Digital twins could help melt the mystery of Alaska’s thawing permafrost

UNIVERSITY PARK, Pa. — Communities around the world have adapted to live on the year-round frozen soil of frigid environments, such as in the Arctic. However, rising temperatures have introduced a new challenge: What happens when the ground under houses and roads begins to melt?

To better understand how these environments are changing, an interdisciplinary team of engineers and geoscientists led by researchers at Penn State has developed a computational framework that can use real-time measurements and artificial intelligence (AI) to predict the physical properties of the frozen soil, called permafrost. In one of the first studies of its kind, published in JGR Earth Surface and featured in Eos Magazine, the team applied their framework to a specific road embankment in Utqiaġvik, Alaska — the northernmost city in the United States. They recreated the permafrost’s thermal properties with chilling accuracy, opening a pathway to more accurate predictions of how climate change may impact permafrost around the world.

According to Ming Xiao, professor of civil engineering at Penn State and corresponding author on the study, rising temperatures on Earth are causing permafrost in many areas of the world to rapidly thaw, with ground temperatures increasing by up to almost two degrees Fahrenheit per decade. This thawing can release dormant strains of bacteria or large emissions of carbon dioxide into the atmosphere, further accelerating global warming. Additionally, Arctic communities and governments could see billions of dollars in infrastructure damage over the coming decades if this trend continues.

“What makes permafrost unique is that it largely consists of ice,” Xiao said. “As this ice melts, it turns to water. This makes thawed permafrost a very weak soil — it becomes very muddy, which can compromise the stability of infrastructure like roads, buildings or pipelines built in or on it.”

Standard methods of predicting permafrost degradation require a ton of computational power and existing data. AI-powered modeling has proven more efficient, but Xiao explained that these models generally perform poorly when applied beyond their initial training data.

However, through what Tieyuan Zhu, associate professor of geosciences at Penn State and co-author on the study, described as an “accidental conversation” at a work party, Xiao’s team realized that there was other research underway at Penn State that could help strike a balance between computational cost and prediction quality.

“My team's research uses advanced seismic temperature sensors built into fiber-optic cable to better study geology,” Zhu said. “Over a conversation at a barbecue, Dr. Xiao recognized how our research could help build an accurate, physics-informed method of understanding and predicting how permafrost is changing in the Arctic.”

This framework is known as a digital twin, which processes terabytes of data to create an extremely accurate, real-time simulation of an area or object. These simulations have become widespread in other fields such as mechanical or biomedical engineering, but Xiao said applying this concept to permafrost monitoring hadn't been thought of before. For this study, the team buried a pair of one-kilometer-long — about two-thirds of a mile or roughly 10 football fields — fiber-optic cables capable of collecting thermal and seismic data from the ground. A short section of these cables is along a road embankment, which collected temperature and seismic data from September 2021 to June 2024. The researchers then used some of the data as a foundation to build the twin.

The framework functions by processing information through two separate models. One uses a variety of complex equations, mathematical functions and AI-powered machine learning to forecast how heat transfers through the ground based on patterns seen in the existing data. The other uses the fiber-optic cables to monitor and collect real-time data, which informs the first model’s calculations. These two models are integrated together, with data continuously updated to create as close to a one-to-one simulation as possible. By informing these predictions with physical measurements, Xiao said that the team can create a simulation that updates with real-time data.

“This is the first attempt to apply a digital twin model to our previous work monitoring infrastructure in the Arctic,” Xiao said. “As new data comes in, our framework updates some key parameters in the mathematical model, including how fast heat transfers in the soil.”

The team found that this approach allowed them to more accurately predict detailed physical properties of the embankment’s permafrost, including unfrozen water content, changes to ground temperature and how heat moved through the ground. As they updated their model over time, they compared the prediction accuracy both to previous readings produced by the digital twin and measurements taken directly from the field over time, observing that as the twin was updated with real-world data, its predictions more closely matched the actual measurements later taken from the embankment.

Although the digital twin simulates a specific road embankment, Xiao said the idea and process can be broadly transferable to other infrastructure monitoring in cold regions. Going forward, the team plans to continue collecting data and refining their framework.

“The fiber optic cables are going to remain in the ground, so it’s a great opportunity to continue to collect data,” Xiao said. “If we have, say, 10 more years of data, that would be extremely valuable to the broader scientific community. It would help us develop even better predictive models.”

Other co-authors on the work include Gabriel Rocha Dos Santos, a geosciences doctoral candidate at Penn State; Zhinong Wang, a postdoctoral scholar of geosciences at Penn State; Xiaohang Ji, who received her doctorate in geotechnical engineering from Penn State and is now an engineer at AECOM; Ling-yun Gou, a civil engineering and environmental engineering doctoral candidate at the University of Wisconsin-Madison who was a research assistant at Penn State at the time of the study; Dmitry Nicolsky, an assistant professor at the University of Alaska Fairbanks; and Eileen Martin, associate professor of geophysics at the Colorado School of Mines.

This work was supported by the U.S. National Science Foundation.

At Penn State, researchers are solving real problems that impact the health, safety and quality of life of people across the commonwealth, the nation and around the world.

For decades, federal support for research has fueled innovation that makes our country safer, our industries more competitive and our economy stronger. Recent federal funding cuts threaten this progress.

Learn more about the implications of federal funding cuts to our future at Research or Regress.