A clearer understanding of the risk of permafrost thawing in Alaska

In the Arctic, permafrost plays an important role in building infrastructure. However, when the area gets warmer and the permafrost is thawed, the infrastructure is threatened as the ground moves under the built environment. Unfortunately, the full scope of risks associated with this process has not yet been realized, but researchers are working to tackle this knowledge gap.

Environmental researcher students Elias Manos and Assistant Professor Chandi Willalana, including UCONN’s Department of Natural Resources and PhD, and Anna Liljedal of the Woodwell Climate Research Center, have developed a way to double Alaska’s mapped infrastructure and more accurately double the economic risks of the project using high-resolution satellite imagery and deep machine learning. Their findings can be found in Communications Earth & Environment.

Withharana says this is the latest in his research group’s long-term research. In this case there is a risk that most of the thawing permafrost due to critical infrastructure such as communities and buildings and roads is unexplained.

“The main focus here was the visual gap in the infrastructure, and we need to create more details to create important information layers for downstream analysis like economic risk. Not for Alaska,” says Withharana.

The motivation behind this study stems from the need to understand the dangers of a changing world, Manos says. However, these assessments cannot occur without a clear understanding of what is causing harm.

“We know that local temperatures are rising and there are changes in the frequency, intensity and timing of extreme weather and dangerous events. We need to understand the potential harms these events can bring, such as rapid onset events such as hurricanes, floods, wildfires, droughts and other slow onset risks,” Manos says.

Manos says permafrost acts as a structural foundation through which the mountains are fixed and the buildings are designed to maintain thermal integrity. Therefore, it is essential that the pile foundation remains stable and fixed in permafrost, but the structural integrity is compromised as this layer thaws.

“As the permafrost temperatures begin to rise, the mountains start to fall out of place. That’s what’s called a loss of bearing capacity, or a decrease in bearing capacity. That was the main danger affecting the building,” Manos says. “After that, there are also transportation infrastructures that were primarily affected by ground subsidence. When ice-rich permafrost thawed, the ground entered the cave, which was a danger that it used to assess road disaster risks.”

Previous studies have created risk estimates based on OpenstreetMap (OSM) data. It is one of the most widely used geospatial datasets available, Manos said. OSM is available in all countries around the world, and information is updated by volunteers who manually enter local data, such as buildings, trails, roads, or other types of infrastructure, from global high-resolution images.

In some regions, such as Europe and parts of the US, the data is accurate, says Manos, but that’s not the case everywhere. Unfortunately, Arctic has a lack of OSM data.

“There are some previous risk studies that relied on this incomplete infrastructure data. It all goes back to the fact that Arctic infrastructure is not fully mapped. It’s a problem if you want to understand disasters.

One of the goals of Withharana’s research group is to improve the way in which large sets of satellite images can be quickly and accurately analyzed. Here, they developed a method to accurately map the thawing risk of infrastructure and permafrost known as high-resolution arctic infrastructure and topographic analysis tools (habitats).

This model uses machine learning and AI to extract road and construction information from high-resolution satellite images from 2018 to 2023. They compared habitat data with OSM data to assess the quality of the new model and looked for potential misclassifications. They then added new information to the OSM, almost doubled the previous amount of information available to Alaska.

“The amount of infrastructure and buildings that were missing from the Open Street Map was really shocking to me, with 47% shortfall,” Manos says. “OpenStreetMap is a powerful volunteer-based resource, but it has limitations, and that’s not a surprise.”

With a large amount of data that had not been considered previously, researchers estimate that by 2050, the cost of permafrost damage to infrastructure will double in low and medium emission scenarios.

“The damage to the infrastructure caused by permafrost thawing is comparable to the average annual cost of all natural disasters in the country, but permafrost thawing is not recognized by the federal government as a natural disaster, some of which,” says Liljedal.

The With Arrana research group and collaborators are working to fill these knowledge gaps and create data that can be used to prepare the community for the future. Manos plans to extend this analysis to assess economic losses using a comprehensive infrastructure map, taking into account the Arctic as a whole.

Witharana adds that access to the NSF Supercomputing Infrastructure, combined with OSM data and thousands of submeter-resolution satellite images provided by the National Science Foundation, researchers can increase the integrity of these datasets.

“We can see the impact and make a better assessment of economic obstacles and risks, so we can prepare ourselves for the necessary policy measures and downstream efforts. That’s a key result. Overall, the integration of AI and big data sets within the application has helped researchers and the community create useful and practical products that can be used now.”