Planetary Cooling Concepts

Scope first, because this slot is defined against its neighbours. Stratospheric aerosol injection — the physics, the volcanic analogue, the cost curve, the moral hazard, the field programmes that were stopped — belongs to Climate Engineering and is not re-run here. The spacecraft engineering of an orbital shade belongs to Artificial Satellites for Climate Control. Treaty machinery belongs to Geoengineering Governance. What is left is the rest of the menu, and it is larger and stranger than the menu's most famous item: cirrus cloud thinning, marine cloud brightening as a physics and efficacy question, space-based sunshades as a mass-and-cost question, surface albedo, and passive radiative cooling materials.

Established These are not four versions of one idea. Three of them reflect sunlight; one of them lets heat out. Established Cirrus thinning is a longwave lever — it works at night and through the polar winter, it reduces cloud rather than adding it, and it is expected to increase precipitation where aerosol injection is expected to reduce it. Frontier That single structural difference is why a prohibition drafted around “reflecting sunlight” does not describe it, and why it sits awkwardly in every assessment framework written so far.

Frontier The organising claim of this brief: every option here is a lever whose sign or whose achievable magnitude is disputed in the peer-reviewed literature. That is unusual. Nobody doubts that a stratospheric aerosol layer cools; the argument there is about where, how unevenly, and at what political cost. Here the arguments are more basic. Does cirrus seeding cool or warm? Does marine cloud brightening work through brightness or through cloud cover — and are the field's delivery designs optimised for the wrong channel? Is a sunshade's cost an engineering estimate or an assumption about launch prices wearing an engineering estimate's clothes? Established On this map, this is the brief where we do not know the sign is the established position rather than a hedge.

Established Cirrus thinning has a mechanism precise enough that its failure mode is visible in the mechanism itself. Thin cirrus warm: they pass sunlight and trap outgoing longwave radiation. Seeding them with efficient ice-nucleating particles is meant to trigger heterogeneous freezing before homogeneous freezing occurs, producing fewer and larger crystals that fall out faster, shortening cirrus lifetime and letting heat escape. Established The failure mode is intrinsic and not an implementation detail: seed too much and you get more numerous, smaller crystals, longer-lived cirrus, and warming. Established The intervention has an optimum with the wrong sign on the far side of it, and Storelvmo, Boos and Herger framed the subject in 2014 with a question mark in the title that has held up better than most of what followed.

Established The founding claim was total, and the field spent a decade arguing with it. Storelvmo and colleagues, in Geophysical Research Letters in 2013, state in the abstract that cirrus cloud seeding has the potential “to cancel the entire warming caused by human activity from pre-industrial times to present day.” Established Then, in order: Penner, Zhou and Liu asked in 2015 whether cirrus seeding could be used for geoengineering at all; Gasparini and Lohmann answered in 2016 with “Why cirrus cloud seeding cannot substantially cool the planet,” reporting that “none of the seeding strategies achieves a significant cooling due to complex microphysical mechanisms limiting their climatic responses”; Tully and colleagues in 2022, with a more physically based ice microphysics scheme, concluded that cirrus thinning is “unlikely to act as a feasible climate intervention strategy on a global scale”; and in 2023 the same group recommended pausing modelling work pending better observational evidence of aerosol–ice-cloud interactions. Frontier A modelling group recommending a moratorium on its own modelling is an unusual datum and it should be read as one.

Frontier But the pull-back is not unanimous, and the honest description is a live sign dispute rather than a refutation. Between the negatives sits a 2020 result reporting that seeding “decreases the damage by about 50% in ECHAM, and by 85% in CESM over the 21 selected land regions” — a large regional benefit, with two models disagreeing by nearly a factor of two about its size. Frontier The pair carries a caveat: the full text was not obtainable here, the abstract defines neither “damage” nor the seeding scenario, and the 50%/85% split should be read as a model-spread signal rather than an efficacy estimate. Frontier Pulling the other way, satellite-constrained radiative-transfer work by Erfani and Mitchell — a 2025 EGUsphere preprint, published in Atmospheric Chemistry and Physics in 2026 — reports that “homogeneous cirrus clouds exhibit stronger radiative effects than heterogeneous cirrus, implying that transitioning could induce substantial surface cooling.” Frontier Authors appear on both sides. This is not a field that changed its mind; it is a field whose answer depends on the ice microphysics scheme, and which knows it.

Established Two physical constraints bound the whole cirrus argument, and both are measurable. First, only cirrus that would otherwise form by homogeneous freezing are susceptible at all; cirrus already dominated by heterogeneous nucleation on natural dust are not. In 2022, ice crystal number concentrations in dust-influenced cirrus were found to exceed ambient dust nucleating-particle concentrations by more than an order of magnitude — evidence that a susceptible population exists, and a demonstration of how hard the two are to tell apart by observation. Established Second, the deliberate effect must be set against the natural one: global model simulations put the radiative forcing of the total ice-nucleating-particle effect on cirrus at −28 to −55 mW m⁻² — between about 0.03 and 0.06 W m⁻². Frontier A widely repeated figure of roughly 2.5 W m⁻² of cirrus-thinning cooling is attributed to the 2013 paper; this brief could not confirm that number against the primary source and therefore does not assert it — the abstract states the claim qualitatively, not numerically. Established For scale on the underlying terms: cirrus near the tropical tropopause warm for optical depths below about 16, midlatitude cirrus over the Qinghai–Tibet Plateau have shown instantaneous regional net effects exceeding 80 W m⁻², and CALIPSO observations put the global longwave warming from cirrus lateral boundaries alone at at least 0.07 W m⁻².

Frontier Marine cloud brightening's evidence base moved sharply in 2024, and it moved for an uncomfortable reason. Chen and colleagues, in Nature Geoscience, used Kīlauea's volcanic degassing over Hawaii as a natural experiment — June to July 2008 and June to July 2018 — with machine learning applied to satellite observations. The aerosol–cloud radiative forcing reported is about −2 W m⁻² (June 2008), −2.8 W m⁻² (June 2018) and −10.5 W m⁻² (July 2018); cloud cover rose by up to 54% in relative terms; cloud-cover susceptibility ranged from 0.38 to 1.6. Frontier The decomposition is the finding: cloud-cover enhancement contributed 65–87% of the cooling, and classical Twomey brightening 25–64%. Established The authors state the corollary themselves and it is not good news: a strong negative aerosol forcing means the observed warming has come from a weaker net forcing than assumed, which implies higher climate sensitivity and a larger response both to greenhouse gases and to air-quality-driven aerosol cuts — while also implying that mitigation via marine cloud brightening is “plausible and is most effective in humid and stable conditions in the tropics.” Frontier The case for the technology and the case for urgency are the same evidence. This is a single natural experiment, one volcano, two summers, one group, and it should be weighted accordingly — but it is the most consequential marine cloud brightening result of the decade.

Frontier A second result attacks the mechanism's name from the other side. Ahlm and colleagues, modelling in 2017, found that “the global-mean clear-sky effective radiative forcing by the injected particles is as large as the corresponding total” effective radiative forcing — meaning that in their model a large share of the cooling is the sea-salt aerosol's own direct scattering, not a cloud response at all. Speculative Three consequences follow: efficacy partly decouples from cloud susceptibility; the technique behaves partly like a low-altitude aerosol injection, inheriting objections it was supposed to avoid; and a field trial measuring cloud brightening could miss most of the forcing it produced. Established The canonical engineering paper — Latham and twenty-three co-authors, 2012 — describes seeding marine stratocumulus with “copious quantities of roughly monodisperse sub-micrometre sea water particles” to raise droplet number and hence albedo. Frontier Every delivery concept in that lineage is optimised for the Twomey channel. If Chen and colleagues are right, the hardware is tuned for the smaller half of the effect.

Established The sunshade numbers are the cleanest in the brief, and the headline is the least useful part of them. Angel's 2006 PNAS design blocks 1.8% of incident solar radiation with metre-sized, roughly gram-mass flyers near the inner Lagrange point — about 20 million tonnes of spacecraft, launched in stacks of 800,000, spread over a cloud roughly 100,000 km long, with a 50-year design life, feasible “within ≈25 years at a cost of a few trillion dollars, <0.5% of world gross domestic product over that time.” Established That cost is conditional on transport at $50 per kilogram, achieved in the paper by an electromagnetic-launch and ion-propulsion architecture that has never been built. Quote the trillions without the $50/kg and you have quoted a conclusion detached from the only assumption that produces it. Established The comparison that settles the concept is with the lunar-dust alternative. Bromley, Khan and Kenyon in 2023 require more than ten billion kilograms of dust for the same 1.8% attenuation — equivalent, they say, to about six days of shading a year — and favour ejection from the lunar surface precisely to avoid Earth's gravity well. Established Angel needs about 2 × 10 billion kilograms of manufactured spacecraft; Bromley needs more than 10 billion kilograms of raw regolith. The mass is set by the physics of scattering, not by the cleverness of the design. Only the gravity well changes.

Established The sunshade is also, quietly, the best-modelled option in climate intervention, because reducing the solar constant is the easiest thing to put in a general circulation model. The GeoMIP G1 experiment is literally a sunshade: under it the Walker circulation shows no significant change in intensity while the Hadley circulation weakens, and a fully coupled “Sunshade World” run concluded that such geoengineering “would inevitably create climate change due to differing forcing patterns compared to natural conditions.” Frontier The option with no deployment pathway supplies the evidence base for the option that has one — a methodological caveat on somebody else's evidence, set out in section 14.

Established Surface albedo is the least glamorous member of the family and the only one being deployed. Akbari, Menon and Rosenfeld's 2008 estimate remains the reference: roofs and pavements are over 60% of urban surface (roofs 20–25%, pavements about 40%); reflective materials raise roof albedo by about 0.25 and pavement albedo by about 0.15, a net urban gain of about 0.1; world-wide, that induces a negative forcing “equivalent to offsetting about 44 Gt of CO₂ emissions,” worth about $1,100 billion at $25 per tonne, with building-scale cooling-cost savings above 20% — and the authors' own statement that the estimate depends on the assumptions used. Established The 44 Gt is a one-time equivalence, not an annual rate, and it holds only while the surfaces stay clean. Frontier Passive radiative cooling materials are the newer entrant: ideal materials reach about 164.8 W m⁻² of cooling power and practical ones about 160.8 W m⁻², and covering 1% of Earth's surface would shift radiative forcing by 1.61 W m⁻², from +0.6 to −1.01. Established One per cent of Earth's surface is about 5.1 million km² — larger than the European Union, and vastly larger than the world's total roof area.

Established The Arctic case supplies the number the whole regional-albedo family should be measured against. A 2018 study modelled hollow glass microspheres on sea ice and reported regional Arctic temperature reductions above 1.5 °C with thicker, more extensive ice; its first author founded the organisation proposing the intervention, which is material to how the result travels. Established Zampieri and Goessling in 2019 found that wind-driven pump intervention could “keep the late-summer sea ice cover at the current extent for the next ≈60 years” while reducing global annual-mean near-surface air temperature by only 0.02 K. Established Those results are compatible, because they are about different quantities — and advocacy that presents regional preservation as planetary cooling is merging them. Frontier The feedback being fought is itself poorly pinned down: July sea-ice albedo across CMIP5 models spans 0.3 to 0.7, a factor of two of spread in exactly the quantity the intervention manipulates.

Frontier Is marine cloud brightening's dominant channel cloud cover rather than cloud brightness? This is the single highest-value open question in the brief, because the answer changes the hardware. The evidence is one natural experiment analysed by one group with one machine-learning attribution, and the cloud-cover channel is under-predicted by the process models the field otherwise relies on. Frontier Independent replication on other degassing volcanoes and on ship-track archives, by a different group with a different method, is cheap, needs no permit, and would either confirm or overturn the decade's most consequential result in this field.

Frontier Are ship tracks a valid analogue at all? Decades of marine cloud brightening inference rest on them. A 2026 preprint compares sulfur-dominated, soot-dominated and salt tracks specifically to ask whether they behave alike; if they do not, the observational foundation needs re-weighting, and the sea-salt case may be the least well constrained of the three. Frontier A related question with a distributional edge: cloud-cover susceptibility measured between 0.38 and 1.6 implies efficacy varying roughly four-fold with meteorological regime, which makes a deployment map a meteorological map. The places where the technique works best are not the places that most need cooling.

Frontier Can susceptible cirrus be identified from orbit, in advance, at delivery-relevant resolution? Everything about cirrus thinning turns on this. Without a discriminator between homogeneous-freezing-dominated and heterogeneous-freezing-dominated cirrus you cannot target, cannot dose, and cannot verify. Frontier The CALIPSO infrared-imaging-radiometer retrievals of cirrus ice-particle concentration are the prototype instrument and the regime-identification methods of the past four years are the prototype method; neither has been turned into a validated global classification. Speculative A second, entirely unwritten requirement: nobody has published a target specification for a cirrus-seeding particle — an ice-nucleating particle that activates inside the cirrus temperature and supersaturation window but not so readily that it over-seeds. The over-seeding threshold is a materials-design specification and it has never been written as one.

Frontier What is the sunshade mass requirement, actually? Angel's 20 million tonnes and Bromley's ten billion kilograms are two independent arrivals at the same order of magnitude, and the study most likely to revise them — Sánchez and McInnes's 2015 optimisation of sunshade configurations near Sun–Earth L1 — could not be obtained in any form for this assessment beyond its bibliographic record. Speculative Reconciling the three into one comparable figure per unit of attenuation is a review-paper task, not an experiment, and it has not been done. Speculative The only route to a materially different answer is a scattering material with a far better cross-section per kilogram than metre-scale films or micron dust — diffractive rather than reflective elements, or engineered nanostructures at the resonance scale. Handwave Nothing reached for this assessment produces a mass estimate below about a billion kilograms, and the burden of proof sits squarely with anyone claiming otherwise. Frontier Meanwhile the concept has stopped being only a thought experiment: a precursor mission design and two roadmap papers entered the peer-reviewed aerospace literature between 2024 and 2026, from a single consortium, and essentially no climate-policy discussion has registered it.

Speculative The most interesting unrun experiment in the category is purely computational. Cirrus thinning is expected to increase precipitation, through enhanced latent heat flux and tropospheric radiative cooling; stratospheric aerosol injection is expected to reduce it. A longwave lever and a shortwave lever have opposite hydrological signatures. Speculative A portfolio combining both could, in principle, hold temperature and precipitation closer to baseline than either alone — and nobody has run it. No assessment framework even distinguishes the two lever types by hydrological sign, which is why the combination has not occurred to the people who would have to authorise it.

Frontier Three smaller live threads. Midlatitude marine cloud brightening is only now being examined separately from the subtropical stratocumulus case the whole literature assumes, and the cloud regimes and the insolation both differ. Frontier The two 2025–2026 global estimates for radiative-cooling materials use incompatible framings — one gives a forcing for 1% surface coverage, the other a “theoretical threshold area” across roughly 2,000 cities, with regional reductions above 6 °C and cooling powers above 80 W m⁻²; reconciling them is open and entirely tractable. Speculative And land radiative management — lighter-leaved cultivars, reflective croplands — changes the surface energy balance, hence sensible and latent heat fluxes, hence boundary-layer convection and precipitation. Frontier That is the same class of problem as an aerosol layer's monsoon effect, arising from an intervention with essentially no governance attached, because planting a different cultivar is agriculture. The magnitudes live in full texts this assessment could not reach and are not stated here.

Frontier The four options in this brief are blocked by four different kinds of thing, and conflating them is the most common analytical error made about the family. Cirrus thinning is blocked by a scientific unknown; marine cloud brightening by an institutional one; sunshades by a market one; surface albedo by ordinary regulation, which is why it is the only one being deployed.

Frontier Cirrus thinning: the binding constraint is that nobody can tell susceptible cirrus from non-susceptible cirrus in real time. Everything else is downstream. Without that discriminator there is no targeting, no dose control below the over-seeding threshold, and no way to detect the sign of the achieved effect quickly enough to stop. Established A technology that can warm needs a fast feedback loop, and the atmosphere's longwave response to cirrus perturbation is precisely the quantity the observing system measures least well.

Frontier Marine cloud brightening: the binding constraint is that no outdoor trial has completed a radiative measurement. The best result on record from a deliberate outdoor programme is a demonstrated increase in aerosol concentration at cloud base height — a delivery demonstration, one step short of a measured cooling. The one US attempt was halted by a city council rather than by physics. Established The science that would settle the central question needs no release at all, which makes the institutional block particularly expensive: it is stopping the wrong work.

Established Sunshades: the binding constraint is launch cost per kilogram, and no climate actor controls it. The mass requirement is fixed by scattering physics at roughly ten billion kilograms for 1.8% attenuation, whether the material is manufactured spacecraft or lunar regolith. Angel's economic case is the $50/kg assumption and nothing else. Speculative That makes the sunshade downstream of a launch-market outcome rather than of a climate result — a strange position for a climate technology, and the reason its engineering belongs with Artificial Satellites for Climate Control rather than here.

Established Surface albedo and radiative cooling: the binding constraints are soiling, area, and procurement. Multi-year albedo retention under field conditions is under-reported and is the whole ballgame for a technology whose benefit is proportional to reflectance maintained; the atmospheric window closes under humid conditions, so performance is worst in hot humid places; reflective surfaces carry a winter heating penalty in cold climates; and the areas involved are set by how much roof and pavement exists. Established None of those is a scientific unknown. They are engineering, maintenance and building code — the only bottleneck in this brief that societies already know how to operate.

Frontier The whole cirrus branch depends on one missing scientific object: an ice-nucleation representation good enough that models stop disagreeing about the sign. The 2013-to-2023 sequence is not a story about evidence accumulating; it is a story about the answer tracking the microphysics scheme. Until competing schemes are made to agree on the same prescribed seeding scenarios, additional model runs generate literature rather than knowledge, which is precisely why the field's most active modelling group recommended pausing.

Frontier The observational dependency is a satellite classification, not a new satellite. Ice-particle-concentration retrievals of the CALIPSO type already exist; what does not exist is a validated global product that labels a given cirrus field by nucleation regime with stated skill. Established Marine cloud brightening depends on a second observational product of the same character: a global, seasonal map of cloud-cover susceptibility, extending the 0.38–1.6 range from one natural experiment to the regimes a deployment would actually encounter.

Established The sunshade depends on nothing scientific at all. Its physics is settled, its climate response is the best-modelled of any option here because reducing the solar constant is the easiest thing to put in a general circulation model, and its dependency is a commercial launch price roughly one to two orders of magnitude below the current one. Speculative Its only scientific escape route is a materials result — a higher scattering cross-section per kilogram — and no programme is pursuing one for this purpose. Speculative The lunar-dust variant depends on something different again and larger: a sustained extraction and ejection capability on the Moon, which is a dependency on an entire industrial presence rather than on any result.

Established Surface albedo and radiative cooling depend on ordinary science that nobody is prioritising, because the deployment does not wait on it. Multi-year albedo-retention data under real soiling, regional circulation studies validated against observation, and a forcing-per-dollar accounting with maintenance included are all cheap and all missing. Frontier That inversion is worth naming: the option with the weakest evidence base is the one being installed, and the options with the most modelling behind them are the ones nobody can deploy.

Frontier The cheapest high-value experiment in this brief requires no release, no permit and no new instrument. Replicate the cloud-cover-dominance result using existing satellite archives, a different volcano or ship-track dataset, and an attribution method that is not the original machine-learning implementation. It would either confirm or overturn the decade's most consequential marine cloud brightening finding and an implied revision to climate sensitivity. Nothing else in the category offers that ratio of information to cost.

Frontier Then, in order, for cirrus. (1) Extend CALIPSO-class ice-particle retrievals into a validated global nucleation-regime classification — this is the binding link and the rest is idle without it. (2) Run a cirrus-thinning model intercomparison with prescribed seeding scenarios and a common ice-nucleation treatment, reported per model, which is the exact analogue of the intercomparison the aerosol-injection literature already has. (3) Publish a target specification for a seeding particle: activation temperature, supersaturation window, and the mass loading at which the sign reverses. (4) Disperse a known mass of ice-nucleating particles along a flight track and measure crystal number, size distribution, sedimentation and cloud lifetime with lidar and in-situ aircraft. (5) Attribute a top-of-atmosphere longwave anomaly to that perturbation. Speculative Step 4 is physically modest — aviation already imposes a large, continuous, uncontrolled cirrus perturbation on the same air — and it is nevertheless the step most likely never to happen, because it is an outdoor climate-intervention release and the completion rate for those is zero.

Frontier For marine cloud brightening, after replication: finish the sulfur-versus-soot-versus-salt track comparison so the analogue is validated or retired; map susceptibility globally by regime and season; get a sprayer or airborne system to produce a specified aerosol size spectrum in operational marine conditions, which no hardware has yet demonstrated; and run the regional-impact assessment for affected non-deploying regions before a trial rather than after one. Established For surface albedo the missing experiments are unglamorous and trivially fundable: multi-year outdoor albedo-retention measurement under real soiling, regional circulation studies over large deployed areas, and a rework of the 44 Gt equivalence into a forcing per dollar with maintenance included. That last one is arithmetic, and it has not been done.

Speculative Cirrus seeding is a dose-control problem wearing an aircraft-dispersal problem's clothes. Delivery would be by aircraft at cruise altitude with a solid ice-nucleating particle — bismuth triiodide is the usual candidate — and the hard part is not dispersal but metering: staying inside an optimum whose width varies with the temperature, humidity, vertical velocity and pre-existing particle loading of the cirrus being seeded, none of which is known in advance. Handwave No proposed system has demonstrated dose control against a moving, heterogeneous, unclassified target, and a system that cannot do that is a system that can produce either sign.

Frontier Marine cloud brightening hardware has a published mass-flux dispute inside a single concept study. The canonical approach is shipborne generation of roughly monodisperse sub-micrometre sea-water particles. An airborne alternative proposes industrially manufactured sodium chloride nanoparticles made by ethanol anti-solvent brine precipitation and delivered by uncrewed aircraft; its preprint claims about 500 times less aerosol mass flux than ship-based spraying and the published version about five times less energy. Frontier Those are different claims about different quantities and the gap between them is large enough to matter. Replication status: one group, concept study, no hardware.

Established The sunshade's engineering requirement is an industrial rate, not a device. Twenty million tonnes of gram-scale flyers, launched in stacks of 800,000, occupying a cloud roughly 100,000 km long, holding station for a 50-year design life. Speculative The lunar-dust variant replaces that with a permanent surface operation ejecting more than ten billion kilograms of regolith on a continuous basis, because dust at L1 disperses on timescales of days to weeks. Speculative That converts a construction project into a sustained industrial presence on the Moon, creates the fastest termination shock of any proposal in the category — the shade evaporates when the supply stops — and raises a hazard nobody has analysed: a persistent dust stream through cislunar space passes through everything else operating there. Established Surface albedo, by contrast, is coatings chemistry, roofing practice and procurement, already commercial.

The nearest brief is Climate Engineering, which owns stratospheric aerosol injection outright, along with the moral-hazard literature, the termination-shock quantification, the funding asymmetry, and the narratives of the field programmes that were stopped. Established This brief is calibrated against it rather than duplicating it: the aerosol-injection cost estimate there is the yardstick against which a few trillion dollars for 1.8% attenuation should be read.

Artificial Satellites for Climate Control holds the spacecraft side of the sunshade — deployment, station-keeping, manufacturing, launch — and the boundary drawn here is that this brief holds the mass, cost and radiative physics while that one holds the vehicle. Established Geoengineering Governance holds the treaty machinery, including the London Convention and Protocol process under which marine cloud brightening is among the techniques being considered for listing, and the two binding sub-national prohibitions now on the books.

Frontier A boundary worth stating rather than assuming: the sunshade's mass, cost and radiative physics are treated here and its vehicle is treated there, because the two questions have different answers and different bottlenecks. Established A reader who wants to know whether a shade can be built goes to the spacecraft brief; a reader who wants to know what it would weigh and why that number will not move stays here.

Established Weather Modification shares this brief's physics almost exactly — ice nucleation, particle delivery, and the attribution problem — and supplies the base rate: an intervention smaller, cheaper and more frequently repeated than any of these has been practised for eighty years and still cannot attribute its effect at the scale it is sold. Frontier Atmospheric Management owns the tropospheric-chemistry framing of aviation-induced cirrus, which is the uncontrolled experiment cirrus thinning would be a controlled version of. Established Arctic Engineering and Polar Development hold the regional context for sea-ice albedo work; Sustainable Megacities holds urban heat, which is where cool roofs and radiative-cooling materials actually earn their keep; and Carbon Capture at Scale holds the intervention class that none of these substitutes for, because not one option in this brief touches ocean acidification.

Established This subject is governed by assessment reports, and assessment reports are assembled from bibliographies — which makes the integrity of the bibliography an institutional question rather than a clerical one. Three concrete hazards showed up in assembling this brief. The 2013 founding paper carries two live DOIs, an agency-canonical form and a publisher-era form; both resolve, both are cited, and a bibliography containing both looks like two papers by the same authors saying the same thing. Established Automated bibliographic tools substituted one document for another repeatedly during this assessment — a units error of nine orders of magnitude in a summarised abstract, a fabricated author list, and an entire query returning an unrelated field's results — all silently, with no error and no low-confidence signal.

Established And the third hazard runs the other way, which is the one worth naming loudest: “could not locate” is not “does not exist.” A prior pass on this subject flagged a cited Erfani and Mitchell cirrus paper as possibly fabricated because its searches came up empty. The paper is real — it appeared in Atmospheric Chemistry and Physics volume 26 in 2026, with an EGUsphere preprint the year before — and the search instrument had simply been blocked from the publisher. Frontier A retrieval failure recorded as a fabrication is a worse error than the one it was hunting, and the correct output of a failed search is unverified in this pass, not phantom. Established That distinction is an institutional requirement, not a nicety: a field whose primary literature sits behind intermittent access, and whose synthesis is increasingly machine-assisted, will generate both kinds of error, and only one of them is currently being looked for.

Frontier The substantive institutional gaps are three. There is no coordinating body for a cirrus-thinning model intercomparison, which is why the field's response to model disagreement was a recommendation to stop rather than a protocol to converge. Established There is no framework that describes a longwave intervention: prohibitions and research governance drafted around reflecting sunlight do not reach a technique that works by letting heat out, and the gap is recorded rather than fixed. Frontier And the field has proposed its own stopping rule — a 2022 opinion piece from inside the marine cloud brightening community arguing that feasibility assessment needs to say in advance what to study and when to stop — which is a healthier institutional instinct than anything the surrounding policy literature has produced, and which nobody has adopted.

Frontier One option here can be deployed in good faith and make things worse, and that is a different ethical object from the rest of the geoengineering menu. An intervention whose sign is uncertain has negative expected value under some reasonable priors. The affected population would have no way to attribute the harm, because the counterfactual is a cirrus field that did not form. Established Every argument that attaches to solar geoengineering attaches here, and the technique is not solar geoengineering — so the objections apply without the protections.

Established Consent has already been the binding constraint twice in this family, and neither case turned on the science. The record on outdoor trials is that they end when a community, a council or an Indigenous body says no, at scales far below any plausible harm threshold. Frontier Arctic intervention has its own rights literature examining sea-ice geoengineering against Indigenous governance of the Arctic Ocean, and it is directly relevant given which outdoor projects are currently funded. Frontier A separate category shift is under way in the same area: hollow glass microspheres proposed as a cooling material are now being modelled as a marine pollutant in a marine-pollution journal. The material has not changed; the question being asked about it has.

Established Surface albedo inverts the ethics of the whole category, and that is the most useful thing about it. The primary benefit is local, immediate, and accrues to whoever pays: reduced heat mortality and reduced cooling load, whether or not the global forcing matters. No consent problem, no transboundary problem, no termination shock. Established It carries one mundane and serious equity point instead — the benefit accrues to whoever is under the roof, and informal settlements have the least roof to work with and the most heat exposure. Frontier Marine cloud brightening carries the distributional problem in a sharper form: if efficacy varies four-fold by meteorological regime, the deployment map is drawn by cloud physics, and the regions that would be cooled are not the regions that chose it.

Frontier The strongest argument against every option here is that it is a distraction from emissions, and the argument has very different strength against each one. Against sunshades it is at its strongest: a few trillion dollars and twenty-five years buys 1.8% of insolation, does nothing about ocean acidification, and is contingent on a launch cost one to two orders of magnitude below today's, from a technology base that has flown nothing. The opportunity cost is enormous and the timeline is longer than the mitigation timeline it is meant to bridge.

Frontier Against cirrus thinning the argument is strong and carries an extra twist: an intervention whose sign is unresolved is not merely a diversion of effort but a candidate for making things worse, which puts it below doing nothing under some priors. Frontier Against marine cloud brightening it is weakest, because the marginal cost of regional deployment for a specific asset — a reef, a heatwave, a fishery — is small and the benefit is local and in principle measurable; as a global lever the standard objection returns in full. Established Against surface albedo the argument largely fails, because the intervention is justified by local co-benefits regardless of its climate effect. A treatment that applies the same distraction argument uniformly across the four is not thinking.

Established None of these is a planetary thermostat, and saying so precisely is the useful contribution of the slot. Frontier The instructive case is cirrus thinning: a mechanism that looked promising, was investigated properly, produced a decade of contradictory results, and had its own practitioners recommend a pause. Established A research programme that can say that about itself is functioning, and a map of frontier science should record that outcome as a success rather than as a gap.

These horizons track when the sign, the mechanism and the cost of each lever might actually be settled — not when anything gets deployed.

• 10 yr: Frontier Independent replication or refutation of the cloud-cover-dominance result, and with it a firmer position on aerosol forcing and on climate sensitivity — the highest-information event available in this decade, and it needs archive data rather than a permit. Frontier A validated satellite classification of cirrus by nucleation regime is achievable in this window and is the thing to watch, because everything about cirrus thinning is downstream of it. Established Cool-roof and reflective-pavement mandates widen; radiative-cooling coatings reach mainstream construction on building-energy grounds alone. Speculative A sunshade precursor mission remains a design on paper unless someone funds it, and no climate programme is structured to.
• 25 yr: Frontier Enough convergent microphysics and observation to say whether cirrus thinning has any deployable window at all, or to retire it — a genuine possible outcome, and a good one. Frontier A marine cloud brightening trial that measures a radiative response rather than a delivery, if any jurisdiction permits one; permission, not physics, is the open variable. Speculative A reconciled sunshade mass figure per unit of attenuation, which is a review-paper task available at any time and therefore likelier to be done late than early. Frontier And a first portfolio study combining a longwave and a shortwave lever, which could be run today by anyone with model time.
• 50 yr: Speculative Surface albedo and radiative cooling as standard urban heat practice across the hot-climate world, with no planetary-scale cooling from any lever in this brief. Speculative If commercial launch approaches the $50-per-kilogram regime, the sunshade becomes an affordability argument rather than a physics one for the first time in its history — and the twenty-million-tonne requirement will not have moved. Handwave A deployed cirrus intervention in this window requires the sign question to have been settled decisively in the affirmative and a real-time verification system to exist; neither is on any funded path.
• 100 / 250+ yr: Handwave A lunar dust operation sustained across generations, replenishing a shade that disperses in weeks, is coherent, and the load-bearing assumption is not the astrodynamics but the continuity of an industrial civilisation on the Moon across centuries. Handwave A diffractive or resonance-scale scattering material that beats ten billion kilograms would change the category outright, and nothing in the reachable literature suggests where it would come from. Established Treating any option in this brief as a substitute for emissions reduction remains unsupported at every horizon, and the ocean chemistry is untouched by all of them.

• Depends on Depends on four things of four different kinds, and the mismatch between them is the structure of the whole subject. A scientific result: an ice-nucleation representation the competing models agree on, without which the sign of cirrus thinning is a property of the scheme rather than of the atmosphere. An observational product: a satellite classification of cirrus by nucleation regime, and a global seasonal map of marine cloud-cover susceptibility, both buildable from instruments already in orbit. An institutional permission: a completed outdoor trial that measures a radiative response rather than a delivery. And a market outcome that no climate actor influences at all — a commercial launch price near $50 per kilogram, on which the entire sunshade cost case rests. Surface albedo depends on none of them, which is why it is the one being installed.
• Requires (not on this map) Cirrus thinning is the only intervention on this map whose sign is genuinely unresolved: with current ice-nucleation representations, models disagree about whether seeding cools or warms, and the disagreement tracks the scheme rather than the evidence. That is a result to be discovered, not a capability to be built — the radiative-cooling materials are already commercial, the sunshade waits on a launch price rather than on a finding, and marine cloud brightening waits on permission to run a trial. None of those is what stands in the way here.
• Enables Enables a genuine portfolio question in place of a menu. A longwave lever and a shortwave lever have opposite hydrological signatures — cirrus thinning is expected to increase precipitation where aerosol injection reduces it — so settling cirrus thinning's sign would open the first serious combination study in climate intervention, and it is a model experiment rather than a deployment. Replicating the cloud-cover result enables an aerosol-forcing constraint that feeds straight into climate sensitivity, which is a contribution to core climate science and not to geoengineering at all. And a forcing-per-dollar accounting across urban albedo, coatings, cloud and orbit would let this whole family be compared on one axis for the first time.
• Adjacent Adjacent to cloud microphysics and ice nucleation, satellite remote sensing and retrieval science, photonic and coatings materials science, astrodynamics and launch economics, marine ecotoxicology by way of the microsphere question, building codes and municipal procurement, and the attribution problem shared with eighty years of weather modification.

Handwave “Cirrus cloud thinning could cancel all anthropogenic warming.” That is the 2013 paper's stated potential, and it is the claim its own field spent the following decade contradicting: cannot substantially cool the planet in 2016, unlikely to act as a feasible climate intervention strategy on a global scale in 2022, and a recommendation to pause the modelling in 2023. Frontier Citing the 2013 headline without the decade after it is this topic's most common error. Frontier The opposite error is now available too, because satellite-constrained work from 2025 and 2026 points back toward substantial surface cooling. The honest description is a sign dispute, not a refutation.

Established “Cirrus thinning is a form of solar geoengineering.” It is not, and the difference is not pedantic. It acts on outgoing longwave radiation, which is why it functions at night and through polar winter when there is no sunlight to reflect; it reduces cloud rather than adding it; and it is expected to increase precipitation where aerosol injection reduces it. Established A prohibition on reflecting sunlight does not reach it, and the frameworks currently being drafted mostly do not mention it.

Frontier “Over-seeding is an implementation detail.” This is the enthusiast's error and it is the central physical fact. The optimum is narrow with warming on the wrong side of it; the more physically based microphysics scheme reduced the maximum global-mean top-of-atmosphere anomalies from over-seeding by more than half and they remained the dominant feature; and no delivery scheme can guarantee staying inside the optimum across heterogeneous real cirrus while no observational method separates susceptible cirrus from the rest.

Frontier “Marine cloud brightening works by making droplets smaller and clouds whiter.” That is the Twomey effect, and it may be the minority channel. The Kīlauea natural experiment attributes 65–87% of the observed cooling to increased cloud cover and only 25–64% to Twomey brightening; a separate modelling study found the clear-sky forcing of the injected particles as large as the total forcing, meaning much of the effect was direct scattering and not a cloud response at all. Frontier The technology may be named for the smaller half of what it does, and the delivery hardware is optimised accordingly.

Established “A space sunshade would cost a few trillion dollars.” That figure is the output of an assumption, not an engineering estimate. Angel's design requires about 20 million tonnes of spacecraft to block 1.8% of sunlight and reaches “a few trillion dollars, <0.5% of world gross domestic product” over about 25 years conditional on transport at $50 per kilogram, via an electromagnetic-launch and ion-propulsion architecture that does not exist. Established At commercial launch prices the mass is unaffordable by orders of magnitude. Quote the cost without the premise and you have quoted a conclusion with its argument removed.

Frontier “Lunar dust would be cheaper, because the material is free.” Do the arithmetic. Bromley and colleagues require more than ten billion kilograms of dust for 1.8% attenuation; Angel requires about twenty billion kilograms of spacecraft for the same 1.8%. Established The mass is set by scattering physics and the two architectures land within a factor of two of each other; the saving is in the gravity well, not in the mass. Speculative And dust disperses, so the operation must run continuously — trading a construction project for a permanent lunar industrial operation with the fastest termination shock in the category.

Frontier “White roofs could offset 44 gigatonnes of CO₂.” The number is real and the shape is wrong. It is a one-time equivalence — the emission whose radiative effect a world-wide urban albedo gain of about 0.1 would cancel — not an annual rate, and it holds only while the surfaces stay reflective, with soiling and maintenance outside the figure. Established “Radiative cooling paint could cool the planet” has the same shape: the one peer-reviewed global estimate obtained here requires 1% of Earth's surface, about 5.1 million km², to move forcing by 1.61 W m⁻² — an area larger than the European Union. The materials are excellent and the building-scale benefit is real. Established And radiative coolers do not violate thermodynamics: they exchange heat with deep space through a transparent atmospheric window, and deep space is cold.

Frontier “Arctic geoengineering could save the sea ice and slow global warming.” It might do the first and does essentially nothing for the second: a pump-based intervention holding late-summer ice at current extent for about sixty years reduces global annual-mean near-surface air temperature by 0.02 K. Established Modelled regional reductions above 1.5 °C are consistent with that, being a different quantity; merging them is the error. Established “The GeoMIP models tell us what stratospheric aerosol injection would do” is a related confusion in the other direction: the G1 experiment reduces the solar constant, so much of the regional-effects literature is sunshade modelling used as a proxy for an aerosol layer that has spectral structure, stratospheric heating and spatial gradients. It is a reasonable first-order proxy and it is not the same thing.

Speculative “The sunshade is science fiction” and “the sunshade costs a few trillion dollars” are opposite errors made by the same people. A precursor system and mission design appeared in Acta Astronautica in 2025, with roadmap papers in 2024 and 2026; the concept has a mission-design literature whose full texts were not obtainable here, so this brief carries their existence and none of their numbers. Frontier Whether it is affordable is a separate question with a clear answer. Handwave And one last correction, aimed at this site's own methods: a citation that a search cannot find is unverified, not fabricated. The 2026 cirrus paper flagged in a prior pass as a possible phantom is real, indexed and published; the search had been blocked from the publisher. In a subject governed by assessment reports assembled out of bibliographies, the difference between could not locate and does not exist is the difference between a caveat and a defamation.