Picture the boreal forest, in northern Quebec, on the morning of June 1st. Bears, moose, and wolves roamed, circling its unusually dry lakes, bogs, rivers, and creeks. In the afternoon, after weeks of sunshine, severe thunderstorms suddenly rolled over the forest’s vast tracts of black and white spruce, subalpine fir, jack pine, and aspens. Lightning struck the ground repeatedly. A hundred and twenty-three fires were lit in a single day. The peat and duff of the forest floor began to smolder. Downed logs crackled and spit. The spruces’ lower branches caught flame. As the fires rapidly grew, smoke billowed into the middle and upper layers of the troposphere, three to five miles above the ground. Meanwhile, a large, coastal low-pressure weather system was stuck over Prince Edward Island, because of an unusually wavy, stagnant jet stream, which research has tied to rapid Arctic warming. Winds wheeled counterclockwise, and, a week after the lightning strikes, swept the vaporized remains of boreal forest into the lungs of New Yorkers.
Hundreds of miles from its Quebec origin, the smoke settled among skyscrapers and schools, blew through fences and alleys, hovered outside brownstones and bodegas. For most of the millions who saw copper skies on June 7th, the smoke’s arrival felt sudden and unexpected. In mid-July, smoke arrived again, this time from Alberta, in Western Canada. In New York, on Sunday, July 16th, Governor Kathy Hochul issued a warning, and the next day, the Air Quality Index (A.Q.I.), which measures air pollution, and usually hovers around forty during the summer in New York City, again shot above a hundred.
The urgency of Hochul’s statement—“we are activating emergency notifications on our roads and public-transit systems and making masks available to counties for distribution,” she said—reflected a better understanding not just of the historic scale of Canadian wildfires this year but of the need to better educate the public on ways to protect themselves from bad air. And yet, why no earlier warning? The wildfires that had started a week earlier weren’t hidden, nor were the weather patterns a secret. I asked a spokesman from New York’s Department of Environmental Conservation (D.E.C.), who said that the agency’s “meteorologists use the best available science to predict daily fluctuations in air quality.” Does that mean the best available science really allows for only one day of prediction?
Solving this mystery starts with bureaucracy. The Environmental Protection Agency (E.P.A.) collects and processes air-quality data from state and local monitors—a consequence of the Clean Air Act—then sends the data to the National Oceanic and Atmospheric Administration (NOAA), which feeds them into its supercomputers’ weather and air-quality models. State and local experts—such as New York’s D.E.C. meteorologists—combine their own analyses with their interpretations of NOAA’s guidance as well as other models, then issue twenty-four-hour air-quality forecasts to their jurisdictions. The E.P.A. also distributes these state and local forecasts nationally.
All this amounts to a one-day forecast, entirely separate from the rest of the weather. But isn’t smoke weather, too? “I would consider smoke a meteorological term,” Matt Sitkowski, the lead science editor-in-chief for the Weather Channel, told me. “Smoke can alter visibility and temperature forecasts.” Weather-station reports, he noted, use an official symbol (resembling a skinny chimney) to indicate the phenomenon: “visibility reduced by smoke.” The American Society of Meteorology (A.M.S.) glossary includes smoke, defined as “foreign particulate matter in the atmosphere resulting from combustion processes; a type of lithometeor.” METAR, a weather code used by aviators and meteorologists, lists smoke, giving it the onomatopoeic Latinate abbreviation FU. And yet there isn’t a tab in my phone’s weather app for smoke. I can check what the dew point, humidity, and barometric pressure will be for the next ten days. I can even check visibility. But FU? Phooey. There’s only a pop-up bubble, offering the real-time air-quality index. That doesn’t tell you whether your asthmatic child will be able to play outside over the weekend, or if elderly wedding guests are going to be safe during an outdoor ceremony. The mystery remains. Why don’t we have smoke forecasts?
In November, 2018, Tina Katopodes Chow, an enthusiastic, bespectacled professor of environmental engineering at U.C. Berkeley, was scouring the Web, wondering the same thing. The Camp Fire was scorching across Butte County, choking the Bay Area with smaze. (See the A.M.S. glossary.) Chow finally found a NOAA Web page featuring a smoke-forecast model. But across the top, in bold letters, it read, “Experimental forecast, use at your own risk.” Still, she checked it every day, and one morning, after a week of bad air, it said that the smoke would clear. “I was like, ‘Oh, it’ll be better tomorrow!’ But it wasn’t.” Chow, obsessed with fluid mechanics—the flow of air and water—since she was an undergrad at Harvard, and not one to let a problem go unprobed, e-mailed the NOAA model’s two developers, Eric James and Ravan Ahmadov. “I was like, ‘Oh, I’m just curious, have you done any validation on your model? Because I noticed this big disparity.’ ”
Forecasting smoke is a hard problem, so tricky that no high-resolution, automatically updating smoke forecast existed in the U.S. until December, 2020. A small group of NOAA scientists had floated the idea of forecasting smoke years earlier, but the opportunity, technology, and, truly, the urgency, hadn’t been present. That started to change around 2014, when NOAA and the National Weather Service launched the High-Resolution Rapid Refresh (or HRRR, rhymes with “fur”) atmospheric model—a real-time, hourly updated, state-of-the-art weather-prediction system. Two years later, the agency asked James—who had a master’s in atmospheric science, but hadn’t even started his Ph.D.—and Ahmadov, an expert in air-pollution chemistry, to write experimental code that could incorporate smoke into a parallel version of HRRR. NASA had launched a series of polar-orbiting satellites, which carried instruments that were gaining the capacity to detect fire hot spots from space. “So that was an opportunity,” James said. The new model was called HRRR-Smoke.
James and Ahmadov agreed to collaborate with Chow on a paper that analyzed HRRR-Smoke’s accuracy during the Camp Fire. Usually, a smoke-forecast model is extremely challenging to validate, since there are often multiple sources of smoke and air pollution at any given moment. But the Camp Fire was the only large wildfire burning at the time in the West, making it a prime case for checking the model’s predictions of where smoke would travel. Chow, and her students, took the project on as an unfunded labor of love. Their paper, published in 2022, found that the model performed fairly well, successfully simulating, initially, the Camp Fire smoke plume’s evolution and its 3-D dimensions over complex terrain. But since HRRR-Smoke relies on satellite observations of fires, along with radiation measurements, it needs to be able to see a fire’s hot spots. The Camp Fire, however, produced such dense smoke that it blocked the satellite’s view. The satellite assumed that the smoke was a cloud, that there was no heat, and that the fire was gone. The paper also noted that the model relied only on polar-orbiting satellites and did not incorporate surface-air sensor data.
James and Ahmadov continued debugging, rewriting, and fine-tuning the model’s code, with help from the work of Chow and her students. (They are now attempting to improve the algorithm so that it knows a giant fire won’t disappear overnight.) The summer of 2020 was especially bad. More than half of California’s population experienced unhealthy or hazardous air (an A.Q.I. above a hundred and fifty) for a month or more. “So many people were impacted by smoke all over the country, but especially in the western U.S.,” James said. In December, 2020, when the N.W.S. launched its fourth version of the all-weather HRRR model, it officially incorporated the smoke-forecast code. HRRR-Smoke was now reliably online, refreshing hourly. No one had to push a button.
Still, more than two years have passed, and the model does not make predictions beyond forty-eight hours. “Forecasting smoke is really hard for a number of reasons,” Marshall Burke, a professor at Stanford University, and a leading expert on wildfire-smoke exposure and impacts, told me. “First, you have to get emissions right. You need a good estimate of the amount of junk that a wildfire is lofting into the atmosphere.” That, in part, is a product of vegetation type. HRRR-Smoke assumes that a forest will produce more smoke than grasslands. But there is immense variability in a forest’s biomass. “The form of biomass and the moisture content are key,” Robert Gray, a wildfire ecologist in British Columbia, told me. The structure of the forest floor depends on the quantity of peat, duff, downed logs, and dead trees—perhaps killed by plagues of beetles, which have thrived as global temperatures have risen. Some tree species, such as aspens, are like asbestos, functioning almost as firestops. Others are highly flammable—boreal firefighters call black spruce “gas on a stick.” Smoldering combustion produces the worst forest-fire emissions. But HRRR-Smoke can’t yet incorporate such details.
It also has trouble determining the “injection height”—how high smoke is being sent into the atmosphere. Is it confined near the surface, meaning it will only affect local air quality, or is it forming a towering plume that extends five miles into the atmosphere, allowing for long-range transport? Or is it going to form a pyrocumulonimbus—a meteorological monster that creates its own weather and injects great quantities of smoke into the stratosphere, where it can circle the entire globe? Smoke from the Quebec fires made it to Europe.
For forecasts extending multiple days into the future, the uncertainties grow simply because it’s difficult to know what the fire will do. Will it be less active in two days because rain is on the way? Or is the temperature rising, and humidity falling? James and Ahmadov are starting to play with modulating the amount of smoke that their model predicts based on the weather forecast. What about the efforts of firefighters and water bombers? Will they control a blaze and slow the smoke? Again, too granular. “That’s way beyond our capabilities at this point,” James said.
Once smoke is in a certain area, it’s not difficult for the model to coarsely predict where it will go. “Transport occurs between high- and low-pressure systems,” James said, “so a low-pressure system to the northeast could bring smoke from the northwest, as has been happening a lot this summer.” Still, if the wind-direction forecast is wrong by even a few degrees, that could mean, across hundreds of miles, that the predicted smoke plume will be wildly off.
The largest challenge is predicting near-surface smoke density. For particulates to severely affect our air quality, they must, of course, be near the surface, in the air we breathe. Even if HRRR-Smoke accurately predicts how much carbon a fire is emitting, Sitkowski, from the Weather Channel, told me, “the smoke can move up and down in the atmosphere as it moves downstream from a wildfire.” This variability in the air column, with patches of dense smoke travelling at different wind speeds and directions, can have a large impact on the resulting air quality in distant locations. In terms of knowing whether the Air Quality Index bubble will say fifty or a hundred and fifty, whether the sky will be simply hazy or apocalyptically barbequed, the forecasts have a ways to go. Smoke forecasts today are about as reliable as weather forecasts were forty-five years ago, Jeff Masters, the co-founder of the site Weather Underground, told me. So we must make do, for now, surrendering to the vagaries of the elements. And, on smoky days, people need to be reminded to close their windows, wear masks, and stay indoors.
The Canadian boreal forest is, like a phoenix, reborn from fire. Several of its common species, like black spruce, and jack and lodgepole pines, cannot procreate without fire. Blazing heat is the only thing that melts the resin holding their serotinous cones closed, releasing their seeds to the forest floor. Serotiny—the name for this process—is evolution’s answer to fire. We too have evolved with fire. Our civilization wouldn’t exist without fire leading us out of the dark, keeping us warm, releasing our food’s nutrients and energy, powering our grand inventions. But all that combustion—particularly the past two centuries of burning fossil fuels—has sent dangerous amounts of carbon dioxide into the atmosphere, fuelling the climate crisis, sparking the worst wildfires our modern society has seen. In the past ten years, in Quebec alone, an average of 554,968 acres had burned by early August. But, this year, 12,670,167 acres have burned. Wildfires are expected to continue burning through September, and possibly later, meaning we can expect more smoky days this summer.
The climate crisis is bringing a new kind of weather, for which smoke forecasts are necessary. “This year is pretty crazy. We’ve never seen anything like it,” Masters told me. “But it’s not gonna be the last one we see like this. With the huge impacts these sorts of events have on our society, there’s gonna be a lot of pressure, I think, to get better smoke forecasts.” (Private companies such as IQAir, are also advancing the field.) As good and necessary as fire may be for the forest, the smoke it emits is never good for us. “It’s one of the worst things you can breathe,” Masters said. Multiday forecasts indicating that smoke is travelling in a given direction might be imprecise, and might not tell us the exact air-quality index a week from now, but they’d still be a good addition to weather broadcasts and apps. If anything, they might serve to remind people about the climate crisis—not just that there are increasingly long and intense wildfire seasons but that everyone will be impacted. It’s no longer a problem only for people who choose to live in wildfire country, or in the West. It’s global, it’s inescapable, it’s everyone’s weather. ♦
