Detecting Landslides from a Few Seismic Wiggles

December 18, 2015
A 200 million ton landslide landed on the toe of Tyndall Glacier and in the water of Taan Fiord on Oct. 17 local time in Icy Bay, Alaska. It was detected by seismologists on the other side of the country. (Geospatial support provided by the Polar Geospatial Center. Imagery Copyright 2015 DigitalGlobe, Inc.)

A 200 million ton landslide landed on the toe of Tyndall Glacier and in the water of Taan Fiord on Oct. 17 local time in Icy Bay, Alaska. It was detected by seismologists on the other side of the country. (Geospatial support provided by the Polar Geospatial Center. Imagery Copyright 2015 DigitalGlobe, Inc.)

The rumbling would have started across Icy Bay at around 8:19 p.m. on Oct. 17, 2015. In the span of about 60 seconds, 200 million metric tons of rock roared down the side of Alaska’s Taan Fiord valley and crashed onto the toe of Tyndall Glacier and into the water, setting off a local tsunami big enough to register at the nearest tidal gauge 155 kilometers away.

No one was near the glacier to witness the massive landslide—the largest detected in North America since the collapse at Mount St. Helens—and it almost went unnoticed. Almost. Its signature appeared almost simultaneously in seismograms monitored by the Global CMT Project at Columbia University’s Lamont-Doherty Earth Observatory on the other side of the country.

Over the last six years, Lamont seismologists Göran Ekström and Colin Stark have been perfecting a technique for picking out the seismic signature of large landslides from the stream of seismic data from earthquakes and other activity around the world. The details they are able to extract could one day help governments sound tsunami warnings, help rescuers find landslide-struck villages faster, and warn of risks such as landslide-dammed rivers that could soon burst through.

When computers detect earthquakes, they look for an abrupt start to seismic activity, typically a sharp, high-frequency burst. Landslides have a different signature that often goes undetected.

“Landslides start very gradually. They grow and they decay over the time period it takes for the mass to go all the way down, so this would be a minute or maybe two,” Ekström said. To pick out landslides, the scientists look for long-period seismic waves—seismic wiggles with a period of 50 seconds or so.

From those waves, Ekström and Stark are now able to discern the size of the landslide, which direction it was moving, its momentum, velocity and acceleration, as well as the precise time that the landslide struck. The data can usually get them within a few kilometers of the precise location, and satellite images—when clouds aren’t obstructing the view—can provide confirmation. Stark presented their new landslide discovery from Taan Fiord and described how they determine the forces involved in greater detail during the American Geophysical Union meeting today in San Francisco.

Comparison of the seismic signatures of a landslide and earthquake from events in the Himalayas on May 5, 2012. (Stark and Ekström)

Comparison of the seismic signatures of a landslide and earthquake from events in the Himalayas on May 5, 2012. (Colin Stark and Göran Ekström)

The team at Lamont has analyzed about 50 massive landslides over the past six years, including several that had never been reported. Landslides associated with earthquakes can be difficult to pick out from all the seismic noise accompanying the slipping fault, but other landslides can be isolated.

A few days before the collapse in Taan Fiord, they detected a landslide in the Yukon—45 million tons of rock and ice had tumbled down Mount Steele, one of Canada’s highest mountains. Just after Valentine’s Day in 2014, they saw the signature of about 70 million tons of debris sliding down Mount La Perouse near Alaska’s Glacier Bay. Ekström and Stark have also used details of the collapse on Mount St. Helens in 1980 and from the deadly Hunza-Attabad landslide of 2010 in Pakistan to fine-tune their ability to calculate the forces. The Pakistan landslide had dammed a river, creating a lake that slowly inundated villages and raised a risk of future dam-break flooding.

“We get a lot of events in the Himalayas, but almost 50 percent seem to come from Alaska,” Stark said. “They are typically in high mountain areas that are actively tectonically deforming, where mountains are being made and destroyed.”

Taan Fiord, near Icy Bay, and the surrounding area along Alaska’s St. Elias, Chugach and Wrangell Mountains is one of the most seismically active in the world. Mountains are still coming up, and the erosion rate is fast because of the fragile nature of the rock, which is often volcanic, and because of its glacier activity. “Weak rock, rapid tectonic deformation, fast uplift rates make high mountains prone to landslides,” Stark said.

An island in Taan Fiord, about 10 km from the landslide, shown by satellite in 2014 (left) and a few days after the landslide and tsunami (right). (Geospatial support provided by the Polar Geospatial Center. Imagery Copyright 2015 DigitalGlobe, Inc.)

An island in Taan Fiord, about 10 km from the landslide, shown by satellite in 2014 (left) and a few days after the landslide and tsunami (right). (Geospatial support provided by the Polar Geospatial Center. Imagery Copyright 2015 DigitalGlobe, Inc.)

In the case of the Taan Fiord landslide, the Tyndall glacier’s retreat from the edge of Icy Bay in 1961 to its current location more than 17 kilometers up the fiord removed buttressing in the valley, leaving the weak rock on the valley wall prone to collapse, Ekström said. The entire valley was once filled with ice as much as 400 meters high, but as the glacier retreated, the ice thinned.

Satellite images several days after the landslide showed the results of the landslide. The images also show what appears to be damage along the sides of Taan Fiord from a tsunami, Stark said. On one recently ice-covered island in the fiord, about 10 kilometers from Tyndall Glacier, the young trees were stripped to the bedrock around the edges and fallen trees formed a pattern that appeared to have been made by a large wave.

“For the first few years, we were a little reticent to say we detected these giant landslides that no one else had noticed. How could they not notice?” Stark said. “It took us a while to be confident, but from analyzing about 50 of these, we have been able to constrain the physics and get the numbers—momentum, kinetic energy—we get dynamics that are impossible to get any other way. We have improved understanding of the physics of landslides.”

Work by Colin P. Stark conducted under NSF EAR award 1227083. Geospatial support for this work provided by the Polar Geospatial Center under NSF PLR award 1043681. Imagery Copyright 2015 DigitalGlobe, Inc.

Media Inquiries: 
Stacy Morford
smorford@ldeo.columbia.edu
845-365-8151