Across Arctic Alaska, towns like Wainwright and Utqiaġvik sit on ground that looks solid from above but is increasingly unstable below. The frozen soil — permafrost — that underpins roads, homes, and pipelines throughout the region is thawing as Arctic temperatures rise two to three times faster than the global average. A 2025 study estimated that permafrost thaw could cost Alaska between $37 billion and $51 billion in building and road damage alone. Now, civil engineers are developing drone-based surveying and computer modeling tools to detect the problem before it becomes catastrophic.
Permafrost is ground that remains at or below freezing for at least two consecutive years. When it contains large bodies of ice, thawing can be devastating to the infrastructure built on top of it. As ice melts and water drains away, the surface subsides. Buildings tilt, roads sink, and pipelines crack. The process is highly uneven — two sites a city block apart can contain very different amounts of ground ice — making broad regional maps insufficient for engineering decisions. In Utqiaġvik, partial bluff collapses in 2023 and 2024 brought eroding edges close to roads and buildings, with wave action removing material from below while thaw weakened the ground above.
To see what lies beneath the tundra without the cost and disruption of drilling dense grids of boreholes, researchers have been working with very low frequency electromagnetic surveying, or VLF-EM. Lightweight sensors carried by drones emit no signals of their own but measure how the ground responds to electromagnetic transmissions from distant sources. Because salty, unfrozen water conducts electricity far better than ice-rich frozen ground, the technique can identify zones where thawing has already occurred. It is particularly useful in coastal areas and on wet, thaw-sensitive tundra that is difficult to reach on foot. While VLF-EM cannot produce a perfect underground image, it tells engineers where to investigate more closely.
Mapping current conditions is only part of the challenge. Infrastructure can remain in service for 50 years or more, so engineers also need projections of future ground behavior. One research group has developed digital twin models — computer simulations that replicate real-world environments and aging infrastructure and update continuously as new sensor data arrives.
At a road embankment in Utqiaġvik, the researchers embedded fiber-optic cables to collect temperature and seismic readings along a roughly 330-foot section. Those measurements feed a model combining heat-transfer physics with machine learning to forecast how permafrost temperature and infrastructure performance will change over time. A separate analysis of Alaska’s Arctic Coastal Plain — which includes Prudhoe Bay and much of the state’s oil industry — projected that under a high-emissions scenario, while less than 10 percent of infrastructure would be at risk by mid-century, that figure could climb to roughly 80 percent of buildings, 60 percent of roads, and 90 percent of pipelines by the 2090s.
Researchers describe permafrost thaw as both an environmental and an engineering problem. The tools being developed — drone surveys, geophysical imaging, sensor networks, and predictive models — are intended to give communities and planners enough advance warning to decide where to monitor, reinforce, repair, or retreat before the ground beneath them gives way.