By any measure, Americans are flooding into avalanche country at a staggering rate. Some estimates suggest a hundred times more people are venturing into the backcountry now than 25 years ago. And yet the number of people dying in avalanches has held roughly constant, at around 26 per year in the United States. The actual rate of death per person in the mountains has fallen dramatically, a quiet, hard-won victory that rarely makes headlines.
“If the fatality rate had held constant, we would be talking about hundreds of people dying every year,” says Dallas Glass, an avalanche forecaster for the Northwest Avalanche Center with two decades in the field.
This winter has put that progress to its most brutal test.
On the morning of Feb. 17, a wall of snow descended on a group of 15 people on a professionally guided three-day backcountry expedition near Castle Peak in the Sierra Nevada. Nine of them died, including three guides from Blackbird Mountain Guides and six mothers from the San Francisco Bay Area. It was the deadliest avalanche in modern California history.
“We can reduce risk in the mountains,” Glass says. “But we cannot eliminate risk.”
The Castle Peak disaster arrived at a moment when the avalanche safety community had arguably never been better equipped, with stronger technology, sharper forecasting and deeper communication across agencies than at any point in the field’s history. It also arrived as a reminder that the mountain, in the end, does not negotiate.
Understanding how close that gap between “reduce” and “eliminate” has gotten requires understanding how the modern avalanche safety infrastructure actually functions.
Andrew Schauer, the lead forecaster for the Chugach National Forest Avalanche Center in Alaska, describes avalanche forecasting as “a season-long process of monitoring trends, developing predictions, collecting data and revising our mental model.”
Each morning, his team has multiple people in the field digging snow pits, investigating recent avalanches, photographing slope conditions, and monitoring weather. They layer in public observations, remote weather station data tracking precipitation, wind and temperature trends, and forecasts from the National Weather Service.
Ethan Greene, director of the Colorado Avalanche Information Center, which maintains the national database of avalanche fatalities, says the analytical side of the work has evolved dramatically in the last two decades. “Our reliance on and use of numerical modeling has changed a lot and continues to change,” he says. “Twenty years ago, it was mostly focused on weather forecasting, and now there’s more happening in snowpack modeling.”
Just as critically, he adds, the days of hunting through scattered databases are over. “Instead of looking at one particular property in a whole bunch of different places, it’s all kind of in one place right now,” Greene says.
By 7 a.m., a forecast is published. Schauer says his center will “occasionally issue Avalanche Warnings or Special Avalanche Bulletins when conditions are particularly dangerous.” A warning had been issued for the Castle Peak area the morning of Feb. 17.
What happens next depends on a web of people and organizations that most skiers never think about.
Glass has worked in every corner of the avalanche safety world. Over twenty years, he’s been a resort patroller, a highway forecaster, a ski guide and a public forecaster. He’s emphatic that these are not the same job.
“They have different objectives, different time scales, different spatial scales, different users,” he says.
What has changed dramatically, he argues, is how well those different worlds now talk to each other. “We communicate with the ski resorts and the Departments of Transportation and the local guide services, and we are all sharing information in a way that allows us to all be better at our jobs and hopefully provide a better product, which ultimately equals better safety.”
On the resort side, that product includes increasingly sophisticated mitigation technology. Brian Roman, a ski patroller at Winter Park in Colorado, describes an operation that has evolved significantly in his years on patrol. When a rescue call comes in, a rapid-deployment team mobilizes immediately, coordinating in real time with regional agencies and weather services. From a helicopter, the team now has situational awareness that earlier generations of rescuers couldn’t have imagined.
“Teams have a better view to see what’s happening in the terrain adjacent to the avalanche,” Roman says. “They can better see possible safe routes in and out of the area, and can help teams better assess if they can enter the area at all.”
The gear has changed too. “We now have long-range avalanche beacons that attach to the bottom of the helicopter,” Roman says. “This allows teams to search an avalanche without having to expose people to the slope.” Recco technology, a passive reflector system embedded in ski gear and clothing that helps rescuers detect buried victims, has also become a standard part of the rescue toolkit.
The window for survival in an avalanche is brutally narrow: according to the Utah Avalanche Center, 93% of victims pulled out within 15 minutes survive. After 45 minutes, only 20 to 30% do. The speed that new tools enable can be the difference between life and death.
Further up the technology curve, remote avalanche control systems are beginning to change the mitigation equation at some resorts and highway corridors. Utah’s Little Cottonwood Canyon, home to the ski resorts Alta and Snowbird, has become the most densely equipped stretch of avalanche terrain in North America, its ridgelines studded with Wyssen Towers — permanent, remotely triggered structures that hurl explosive charges into avalanche starting zones without putting a single worker on the slope.
Drones are moving into the picture as well. Alaska’s Department of Transportation has been particularly aggressive, using drone-delivered explosives to trigger controlled slides in avalanche paths while keeping workers out of harm’s way.
“Now I don’t have to walk up there and do it by hand,” Glass says. “Or we don’t have to get a helicopter to do it.”
Snowpack modeling, using computers to simulate what is happening inside the layers of snow on a given mountain, has gone from science fiction to operational reality in Glass’ career. He laughs recalling his early skepticism. When someone asked him whether computers would ever model snowpacks, “I actually remember saying, ‘No, it’s so complex. We won’t be modeling this stuff anytime soon,’ ” he says. “And now, even as we’re talking, I have a computer model pulled up of what one computer thinks the snowpack looks like right now.”
Schauer describes the potential as significant. “There are now tools being developed that can simulate the snowpack on the ground,” he says, “estimating the likelihood of an avalanche failing on some layer in the snowpack given the current snowpack structure and predicted weather patterns. That has the potential to dramatically change the way we predict avalanches.”
Glass is careful about how much weight to put on any single tool. “A model is not reality,” he says. “A model is one computer’s opinion of what things may or may not look like.” The old forecaster’s saying still applies: all models are wrong. Some are useful.
Glass’ center has used AI to sort through years of forecasting data, identifying where their predictions have historically been least reliable. “It’s done a pretty good job of helping us identify, ‘Here’s a spot within the forecasting process that y’all seem to have the most uncertainty,’ ” he says. “So it’s helping us narrow down the question, then we can focus on how we answer it.”
Schauer, who has been forecasting professionally for seven years, finds it almost dizzying to inventory what has changed in even that short time. “Snowpack modeling, AI tools to support writing and weather prediction, drone-delivered explosive programs, modern avalanche detection systems, machine learning tools to predict avalanche danger,” he says. “A lot of this is still in its development phase, but many of these tools are currently being implemented at an operational level. It’s impressive how quickly things are improving, and it will be interesting to see what avalanche forecasting looks like ten years from now.
For all the towers and drones and models and beacons, there is a fundamental mismatch between the scale at which humans move through mountains and the scale at which avalanche risk actually operates, and Glass doesn’t think it will ever fully close.
“Some of our forecast zones here in the Northwest are the size of Rhode Island,” he says. “And now I’m going to go run around out there and touch just a couple hundred square yards. Those are two really different scales.”
Greene is candid about where the responsibility of his center ends and the individual’s begins. “What we’re doing is providing an assessment of the hazard and a lot of the characteristics contributing to that hazard,” he says, “but how people actually manage their risk is up to them.”
And that includes the forecasters and patrollers and researchers on the front lines. Roman, on patrol in Colorado for four decades, has watched the mountains shift in real time. “We had rain in December that went to the top of the mountain this season,” he says. “And we have also had the equivalent of a Category 2 hurricane wind hit the mountain for a 12-hour period. It really changes how we assess circumstances for teams about to go on a deployment. We don’t want to end up in a situation where, as the rescuer, we suddenly become the one needing rescue.”
Glass knows that feeling firsthand. The day before our interview, he’d been skiing in the Northwest with two colleagues under conditions he described as tricky. “The snowpack here is a little scary right now,” he says. “We were actively avoiding a lot of avalanche slopes yesterday because we’ve seen the consequences.”
That is the central tension of this work. The tools keep getting better, the communication keeps improving, the fatality rate stays remarkably, stubbornly low. And still, no amount of progress has made the mountains fully safe, as this February’s disaster in the Sierra Nevada made devastatingly clear.
But it’s not enough to keep people like Glass away.
“There’s a level of respect for the mountains that comes with working in this industry,” he says. “Both positive and negative. We love the snow, we love winter, even with all the dangers. We didn’t fall into this line of work. All of us picked it because it’s a passion.”
After this interview, he was back out on skis the very same day.
