instead of just along the forest floor. Once all those pine needles fell, which appears to be happening now, risk of severe fire was expected to decrease for a while. The scariest part was thought to lie at least 10 or 15 years in the future, when all 150 million dead trees—an estimated 95 million bone-dry tons of firewood—were expected to fall on top of an already deep kindling pile of fine conifer duff heaped with small twigs and ever larger tree branches. At that point, we would have collectively prepared the entire western slope of the Sierra Nevada, through more than a century’s work with taxpayer dollars ostensibly aimed at preserving wilderness and the economic value of wood, to incinerate in the greatest firestorm ever seen by human beings.

Neither this horrifying long-term risk nor the overall trend toward increasingly destructive fires has been lost on the California state government, which is how Pyregence came into being. Coordinated by Saah from the University of San Francisco, Pyregence has set out to create an entirely new software ecosystem, including for mass fires and plume-driven megafires. The idea is partly to help firefighters respond and partly to help the rest of us make smart decisions about urban planning and fuel treatments like prescribed burns. The overall challenge is too big and urgent for any single lab, so Pyregence has divided it up into a sort of distributed Manhattan Project of collaborative fire-modeling research.

Finney has joined a Pyregence working group studying the behavior of large woody fuels piled deep, like in our National Forests out West. Field researchers have gone out and taken detailed measurements of wildfire fuel beds, while Finney, back in Montana, has commissioned the construction of a new burn chamber the size of a grain silo. Once complete, that chamber will let him replicate wildfire fuel beds by piling logs and other material as much as a few feet deep. He will then ignite them, hit them with wind and moisture, and quantify their burn rate and energy-­release rate—what he calls the “heat-engine part of mass fires.”

“Really what we’re looking for,” Finney says, “is how these things transition to flaming. Instead of just smoldering on the forest floor, how do they become actively involved in these large fires?”

If all goes well, Finney’s working group will eventually code three-dimensional digital simulations of various wildland fuel beds—digital cubes, in essence, not unlike Minecraft voxels—that can be stacked and arranged in infinite variation across landscapes generated by GIS mapping data.

Yet another group, led by Janice Coen of the National Center for Atmospheric Research, has split California into eight fire regions and studied severe past blazes in each. By analyzing how and when those blazes spread, Coen’s team has identified days when fire grew at exceptional speeds, then combed weather-station and satellite data for two related sets of data: local weather conditions like hot local winds that are consistently associated with extreme fire growth; and large-scale weather patterns 500 miles wide and more that are consistently associated with those local conditions. The hope is to create a meteorological early warning system for extreme fire weather in every region. Coen has already run proof-of-concept testing with an experimental model called Coupled Atmosphere Wildland Fire Environment, or CAWFE (pronounced coffee). An atmospheric weather simulator coupled with fire-spread algorithms, CAWFE has allowed Coen to plug in the precise local and large-scale weather that occurred around past events like, say, the Carr Fire. She has even triggered fire ignition at the exact point where the Carr Fire began and watched the fire tornado spin up on its own. The hope, according to Saah, who serves also as managing principal of the environmental think tank Spatial Informatics Group, is someday to supplement the fire-spread component of CAWFE with a fuel model like the one Finney hopes to produce, accounting for the enormous additional heat contributed by long-­burning heavy fuels under open flaming combustion. By then feeding in live real-time weather data, Pyregence should someday be able to produce, for the first time, accurate near-term predictions of extreme plume-driven mass fire all over California.

At UC Merced, meanwhile, a climate researcher named LeRoy Westerling leads a Pyregence group tackling the crucial long-term problem of how to prevent apocalyptic fires in the future. This becomes especially pressing, says Westerling, when you consider that every future fire season in the American West is likely to be worse than the last, on average. “How do you adapt to that? It’s not just California,” he says. “It would be the whole West Coast and the Rockies and parts of Canada and Alaska all going off on a regular basis. So just the magnitude of managing fires over that geographical scale simultaneously is staggering, right down to the psychological impact of living with that.” By way of solution, Westerling’s group is even now developing what Saah calls “statistical machine-learning monstrosities”—big simulation engines that will allow researchers to run various long-range climate scenarios in which ground fuel, regular fire, and even land-management practices like prescribed burns interact with each other. In an ideal world, this will let policymakers ask questions like, If we get stuck with doomsday-level climate change but do lots of smart prescribed burning while allowing only fire-hardened home construction in the mountains, what might the firestorms of 50 years from now look like?

The aftermath of a blaze started by a dry thunderstorm in August, near Napa, California.

Photographs: Ian Bates

California’s catastrophic 2020 wildfire season kicked off midway through the hottest August on record with a dry thunderstorm in which 12,000 lighting strikes ignited hundreds of fires over the course of a week. Three were among the state’s largest of all time by early September, when hard northeasterly winds blew them into an entirely new realm of superlatives. Near the Blodgett Forest, those northeasterlies pushed the relatively small Bear Fire into a giant pyrocumulonimbus storm; in the space of 24 hours, it ripped across 230,000 acres, one of the largest single-­day fire spreads ever observed, destroying hundreds of structures and killing 15 people. Across the Sacramento Valley, those same winds fused other wildfires into the gargantuan August Complex, the state’s all-time biggest fire by nearly a factor of two, at more than 850,000 acres.

Still more astonishing is the Creek Fire, which ignited on September 4 in an area with a lot of dead trees in the southern Sierra Nevada. By the very next day, a huge pyrocumulonimbus formed and helped burn 115,000 acres through so many popular lakes and cabins and campgrounds—somehow tearing gigantic live trees out of the ground and hurling them across roads—that more than 360 people and 16 dogs got trapped on the shores of the Mammoth Pool Reservoir. That, in turn, forced the California National Guard to rescue hundreds of people overnight in military helicopters, something that had never been done before.

“That’s a weird beast,” says Saah of the Creek Fire. “In our research group, there’s so much conversation around that specific fire, because it’s doing things that are just out of the norm.” Among the most peculiar was the fact that energy release across the Creek Fire’s vast center remained just as hot and high as along its periphery. This classic hallmark of mass fire may well mean that the scary part—the future in which 150 million dead trees go up in flames—is already upon us. “If you look at satellite images of the Creek Fire,” says Saah, “it looks like a nuclear bomb went off. It’s crazy—just its behavior and its intensity and how fast it grew.”

Lareau, the atmospheric physicist, was dumbfounded too. “I’m just kind of at a loss for words,” he says. “Having looked at plenty of big fires that produced big pyrocumulonimbus clouds in the Sierra, I mean, this thing just blows everything out of the water. Instead of the cloud going to 40,000 feet, it’s going to more than 50,000 feet. It’s producing long-lived tornado-­strain vortices for periods of hours.”

Those vortices knocked huge live trees to the ground in circular patterns, some inside a campground and others onto roads, blocking escape routes. The fire’s plume also generated lightning on and off for 12 hours, and another unusual behavior known as plume collapse in which all that hot rising air, upon cooling up high, suddenly reverses direction into a powerful downdraft toward the middle of the blaze, forcing fire outward in all directions, igniting huge new swaths of land.

“It stands out to me as potentially one of the most intense firestorms we’ve ever seen,” Lareau says. “I think it’s in many ways a way more intense fire than the Carr Fire was.”

For Knapp, of course, no fire is ever likely to be more intense than the Carr. Especially that moment when he found himself in the middle of a blazing patch of bark chips while burning firebrands ignited homes all around. At that point, Knapp told me recently, “I just had to recognize I didn’t have all my safety equipment, I wasn’t attached to any firefighting resource”—there was nobody to call for help—“and I had a family on the other side of town.”

Heading for his car, Knapp drove right into a traffic jam of terrified neighbors. Slowly, with that tornado roaring overhead and their own homes afire all around, they inched their way down the road to safety. The next day, Knapp drove back to look at Derksen’s house. More than 60 homes in her neighborhood had been destroyed overnight, including the one right next door. A single ember made it through a ground-level screen vent at Derksen’s place, slowly igniting a floor board. It appeared that, before this fire could burn out of control, passing firefighters had snuffed it out.

The scene was “intense and sad,” as Knapp put it, not least because he and everyone else—unable to see the forest for the trees—had been so unaware of how much danger they were in.


Daniel Duane (@Danielduane) is the author of six books. He’s at work on the next, about the Sierra Nevada. His last story for WIRED, about San Francisco’s response to the pandemic, was in issue 28.09.

Marcus Yam photograph: Copyright© 2017. Los Angeles Times. Used with Permission.

Cover photograph by Kevin Cooley/Redux

Lettering by Cymone Wilder

This article appears in the November issue. Subscribe now.

Let us know what you think about this article. Submit a letter to the editor at mail@wired.com.


If you buy something using links in our stories, we may earn a commission. This helps support our journalism. Learn more.


More Great WIRED Stories