The researchers asked how much mid-century permafrost degradation could damage Arctic buildings when risk calculations count building use and multiple stories rather than treating every structure as a flat footprint.
The research question and why it matters
The researchers asked how much mid-century permafrost degradation could damage Arctic buildings when risk calculations count building use and multiple stories rather than treating every structure as a flat footprint.
Earlier circumpolar studies estimated permafrost-related losses using government inventories or OpenStreetMap and generally treated buildings as two-dimensional footprints. One recent estimate placed high-emissions building losses near $110 billion. The new work extends an earlier Alaska mapping system across much of the circumpolar Arctic, adds occupancy classification and story counts, and then runs the expanded inventory through the established geotechnical damage model.
What researchers found
The mapped Arctic residential stock contained an estimated 301 million square meters of floor area—75% more than its two-dimensional footprint area. The median projected damage was $76 billion under SSP2-4.5 and $261 billion under SSP5-8.5 in purchasing-power-parity-adjusted 2024 U.S. dollars. The corresponding 5th-to-95th-percentile intervals were $2.88–$259.48 billion and $17.01–$379.11 billion. The higher high-emissions median was more than twice a previous $110 billion estimate that lacked the new three-dimensional building inventory.
Key results from the tested systems
communities mapped from satellite imagery
The study region covered Arctic settlements on permafrost in Alaska, Canada and Russia.
estimated residential floor area
Adding likely upper stories increased exposed area by 75% over flat building footprints.
median mid-century damage estimates
The two values correspond to intermediate and very high emissions pathways, in PPP-adjusted 2024 U.S. dollars.
high-emissions 5th–95th percentile range
The broad interval shows why the $261 billion median is not a precise forecast.
How the research worked
The team developed a deep-learning workflow to detect buildings in commercial satellite imagery and classify them as residential or non-residential. They combined new detections with OpenStreetMap, estimated building heights from the two-meter ArcticDEM surface model, and converted heights to likely story counts using building codes and construction references. Regional replacement-cost estimates were then joined to a previously published geotechnical model of pile-foundation bearing-capacity loss. Monte Carlo simulations propagated uncertainty across intermediate SSP2-4.5 and very high SSP5-8.5 emissions pathways for the mid-century period of 2055–2064.
How to interpret this design
The result is conditional on the model structure, inputs, boundary conditions and scenarios chosen by the researchers. Agreement with known observations strengthens confidence, but a projection is not a direct observation of the future or the inaccessible past.
The evidence is produced by computation rather than direct experimental manipulation of the target system. Its value depends on transparent assumptions, realistic inputs, sensitivity testing and comparison with independent observations.
What strengthens or limits the finding?
The study improves the exposed-building inventory with sub-meter imagery, validates model components against authoritative data, reports Monte Carlo uncertainty, and openly shares most code and derived data. Its dollar estimates remain projections rather than observed losses, and their 5th-to-95th-percentile ranges are extremely broad because future climate, engineering practice and mapping errors remain uncertain.
The result explores a plausible explanation or scenario. Its reliability is conditional on assumptions and should be tested against new observations or experiments.
Funding and disclosure context
The recorded funding source is: U.S. National Science Foundation grants RISE-1927723, 2052107, 2019691 and 2022504; Google.org Impact Challenge on Climate Innovation; Texas Advanced Computing Center award DPP20001. The recorded conflict information is: The authors declared no conflicts of interest relevant to the study. Funding or a disclosed relationship does not by itself invalidate a result, but it is relevant when judging design choices, analysis and the need for independent replication.
What it means
Accounting for upper floors changes the scale and geography of assets exposed to thawing ground, especially in Arctic settlements with multi-story housing. The dataset can help communities and governments identify where engineering surveys and adaptation planning deserve attention. The range of plausible costs—not only the $261 billion midpoint—is essential for budgeting because it shows that the model cannot predict a single future bill with precision.
Deeper analysis
A third dimension changes exposure before it changes hazard
The permafrost process did not become more severe merely because the researchers counted floors. What changed was the inventory of assets sitting above the hazard. A multi-story apartment block and a one-story building can occupy similar ground footprints while representing very different amounts of material and replacement value.
The uncertainty interval is the central result
The very high-emissions median of $261 billion is memorable, but the study's own plausible interval stretches from $17.01 billion to $379.11 billion. Climate variability, engineering assumptions and mapping errors dominate different parts of that range. Planning can use the estimate as a risk envelope; it should not treat the midpoint as an invoice.
Open outputs improve scrutiny, with one important exception
The authors released code, public inputs, derived building products and damage estimates through GitHub, Zenodo and the Arctic Data Center. That supports independent checking and reuse. The underlying commercial satellite scenes remain restricted, so an outside team cannot reproduce every mapping step without equivalent imagery access.
Risk maps support—but do not replace—community decisions
A circumpolar model can reveal where the aggregate problem is large and where detailed engineering work may be urgent. It cannot decide which homes communities value, which adaptations are acceptable or how cultural and relocation costs should be weighed. Those choices require local knowledge alongside the technical map.
What it does NOT prove
- It does not show that $261 billion of damage has occurred or is certain to occur. That value is the median of a modeled very high-emissions scenario for mid-century.
- It does not predict the repair cost for a particular town or building. Local ground ice, foundations, maintenance and adaptation can change risk substantially.
- It does not estimate every Arctic infrastructure loss. The floor-area calculation focused on residential buildings and the damage model addressed loss of pile-foundation bearing capacity.
- It does not establish that artificial intelligence made the forecast inherently accurate. Deep learning helped map buildings but also contributed error included in the uncertainty analysis.
- It does not include the value of displacement, cultural loss, business disruption or all roads, pipelines and industrial facilities.
Important limitations
- The analysis is a linked model, not a record of observed future damage. Errors in building detection, classification, height, replacement cost, climate projections and geotechnical response can compound.
- The 5th-to-95th-percentile intervals are very wide and overlap substantially between emissions pathways, limiting precision for any single cost estimate.
- Cloud- and snow-free commercial imagery was unavailable for 176 of the 1,162 identified communities; the team modeled 986 and supplemented coverage with OpenStreetMap.
- Training footprints came from about 10,000 buildings in 18 communities and one industrial site, while authoritative validation data were not equally available across countries, particularly Russia.
- ArcticDEM has approximately four-meter horizontal and vertical accuracy, so estimated heights and story counts can be wrong for individual buildings.
- The study did not model demographic change or determine which buildings will still be occupied by mid-century.
- Replacement costs combine regional data and assumptions expressed in purchasing-power-parity-adjusted 2024 dollars; actual labor, logistics and material prices can differ sharply in remote communities.
- Some commercial satellite imagery cannot be shared publicly, although the paper, code, public inputs, derived products and damage estimates are available through repositories.
How this fits with previous research
Earlier circumpolar studies estimated permafrost-related losses using government inventories or OpenStreetMap and generally treated buildings as two-dimensional footprints. One recent estimate placed high-emissions building losses near $110 billion. The new work extends an earlier Alaska mapping system across much of the circumpolar Arctic, adds occupancy classification and story counts, and then runs the expanded inventory through the established geotechnical damage model.
Questions still unanswered
- How much will local adaptation—foundation retrofits, cooling systems, drainage and managed relocation—reduce the modeled losses?
- Can community-supplied records improve building-use, occupancy and replacement-cost estimates in places with sparse validation data?
- How would population change and abandonment alter the amount of occupied floor space exposed in 2055–2064?
- What costs emerge when roads, utilities, pipelines, industrial sites and indirect social disruption are modeled alongside buildings?
- How well do the projected damage patterns match observed foundation failures as longer monitoring records accumulate?
Relevant U.S. government resources
These resources serve different purposes. A registry can verify what researchers planned, a repository can locate government-funded work, and an agency page can supply authoritative background. None automatically proves that this paper's conclusion is correct.
USGS Publications Warehouse ↗
The authoritative catalog of USGS scientific publications, used to check related government research and long-term observational context.
NOAA research and data ↗
Federal observations and research on climate, oceans, atmosphere and ecosystems provide context for environmental claims. They do not automatically validate a separate model or paper.
Reuse note: Facts and discoveries are summarized here in original language. We link to government material instead of copying it wholesale, and we do not reuse agency logos, photographs, charts or third-party material unless the specific reuse rights are verified.
A three-dimensional Arctic building map raised estimated permafrost damage costs
This review was developed from the source record below and, when separately available, the primary paper or government report. The summary and analysis on this page are original editorial writing.
- Source organization
- University of Connecticut
- Source type
- University
- Authors
- Elias Manos, Dmitry A. Streletskiy, Chandi Witharana, Elizabeth Haggerty and Anna K. Liljedahl
- Journal / report
- Earth's Future
- Publication date
- June 27, 2026
- DOI
- 10.1029/2026EF008578
- PMID
- Not available
- Institution
- University of Connecticut-led collaboration with George Washington University and Woodwell Climate Research Center
- Funding
- U.S. National Science Foundation grants RISE-1927723, 2052107, 2019691 and 2022504; Google.org Impact Challenge on Climate Innovation; Texas Advanced Computing Center award DPP20001
- Conflicts
- The authors declared no conflicts of interest relevant to the study
- Open access
- Yes
- Reuse approach
- Facts summarized in original language from the University of Connecticut report, open peer-reviewed paper and linked research repositories; no source wording, maps, figures, tables, satellite imagery, photographs or code reproduced.
AI-assisted editorial process: AI tools helped organize sources and draft this review. The linked research records—not AI output—are the evidence. Publication standards and corrections are publisher-directed. Read our AI transparency policy.