Ice Sheet Stabilization

Speculative If a seabed curtain cut warm-water delivery to the Amundsen Sea grounding lines by the amount its authors model — which no experiment at any scale has tested, and section 7 names what would — the thing purchased is time on 65 cm of sea-level equivalent held in Thwaites Glacier and a brake on the 3.3 m behind it in the West Antarctic Ice Sheet. The condition is the whole proposition, so it goes first. Every number below is conditional on it.

Established The baseline is that ocean warming at the Amundsen grounding lines is now committed across emissions scenarios. Naughten, Holland and De Rydt project 0.8–1.4 °C per century of continental-shelf warming over the twenty-first century against a historical mean of 0.25 °C per century, with the 1.5 °C, 2 °C and mid-range pathways statistically indistinguishable from each other until about 2045. Speculative Decarbonisation is the only lever that works on everything else and it does not reach this variable on this timescale; that is the space a mechanical intervention claims, and it is a narrow one.

Speculative The costed claim is US$40–80 billion of construction plus US$1–2 billion a year of maintenance for an 80 km curtain in 600 m of water, set against roughly US$40 billion a year of global coastline protection attributed to the collapse the curtain is meant to delay. Frontier Those are the proponents’ own figures and section 15 takes them apart; the assessment that disputes them puts the real cost far higher on a per-kilometre comparison with the Thames Barrier. The honest statement is that the two published cost figures differ by roughly an order of magnitude and neither rests on a built structure.

Handwave The gains, if any, accrue to every coastline and the costs land on the Southern Ocean and on whoever pays, which is a distribution no existing instrument matches. Established The programme pursuing it is raising US$10 million — four to eight thousand times below its own deployment estimate — and third-party assessment puts the concept at technology readiness level 1. Speculative The upside here is narrower than the coverage suggests: what is actually on offer in the next decade is not a curtain but the measurement that would tell you whether one is worth designing, and section 5 is the reason that is the correct ambition.

Established The intervention is mechanical, local and aimed at a fluid, which is what separates it from everything else filed under climate engineering. Solar radiation management changes the planet’s energy balance and reaches the ice sheets by a long chain through the atmosphere. A seabed curtain changes nothing about the radiation budget: it is a fabric barrier moored across a bathymetric trough to keep a specific layer of warm, dense seawater from reaching the underside of a specific ice shelf. The target is heat delivery in a channel tens of kilometres wide.

Established The physical setting is a marine-based ice sheet whose bed deepens inland. The grounding line is where ice stops resting on rock and starts floating; seaward of it a floating shelf presses against seabed highs and valley walls and holds back the grounded ice behind. Where the bed slopes downward inland, retreat into deeper water thickens the ice at the line, increases the flux across it and drives further retreat — the marine ice sheet instability set out formally by Schoof in 2007, and the reason Thwaites and Pine Island are treated as a threshold system. Frontier The 2018 modelling that launched this line of work is explicit that its calving scheme cannot represent marine ice cliff instability, the faster and more contested cousin of that mechanism, so the intervention literature is silent on the failure mode that would matter most if it is real.

Established The heat arrives as modified Circumpolar Deep Water, and it arrives along routes. Nakayama and colleagues, running the MITgcm at 200 m horizontal resolution, find water reaching both the Pine Island and Thwaites grounding lines passing through a single small area near 104°W, 74.3°S along pathways 10–20 km wide, with the warm layer at roughly 0.5–1.5 °C below about 300–500 m. Speculative That topographic funnelling is the entire premise of the curtain: a forcing delivered through a narrow channel is, in principle, a forcing you can put a valve on.

Speculative The proposed object is a buoyant, flexible fabric barrier rather than a wall. Keefer, Wolovick and Moore describe buoyantly tensioned panels of low-permeability fabric with vertical tensile members and adjustable buoyancy elements, moored to alluvial sediment, designed to bend and flex under iceberg contact rather than resist it; the stated reasons for choosing fabric over rock or steel are cost, robustness to collision, and reparability or removal if side effects turn out badly. Frontier An earlier and cruder version — a gravel and sand sill, or isolated artificial pinning points — is what the 2018 ice-sheet modelling actually simulated, and the distinction matters when those model results are quoted in support of the fabric design.

Established The claimed mechanism has two steps and only the first is about the ocean. Step one: less warm water under the shelf means less basal melt. Step two: a thicker shelf flows outward, re-grounds on a seabed high, increases buttressing and reduces the ice flux across the grounding line. Speculative Step two is where the sea-level benefit lives and it depends on ice dynamics rather than on the barrier working; section 3 separates what is measured in each step from what is modelled.

Established Thwaites Glacier is losing about 50 billion tonnes of ice a year and supplies about 4 per cent of present-day global sea-level rise. The International Thwaites Glacier Collaboration puts the basin at 192,000 km², grounding-line water depths at 800–1,200 m, cumulative loss above 1,000 billion tonnes since 2000 and a doubling of the loss rate over thirty years. The sea-level equivalent held in Thwaites alone is 65 cm; the West Antarctic Ice Sheet as a whole is 3.3 m. Those four numbers are the reason anyone is proposing to build anything at all in the Amundsen Sea.

Established The ocean heat is real, it is measured, and it is delivered along a small number of deep channels. Nakayama and colleagues resolve pathways 10–20 km wide converging on a single area of the shelf, with modelled melt integrated over the Pine Island and Thwaites ice shelves at 64.8 and 65.1 Gt a year and peak rates of 70–80 m a year. Established The same group traced the water upstream: Circumpolar Deep Water released across 67°S reaches these shelves after three to five years, and on- and off-shelf water was 0.1–0.2 °C warmer in 2009–2014 than in 2001–2006, with Ross Gyre strength at 21 Sv against 13 Sv and the source region displaced about 900 km west. Frontier A curtain is therefore a local valve on a supply set thousands of kilometres upstream, on a three-to-five-year lag, by a gyre nobody proposes to manage.

Established The single strongest piece of evidence that removing ocean heat changes the ice is not a model — it is the 2011–2013 cool phase. Jenkins and colleagues report that melt in the Amundsen changes “by a factor of about four between cool and warm extremes” through a nonlinear relationship with ocean temperature, and that the Pine Island grounding line “re-advanced during a post-2011 cool phase”. Webber and colleagues put numbers on the same event: heat content between 400 and 700 m in Pine Island Bay fell 62 per cent from spring 2009 to spring 2012, from 3.28 GJ to 1.25 GJ, with interannual reversals in the currents carrying heat into the bay. Frontier This is the steelman for the whole proposal, and it is stronger than anything its proponents have modelled: the natural system has already run the experiment of withholding warm water from an Amundsen grounding line, and the grounding line moved the way the intervention predicts. Speculative What the natural experiment cannot tell you is whether a fabric barrier can produce the same heat reduction, or hold it for centuries rather than eighteen months.

Established The counterweight is that basal melt is not a simple function of how much heat you deliver. Two Nature papers from the same January 2020 borehole campaign made this concrete. Davis, Nicholls, Holland and colleagues measured melt at the eastern Thwaites grounding zone at “no more than 2.0–5.4 m year−1” despite water well above the in-situ freezing point, because density stratification in the ice-ocean boundary layer suppressed vertical heat transport; they concluded the canonical melt parameterisation used for sea-level projections cannot reproduce observed rates and substantially overestimates boundary-layer transport. Established Schmidt and colleagues, working the same site with the Icefin vehicle, found melting concentrated by geometry rather than bulk heat: about 5 m a year on flat surfaces, approaching 30 m a year on vertical faces, up to 43 m a year in crevasses, with slopes steeper than 30° contributing roughly 27 per cent of total melt from 9 per cent of the area. Frontier Together these say the transfer function from heat delivered to the cavity to metres of ice removed is controlled by stratification and ice-base shape, and is the part of the system the models get wrong. A curtain acts on the input to a function nobody can yet write down.

Frontier The design depth band is warming now. Wild and colleagues, reporting three years of borehole instruments under the Thwaites Eastern Ice Shelf from January 2020 to January 2023, document sustained warming and thickening of the modified Circumpolar Deep Water layer near the seabed at 745–784 m: about 0.04 °C a year to July 2021, a plateau, then 0.01–0.02 °C a year through 2023. Speculative A curtain designed to block a layer that is simultaneously warming and thickening has a moving target, and no published design states the temperature and thickness it is specified against.

Frontier The committed-warming result is the strongest argument the proponents have, and it was not written by them. Naughten, Holland and De Rydt project 0.8–1.4 °C per century of continental-shelf warming against a historical 0.25 °C per century, conclude that mid-range and Paris-aligned scenarios are statistically indistinguishable given internal variability until RCP 8.5 diverges around 2045, and write that “the opportunity to preserve the WAIS in its present-day state has probably passed”. Speculative If that holds, the choice on this variable is not between intervention and mitigation but between intervention, adaptation and acceptance. This brief does not discount it.

Speculative The efficacy case rests on one 2018 run in a deliberately simple model. Wolovick and Moore simulated Thwaites with a quasi-two-dimensional ice flowband coupled to a turbulent buoyant plume ocean model and reported regrounding in 30 per cent of runs for isolated pinning points, 70 per cent for a small sill blocking half the warm water, 90 per cent for a small sill blocking all of it and 100 per cent for a large sill blocking all of it, with aggregate volumes of 0.1–1.5 km³ for pinning points and one to two orders of magnitude more for continuous sills. Frontier The authors describe their model as “the least complex model that can address these questions”, say it “may miss important elements of the ice-ocean system” including lateral buttressing and ocean circulation, note that it cannot include marine ice cliff instability, and state that “exact values of collapse timing, sea level rise rate, success probability” will change with better models. Those percentages are ensemble fractions inside a stated-to-be-crude model, and they are quoted in secondary coverage as probabilities that an intervention would work. They are not that.

Speculative The two 2023 curtain papers used no ocean model at all. Wolovick, Moore and Keefer sited four candidate routes across the Amundsen Sea Embayment — Trough 2 at 4.3 km, Inner Bay at 35 km, Trough 3 at 50 km, Outer Bay at 82 km — with mean curtain heights of 79–130 m, maxima of 142–455 m, maximum water depths of 679–971 m and an optimal blocking depth of 500–550 m for every route. The oceanography behind the siting is World Ocean Database conductivity, temperature and depth casts, not a circulation model; the paper states that “the detailed fluid-structural modeling that would be required to accurately characterize ocean heat flux over and through the curtain is beyond the scope of this paper”, and names “leakiness” as one of the largest unknowns affecting effectiveness. Frontier The companion engineering paper carries the costed claim: an 80 km curtain in 600 m of water on alluvial sediments at US$40–80 billion plus US$1–2 billion a year, with polar installation presenting severe challenges that “can however, be overcome with present-day technology”. Frontier Note that the author order differs between the two papers: the modelling paper is Wolovick, Moore and Keefer, the engineering paper is Keefer, Wolovick and Moore, and the 80 km figure belongs to the second.

Frontier In 2025 a large multi-author assessment concluded that the concept does not survive scrutiny, and it said so about the whole family. Siegert and colleagues, writing in Frontiers in Science, assessed stratospheric aerosol injection, sea curtains and walls, sea-ice management, basal water removal and ocean fertilisation, and concluded: “According to our expert assessment, none of these geoengineering ideas pass scrutiny regarding their use in the coming decades”, and that “the assessed approaches are not feasible, and that further research into these techniques would not be an effective use of limited time and resources”. Established On curtains the assessment reports that 56 per cent of research cruises to the region experienced at least partial disruption from sea ice or significant difficulty entering or leaving, 22 per cent could not access the region at all, and the Thwaites Inner Bay site “has only been accessed once in the last 40 years despite frequent expeditions”; that each transit for a single curtain section “would probably take well over a month, with a likely maximum of four to five round trips per year being possible”; and that installation “is likely to have far-reaching detrimental marine environmental consequences”. Frontier On cost it argues the real figure will “probably far exceed” US$80 billion, citing the Thames Barrier at GBP 2.4 billion, or US$6.2 billion per kilometre at 2024 prices, for a shallower and simpler structure. This brief treats that assessment as a serious finding, not as an obstacle to be routed around.

Established The programme that would build it is a philanthropically funded consortium at technology readiness level 1. The Seabed Curtain Project lists Aker Solutions as engineering partner alongside Cambridge, the University of Chicago, the Alfred Wegener Institute, New York University, Dartmouth, NIVA, the Korea Polar Research Institute and the Arctic Centre at the University of Lapland, funded by the Tom Wilhelmsen Foundation, Outlier Projects and the UArctic Foundation, with a stated fundraising goal of US$10 million (programme’s own figures). The registry climateinterventions.org scores the concept at technology readiness level 1 and states that it “currently does not fit Antarctic governance structures”. Frontier Aker Solutions’ own 2022 announcement describes feasibility and front-end engineering consultancy for a Greenland application and names no dimensions or costs (vendor). The gap between a US$10 million research budget and a US$40–80 billion deployment estimate is itself a finding: nothing at the funded scale can test the deployed claim.

Frontier The first open question is whether the Amundsen trend is forced or oscillatory, because a curtain is a different proposition against each. Jenkins and colleagues argued that “the strength and timing of decadal extremes is more influential than changes in the longer-term mean state”; Naughten and colleagues project committed warming that swamps scenario differences until 2045. Frontier Both can be true, and the design consequence is sharp: a structure sized against the mean is under-specified for the excursions, and one sized against the excursions is over-built most of the time. No published design states which it is sized for.

Frontier The second is leakiness, which the proponents name as their own largest unknown and nobody has quantified. A buoyant fabric barrier in a stratified flow does not simply stop water: it diverts some, spills some over the crest, passes some through seams, and generates its own mixing. Speculative The fraction of the warm layer actually withheld at the grounding line — the quantity the 2018 ice model parameterises as 50 or 100 per cent blocking — is a free parameter, and section 7 argues it is the number a field programme should be organised to measure.

Frontier The third is rerouting. The curtain papers cite work indicating that blocked warm water might reroute to other ice shelves, reducing the net benefit. Speculative Given that the heat arrives on a gyre-scale supply with a three-to-five-year transit, the interesting version of the question is not whether some water goes elsewhere but whether the whole shelf-break exchange adjusts — whether a barrier at 500–550 m changes the density structure driving the intrusion. Nobody has run that experiment in a model, let alone in water.

Frontier The fourth is whether Thwaites is still ocean-controlled. This corpus’s Polar Development brief reports a conference abstract by Williams and colleagues for the 2026 European Geosciences Union assembly claiming that with ocean melting set to zero for 150 years Thwaites continues to lose ice and cannot re-advance under any ocean cooling. Speculative This pass could not retrieve that abstract, so this brief reports it only as its sibling reports it and attaches no weight to a claim it has not read. It does define the test: if the glacier’s trajectory is set by internal dynamics and bed geometry rather than ocean forcing, the intervention fails on its founding premise rather than its budget.

Frontier The fifth is the melt parameterisation, and it is a shared dependency of the proposal and its critique. If the canonical melt law overestimates boundary-layer transport, then projections of unmitigated melt and projections of a curtain’s benefit inherit the same error in the same direction.

Speculative The sixth is what a curtain does to the ecosystem it sits in, where the literature is empty. Siegert and colleagues list the mechanisms — ocean circulation, sea ice, delivery of nutrients and most notably iron, primary productivity, carbon sequestration, and the seabirds and marine mammals that feed at depth — and note that the Commission for the Conservation of Antarctic Marine Living Resources may also have concerns. Handwave Neither curtain paper contains an ecological assessment; both call for one. A barrier at the optimal blocking depth for warm water is, by construction, placed across the upwelling that fertilises the surface ocean above it, and nobody has estimated the sign of that effect.

Established The binding constraint is not fabric, cost or politics. It is that no coupled ice-ocean model has ever been run with a curtain in the domain. The published efficacy chain runs from CTD casts, to an assumed blocking fraction, to a 2018 flowband ice model its authors call the least complex that can address the question. Frontier Every other bottleneck below is downstream of that gap: until the chain is closed nobody knows what the structure has to achieve, so nobody can say what it has to survive.

Frontier Bathymetry is the second constraint and it is the one that could move fastest. The siting of all four candidate routes depends on seabed depth and trough geometry in a sector mapped largely from airborne gravity inversion and sparse ship tracks — Millan and colleagues’ Operation IceBridge compilation is the standard source and this brief could not retrieve it this pass. Nakayama and colleagues state directly that accurate and high-resolution ocean bathymetry and subglacial topography remain essential for determining pathways and melt rates. Speculative A curtain is a structure whose height specification is the difference between two poorly known surfaces, and a 100 m bathymetric error over a 79 m mean curtain height is not a refinement.

Established Access is a hard physical limit and it is quantified. Over four decades, 56 per cent of research cruises to the region were at least partly disrupted by sea ice and 22 per cent could not get in at all; the Inner Bay site has been reached once in forty years; a single curtain section’s transit takes well over a month and four to five round trips a year is the likely maximum. Frontier The arithmetic follows: a curtain assembled from N transported sections takes at least N divided by four-and-a-half years to place. Neither curtain paper publishes N, so neither publishes an installation schedule, and the US$40–80 billion figure has no duration attached beyond “a decade of construction”.

Speculative Iceberg scour is the gap with the least literature and the most load-bearing role. The design case for fabric over rigid structures is that it bends under iceberg contact. Lockley and colleagues produced the only geotechnical treatment this brief could find and it offers no scour-depth analysis; the 2025 assessment notes steep slopes and strong scouring currents without giving keel statistics for the candidate routes. Handwave The central survivability argument for the chosen material rests on no published measurement of the load it is chosen to survive.

Speculative Anchoring into soft sediment at 600–1,000 m under sustained hydrodynamic load for centuries has no precedent. Offshore moorings at these depths are routine, but they are designed for decades, inspected by vessels that can reach them most of the year, and carry point loads rather than a continuous line load along tens of kilometres. Frontier Lockley and colleagues, working the adjacent problem of ice-shelf tensile reinforcement, concluded that cable networks with self-embedding plough devices would be “extremely expensive in terms of material, installation and continuing operational costs” — the nearest published anchoring assessment in the family.

Established The last constraint is perpetuity, and it binds institutions rather than materials. US$1–2 billion a year, indefinitely, for a structure whose benefit is diffuse and whose failure would be discovered slowly. Handwave No mechanism has ever committed spending at that level, for that duration, for a good with no owner; section 10 is where that becomes the actual blocker.

Established The first dependency is already being addressed and is the cheapest thing in this brief. The Seabed Curtain Project states that mooring robots are being deployed in the Pine Island Thwaites trough, with a mooring expedition in early 2026 and a Svalbard natural-analogue report expected in September 2026 (programme’s own figures); secondary reporting describes two underwater robots travelling up and down wires to measure water speed, temperature and salinity over a two-year assignment. Frontier Time series of warm-layer thickness, temperature and transport at the candidate sites are a prerequisite for any design and are useful science whether or not anything is built — the honest test of whether an intervention research programme is worth funding.

Frontier The second is a coupled ice-ocean model with a structure in it. Three components exist separately: a cavity-resolving ocean model at the resolution Nakayama and colleagues used, an ice-sheet model with grounding-line dynamics, and a fluid-structure representation of a permeable, compliant barrier in a stratified flow. Speculative The third is the missing piece, and the 2023 modelling paper says so in terms.

Frontier The third is a melt parameterisation that matches observations. Both the suppressed-melt and heterogeneous-melt results are three years old and no replacement has become standard; the modelling infrastructure question belongs with Earth-System Digital Twins.

Speculative The fourth is an ecological baseline for the Amundsen Sea at curtain depth. The objection that a barrier at 500–550 m would interfere with iron delivery, productivity and deep-feeding predators cannot be answered either way without a benthic and pelagic baseline that does not exist. Handwave This is the dependency most likely to be skipped and least recoverable if it is.

Established The fifth is materials qualification. Keefer, Wolovick and Moore state that suitable materials are commonly available. Frontier That is a claim about existence, not qualification: a low-permeability fabric with tensile members holding design tension at 600 m, at freezing temperatures, under abrasion from ice and sediment, for a service life in centuries, has no qualification standard and no test article. The nearest mature discipline is deepwater mooring, covered at Floating Offshore Wind, and its design lives are decades.

Speculative The sixth is a legal instrument, and it is a dependency rather than a consideration: the registry states the concept does not fit Antarctic governance structures as they stand. No amount of oceanography removes that, and section 10 treats it as the design problem it is.

Speculative The decisive experiment is a kilometre-scale curtain moored across an instrumented fjord sill and measured for one year on two quantities: the reduction in warm-water flux past the sill, and whether the moorings survive a winter of iceberg scour. Everything else in this brief is a model, a cost estimate or an argument. This is the one result that would move the assessment in either direction, because it converts the two largest free parameters — leakiness and survivability — into measurements.

Established Something close to it is scheduled, and it is not close enough. The programme states a staging sequence of tank tests, field tests in a Norwegian fjord, a Svalbard pilot, then possibly Greenland and Thwaites, and secondary reporting gives the fjord test as Ramfjorden near Tromso in 2027 (programme’s own figures). Frontier A test in a sheltered, largely iceberg-free Norwegian fjord can measure hydrodynamic blocking, mooring loads, deployment method and fabric behaviour; it cannot measure scour survival, because the scouring agent is not present. Speculative The scour half needs a fjord with iceberg keels — the Svalbard pilot or a Greenland site — and the programme’s sequence puts that after the Norwegian test rather than alongside it.

Frontier The design of the sill experiment matters more than its scale. What counts is the difference between warm-water flux across the sill with and without the curtain, resolved over the seasonal cycle, with enough moorings on both sides to close a heat budget rather than record a temperature. Speculative A single-season before-and-after comparison, in a fjord with interannual variability of the size Webber and colleagues documented in Pine Island Bay, will not separate the structure’s effect from the weather. The experiment needs a control sill or a removal-and-replacement cycle, and neither appears in any published plan.

Frontier A second experiment is already running and nobody designed it as one. The Svalbard natural-analogue study compares two fjords with different warm-water access, the closest available substitute for a controlled blocking trial. Speculative Its value depends on whether the two fjords differ in sill depth alone; if they differ in glacier geometry, runoff or circulation as well, it answers a different question. The September 2026 report would settle that, and this brief could not obtain it.

Frontier The third experiment is computational and is the cheapest decisive test available. Run a cavity-resolving ocean model of the Amundsen with an explicit permeable barrier at the published Outer Bay geometry, and report the fraction of warm-layer heat transport withheld at the grounding line as a function of curtain permeability and crest depth. Speculative If a realistic barrier withholds a small fraction, the 2018 ice-model runs that assume 50 or 100 per cent blocking stop being relevant and the proposal is finished on physics rather than cost. If it is large, the engineering question is worth its budget. Nobody has published this run.

Frontier The fourth is the one the critics should want: a published bottom-up cost model — materials, vessel days, mooring count, installation rate — which would either confirm the order-of-magnitude gap or close it. Speculative That both sides are arguing from top-down analogies is why the cost dispute has not converged.

Established The natural experiment that already ran is the benchmark any curtain must beat. The 2011–2013 cool phase reduced 400–700 m heat content in Pine Island Bay by 62 per cent and the grounding line re-advanced. Frontier Any proposed curtain should state what fraction of that natural heat reduction it expects to reproduce, and for how long. No published design states it, and the omission is the clearest indicator of how early this work is.

Handwave No experiment currently planned would settle whether Thwaites remains ocean-controlled. That question is answered by ice-sheet modelling and grounding-zone observation, not by curtain tests, and it sits upstream of all of them. Speculative A programme spending its first decade on structures rather than on that question is optimising the wrong stage.

Established Start with the area of fabric, because it is computable from the published geometry and nobody quotes it. The Outer Bay route at 82 km with a mean height of 79 m is about 6.5 km² of barrier; the Inner Bay route at 35 km and a mean 127 m is about 4.4 km². Frontier If instead the curtains are the 150–500 m structures the 2025 assessment describes, an 80 km line is 12–40 km². That discrepancy is unresolved in the literature, and a factor of six in area is a factor of six in material, vessel days and mooring count.

Frontier The structural concept is a tensioned membrane, not a dam, and that choice sets every other requirement. Buoyancy elements at the crest hold the fabric up; vertical tensile members carry load to seabed anchors; the membrane is low-permeability rather than impermeable. Speculative A membrane in a current develops drag proportional to blocked area and transfers it to the anchors as a distributed line load, with the design condition set by peak current during a warm intrusion. No published figure gives anchor load per metre for any of the four routes, so none gives anchor count, so the installation estimate has no unit of work behind it.

Established Installation is vessel-limited and season-limited, and both limits are measured. Four to five round trips a year, each transit over a month, into a sector where more than half of historical cruises were disrupted. Frontier The realistic comparison is not a Thames-style civil project but deepwater mooring campaigns for floating production and floating wind, where a single suction-anchor leg is a multi-day vessel operation in temperate water — covered in this corpus at Floating Offshore Wind, and the transfer of its cost curves to an ice-covered sea with a three-month season is the unexamined step.

Speculative Reparability is claimed as a design virtue and is untested at depth. Handwave Nothing of this kind has been repaired or removed at 600 m anywhere, and removal does not reverse the ecological effects of years of installation. Reversibility that depends on an unperformed operation is an assertion.

Frontier The Greenland variant is a different engineering problem with much larger published claims. Hunt and Byers costed submerged steel barriers across ten Greenland glaciers at US$68.9 billion and rock-filled dams at US$337.1 billion, assuming 200 cm marine-grade steel, a 500 m maximum depth limit, a 150 km affected glacier width and 1,000 m ice thickness, and projected up to 5.3 m of avoided sea-level rise by 2300. Speculative Those assumptions are uniform across glaciers by the authors’ own admission and the ice-dynamic coupling is deferred to future work, so the figure should not be quoted alongside the Antarctic numbers as the same kind of estimate. Frontier A much smaller Greenland concept — a roughly 100 m high, three-mile barrier at the mouth of Ilulissat Icefjord at about US$500 million — is the one Aker Solutions was engaged on, and it sits at the entrance to a UNESCO World Heritage site.

Established Two Greenland-specific facts change the engineering trade. Roughly half of Greenland’s mass loss is attributed to atmospheric rather than oceanic forcing, so a barrier that perfectly blocked warm water addresses at most half the problem. Frontier And fjord sills are shallower than Antarctic troughs, which makes the structure easier and the natural system already partly self-blocking. Speculative Greenland is therefore the better test site and the worse deployment site, an awkward position for a programme that needs both.

Frontier The comparison class for schedule and cost behaviour is megaproject delivery, covered at Megaproject Governance. Speculative The proponents’ US$40–80 billion and the assessment’s much higher implied figure are not competing estimates; they are one top-down analogy against another, and section 7 names the bottom-up cost model neither side has published.

Established The closest sibling in this corpus is the family of cooling levers, and the relationship is contrast rather than overlap. Planetary Cooling Concepts owns every cooling approach that is not stratospheric aerosol injection, and it is about the radiation budget; this brief is about a fluid in a channel. Frontier The two share a governance structure and a moral-hazard argument and no physics: a curtain changes no radiative forcing anywhere, produces no termination shock in the atmospheric sense, and its failure mode is a structure on a seabed rather than a rebound in temperature.

Frontier The governance neighbour is the one that decides the outcome. Geoengineering Governance carries the record of failed multilateral attempts and of the consent veto that stopped an outdoor experiment before any regulator ruled, and Climate Engineering carries the general case. Speculative What this brief adds is that Antarctica has no residents, so the consent question that has been decisive elsewhere is answered by treaty proxy — and the Greenland variant inverts that.

Established The demand side is coastal defence. Coastal Defense Systems consumes Antarctic sea-level projections as a design basis, which makes it the brief revised first by either outcome here. Frontier The proponents’ central cost comparison is a claim about that brief’s subject, not this one, and it is doing a lot of the persuasive work.

Frontier The timing question belongs with overshoot and with tipping-point detection. Climate Overshoot and Lock-In owns the asymmetry that temperature reverses and sea level does not, which is the reason this proposal exists at all; Tipping Point Detection owns the question of whether anyone could tell in advance that the threshold had been crossed. Speculative If the detection machinery cannot make a scored forecast, then the decision to build is a decision under a threshold nobody can time, which changes the value of the option rather than the physics.

Established The engineering neighbours are marine. Ocean Engineering covers seabed operations and their regulatory vacuum; Floating Offshore Wind is installing deepwater moorings at volume; Arctic Engineering carries cold-region foundations. Frontier The adjacency missing from all three is scour: an iceberg-keel load spectrum for a fixed subsea structure is a requirement none of them produces.

Frontier The observational neighbour is autonomous under-ice vehicles and fibre sensing. The two Nature grounding-zone results came from a hot-water borehole, a torpedo-shaped vehicle and distributed temperature sensing cable, as did the three-year cavity record. Speculative Every measurement this brief says is missing is one that hardware of this kind could make; the constraint is ship time and drilling seasons, not sensors.

Established The Antarctic instrument is the Protocol on Environmental Protection, and the provision that bites is its environmental impact assessment requirement. The 2025 assessment cites Article 8 of the Protocol as requiring a Comprehensive Environmental Evaluation, and notes that the Commission for the Conservation of Antarctic Marine Living Resources may also have concerns. Frontier This brief could not retrieve the treaty text this pass and reports those provisions as that assessment cites them; the wider structure is covered at Geoengineering Governance.

Speculative No Comprehensive Environmental Evaluation has ever been prepared for a permanent industrial structure in the Antarctic marine environment, which makes the precedent the real instrument. The evaluations that exist cover stations, runways, drilling and traverse routes. Handwave An 80 km barrier is a different category of object, and the first one approved would define what the Treaty system permits for every subsequent proposal.

Established The consent structure is unusual and it cuts both ways. Secondary reporting describes the Antarctic Treaty System as a network of 57 nations; there is no resident population whose consent is at stake, which removes the veto that has repeatedly stopped outdoor climate experiments elsewhere. Frontier Greenland is the opposite case: Ilulissat Icefjord is a UNESCO World Heritage site with a town at its mouth, and the University of Lapland’s GRISCO project — a EUR 54,000 Nordregio-funded study running from August 2021 to October 2022 under John Moore — was built around knowledge co-production with the community precisely because an intervention affecting the fjord cannot be designed without fishing and tourism livelihoods in it.

Frontier The funding structure is philanthropy plus a university network, and that shapes what gets asked. The consortium runs through the University of the Arctic’s Frozen Arctic Conservation thematic network, led by John Moore, with money from the Tom Wilhelmsen Foundation and Outlier Projects. Speculative Philanthropy is what allows an unpopular question to be asked at all, and also what allows it to be asked without the adversarial review a national programme would impose. The field’s disagreement is largely about which of those dominates.

Frontier The critique came with an explicit funding recommendation, which is why it is contested as strongly as it is. Siegert and colleagues did not only judge the concepts infeasible; they wrote that further research into them “would not be an effective use of limited time and resources”, that limited public funding should focus on Earth-system tipping points, and they note the 2022 proposal by 400 scientists for a non-use agreement on solar geoengineering with five principles including no public funding and no outdoor experiments. Frontier Expert reaction collected by the Science Media Centre divided sharply: van de Flierdt held that all the methods are “scientifically flawed, unproven, dangerous, or logistically unfeasible” and Davies called the assessment long overdue, while Watson objected that it reads as “a group of concerned scientists presenting a view with unreasonable certainty”, Irvine called it one-sided and Hunt called continued research essential and urgent.

Established There is no payer. The benefit is avoided sea-level rise distributed across every coastline over centuries; the cost is concentrated, immediate and continuing. Handwave Nothing in this literature proposes a mechanism — not a treaty fund, not a coastal-cities compact, not an insurance pool — and section 14 records it as a requirement rather than a detail.

Frontier The moral-hazard argument is the sharpest and the least settled. Siegert argues that “if there is a sense of a solution other than decarbonizing, for some people, that will be appealing” and calls that a dangerous message. Frontier Hunt’s reply in the Science Media Centre round-up is that there is “no evidence” supporting the claim that geoengineering research deters emissions reduction. Speculative Both are claims about an empirical question — does intervention research measurably reduce mitigation effort — and neither cites a measurement of it. Stating that is more useful than picking a side.

Frontier The proponents answered with a reframing rather than with evidence. Moore, Macias-Fauria and Wolovick published a viewpoint alongside the assessment arguing for a “compassionate harm reduction paradigm”, that intervention research and decarbonisation are not mutually exclusive, and that “only more research can answer” whether interventions are viable. Speculative It is an argument about the ethics of inquiry, not about curtains, and it does not engage the engineering critique: the most visible proponent response to the most detailed technical objection is a position paper on research freedom.

Established The published rebuttal that does exist is about a different technology. Henry and Duffey’s commentary disputes the assessment on stratospheric aerosol injection — winter polar cooling, injection-latitude specificity, and the omission of quantitative side-effect assessments — and does not address sea curtains. Frontier As of this writing no peer-reviewed quantitative rebuttal of the curtain-specific engineering and access critique has been published. That absence is a finding about the state of the debate, not a verdict on the technology.

Speculative The intergenerational commitment is the ethical problem the cost figures hide. US$1–2 billion a year for centuries binds every subsequent generation, because abandonment is worse than never starting once coastal investment has assumed the structure holds. Handwave No published design states what happens on default — a point this corpus’s Polar Development brief makes about the same proposal from the other direction.

Frontier The ecological objection has moral weight independent of its magnitude. The assessment holds that curtain installation is likely to have far-reaching detrimental consequences for circulation, sea ice, iron delivery, productivity, carbon sequestration and deep-feeding predators. Speculative One defensible position is that an intervention intended to protect a system should not be the largest physical disturbance ever imposed on it; the counter-position is that unmitigated warming will disturb it more. Neither has the quantified comparison that would settle it.

Established The Greenland case has a rights-holder and the Antarctic case does not, and that is the most consequential ethical asymmetry in the subject. The GRISCO work ran community workshops, school collaboration, interviews and art across three visits in 2021–2022 in Ilulissat and Nuuk, concluding that an intervention affecting the fjord cannot be planned without local sustainability in the design. Frontier Antarctica offers no such check. A programme that learns its consent practice in Greenland and then deploys where consent is a treaty formality has moved to a weaker ethical regime, and should say so.

Speculative If ice-sheet interventions worked, sea level would become a managed variable rather than a boundary condition, and that is a category change in what a coastline is. Insurance, mortgage terms, port design life and municipal bond horizons all currently treat sea level as exogenous. Handwave A world in which it is a policy output has different property law, and whoever operated the structures would hold a concentration of leverage with no precedent and no allocation rule.

Frontier The likelier civilizational effect is not the structure but the belief in it. Naughten’s point that fifty years of lead time makes all the difference to a coastal region that must be abandoned or re-engineered is the adaptation case, and it competes for the same decades. Speculative An intervention fifteen years from a decision point can defer adaptation decisions that have to be made now, and that deferral is borne by the places least able to recover it.

Established The precedent effect inside the Antarctic Treaty system is real regardless of whether anything is built. Frontier A Comprehensive Environmental Evaluation for an 80 km barrier would be the largest test the Protocol has faced, and the answer either way becomes the template for the mineral, fisheries and infrastructure questions queued behind it.

Speculative There is a version of this that is quietly valuable and does not involve building anything. The observational programme the proposal requires — moored time series at the candidate sites, cavity-resolving ocean modelling with structures, an Amundsen ecological baseline, a measured melt law — is the programme Antarctic science would want anyway. Frontier If the intervention question is what releases funding for measurements that decarbonisation arguments have not, the net return may be scientific rather than structural. That is a weaker claim than the programme makes and a stronger one than its critics allow.

Handwave The failure mode worth naming is a half-built structure. A decade of construction, four to five vessel round trips a year, a novel legal instrument and a philanthropic funding base has an obvious discontinuity: partial installation that blocks some flow, alters some circulation, then is abandoned when the money or the politics changes. Speculative Nobody has modelled a partially complete curtain, and it is the configuration a real project would spend years in.

These horizons track the decision points the programme and its critics have both put on the record, not the date a structure could exist.

• 10 yr: Frontier Moored instrument records from the Pine Island Thwaites trough, tank tests, the Svalbard natural-analogue comparison and a Norwegian fjord field test stated for 2027. Speculative Expect the first published fluid-structure model of a permeable barrier in a stratified flow, and expect the cost dispute to stay unresolved until someone publishes a bottom-up estimate. Speculative Do not expect an iceberg-keel load spectrum for the candidate routes unless a funder asks for one specifically.
• 25 yr: Speculative The programme’s own fifteen-year decision horizon lands in the early 2040s and requires peer-reviewed science, proven technology and a legal framework, of which the third does not exist. Handwave A Comprehensive Environmental Evaluation would have to be drafted, reviewed and accepted inside this window, and no state has volunteered to sponsor one. Frontier Coastal planning at Coastal Defense Systems will have made its design-basis decisions on this horizon whether or not the curtain question is settled.
• 50 yr: Speculative The earliest horizon on which an Amundsen installation could plausibly be complete, given a decade of construction preceded by two decades of qualification and consent. Handwave Naughten’s committed warming is fully realised well inside this window, so any structure is operating against a warmer and thicker warm layer than any published design assumes.
• 100 / 250+ yr: Handwave The maintenance regime, if it exists, is the oldest continuously funded engineering commitment in history, and the institution operating it does not yet exist. Speculative On this horizon Thwaites is inside its own collapse window under most projections, so the intervention is either long since irrelevant or is the reason the window did not close — and no measurement planned today would let a future reader tell which.

• Depends on the sea-level projections that Coastal Defense Systems consumes as a design basis and that Climate Overshoot and Lock-In shows do not reverse when temperature does; on the detection question owned by Tipping Point Detection, because a threshold nobody can time is a threshold nobody can schedule against; and on the consent and treaty record held by Geoengineering Governance. Polar Development reaches the same proposal from the Arctic side and reports the zero-ocean-melt objection this brief could not retrieve.
• Requires (not on this map) first, a cavity-resolving ocean model run with an explicit permeable, compliant barrier in the domain, which is the missing link between the CTD siting work and the 2018 ice-model runs; second, a measured distribution of iceberg keel depths and scour frequency along the candidate routes, because bending under iceberg contact is the stated reason for choosing fabric and no published measurement supports it; third, vessel and mooring installation capacity able to work a three-month season behind sea ice that turned back 22 per cent of historical expeditions; fourth, an accepted Comprehensive Environmental Evaluation for a permanent marine structure under the Antarctic environmental protocol, of which there is no precedent; fifth, an actual payer for a benefit distributed across every coastline over centuries, which no treaty fund, insurance pool or coastal compact currently provides; and sixth, a qualified supply chain for low-permeability fabric and tensile members with a service life in centuries at 600 m, where the nearest commercial standards are written for decades.
• Enables nothing else on this map directly: this is a terminal technology, not an enabling one. Its indirect product is measurement — moored ocean time series at the grounding lines, an Amundsen ecological baseline and a melt law that matches observation — which would improve the projections that Coastal Defense Systems and Earth-System Digital Twins depend on, whether or not anything is built.
• Adjacent to Planetary Cooling Concepts, which owns every non-aerosol cooling lever and shares this brief’s governance argument while sharing none of its physics; to Climate Engineering for the general case; to Ocean Engineering and Floating Offshore Wind for the seabed and mooring disciplines that would have to supply it; to Arctic Engineering for cold-region foundations; and to Megaproject Governance, which is the literature that predicts how an 80 km first-of-a-kind subsea structure with no unit-cost basis actually behaves.

Frontier “Modelling shows a curtain has a 70 to 100 per cent chance of saving Thwaites.” Those figures are real and mean something much narrower. They are the fraction of runs in which the grounding line regrounded, in a quasi-two-dimensional flowband ice model coupled to a plume ocean model, under an imposed assumption that 50 or 100 per cent of warm water is blocked. Established The authors call it the least complex model capable of addressing the question, say it may miss lateral buttressing and ocean circulation, and state that it cannot include marine ice cliff instability. Speculative An ensemble fraction conditional on perfect blocking in a simplified model is not a probability that a structure would work, and the 2023 curtain papers that inherited the claim used no ocean model at all.

Frontier “It is cheap — US$40–80 billion against US$40 billion a year of coastal protection.” This is the proponents’ framing and it has three problems. Established First, the 2025 assessment’s per-kilometre comparison — the Thames Barrier at US$6.2 billion per kilometre at 2024 prices — implies roughly US$500 billion for 80 km; that arithmetic is an order-of-magnitude illustration rather than a costing, because a movable steel gate across a tidal river is not a buoyant fabric membrane, and this brief presents it as the scale argument it is. Frontier Second, the comparison is not like for like: coastal protection defends against all sources of sea-level rise, not against Thwaites alone. Speculative Third, neither figure is bottom-up. There is no published anchor count, no mooring unit cost, no fabric quantity and no installation rate for any of the four routes.

Established “This is solar geoengineering by another name.” It is not, and the distinction is load-bearing. A curtain alters no radiative forcing, has no global temperature signature, produces no termination shock in the atmospheric sense, and its effects are confined to one sector of one ocean. Frontier What it shares is the governance vacuum, the moral-hazard argument and the pattern of being assessed as part of a family: the 2025 paper judged five concepts together and the most visible published rebuttal disputes only the aerosol chapter. Speculative Treating the family verdict as a verdict on this member is the error both sides make, in opposite directions.

Frontier “The 2025 assessment settles it.” It is the most detailed technical critique in existence and its access and logistics numbers are specific, checkable and damaging. Frontier It is also contested by named scientists on methodological grounds: Watson called it a view presented with unreasonable certainty, Irvine called the analysis one-sided, Fitzgerald objected that it presents a single viewpoint, and Williamson, who welcomed scrutiny of specific proposals, criticised its use of the word geoengineering as inconsistent. Established Against that, van de Flierdt, Davies and Henley endorsed it squarely. Speculative A reader should take the verdict on curtains seriously and treat the assessment’s funding recommendation — that further research would not be an effective use of limited resources — as the contested part, because it is a judgement about research portfolios rather than about seabed structures.

Frontier “Blocking warm water has never been shown to help.” This is the claim this brief overturns. Established The 2011–2013 cool phase in Pine Island Bay reduced heat content at 400–700 m by 62 per cent, melt across the Amundsen varies by a factor of about four between cool and warm extremes, and the Pine Island grounding line re-advanced during that cool phase. Frontier The physics premise is not speculative; it is observed. What remains speculative is whether a structure can reproduce any useful fraction of that reduction, and for how long.

Frontier “Less ocean heat means proportionally less melt.” The measurements say otherwise. Established At the eastern Thwaites grounding zone melt was 2.0–5.4 m a year despite water well above freezing, because stratification suppressed vertical heat transport; a kilometre away, melt on vertical faces approached 30 m a year and reached 43 m a year in crevasses, with slopes above 30° delivering 27 per cent of the total from 9 per cent of the area. Speculative The benefit of removing heat depends on a transfer function nobody can currently write — a problem for the proposal and equally for the projections its critics rely on.

Speculative “It is reversible, so the risk is bounded.” Handwave Nothing comparable has been removed at 600 m anywhere, the Inner Bay site has been reached once in forty years, and removal does not undo the circulation, sea-ice and nutrient effects of the installed period. Frontier Reversibility that has never been demonstrated is a design intention, and the partially built and abandoned case has not been modelled by anyone.

Frontier “Greenland is the easy version.” It is the easier engineering and the harder case. Established Roughly half of Greenland’s loss is driven by atmospheric rather than oceanic forcing, so the ceiling on a perfect barrier is about half the problem; Ilulissat is a UNESCO World Heritage area with a resident community whose consent has been actively sought; and published Greenland cost claims range from about US$500 million for one fjord barrier to US$68.9 billion for ten submerged steel barriers and US$337.1 billion for dams. Speculative The US$68.9 billion study assumes uniform 150 km glacier widths and 1,000 m ice thickness across all ten sites and defers ice-dynamic coupling to future work; its headline of up to 5.3 m of avoided sea-level rise by 2300 should not be cited beside the Antarctic numbers as the same class of estimate.

Established “Seventy scientists signed it” — or forty-two. Secondary accounts of the 2025 assessment give different author counts, one describing more than seventy co-authors and another forty-two, and this brief could not reconcile them. Speculative The disagreement does not touch the findings, which are quoted here from the paper itself; it is recorded because the author count is the number most used to convey the paper’s authority.

Handwave “A US$10 million research programme can answer this.” Theprogramme’s stated fundraising goal is US$10 million against its own deployment estimate of US$40–80 billion, a ratio of roughly one to four thousand or worse. Frontier What US$10 million buys is moorings, tank tests and one sheltered fjord trial — genuinely useful, and section 7 argues it is the correct spend. Speculative What it does not buy is an answer to the deployed question; framing it as ten million against trillions of avoided damage compresses a twenty-year qualification programme into a decision nobody is actually being offered.