From the air, Wainwright, Alaska, looks like a perfectly stable band of homes and roads between the Chukchi Sea and a vast, flat tundra. But just below the surface, the frozen ground that this town was built on contains large bodies of ice. Along the eroding coastal bluffs, that hidden ice is becoming exposed and literally melting away.

This is more than a landscape change. When ice within frozen ground, known as permafrost, melts, roads can sink, buildings can tilt, pipelines can crack and leak, and shorelines can retreat, leaving homes precariously close to the ocean’s edge. A study in 2025 estimated that permafrost thaw could cost Alaska US$37 billion to $51 billion just in building and road damage.

As a civil engineer, I have spent the past decade studying how permafrost and infrastructure respond as Arctic temperatures rise two to three times faster than the global average.

A regional permafrost map alone cannot tell a builder how a particular foundation or road will survive as temperatures rise. Engineers need methods that can help them see beneath the surface and find the weaknesses. Using drones and digital twin technology, my colleagues and I have been developing those techniques.

When frozen ground is part of the foundation

Permafrost is ground that stays at or below 32 degrees Fahrenheit (0 degrees Celsius) for at least two consecutive years. Where permafrost is rich in ground ice, thawing can fundamentally change its ability to support infrastructure.

Frozen soil that has held buildings, roads and pipelines for decades can start to sink as temperatures rise and the ice that bonds soil particles weakens. When ground ice melts and water drains away, the surface can subside. The weight of homes, roads and other infrastructure can cause the ground to sink even more.

This process is highly uneven. Locations a city block apart can contain very different amounts of ice and unfrozen water. Roads and buildings also change how snow accumulates and water drains.

Along the coast, thawing permafrost can leave coastal areas more vulnerable to waves and erosion. In Utqiaġvik, a bluff’s partial collapse in 2023 and 2024 brought the bluff’s eroding edge close to roads and buildings. Waves removed material from the bottom while thaw weakened frozen ground above, causing the edge to slump.

Seeing what lies beneath the tundra

Much of the critical information that developers need to know before building a home or road is underground. However, drilling dense grids of boreholes in frozen soil to take samples of the ground below is expensive, disruptive and extremely labor intensive.

Geophysical imaging can help fill the gaps.

One method my team is working with is very low frequency electromagnetic surveying, or VLF-EM. It uses electromagnetic signals from distant transmitters and measures how the ground responds. Because it does not require direct ground contact, a lightweight sensor can be carried by a drone over wetlands and tundra that are susceptible to thawing.

Electromagnetic surveying is especially useful along Arctic coasts. Salty water that isn’t frozen conducts electricity much better than ice-rich frozen ground, so the differences that show up can reveal areas where the ground has thawed.

VLF-EM does not produce a perfect underground picture – depth estimates can be uncertain – but it can identify where subsurface conditions have changed so engineers know where to look closely.

From measurements to forecasts

Infrastructure may remain in service for 20, 50 or more years, so mapping today’s underground conditions is not enough. Builders need to know what conditions are likely to look like decades into the future.

My research group has been working on ways to project how thaw affects two quantities that engineers care about: how much the ground settles, and how much load a foundation can support.

In a 2026 study, we looked at permafrost stability across Alaska’s Arctic Coastal Plain, the northern coastal region that includes Prudhoe Bay and much of Alaska’s oil industry operations.

We found that if the world emits greenhouse gases at a high rate that continues to raise temperatures quickly, the proportion of infrastructure at risk remains below 10% around mid-century, then rises rapidly between the 2060s and 2080s. By the 2090s, about 80% of buildings, 60% of roads and 90% of pipelines are projected to be at risk of damage from sinking land.

We are also developing computer models called digital twins for monitoring cold regions’ infrastructure. The digital twins replicate the real-world changing environments and aging infrastructure and can be used to predict what is likely to occur.

At a road embankment in Utqiaġvik, our team used fiber-optic cables to collect temperature and seismic data. In a 2026 study, we used measurements from a roughly 330-foot (100 meter) section to update a computer model that combines heat-transfer physics and machine learning. As new data arrives, the digital twin adjusts to improve predictions of how permafrost temperature changes and how infrastructure performs.

Our goal is to use drones and geophysics to map the land, sensors to track changes, and computer models to forecast how those changes affect infrastructure.

Preparing for the future

Communities need easy-to-use tools showing where hazards may develop, how those hazards may change over coming decades, and what infrastructure could be affected.

Permafrost thaw is often described as an environmental problem. In the Arctic, it is also an engineering problem. The better that engineers can see what is happening beneath the surface – and anticipate what comes next – the more time communities have to decide where to monitor, reinforce, repair or adapt.

Ming Xiao receives funding from the U.S. National Science Foundation and the U.S. Department of Defense.