Researchers in Duke University’s Department of Civil and Environmental Engineering have developed a new mathematical model that uses temperatures within a deep-seated landslide to help predict the sudden, catastrophic failure of the moving land mass. A deep-seated landslide is one in which the sliding surface of rock is deep underground, roughly 30 m or even much deeper, notes Manolis Veveakis, Ph.D., an assistant professor of civil and environmental engineering at Duke.
Although deep-seated landslides are less common than shallow landslides — which involve material located closer to the surface, often in a slurrylike state — the deep-seated events involve massive amounts of rock, often moving as rigid blocks, that generally cause greater damage, Veveakis says. Deep-seated landslides can move slowly for decades or longer, creeping just a few inches each year, before suddenly accelerating and collapsing without warning. “When (a deep-seated landslide) moves, it takes the entire mountain with it — all the infrastructure, the people living there,” Veveakis explains.
One such landslide struck northern Italy in 1963 at the Vajont Dam (or Vaiont Dam), at the base of the Monte Toc mountain. The dam operators had tried for several years to mitigate the slow-moving event, located farther up the mountain, by lowering the water level of the lake behind the dam. Although these efforts initially seemed to help slow the landslide, in the end it collapsed anyway, suddenly sending approximately 10 billion cu ft of material plummeting into the lake at nearly 70 mph. The failure of the landslide generated a tsunami more than 800 ft tall that crashed over the dam, destroying several small towns and killing nearly 2,000 people, according to a Duke press release on Veveakis’ research.
Carolina Seguí, Ph.D., a geological engineer at Duke, and Hadrien Rattez, a postdoctoral researcher, applied the new temperature-based model to the Vajont Dam scenario in a paper, written with Veveakis, that was published online in June in the Journal of Geophysical Research: Earth Surface. The model accurately recreated the movements of the Vajont landslide and explained the mechanisms underpinning its motion over a period of more than two years, Duke reported in a June 15 press release.
That paper also applied the model to the Shuping landslide, an ongoing, active, deep-seated landslide near the Three Gorges Dam in China, and again accurately reproduced that landslide’s movement over a period of more than a decade. The research was funded by the National Science Foundation.
The Shuping landslide is also associated with a dam and a lake. But its behavior is the opposite of what happened at the Vajont Dam: The Shuping landslide accelerates when the level of the nearby lake is lowered, Veveakis notes. Every landslide is different, he explains, and so “we wanted to see how the model would perform in two opposite scenarios.”
The model works by taking temperature readings in a relatively thin layer of soft rock known as the shear band, which is generally found at the base of a landslide. Although the model specifically examined clay in the Vajont and Shuping examples, the approach “is generic for rocks that are sensitive to shearing, creeping, and temperature variations,” Veveakis says. Because actual measurements from the shear bands in the Vajont and Shuping landslides were not available, the researchers estimated the friction and internal temperatures, Duke noted in the same release.
Veveakis and Seguí are applying the model to an active landslide in Andorra known as El Forn, which is being closely monitored by the Andorran government. Movement of the El Forn landslide is not associated with a dam and lake, but rather it is affected by snowmelt, precipitation, and groundwater levels, says Seguí. A bore hole drilled into the El Forn site provided access for a thermometer to measure temperature as well as piezometers to record water pressure and an extensometer to measure horizontal displacement. Core samples were also extracted, which means the Duke team now has several years’ worth of site-specific data and material to work with. Because each landslide is unique, even the changes in temperature that the researchers are studying are subjective rather than absolute.
“You have to test the material in a laboratory and see its (thermal) sensitivity,” explains Seguí. “You take months or years of data to train the model, and then see the evolution of, say, the velocity or the water pressure linked to the response of the temperature of the material. Then you can see that, say, a 0.2-degree change in temperature is a lot for that material, or maybe you need a change of 10 degrees.”
So far, the results are promising. In a paper scheduled for an upcoming issue of the journal Geophysical Research Letters, Seguí and Veveakis conclude that the combined data offer “favorable results of the mathematical model. This will allow scientists to “forecast… the behavior (i.e., displacement) of a deep-seated landslide.”
The goal is to provide advance warning of an imminent collapse as well as a stability threshold, Veveakis says, “so that we can tell the engineer in the field not only when the landslide will fail but also at what point it is still susceptible to human intervention.” If the model’s calculations are correct, Veveakis adds, it could have predicted the Vajont collapse a few weeks in advance. In other cases, the model could predict a collapse as much as a year or more ahead of the event. That would give engineers time to increase or lower reservoir levels, dig wells to pump out groundwater, evacuate populations, or take other actions “before the landslide reaches the point of no return,” Seguí says.
This article first appeared in the December 2020 issue of Civil Engineering as “Duke Researchers Test Landslide Temperatures To Predict Collapses.”