Climate Engineering

Climate engineering here means solar radiation modification: reflecting a small fraction of incoming sunlight to reduce surface temperature. The two candidate planetary levers are stratospheric aerosol injection, which mimics the cooling that follows a large volcanic eruption, and marine cloud brightening, which raises the reflectivity of low marine clouds by seeding them with sea salt. Carbon removal is a different intervention with a different risk profile and is treated in Carbon Capture at Scale; the two meet only where the moral-hazard argument links them.

This slot asks the question the neighbouring briefs deliberately do not. Planetary Cooling Concepts covers cirrus thinning, surface albedo and radiative-cooling materials, and records that the cirrus sign is not robust; Atmospheric Management covers methane removal and the ozone regime, and records that the ozone precedent generalises poorly. Neither asks whether deliberate intervention in the climate system, as a class of action, is a usable option for anybody. That is a question about physics, cost, models, consent and the failure record at once, and it resolves differently from any single lever.

The short answer this brief supports, before any of the evidence: the physics is the least contested part; the cost is low enough to be destabilising rather than self-governing; the regional consequences are the genuine scientific dispute; and the outdoor experimental record is very nearly empty. Established One boundary holds throughout and defeats the strongest version of the framing on its own: SRM treats a symptom. It reduces temperature without reducing CO₂, so ocean acidification proceeds regardless — the EU's Scientific Advice Mechanism states flatly that stratospheric aerosol injection does not address the direct impacts of elevated greenhouse gas concentrations, such as ocean acidification.

Established Reflecting sunlight cools the planet, and nobody disputes it. The European Commission's Scientific Advice Mechanism evidence review of December 2024 concludes that global cooling is in principle achievable while adding that uncertainties in required amounts and deployment strategies are still considerable, and puts observed global surface cooling after Mount Pinatubo at approximately 0.3 °C. Frontier What that establishes is narrower than it sounds. It establishes that a sulphate burden of volcanic magnitude produces a global-mean response of roughly the expected sign and size for a year or two. It does not establish that a sustained, engineered, controlled layer behaves the same way, because a volcano is an impulse and a deployment is a forcing held constant for decades against a rising greenhouse background — a different problem in stratospheric dynamics, aerosol microphysics and chemistry. Established The 2026 Scientific Assessment of Ozone Depletion is, per NOAA's Chemical Sciences Laboratory, still in preparation and will for the first time prominently include the effect of aerosol injection on the stratosphere and ozone. It has not reported. Anyone citing a settled injection-and-ozone number is citing something that does not exist. Frontier The EU review's list of expected side effects is the honest short version: regional temperature and precipitation disparities under uneven hemispheric injection; precipitation decreases relative to baseline even where global-mean temperature is matched; termination shock; and ozone depletion risk from sulphur-based approaches.

Established The cost numbers are real, and they are why this capability cannot be governed by its own difficulty. Smith and Wagner's engineering study in Environmental Research Letters remains the reference estimate: a purpose-designed high-altitude tanker — the SAIL, roughly 46 m long, 55 m wingspan, about 25 t of payload, 100 t maximum take-off weight, four wing-mounted engines — injecting at about 20 km, ramping from 0.1 Mt of sulphur in year one by roughly 0.1 Mt a year to 1.5 Mt of sulphur in year 15, targeting a −0.25 W/m² forcing offset — which halves the rate of forcing increase, not the accumulated forcing. Average cost about $2.25 billion a year, $36 billion over 15 years, with about $3.5 billion of pre-deployment capital. Established The authors' own caveat is load-bearing and is routinely dropped when the number is quoted: the estimate excludes any indirect costs such as monitoring and measuring the impacts of deployment, and they cite a suggestion of adding roughly $3 billion a year for research oversight, which would more than double the figure. Frontier The EU review reports a wider literature range of $10–107 billion a year per degree of warming prevented. Even the high end sits inside the discretionary budget of about thirty states; the low end sits inside that of a few dozen individuals.

Established Against that capability, the knowledge base is thinner than most readers assume. Total research funding for solar radiation modification from 2007 to 2025 was approximately $340 million, of which about $170 million was public, with more than $60 million of commercial and philanthropic money entering in 2025 alone, per the UK Parliamentary Office of Science and Technology. Cumulatively across two decades the field has spent roughly one-seventh of what a single year of deployment would cost. Frontier That asymmetry — a capability cheaper than the research programme that would characterise it — is the structural fact from which every governance problem in Geoengineering Governance follows.

Frontier The models agree on temperature and disagree on the thing people actually live under, and this is the real scientific frontier. Irvine and colleagues, in Nature Climate Change in 2019, modelled half-SG — a solar reduction offsetting half of CO₂-induced warming — and found that neither temperature, extreme heat nor extreme precipitation was exacerbated when averaged over the IPCC SREX regions, that less than 0.4% of the ice-free land surface saw exacerbation of extreme precipitation or water availability, and that a high-resolution model offset most of the CO₂-induced increase in tropical cyclone intensity. The authors state that distributional concerns are appropriate, and that this is a half-deployment idealisation rather than a policy. Frontier The pessimistic result comes from the region that would bear it. Bonou and colleagues, in Earth's Future in 2023, ran the GeoMIP G3 and G4 experiments across eleven models for West African monsoon precipitation. Under RCP4.5 warming the Sahel signals are small and non-significant, −0.86% northern Sahel and −0.80% southern. Under G3, with injection ramped to hold global temperature at 2020 levels, northern Sahel precipitation falls −17.4%, southern Sahel −8.47% and the West African region −3.71% — in absolute terms about 0.10 and 0.36 mm a day. The mechanism is dynamic rather than thermodynamic: injection reduces the land–sea thermal contrast, weakening monsoon circulation and shifting the intertropical convergence zone.

Frontier And the models disagree with each other, which is the part that matters most and is almost never reported. Most models showed decreases under G3 — except GISS-E2-R and HadGEM-ES. Under G4 the MIROC-ESM variants showed precipitation increases while BNU-ESM and CSIRO-MK3L showed decreases. The ensemble mean masks the divergence, particularly over coastal regions. Frontier Reading the three results together is the honest terminal position, and it is a declared tie. A global-mean-optimised half-deployment looks benign in a coarse regional average; a temperature-holding deployment looks severe for the Sahel in most models; and the models do not agree on the sign in some regions. The claim that SRM has been shown to be regionally acceptable is not supported. Neither is the claim that it has been shown to wreck the monsoon. The literature declares a tie at regional scale, and the tie is the finding rather than a failure to reach one.

Frontier Termination shock has been quantified, and disputed by people working from the same physics. Trisos and colleagues, in Nature Ecology & Evolution in 2018, used the G4 scenario — 5 Tg of sulphur dioxide a year from 2020, terminating abruptly in 2070 — and computed climate velocity, the speed at which a species must move to track its thermal niche. Post-termination velocities reach 12.8 km a year over ocean and 10.4 km a year over land, more than double the velocities for recent and projected climate change; ocean velocities in tropical marine biodiversity hotspots exceed 38 km a year. Median mammal dispersal is about 1.7 km a year, and termination velocity exceeds dispersal capacity for 93% of mammal species. On 29% of the land surface the temperature and precipitation velocity vectors diverge by more than 90 degrees — the niche fragments in two directions at once. Speculative Parker and Irvine argue the risk is overplayed, on the grounds that termination shock requires three things at once: large accumulated cooling, sudden and complete cessation, and a disruption persisting for months. They hold that a deployment system with distributed infrastructure, spare capacity and multiple capable nations should be easy to build as a resilient system, that gradual phase-down is the expected behaviour, and that additional warming on cessation would be of the order of a few tenths of a degree per decade; their named triggers of concern are terrorism against infrastructure, global catastrophe and unilateral political termination. Frontier The dispute is not about physics; it is about institutional reliability over a century. Trisos assumes a discontinuity, Parker and Irvine assume competent redundancy, and neither assumption has an evidence base, because no such system has existed. The velocity computation is a valid conditional result whose conditional is contested; the resilience claim is an argument from an unbuilt system, and is flagged the weaker of the two here for that reason.

Frontier The empirical moral-hazard literature does not find the effect the argument assumes, and this is the most under-reported finding in the subject. Moral hazard — that acknowledging SRM reduces support for emissions cuts — is the central ethical objection and is usually asserted as though established. Eggers and colleagues ran a pre-registered factorial experiment in the Journal of Experimental Political Science with n = 2,509 US residents across four arms — control, decarbonisation message, balanced information on aerosol injection, and both — and report no evidence that informing Americans about stratospheric aerosol injection affects average support for emissions reductions; balanced information increased perceptions that climate change is serious and requires decarbonisation, with weak heterogeneity by ideology at p = 0.059. Frontier Merk and Wagner, in Climatic Change in 2023, ran a revealed-preference field experiment across approximately 340,000 Facebook users and found little to no support for the hypothesis, with the exception that extreme messengers and framings did move responses — how it is said mattering more than whether it is said. Frontier Mark the interests in both directions. Gernot Wagner is an SRM economist and Christine Merk works on SRM perception, so their interest arguably runs with a null, which lowers the weight; Eggers and colleagues are political scientists with no field stake, which raises it. The Eggers result is also a replication failure against earlier studies finding either moral hazard or a “clarion call” effect in the opposite direction, which is the ordinary shape of a small-effects literature. Speculative The honest statement of the state of play is a split. The structural version of the argument — that a cheap temperature fix changes what politicians and firms can plausibly promise — survives untested. The individual-attitudes version has been tested repeatedly and mostly comes back null. Conflating the two is the commonest error in the discourse, on both sides.

Established The outdoor experiment record is nearly empty, and every attempt to fill it has failed on consent rather than on science. SCoPEx, proposed at Harvard in 2017 for Tucson, would have released a few kilograms of calcium carbonate or sulfuric acid from a stratospheric balloon to measure dispersal and optical properties. Relocated to Esrange in Sweden in 2021, its Swedish launch — an equipment test with no particle release — was cancelled in March 2021 after the advisory committee recommended against it, following opposition from the Saami Council and environmental organisations. David Keith left Harvard for Chicago in spring 2023, the team suspended work that summer, and Frank Keutsch confirmed in March 2024 that he was no longer pursuing the experiment. Funders included Bill Gates and the Hewlett and Sloan foundations. Established The University of Washington's marine cloud brightening trial sprayed sea-salt aerosol from the flight deck of the museum ship USS Hornet at Alameda from early 2024; the mayor learned of it from a newspaper in April 2024, the city halted it for violating the museum's lease and for inadequate notice — the city had received only vague advance mention of climate change science and misting — and the council voted unanimously in June 2024 against resumption after an hours-long meeting dominated by resident health concerns. Established The one continuous outdoor programme is the Reef Restoration and Adaptation Program's cooling and shading work on the Great Barrier Reef, running since roughly 2020; its reported 2025 field result is that a first-generation outdoor marine cloud brightening trial increases aerosol concentration at cloud base height — a demonstration that a sprayer puts particles where intended, which is a necessary first step and not a cooling measurement. Established The UK's ARIA committed £56.8 million across 22 projects in spring 2025, five involving outdoor experiments strictly limited in size and duration; at the one-year review two outdoor projects are active — sea-ice thickening in Nunavut and marine cloud brightening on the Great Barrier Reef — and three remain in technical design with no site confirmed. Established The scale gap is the decisive number. Per the UK POST briefing, small-scale outdoor experiments involve one kilogram to one tonne of material and intermediate-scale up to 1,000 tonnes; a year-15 deployment injection is 1.5 million tonnes of sulphur. The largest experiment anybody has proposed is three orders of magnitude below the smallest deployment. In roughly two decades, the number of completed outdoor stratospheric SRM experiments by a research institution is zero. Two were proposed and cancelled, both on governance and consent, neither on science.

Established The unilateral actor stopped being hypothetical, and is smaller and stranger than the literature predicted. Make Sunsets, a US startup, releases sulfur dioxide from small balloons and sells cooling credits against it; its April 2023 release in Baja California prompted Mexico's government to announce an intention to ban all SRM experiments including research, though the available analysis does not establish that the announcement became binding law. By April 2025 the US EPA, invoking Clean Air Act section 114, demanded information from the company, noting more than 124 deployments to date, that sulfur dioxide has been a criteria pollutant under the national ambient air quality standards programme since 1971, and that it was unclear whether the company had obtained permits or contacted any air quality agency. The administrator's framing was political; the regulatory point stands, which is that nobody knew and no mechanism existed to have known. Established Stardust Solutions raised $60 million in October 2025 led by Lowercarbon Capital, on top of a $15 million seed, with 25 scientists and engineers based in Ness Ziona, Israel and a planned US headquarters; its chief executive is a former deputy chief scientist of the Israeli Atomic Energy Commission and its lead scientist heads particle physics and astrophysics at the Weizmann Institute. To date it has flown lower-atmosphere tests using white smoke as a particle simulant and run indoor wind-tunnel dispersion tests, with no outdoor particle release. Handwave The proposed particle is proprietary and undisclosed, and its claimed properties are unverifiable until published; Cornell's Douglas MacMartin, himself an injection researcher, said the company has ignored every recommendation from everyone. Mark the interest carefully in both directions: established researchers criticising a for-profit entrant are competitors for legitimacy in the same field, which is a reason to discount — but the criticism is about disclosure, and the company concedes the disclosure gap and promises publication, which raises the weight of the criticism rather than lowering it.

Frontier Does the regional sign resolve, or is it model-structural? This is the field's first-order open question and it is answerable. A GeoMIP-successor intercomparison with common injection strategies at higher resolution, evaluated on monsoon dynamics rather than global mean and reported per model rather than as an ensemble mean, would establish whether the Sahel result is robust physics or an artefact of how a subset of models represents the intertropical convergence zone. Nobody is running it. Speculative A regional or hemispheric deployment aimed at a specific national objective — Arctic-only injection, or a deliberate monsoon adjustment — is flagged as a scenario type in the US research plan and is the least-studied high-consequence case in the whole subject; the strategic literature notes that Arctic and tropical deployments can cancel each other, and no state has proposed one.

Frontier Is SRM testable in principle? This is a genuine standoff and not a fringe position. The non-use argument holds that nothing below deployment scale generates information about deployment-scale effects, so research cannot reduce the decision-relevant uncertainty and functions only to normalise the option and build constituencies. The microphysics community holds the opposite: plume-scale processes — coagulation, size-distribution evolution, local heating — are the parameterisation uncertainty, so a kilogram-scale release would constrain models directly. Frontier Both positions are defensible, neither has been tested, and the dispute is presently unresolvable by construction, because settling it would require running the experiment whose value is in dispute.

Frontier Which particle, and does the choice matter as much as its advocates claim? Sulphate is the default because the volcanic analogue is sulphate. Speculative Calcium carbonate and other non-sulphate candidates are proposed to avoid the ozone and stratospheric-heating penalties, and the evidence for them is laboratory and modelling only — SCoPEx was designed to test exactly this and never flew, so the case rests on an experiment that was cancelled. Handwave A third position, that a proprietary engineered particle outperforms both, is held by one venture-funded company and supported by nothing in the public record.

Frontier Marine cloud brightening splits into two hypotheses that deserve different flags. As a global lever it rests on the Twomey effect, which is established physics, and on a scale requirement of at least 10,000 ocean-going sprayers across half the world's oceans at tens of billions of dollars a year, with roughly 50-nanometre sea-salt particles; aerosol–cloud interaction remains the single largest uncertainty in the entire forcing budget, and whether cloud susceptibility declines in a warmer climate is unresolved. Frontier As a regional reef intervention it is a much smaller and more testable proposition with the strongest outdoor record in the field — which is a first-generation aerosol delivery demonstration, not a measured radiative response. Conflating the two is how the field's thinnest result gets read as its strongest.

Speculative Arctic-focused intervention by sea-ice thickening — pumping seawater onto ice in polar winter to grow thickness and delay summer melt — is now in field trial under ARIA funding and through at least one other group. Energy and logistics scaling are unresolved and whether a thickness gain survives a melt season over any meaningful area is unknown; the engineering side belongs to Arctic Engineering. Speculative A space-based sunshade or L1 reflector swarm is the least atmosphere-invasive and least timely option in the menu: the physics is trivial and the engineering is not, at 10⁷–10⁸ tonnes of structure or roughly 1.5 billion sailcraft, trillions of dollars, and viability only at launch costs approaching $50 per kilogram. It is owned by Artificial Satellites for Climate Control; its presence in the menu is what makes “SRM” an ambiguous word.

Frontier Is unilateral state deployment the governing risk, or a myth? Martin and Moore, in the Harvard International Law Journal in 2025, argue that existing law provides only weak constraints and that a state might perceive another's injection as a national security threat and respond with force. Horton argues the opposite from structure rather than law: competing injection schemes interfere, the termination problem locks a deployer into indefinite commitment, and opponents hold cheap countermeasures, so cooperation follows from the physics. Speculative Horton's argument has an untested hole, and it is a striking one. The counter-geoengineering capability he names — fluorocarbon release, strategic black carbon, a competing scheme in the opposite hemisphere — has never been simulated in a coupled model. Nobody has established whether a deliberate counter-deployment neutralises the original perturbation or merely doubles it. An entire strategic-stability argument rests on a capability nobody has modelled.

Established And both arguments address the wrong actor. The unilateralism literature was built for a state deploying at climate-altering scale. What has actually occurred is a two-person company launching balloons and selling credits, and a venture-funded firm with a secret particle. The unilateral-actor problem arrived below the scale the theory was built for, which is a fact about the theory rather than about the risk. Frontier A related and largely unfunded question follows: whether a deployment would be detectable and attributable at all. The signatories of the 2023 climate-intervention research letter list detection and attribution as an explicit research need, stating that current knowledge is not sufficient to detect, attribute or project the consequences of SRM interventions — and without attribution no liability regime has anything to attach to.

Frontier And the framing question nobody asks directly: is this a tourniquet or a substitute? The Climate Overshoot Commission, the US research plan and most institutional assessments contemplate SRM only as a temporary overshoot measure alongside deep mitigation. Speculative That framing has never been operationalised in any instrument — no document anywhere defines what would count as the emergency, who declares it, or what ends it — so the tourniquet exists as a policy intention with no mechanism attached. Handwave The permanent-substitute version is held by essentially nobody in the literature and by a nonzero fraction of public discourse, and the ocean acidification point defeats it on its own.

Established The first bottleneck is an airframe that does not exist. No current aircraft combines the required payload with sustained flight at 20 km; commercial aviation cruises at 10–12 km, in the troposphere, where aerosols rain out in days to weeks rather than persisting for a year or more. The Smith and Wagner design is a bespoke tanker with a development programme attached, and its cost estimate is a design study rather than a quotation. Frontier A sulphur supply chain at megatonne scale, a fleet operations tempo, and the logistics of a continuous injection sustained for decades are all engineering problems nobody has been asked to solve at a level beyond paper.

Frontier The second is aerosol microphysics, which the models parameterise rather than resolve. Coagulation, size-distribution evolution under sustained injection, local stratospheric heating from absorbing particles and the resulting circulation feedbacks are the quantities that separate a volcanic impulse from an engineered layer. These are precisely what a kilogram-scale plume experiment would constrain, and precisely what has never been measured outdoors.

Frontier The third is model resolution at the scale of the monsoon. The GeoMIP experiments run at resolutions that represent the intertropical convergence zone coarsely, which is one candidate explanation for the inter-model sign disagreement over coastal West Africa. Whether the disagreement is physical or numerical is not currently separable from the published output.

Established The fourth is monitoring, and it is the one the cost estimate explicitly excludes. A deployment would require continuous stratospheric aerosol optical depth measurement, independent of the deploying party, at a fidelity sufficient to attribute regional anomalies. The suggested addition of about $3 billion a year for oversight is larger than the injection programme itself, which inverts the usual relationship between doing a thing and watching it being done. Frontier For marine cloud brightening the equivalent bottleneck is measuring a radiative response rather than an aerosol concentration — the step the field's best outdoor result stops one short of.

Established And the fifth is not technological at all, which is why it binds. Two institutional outdoor experiments were proposed and cancelled; three ARIA outdoor projects cannot name a site. No outdoor SRM experiment has been stopped by a regulator applying a rule; every one has been stopped by a community objecting to a process. Consent is functioning as the binding constraint in the absence of law — effectively, but with no standard, no appeal and no way for a researcher to know in advance what would satisfy it.

Established Nothing on this map produces the result this brief waits on. Planetary Cooling Concepts owns the levers this slot excludes and has already retired cirrus thinning on the evidence; Atmospheric Management owns the ozone regime and the composition side; Planetary Stewardship owns the question of whether any institution can manage a planet. None of them produces a number that would change this brief's verdict. What this slot waits on are two facts about the world and one missing scientific result, recorded as typed requirements below.

Frontier The scientific dependency is specific and nameable, which is the only condition under which it belongs on the map. It is not “we do not know enough about regional effects.” It is that eleven GeoMIP models disagree on the sign of monsoon precipitation change under the same prescribed experiment, and no published work isolates whether that disagreement comes from convective parameterisation, from resolution at the intertropical convergence zone, from stratospheric transport, or from the injection strategy each model implements. Until the source of the divergence is identified, the ensemble mean is an average over an unexplained disagreement, and neither an authorisation nor a prohibition can rest on it.

Established The institutional dependencies are the ones actually binding today. There is no decision rule anywhere in the world that could authorise or forbid a deployment, and there is no consent standard an outdoor experiment could satisfy in advance. Both are things a legislature or a treaty body could choose to supply and has so far chosen not to; neither is a research result, and the record of the last five years is that the second one has stopped more experiments than the first. Their construction is assessed in Geoengineering Governance, which is where the instruments live.

Frontier The decisive experiments are known, cheap by the standards of the subject, and unfunded. First, a GeoMIP-successor model intercomparison with common injection strategies at higher resolution, evaluated on monsoon dynamics and reported per model rather than as an ensemble mean. It requires no release, no permit and no consent, and would settle whether the Sahel result is robust or model-structural. It is the highest-value experiment in this brief and the one nearest to being runnable today.

Frontier Second, a stratospheric perturbation experiment with kilogram-scale release and full plume instrumentation. This is SCoPEx's design, unbuilt. It would constrain the aerosol microphysics that models currently parameterise. Its value is genuinely contested — the untestable-in-principle argument holds that kilogram-scale microphysics does not extrapolate to megatonne scale — and this brief records that both positions are defensible and neither has been tested. Established It is worth being precise about what was cancelled: the Swedish flight was an equipment test with no particle release, and it was still stopped.

Frontier Third, a marine cloud brightening trial with a measured radiative response rather than a measured aerosol concentration. The reef programme's result is the state of the art and it is one step below the question. The trial would also test whether cloud susceptibility behaves as the models assume in a warmer boundary layer.

Speculative Fourth, a coupled-model simulation of deliberate counter-geoengineering. Nobody has run one. The strategic-stability literature asserts that fluorocarbon release or strategic black carbon would make unilateral deployment futile; whether a counter-deployment neutralises or compounds the original perturbation is a straightforwardly simulable question that no group has published. Speculative Fifth, a prospective institutional moral-hazard study — tracking actual mitigation policy or corporate abatement decisions after SRM enters a jurisdiction's policy discourse. Every existing study measures individual attitudes; the version of the argument that survives is about institutions, and it has never been measured.

Frontier Sixth, and least discussed: a detection-and-attribution capability exercise. Building the ability to detect a stratospheric perturbation and attribute a regional anomaly to it requires no release at all, and it is a precondition for enforcing any regime, permissive or prohibitive. It is bundled with efficacy research under the single label “SRM research” by both sides of the petition dispute, and separating them is the most obviously available move nobody has made.

Established The reference deployment architecture is specific enough to price, which is unusual for this map and is the source of most of the trouble. The SAIL airframe: about 46 m long, 55 m wingspan, 25 t payload, 100 t maximum take-off weight, four wing-mounted engines, injecting at roughly 20 km. The programme: 0.1 Mt of sulphur in year one rising by about 0.1 Mt a year to 1.5 Mt in year 15, for a −0.25 W/m² offset. The bill: about $2.25 billion a year, $36 billion over 15 years, $3.5 billion of pre-deployment capital — and no monitoring, which the authors say explicitly and which would roughly double it.

Frontier Marine cloud brightening's engineering is worse specified and larger. A global lever needs at least 10,000 ocean-going sprayers producing roughly 50-nanometre sea-salt aerosol across half the world's oceans, at tens of billions of dollars a year, with the delivered particle size distribution as the critical parameter and no operating fleet to calibrate against. The reef-scale version is a small number of vessels and land-based sprayers, and has demonstrated that it can raise aerosol concentration at cloud base.

Established Sea-ice thickening is a pumping problem with a polar-winter logistics problem attached, currently in first field trials; the energy budget for any climatically meaningful area has not been published.

Frontier The engineering requirement nobody costs is the observing system. A deployment implies continuous, independent stratospheric aerosol measurement at sufficient fidelity to detect a perturbation, distinguish it from volcanic and wildfire injections, and support attribution of regional anomalies — maintained by parties other than the deployer, and sustained for the century of the commitment. Speculative No such system is funded, designed or proposed as an instrument. It is the one piece of hardware that every position in the debate needs, including the position that deployment should never happen.

Within this map: Planetary Cooling Concepts, which owns the cooling levers this slot excludes — surface albedo, radiative-cooling materials and cirrus thinning, whose sign it records as not robust and which this brief does not re-run; Atmospheric Management, which owns atmospheric composition where this slot owns the energy budget, and which holds the ozone-regime side of the injection-and-ozone interaction; Geoengineering Governance, which takes every institutional question this brief raises and none of the physics; Carbon Capture at Scale, the other intervention, and the only one that acts on ocean acidification; Artificial Satellites for Climate Control, which owns orbital sunshades and L1 reflectors; Weather Modification, a different mechanism on a different timescale and the home of the cloud-seeding record; Arctic Engineering and Ocean Engineering, which own the platforms two of these levers would run from; and Planetary Stewardship, the capstone accounting layer, whose finding that both successful stewardship episodes worked through identifiable producers with substitutes is exactly why this case is different — SRM has no producer to regulate, because the actor and the substance are whoever decides to fly.

This brief supersedes the carbon programme's treatment as this slot's coverage; that page remains as the programme's shorter narrative entry point and is the better first read for someone meeting the subject for the first time. Outside the map: stratospheric chemistry and dynamics; aerosol microphysics; the volcanic climate record; agricultural and hydrological modelling of monsoon systems; and the political science of technology assessment, which supplies the moral-hazard experiments.

Established The institutional finding of the last five years is an inversion, and it is the most important thing on this page. Legitimate actors cannot currently obtain consent to test at a scale of a few kilograms in a single location; an unsanctioned actor asking nobody has flown more than 124 times. A constraint that binds the transparent and not the opaque is the exact inverse of what governance is for, and it is what exists today.

Established Several of this brief's best sources are interested parties, and it is load-bearing to say which way each one runs. ARIA is a funder describing its own programme. The Reef Restoration and Adaptation Program reports its own field result. SRM360 is an SRM-focused non-profit whose funders are not disclosed on the pages cited. The Degrees Initiative reports on itself, and Geoengineering Monitor campaigns against all geoengineering — the two are cited against each other deliberately. Frontier Two findings run against their authors' apparent interest and are weighted up for it: the EU Scientific Advice Mechanism stating that injection does nothing for ocean acidification, and Smith and Wagner excluding monitoring costs from their own headline figure and saying so.

Established Research governance schemes reach the wrong actors. ARIA operates the most elaborate approval architecture actually in force — multi-stage approval, community engagement, environmental and legal assessment, an independent oversight committee and chief-executive sign-off. It is a condition of public funding. In 2025 alone more than $60 million of commercial and philanthropic money entered the field against roughly $170 million of cumulative public funding across two decades. A code of conduct enforced through a grant does not reach a venture-funded company with a proprietary particle.

Frontier Every expert body that has reported converges on the same narrow position, and none of it binds anyone. The US National Academies recommended a federal research programme of $100–200 million over five years with codes of conduct as a funding condition and patent pledges, supporting research and not deployment. The US Office of Science and Technology Policy's congressionally mandated report sets five research goals, states that issuing the report signals no policy decision and that current research is not a preparatory measure for deployment — while also stating an aim to prepare the United States for possible deployment by other public or private actors, which is the first US government acknowledgment that the relevant scenario may not be one it controls. The EU review presents options rather than recommendations and declines to recommend a moratorium. Research yes, deployment no, outdoor work only under consent and assessment — and nothing in that consensus is enforceable.

Frontier The distributional question is the sharpest one, and the honest answer is that it is unresolved rather than reassuring. A deployment optimised on global-mean temperature can look benign in a coarse regional average and still impose a large precipitation change on one monsoon system; the Sahel result under a temperature-holding scenario is −17.4% in the northern Sahel in most models, and the models that dissent do not dissent by a little. The populations with most exposure to that outcome have the least leverage over whether it is attempted. Established That is not a hypothetical asymmetry: it was the observable structure of the 2024 UN Environment Assembly negotiation, where the states most able to deploy resisted governance and the states most likely to be affected demanded it.

Established Consent is where the ethics has actually bitten, and it did so through communities rather than through ethicists. The Saami Council's objection stopped a balloon flight that would have released nothing. Alameda's council stopped a sea-salt sprayer after residents discovered it from a newspaper. Both were procedural objections about being told nothing, not scientific objections about aerosols, and both worked. Speculative Whether that is a defensible standard is a separate question from whether it is the operative one: it has no threshold, no appeal, no reach over commercial actors and no way for a researcher to know in advance what would satisfy it.

Frontier The moral-hazard objection deserves to be stated correctly rather than repeated. Its individual-attitudes version has been tested at n = 2,509 and across roughly 340,000 users and mostly comes back null — and balanced information raised stated concern about climate change in the pre-registered study. Its structural version, that a cheap temperature fix changes what governments and firms can credibly be held to, is untested and remains a serious argument. Speculative Asserting the tested version as established, which both advocates and opponents do, spends the argument's credibility on the half that has been disconfirmed.

Frontier Termination is an intergenerational commitment problem with a number attached. A deployment large enough to matter creates an obligation to continue it for as long as the greenhouse forcing persists, discharged by institutions that do not yet exist, against a computed post-termination climate velocity that exceeds the dispersal capacity of 93% of mammal species. The counter-argument — build in redundancy and phase down gradually — is an argument about the competence of a century of future institutions, and it is not evidence.

Frontier And research capacity in the Global South is itself contested, with interested parties on both sides. The Degrees Initiative reports five new country teams, fourteen renewals, roughly 60 papers and stated neutrality on whether SRM should ever be used. Geoengineering Monitor alleges that more than 90% of funded studies model injection under the highest-emissions scenarios, that 80% of authors are at Global North institutions, and that funders overlap with marine cloud brightening experiments. The underlying counts are largely not disputed; the interpretation is. Declare the tie and record both.

Established The terminal position on the framing is that it fails, and it fails somewhere unexpected. “Usable option” implies a capability that can be exercised. The record shows a capability that cannot currently obtain permission to be tested at a scale of a few kilograms in one location, while an actor who asks nobody has exercised it 124 times. The binding constraint on SRM is not that it does not work. It is that legitimate actors cannot get permission to find out, and illegitimate ones do not need it.

Established This is the only planetary-scale intervention on this map whose cost does not govern it. A space elevator, a fusion economy, a continental irrigation scheme — each is disciplined by its own price. A capability at $2.25 billion a year sits inside the discretionary budget of dozens of states and, at the low end of the literature range, of individuals. Frontier Every governance pathology in this subject follows from that single arithmetic fact, and no amount of institutional design changes the arithmetic.

Frontier The second-order finding is quieter and cuts the same way. The moral-hazard objection, which is where the field spends most of its argumentative energy, is the one that repeatedly fails to replicate. The objection that is holding — procedural consent — receives comparatively little research funding, has no dedicated methodology and appears in no framework as an operational standard. The field is optimising against the wrong constraint.

Speculative The civilizational reading, stated as the speculation it is: a technology whose research programme costs more than its deployment inverts the usual sequence in which a civilisation learns before it acts. In every other case on this map, capability arrives after understanding because building the thing is the hard part. Here understanding is the expensive half. Handwave Whether a civilisation can be governed into learning first, when acting is cheaper than learning, is not a question the historical record answers — the two completed planetary interventions in the record both worked through identifiable producers with substitutes, and SRM has neither.

These horizons track experiments, funding cycles and reporting dates rather than deployment, because no deployment schedule exists to track:

• 10 yr: Established Two hard dates dominate the near term: the 2026 Scientific Assessment of Ozone Depletion, which will for the first time treat stratospheric aerosol injection prominently and has not yet reported, and ARIA's outdoor programme, where two of five projects are active and three cannot name a site. Frontier Expect the model intercomparison question to be resolvable in this window and probably not resolved, because it is nobody's funded priority. Frontier Expect at least one further commercial actor and at least one further national or sub-national prohibition; the observed rate over 2023–2026 supports both. Speculative Whether any research institution completes an outdoor stratospheric release with particles in this window is genuinely open, and the base rate over two decades is zero.
• 25 yr: Speculative If a deployment is ever attempted, this is the earliest window in which a state-scale programme could be built, and the constraint is the bespoke airframe rather than the decision. Speculative A more likely path is asymmetric: modelling improves, the regional sign question resolves one way or the other, and the outdoor record stays thin because consent remains the binding constraint. Handwave Forecasting which of those obtains is asserting a political outcome, not extrapolating a measurement.
• 50 yr: Speculative On a high-overshoot trajectory the tourniquet framing — temporary deployment alongside deep mitigation — is the scenario most institutional assessments contemplate, and it has never been operationalised in any instrument, so what it would look like in practice is unknown. Speculative A deployment begun and sustained through this horizon implies an unbroken institutional commitment longer than most constitutions have lasted. Handwave Assigning a probability to that is not a measurement of anything.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The only relevant datum at this horizon is that a deployment initiated today and terminated at any point within it produces the computed climate-velocity result — and the computation is conditional on an abrupt cessation whose likelihood is exactly the thing nobody can estimate. Handwave The ocean acidification point is the one claim that survives at any horizon: reflecting sunlight does not touch it, so on this timescale the carbon problem is unchanged by anything in this brief.

• Depends on Nothing on this map. This brief waits on no result another brief produces. Planetary Cooling Concepts has retired cirrus thinning on its own evidence and its adjudication already records the ice-nucleation result that would settle that sign; Atmospheric Management holds the ozone regime; Planetary Stewardship holds the accounting layer. None of them produces a number that changes this verdict. No typed depends-on edge is claimed.
• Requires (not on this map) Three things, none of which is a brief on this map. First, a decision rule that could authorise or forbid a deployment: no instrument anywhere assigns anyone the power to permit or prohibit stratospheric aerosol injection, and the two enacted prohibitions in the world are US state statutes covering conduct within one state's borders. Second, a consent standard an outdoor experiment could satisfy in advance: SCoPEx was cancelled in Sweden after the Saami Council objected and abandoned in the United States; the Alameda marine cloud brightening trial was stopped by a unanimous council vote; three of ARIA's five outdoor projects still cannot name a site — and in none of those cases was there a threshold a researcher could have aimed at, because none exists. Third, and this is a scientific result nobody is producing rather than a fact anyone could choose to supply: eleven GeoMIP models disagree on the sign of monsoon precipitation change under the same prescribed injection experiment — most show Sahel decreases under G3, GISS-E2-R and HadGEM-ES do not, and under G4 the MIROC-ESM variants show increases where BNU-ESM and CSIRO-MK3L show decreases — and no published work isolates whether the divergence comes from convective parameterisation, from resolution at the intertropical convergence zone, from stratospheric transport, or from differences in implemented injection strategy. Until that is identified, the ensemble mean is an average over an unexplained disagreement, and no authorisation or prohibition can honestly rest on it.
• Enables In principle, any overshoot-management pathway that assumes a fast temperature lever depends on this question resolving. No typed enabling edge is claimed, and the reason is the finding: nothing here has reached the point where another brief could rest a result on it. The one place a genuine enabling relationship might form is detection and attribution, which every governance proposal in Geoengineering Governance presupposes and which no programme funds.
• Adjacent Stratospheric chemistry and dynamics; aerosol microphysics; the volcanic climate record; monsoon hydrology; the political science of technology assessment, which supplies the moral-hazard experiments; and within this map Planetary Cooling Concepts, Atmospheric Management, Geoengineering Governance, Artificial Satellites for Climate Control and Weather Modification.

Handwave “A secret large-scale atmospheric spraying programme is already operating.” It is not, and treating the claim as beneath discussion is a mistake, because the reason it fails is a genuinely instructive lesson in atmospheric science and in what a covert planetary intervention would actually require. A reader who arrives believing it should be able to leave understanding why it does not work. Established The direct test exists. Shearer, West, Caldeira and Davis surveyed two expert groups — atmospheric chemists specialising in condensation trails, and geochemists working on atmospheric deposition of dust and pollution. 76 of 77 participating scientists reported no evidence of a secret spraying programme. The single dissent concerned atmospheric barium at a location with low soil barium. The experts assessed that the photographic and sampling evidence offered by proponents is explained by ordinary contrail formation and by poor sampling practice.

Established The observation driving the belief is real and has a complete, unglamorous explanation. Whether an exhaust plume dissipates in seconds or spreads into cirrus for hours is governed by ambient temperature and ice supersaturation at flight level, which is why two aircraft minutes apart can leave completely different trails. That is contrail physics, not a difference in cargo. Established Soil and water samples reported as showing aluminium, barium and strontium are showing three of the most abundant crustal metals in ordinary proportions; a detection without a matched control and a background baseline is not a detection.

Established The scale argument is the decisive one, and it is where the contrast with the real record does the work. The engineered deployment in the literature requires a bespoke aircraft that does not exist, because no current airframe combines the payload with sustained flight at 20 km; commercial aviation cruises at 10–12 km, in the troposphere, where aerosols rain out in days to weeks. A covert programme would need a purpose-built high-altitude tanker fleet, a sulphur supply chain at megatonne scale, and it would produce an unmistakable and continuously rising stratospheric aerosol optical depth signal in the satellite and lidar record maintained by dozens of independent institutions in mutually hostile countries. Stratospheric aerosol loading is one of the best-monitored quantities in Earth observation, and it does not show it. Established Set that against the actual outdoor experimental record: zero completed institutional stratospheric releases in two decades, two proposals cancelled on consent, and the largest experiment anybody has proposed involving one kilogram to one tonne of material against a deployment year-15 figure of 1.5 million tonnes. The discourse implies a field with a testing programme. It does not have one.

Frontier The part worth a reader's time is that the believers are right about one structural thing and wrong about everything else. They are right that a small number of actors could, cheaply, alter the atmosphere over everyone else without asking — that is precisely the finding of this brief's cost and unilateral-actor sections. They are wrong that it has happened. The genuine scandal is not concealment; it is that the thing they fear is legal, cheap, has been attempted at small scale by a startup, and no framework exists to stop it. Conflating the false claim with the documented one makes the real gap harder to discuss, which is a cost borne by everybody, including them.

Established “Contrails are irrelevant to climate.” This is the mirror-image error and it is also wrong. Aviation-induced cirrus is a real, measurable, net-warming forcing, and contrail avoidance by flight-path adjustment is an active mitigation proposal. Frontier Its efficacy is contested and the forecasting skill required to route around ice-supersaturated regions is the open question. Aviation is an inadvertent climate intervention; it is not a covert one, and giving the contrail observation its due is part of answering the claim rather than a concession to it.

Frontier “SRM has been shown to be regionally acceptable” and “SRM has been shown to wreck the monsoon” are both over-readings of the same literature. One rests on a half-deployment idealisation evaluated over coarse regions; the other on a temperature-holding scenario in a subset of models, two of which go the other way. Frontier “Termination shock is settled.” It is not settled in either direction: one paper computes a severe outcome conditional on abrupt cessation, another argues abrupt cessation is engineerable away, and no system exists to test either. Citing Parker and Irvine to support the severity of termination shock, as summaries sometimes do, cites the paper arguing the opposite.

Frontier “Moral hazard is established.” The individual-attitudes version is closer to disconfirmed than confirmed. Handwave “Stardust has a particle with the properties claimed.” Nothing has been published; the claim that a proprietary particle could counteract all the greenhouse gases emitted over the last 150 years is unsupported by anything disclosed, and the flag stays here regardless of the team's credentials. Speculative “Counter-geoengineering would deter a unilateral deployer.” Asserted in the strategic literature, never simulated in a coupled model. Established “Marine cloud brightening has demonstrated cooling in the field.” The best result to date is a demonstrated increase in aerosol concentration at cloud base height. Established “A state has deployed, tested or announced intent to deploy at climate-altering scale.” No such case exists.

Handwave And the claim that defeats the framing outright: “SRM can substitute for emissions reduction.” No credible pathway supports it, and essentially nobody in the literature holds it — but a nonzero fraction of public discourse does. The ocean acidification point alone is sufficient: reflecting sunlight lowers temperature and leaves the carbon in the ocean, which is why the only intervention in this category that acts on acidification is the one in Carbon Capture at Scale, not this one.