Opening a window to the inner 'cooking’ of a volcano
Earth Sciences Professor thomas Giachetti is developing protocols and machine learning tools to help scientists accurately model volcanic eruptions, such as the 1980 eruption of Mount St. Helens. Photo by Anaïs Férot
Detailed Distributions of Tephra Fall Characteristics: Insights into Magma Fragmentation and Transport Via Volcanic Plumes
Surrounded by 10 of the nation’s most dangerous volcanoes, scientists at the University of Oregon and across the Pacific Northwest are working to predict not only when they might erupt, but how. Will they explode in a giant plume of ash and magma? Or will they seep lava in a slow, snaking flow?
The answers to these questions can help determine how communities respond to an eruption. But first volcanologists need more data—data that must be painstakingly extracted from thousands of ash and rock samples collected from both past and current eruptions.
The eruption of Mount Mazama more than 7,000 years ago formed the caldera that holds Crater Lake.
“We have instruments on these volcanoes that tell you something is cooking, but what’s next? What will the eruption look like? How will it progress with time? You can find clues in the particles that are ejected whether it’s about to turn into something less violent and more effusive,” says Thomas Giachetti, a professor of Earth sciences in the College of Arts and Sciences. “We then try to reproduce these eruptions with models, and we improve those models based on past eruptions.”
Giachetti has a plan for accelerating the scientific community’s ability to accurately model volcanic eruptions by standardizing and streamlining the process for analyzing samples. With funding from a Faculty Early Career Development Program (CAREER) award from the National Science Foundation (NSF), he and his team are developing protocols and machine learning tools to help volcanologists analyze the shape, size, and composition of particle samples at a faster rate.
Earth Sciences Professor Thomas Giachetti
To get a more complete picture of an eruption, we need to analyze samples from many different locations around a volcano. We have not done that very well in our discipline for the past few decades. It’s a time-consuming process requiring many different instruments in the lab, and the more detail you want the more time you need.
Thomas Giachetti, professor of Earth sciences
With the help of a neural network, similar to the image recognition technology used by Google to search images, Giachetti’s team is training an artificial intelligence algorithm to recognize different types of volcanic rocks based on images uploaded to the software, allowing scientists to quickly identify new samples.
Swift sample analysis is particularly critical during an eruption, says Giachetti, who will work with the Cascade Volcano Observatory, among other collaborators, to help predict how volcanic clouds formed by explosive eruptions evolve with time.
“If an eruption happened tomorrow, everyone would be trying to figure out where the ash would go,” he says. “Will it be very explosive, where we have to evacuate people and shut down aviation in the Pacific Northwest, or will it be a lava flow, which requires far different preparedness and mitigation? By streamlining and standardizing these methodologies, we can have much faster communication with the people who are running the model live.”
Giachetti (right) and postdoctoral researcher Josh Wiejaczka collect samples near Crater Lake. Photo by Anaïs Férot
Because ashfall from an eruption can be so widespread—in some cases covering a third of a continent—Giachetti hopes the tools and procedures he’s developing can be used to enlist the help of citizen scientists in collecting samples from across the Pacific Northwest and beyond. To that end, he’s also partnering with schools built on top of deposits from past eruptions to develop a hands-on science program that will engage students in digging up and analyzing samples from their own schoolyards.
“They can take a shovel and dig in the backyard of their school to find deposits from past eruptions,” he says.
Using the tools Giachetti provides, including sieves and microscopes, students will have the opportunity to devise their own methods for analyzing their samples, characterizing the ash and comparing their results with other schools participating in the program. At the end of the project, they’ll present their findings to their community in local public libraries and community centers.
“The idea is really for them to understand how you can have particles from a volcano so far away and what that implies for potential future eruptions in the Pacific Northwest—and to raise awareness about it,” he says. “The ultimate goal with the schools is not to produce any scientific data I can directly plug into the model, but rather to have the students realize what they are living on, why this volcanic product is there, and what we need to know to better understand how it got there.”
Giachetti hopes his research will help push the field forward in its ability to accurately predict how future eruptions will progress.
“We’re using whatever the volcano is giving us to understand how the eruption is unfolding,” Giachetti says. “We want to understand what’s happening inside the volcano—what we can’t see and will never be able to see.”
