Science1 distinct publisher2 min readPublished
Mapping Arctic buildings by height rather than footprint lifts the projected mid-century rebuilding bill from $110 billion to $261 billion, a correction that could only run upward once floors entered the exposure model.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
A footprint model prices a building as an area. Rebuilding prices it as a volume, and the gap between those two accounting choices widens with every story above the first. That gap is what the revision corrects: it is a change in specification, from counting footprints to counting floors, not a dispute over how anything was measured [6].
The arithmetic is plain. Against the earlier $110 billion, the new $261 billion is a $151 billion increase [17], or 2.37 times the figure it replaces [18]. Adding stories to a footprint cannot subtract floorspace, so the direction of that correction was knowable before anyone ran it.
Height is the hard part, and the team got it by differencing ArcticDEM rooftop and ground elevations, which yields meters rather than stories [9]. The use-type classifier trained after the building-detection model [8] is what turns meters into floors, together with residential building codes from the U.S., Canada and Russia that fix an average floor height [10]. That conversion is also why nonresidential floorspace is missing: shops and industrial spaces vary too much for a single factor to apply [11].
Two gaps in the estimate push opposite ways. The absent nonresidential stock can only add exposure, so on its own method the $261 billion behaves as a floor. Working against it, the paper carries no demographic trends, so a depopulating resource town's empty five-story block enters the map at full floorspace [13]. That matters most where the revision is concentrated, in the Siberian communities the Soviet Union built upward to service resource extraction [7].
The uncertainty in that total comes from two separate sources, the building inventory and the permafrost degradation model from co-author Dmitry Streletskiy at George Washington University [12], and the paper does not say how much of the total's error belongs to each. A better building inventory does nothing for error in the thaw projection, and the account of the work puts a confidence interval on neither side. It also covers only the built-environment channel; the greenhouse gases released as permafrost thaws are a separate cost this number does not carry [19].
For a village council, what matters is the per-building floorspace layer underneath the hemispheric total, produced from 10 terabytes of satellite imagery on supercomputing time and commercial data that Chandi Witharana says the work would not have been possible without [8][15]. Elias Manos frames the audience as communities lacking the technical and administrative capacity to generate these estimates themselves [16]. If that is the use case, the layer is the deliverable and the headline sum is a summary of it.
Ranked by verification strength, evidence, and original report placement.
A study published in Earth's Future estimates that permafrost melt will cost $261 billion in infrastructure damages by the middle of the 21st century.
A previous study estimated damages at $110 billion, considering only buildings' two-dimensional structure, which led to significant underestimates of damage and rebuilding costs.
The work was led by Elias Manos, a doctoral candidate in the Department of Natural Resources and the Environment, and Chandi Witharana, assistant professor of NRE, alongside collaborators at George Washington University and the Woodwell Climate Research Center.
Permafrost covers about 24% of the Northern Hemisphere, including large areas of Alaska, Canada and Siberia.
Permafrost is ground that permanently remains below freezing for at least two consecutive years.
The updated estimate is much higher because the material costs to rebuild multistory buildings are much higher than those for single-story buildings.
Distinct publishers with included, body-backed reporting in this cluster.
phys.org
1 article · September 3, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
science
156 sites, 24 years: permafrost thaw is now a documented global trend, not a set of hotspots1 distinct publisher
invest
Alaska's disappearing lakes make permafrost thaw an asset-level observable1 distinct publisher
science
A pole-on magnetar hands vacuum birefringence its first astrophysical candidate1 distinct publisher
product
A magnetar may have caught empty space bending light, if its tilt is what the authors say1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Peer-reviewed paper, single institutional retelling
There is a real paper with a DOI behind this, and the method is described in enough detail that a specialist could argue with it — that is the floor under the score. The ceiling is that every checkable element reaches the reader through one write-up sourced to the authors, the $261 billion carries no range or scenario, the damage projection comes from a co-author's own model, and the earlier $110 billion study is never even named, so the comparison that makes the story cannot be inspected.
Posted to a gateway, used by no one yet
Two things have actually happened: the paper is out, and the outputs sit on the Permafrost Discovery Gateway. Nothing in this reporting shows a community, utility or agency acting on the numbers, and the authors' own account of why — the data and compute required, and the thin administrative capacity in the settlements most exposed — is the best available reason to expect uptake to lag the publication by years.
Candid about limits, silent about error bars
This reporting volunteers its own weak spots — nonresidential floorspace uncounted, occupancy trends absent, future work flagged — which is more than most research coverage does. What tilts it modestly overstated is a single point estimate doing headline duty with no range attached, and a revision whose direction was fixed before the modelling started: adding storeys to a footprint cannot lower a rebuilding bill. That the correction could only run upward is treated as a finding rather than as a property of the change in method.
The authors' institution is the only narrator
A university communications piece announcing that its own researchers have more than doubled the field's damage estimate is not a neutral vantage point, and phys.org carries it essentially intact — quotes, structure, the closing pitch that the method could extend to industrial assets. Witharana's remarks about needing supercomputers and commercial data access read as much as a case for continued funding and data agreements as a methods caveat. None of that makes the number wrong; it does mean nobody in this story had an interest in finding it smaller.
Firm on method, thin on corroboration
I would bet on the description of what the team did — it is specific, internally consistent and tied to a citable paper. I would not yet bet on $261 billion as the number the field settles on, because one publisher, one institution and one co-authored hazard model stand behind it, and the pieces most likely to move it (model accuracy, excluded nonresidential stock, occupancy) are exactly the pieces left unquantified.