Avalanche forecasting has come a long way since the 1950s, when forecasters relied solely on weather to predict when and where snow might slide. But it still requires scientists skiing and digging into the snowpack. That’s changing as satellites, aircraft-mounted sensors and ground-based remote monitoring fast-track the evolution of snow science, giving experts comprehensive insight into the uncanny nature of avalanches.
The Colorado Avalanche Information Center has been testing satellite imagery to detect avalanches. The technology is building a more accurate library of avalanche activity over a winter season, and year over year. And not just for the most trafficked zones, said Mike “Coop” Cooperstein, the center’s lead forecaster for the northern mountains.
“We have really good information along the highways, in the really popular recreation spots — Berthoud Pass, Loveland Pass, Red Mountain Pass. But it’s pretty close to the road,” Cooperstein said. “So we wanted to look into those deeper areas, a few miles from the trailhead, and see what’s happening, because we are forecasting for those areas.”
With 11 avalanche fatalities in Colorado this winter, and 32 nationwide, avalanche forecasters like those at CAIC need all the resources they can get to create accurate forecasts for backcountry regions. But methods of gathering good information are decades old. Emerging technologies may help, but it could be years before they are operational or affordable enough for avalanche forecast centers to use on a daily basis.
Relying on observations shared by travelers on roads and skintracks yields only a partial picture of avalanche activity, and doesn’t necessarily reflect the hazard spread across entire ranges. As a result, during any given avalanche cycle, forecasters may miss part of the avalanche activity because it wasn’t witnessed, and wouldn’t be able to warn their audience of backcountry goers. The other issue is not being able to verify whether their forecast was correct after the fact, making it difficult to identify patterns of inaccuracy and improve forecasts over time.
An avalanche record biased toward easy-to-access areas also leaves researchers in the lurch as they have only a fuzzy quantitative idea of what “normal” is as far as avalanches go. This could complicate understanding how avalanche activity shifts away from historical trends because of climate changes.
To balance and fill in the record, researchers, avalanche forecast centers and private companies are leaning on new technologies.
Late last year, Norway began using satellite-mounted radar to detect avalanches across the country.
The system was developed by a team from the Norwegian Research Center, known as NORCE, using synthetic aperture radar, or SAR, mounted on the European Space Agency’s two Sentinel-1 satellites. When a slide was detected it was automatically reported to Norwegian avalanche forecasters. Because the SAR device emits microwaves toward the Earth’s surface, rather than sensing natural light, this method allows the satellites to capture images on cloudy days or at night.
When an avalanche releases, the debris pile left at the slope’s base is both rougher and more dense than the surrounding snow, which scatters the microwaves emitted by the SAR so that fewer of them return to the sensor on the satellite. The NORCE team has built what’s called a data processing chain that recognizes this increased “backscatter” and flags it as potential avalanche debris.
The method is not perfect — with accuracy varying on how much water is in the snowpack — but technology is helping the Norwegian avalanche forecasting agency identify thousands more avalanches than any field-based efforts would, said Markus Eckerstorfer, the NORCE researcher who developed the data processing chain.
“In general, the method is very promising,” Eckerstorfer said. “It’s probably better than anything else. If you have a certain region, you could never cover the entire region as you do with a satellite. But you’re still not able to detect everything.”
In 2018, Eckerstorfer received funding from the ESA to test the method in Colorado, teaming up with CAIC. But the test fell flat.
Colorado’s mid-latitude played a role in the poor results: Because the Sentinel-1 satellites are polar-orbiting, moving between the North and South poles, the closer a region is to the equator, the less coverage it receives. Norway, being so far north, receives daily passes, while Colorado is passed every six or so days. Additionally, during the winter of 2018 one of the pair of satellites stopped sending images of Colorado because it needed time to recharge for imaging Europe and countries that paid to put the constellation into orbit.
An image every 12 days was far less useful for CAIC, and interest in the project dropped off.
“We wanted to have pretty close — a day or two at the most — of detection so that we can, after our forecast, look and see if what we think happened, happened,” Cooperstein said. “For that, there just isn’t enough coverage. There aren’t enough passes over Colorado to do that kind of work right now.”
Research using the method continues elsewhere in the states, though, which may one day come back to benefit Colorado.
An avalanche forecasting center in Idaho and another in Montana are using a rudimentary version of NORCE’s automated avalanche detection from Sentinel-1s, led by Zach Keskinen, a snow science master’s candidate at Montana State University.
Keskinen said the technology would not replace forecasters on skis, but it can be yet another tool in the high-consequence science of predicting avalanches.
“It’s starting to show promise to where [an avalanche forecaster] could say, ‘Hey, this is where you should go for your field day.’” he said. “Hopefully it will help to guide field days for forecasters and just make their field time more powerful.
Keskinen expects satellite coverage in the states to improve. The U.S.-Indian NISAR satellite scheduled for launch in 2022, will have SAR sensors.
Despite the low utility of radar in Colorado to date, satellites are still in the mix at CAIC. The organization is using optical imagery instead of radar to review the once-in-a-century avalanche cycle from March 2019, when historic slides swept over Interstate 70, blew out massive swaths of forest and spiked four of Colorado’s avalanche advisory zones into “extreme” avalanche danger.
“We can pick out avalanches from that cycle pretty well with, you know, 90% confidence that we’re picking out avalanches,” Cooperstein said.
CAIC staff can look at destruction of vegetation, or signatures of decaying vegetation, to see where an avalanche has cleared forest, documenting slides that may otherwise have been missed.
“The thing there is it takes such a large avalanche to destroy trees,” Cooperstein said. “That avalanche cycle, I’ll probably never see another one of those in my career, that’s that big. But [optical satellite imagery] has really good promise and we have pretty good results from doing that.”
Lasers can detect avalanches, too, specifically light detection and ranging, or LiDAR.
Jeffrey Deems, a researcher at University of Colorado and co-founder of Airborne Snow Observatory (an offshoot of NASA’s Jet Propulsion Laboratory), uses LiDAR mounted on aircraft to measure snow in basins across the western U.S. While the goal of these flights is to map how much snow—and therefore water—a river basin holds so that water managers can better plan how to allocate runoff, they also detect avalanches.
The method works by “mowing the lawn,” as Deems puts it, flying a plane with a LiDAR instrument scanning every foot of an entire river basin. Such a flight path is done in the summer without snow to create a baseline, then again in the winter, allowing Airborne Snow Observatories to subtract the summer image from the winter one to determine snow depth.
In 2019, flights in the Upper Gunnison and Roaring Fork river basins revealed large avalanches that had not been documented. LiDAR offers a more detailed look into an avalanche than other optical or radar imagery.
“With the LiDAR technology, we can actually see the change in snow depth, so we can actually get a better metric, or a better understanding of the volume of snow that had to move, rather than just mapping the extent of it,” Deems said.