NEAR POLEBRIDGE — A steady breeze teased Erich Peitzsch’s drone last Friday. But high above Flathead National Forest, the tiny aircraft held steady.
“It’s really stable given these winds,” remarked Peitzsch, a physical scientist with the U.S. Geological Survey, steering it over a low cliff beside the Flathead River’s North Fork.
As he watched the drone’s video feed on his controller, physical scientist Caitlyn Florentine watched for nearby birds, and physical science technician Zachary Miller kept his eyes on the drone itself, a black speck against the blue.
All around them towered Glacier Country’s rounded mountains, terrain that can shed its snowpack in sudden, sometimes deadly avalanches. Forecasters spend each winter and spring assessing the risk of these events, and these scientists aim to augment their skills with drones. Peitzsch and his colleagues are using them to track one of avalanche forecasting’s key metrics: snow depth.
As one winter storm after another hits a slope, the snowpack takes on layers of different weight, strength and thickness. Avalanche forecasters study these layers — identifying weak ones that might crumble, and watching the snow load building up on top of them — to determine avalanche risk and let backcountry travelers know what to watch for.
The work is low-tech. Forecasters ski into the backcountry, digging pits to examine the cross-section of layers, and sometimes starting avalanches themselves to feel how the snow gives way. “A lot of our job is interpolating beyond single point data,” explained Zach Guy, director of the Flathead Avalanche Center.
But these isolated spots give an imperfect view of the entire slope, said Peitzsch.
“Things are so variable across the slope, it's really sort of hard to understand how shallow or how deep that weak layer is, and how much new snow has fallen upon that weak layer,” based on just one or a few test sites, he said.
Enter drones. On the way to the test site, Peitzsch explained how he and his colleagues are using these aircraft to monitor an avalanche’s entire start zone.
First, they fly back and forth over the area, using its camera to systematically snap photos of the entire slope.
“What we’ll do here today is, we'll go out and fly, and we'll collect these images that are overlapping,” he explained. Once they’ve taken overlapping images of the region, they put it into processing software that analyzes them, and creates a “point cloud,” a 3-D mesh of data points in the shape of the terrain. Each data point’s exact position, including elevation, is recorded.
The cloud is used to create a digital surface model of the start zone. By comparing new digital surface models to previous ones, forecasters can see how the snowpack has changed after a storm, thaw or windy day — and identify potential trouble spots.
“That’s sort of the advantage there is” to this method, Peitzsch said. “You already know the depth of that weak layer.” Aerial analysis enables “understanding how much new snow (has fallen, and) how much that load above it varies across the slope.”
Peitzsch and his fellow researchers have been honing this method since 2016. Last Friday, he, Florentine and Miller headed to a low cliff near their main test site, Glacier View Mountain, to test a new pair of drones.
“Why we're going here is that it's a smaller path,” he said. “If things aren't going as we expected with the drone, it's easy to abort the mission and fly the drone back.”
The scientists had charted a 6-acre patch to test its mapping capabilities. But first, they had to learn how the drones performed.
The model they tested first had a propeller span about as wide as a pizza box. But it still came with a lengthy pre-takeoff checklist. The scientists filed a flight plan; clocked the wind speed with a handheld gauge; calibrated the drone by rotating it 360 degrees, faceup and facedown; and finally, stepped several feet back from an orange landing pad they had set out on the snow.
When the propeller blades spun, the quadricopter rose fast. Its buzz echoed off the hillsides, as Miller watched the elusive speck.
“You got plenty of space,” he told Peitzsch, whose eyes were glued to the iPad connected to the controller. “The video feed is great right now,” he said.
They brought it back after seven minutes, and a maximum height of 400 feet. “Right now, I’m just trying to get a feel for how it handles,” Peitzsch explained. He cut the second flight short when Florentine spotted a bald eagle cresting over the cliff. They’re careful to give birds their space. “We’ve been thwarted by bald eagles a couple of times,” Miller said.
They soon had another avian incursion. A flock of small birds, likely sparrows, started gyrating around the drone. “They did not like that thing,” Erich observed as he pulled the drone in, then sent it back out once they passed.
Issues like these are part of the process, Miller said. “We end up spending a lot of time in the field like ‘Oh, no, not quite.’” Wind and birds aren’t the only hurdles: the team ended up not scanning their planned area Friday, realizing that they first had to fix a software issue. And the federal government shutdown grounded them for more than a month this winter.
“We were hoping to collect throughout most of the year, and we weren't able to collect during that period,” Peitzsch said. “I think at this point we'll certainly have a chunk that's missing from those days. … We’ll try and get what we can from the rest of the season.”
He’s hoping to measure the snowpack that Glacier National Park’s plowing crews will face when they tackle Going-to-the-Sun Road this spring. Beyond that, he’s confident that a drone-based snowpack monitoring method can be ready by next winter.
Avalanches pose a very real threat across Northwest Montana, from Missoula backyards to the rail line that follows the Flathead River’s Middle Fork. Avalanches in this state have killed 77 people since 1998, the earliest year for which the American Avalanche Association keeps records.
The Flathead Avalanche Center’s Zach Guy aims to keep that number down, studying the snowpack and informing backcountry travelers of what they’ll face. Since much of his data comes from single points, he said that “having a more two-dimensional piece like what Erich’s working on would be nice.”
Whether it’s practical is another story. “It would be amazing if I could launch a drone from my house every morning and fly it over every mountain range …[but] there’s some limitations at this point.”
The Flathead Avalanche Center, Guy explained, monitors four mountain ranges, stretching from the Canadian border to just east of Flathead Lake. Much of that territory is designated wilderness where drones cannot fly.
“Operationally speaking, I think where I see the most use for something like that is operations that forecast on a slope-specific scale, like ski area forecasters or highway forecasters or railway forecasters,” Guy said.
But “even if it sounds like we’re not using his stuff operationally, advancing the snow science is a good thing, so I’m all in favor of it,” Guy said. “It’ll be useful for us to kind of learn more about patterns, things that are unknown to us like … how does the snow distribute across a start zone?”
In this volatile terrain, that's vital knowledge, said Peitzsch.
“We'll still need to go into the field, no doubt, we'll still need to dig pits and do stability tests,” he said. “But the value here is actually being able to look at the variability of that change across the landscape on a big scale.”