Why we created this roadmap
The world needs to be able to make a legitimate, evidence-based decision about if, when or how to pursue sunlight reflection—fast enough to matter.
SAI—releasing reflective particles into the stratosphere that reflect sunlight back to space—may be the only practical near-term option to mitigate many potentially devastating climate impacts in the next few decades. We know it can cool the planet, which could mitigate near-term climate risks while the world pursues the necessary work of decarbonization. But today, we lack the data, tools, and infrastructure required to judge whether, how, or under what conditions it can or should be deployed responsibly. This is a plan to build that evidence base: which questions must be answered, in what order, and at what cost.
The SAI Roadmap, built for policymakers, funders and scientists, builds on Reflective's Uncertainty Database. Launched in 2025, the Uncertainty Database took the first step: identifying which questions need answering before a well-informed deployment decision is possible. The SAI Roadmap takes the next step: scoping, sequencing, and costing out the research activities that generate the evidence those open questions require. For each research activity in the Roadmap, we give a scientific justification, scope of work, estimated budget, and timeline: resulting in a specific, legible plan for the research program needed for informed decision making.
We estimate that an informed decision about whether to begin monitored cooling, closely observed and adaptively managed, is achievable in 10-22 years for $370M-$1.36B. How we arrive at these figures is described in the Methodology.
A clinical-trial approach to SAI research
The medical field has already solved a version of the problem that SAI research faces today: how to build an evidence base before an intervention reaches the public, and decide in advance what would count as a reason to stop.
Medicine developed the process of clinical trials slowly, over decades, largely in response to crises that claimed human lives and forced the issue. We can apply that same logic—and its phased structure—to SAI research without waiting for a crisis. Each phase is a coordinated research program across impacts, aerosols, engineering, and observations. Phase 0 builds foundational knowledge. In Phases 1 and 2, outdoor experiments anchor each phase and provide critical new data, but they are not the whole phase. Phase 1 constrains aerosol microphysics; Phase 2 measures stratospheric mixing and distribution; and Phase 3 monitors deliberate cooling.
Two stage gates separate the phases. At each gate, the full body of evidence—not just the result of an experiment—is assessed against criteria set in advance before the next phase can be considered–the same kind of go/no-go checkpoint that governs whether a clinical trial advances to its next phase.
Reflective's Uncertainty Database maps how well-understood each open question about SAI is, rating it high, medium, or low uncertainty, and tracks how that improves over time. Passing each gate means the activities in the previous phase have measurably reduced uncertainty. To move from Phase 1 to Phase 2, every high priority uncertainty must come down to medium priority, and every medium priority uncertainty about impacts must come down to low priority. To move from Phase 2 to Phase 3, every remaining uncertainty — across engineering, aerosols, observations, and impacts — must sit at low priority.
Phase 0: Building Foundational Evidence
In Phase 0, where we are today, models, laboratory studies, and analysis of existing observations—including data from past volcanic eruptions—help estimate the potential benefits and risks of SAI and identify the uncertainties that matter most. This work identifies which uncertainties cannot be resolved without outdoor data, determines what an experiment would need to measure, and assesses whether a feasible experiment design could produce the needed evidence.
Phase 1: Constraining Aerosol Microphysics
Phase 1 is designed to gather real-world observations of how sulfate aerosols form and grow in the stratosphere in the days and weeks after release. While existing models predict how these particles will evolve, different models make different assumptions about aerosol evolution in the stratosphere, and no one has measured this directly. As a result, today, our models disagree by roughly a factor of two on how much cooling would result from a given amount of sulfur dioxide. The aerosol microphysics experiment—which would release around 10 tonnes of sulfur dioxide into the stratosphere and have no cooling effect—is expected to cut that uncertainty by roughly one quarter.
The experiment anchors Phase 1 but is not all of it. The phase also builds the modeling tools that turn experimental data into improved projections, advances research on SAI's climate and Earth system impacts, and develops the observing systems that any larger experiment—and any potential deployment—would depend on.
Gate: Proceeding to Phase 2 requires all three of the following scientific decision criteria to be met—these are the roadmap's proposed evidence tests; the final criteria, and the process for judging whether they've been met, should be agreed by independent stakeholders before Phase 2 begins.
- Are projections of SAI's climate and Earth-system effects meaningfully better constrained?
- Do those projections point to a potential deployment scenario that would reduce climate risks and improve lives?
- Are the important uncertainties that remain ones a larger controlled experiment would resolve?
All three must be true to proceed to a Phase 2: a larger controlled experiment using more material over longer periods of time. Anything less means repeating work, or stopping here.
Phase 2: Measuring Stratospheric Distribution
Phase 2 is designed to gather data from larger releases of sulfur dioxide—approximately 250 tonnes of sulfur dioxide per day, sustained over a full season—to understand how stratospheric winds transport sulfate aerosols across the globe. An aerosol evolution and transport experiment at this scale would remain measurable in the stratosphere for many months, allowing it to test model predictions of aerosol growth, transport, and lifetime over months to a year. Together, these three properties determine the effects of SAI:
- Growth: how large the aerosol particles become governs how efficiently they reflect sunlight back to space.
- Transport: stratospheric winds carry the aerosols around the globe, often to locations far from where they were released—shaping which regions see the surface climate response.
- Lifetime: how long the aerosols remain in the stratosphere before falling out determines how long a given release keeps cooling the planet (larger particles fall out faster, so growth and lifetime are linked).
With this experimental data, scientists can model different injection strategies and get more precise estimates of the climate and Earth system effects each would produce.
Gate: Proceeding to Phase 3 requires all of the following scientific decision criteria to be met — these are the roadmap's proposed evidence tests; the final criteria, and the process for judging whether they've been met, should be agreed by independent stakeholders before Phase 3 begins.
- Do the improved projections point to at least one deployment scenario that would reduce climate risks and improve lives?
- Have the engineering and observing systems needed to conduct, monitor, and evaluate a gradual deployment—and to pause or change course if needed—collected enough baseline data?
Yes to both could unlock Phase 3: monitored cooling. The decision to begin monitored cooling belongs to a parallel governance process this roadmap informs but does not replace.
Phase 3: Monitored Cooling
Phase 3 is designed to monitor deliberate cooling as it happens—much like a Phase IV clinical trial monitors for effects that are rare or slow-emerging. If evidence from Phases 1 and 2 supports moving forward, governments would begin with a slow ramp-up rather than deploying at full scale immediately, and the climate response would be continuously measured against model predictions—with any unexpected or harmful effects able to trigger a pause, a change in approach, or a gradual phase-out. Phase 3 monitoring draws on observations of both the stratosphere and the slower-emerging climate effects, using them to keep reducing uncertainty and to confirm that any changes we detect are actually caused by SAI, rather than by natural variability or other factors.
Focus of the Roadmap
The SAI Roadmap does not make the case for deployment—it lays out the evidence base decisionmakers will need to decide for themselves.
The SAI Roadmap focuses on SAI using sulfate aerosols. We focus on sulfate SAI because it is the sunlight-reflection approach best suited to meaningful global cooling, has a real-world precedent in volcanic eruptions (Pinatubo's 1991 eruption cooled the planet by roughly 0.5°C for a year or two), and offers a tractable path to reducing key uncertainties through small outdoor experiments. This is a prioritization under time and resource constraints, not an endorsement—other approaches, like marine cloud brightening, may still matter for more targeted objectives.
Within sulfate SAI, where relevant, we assume a deployment scenario broadly aligned with the one described in our Uncertainty Database methodology: a hemispherically balanced deployment of sulfur dioxide that ramps up gradually and is released from aircraft. While experiments and early-stage deployments could start at high latitudes and low altitudes (13–15 km) using modified existing aircraft, we assume deployment would move to high altitude (>20 km) and predominantly subtropical latitudes within the first decade of deployment, before reaching 0.5°C of cooling.
The SAI Roadmap considers only technical research—that is, research in the physical sciences and engineering. Research into the governance of SAI and its social and political aspects, while vital, is therefore out of scope for this roadmap.
More information on funding scenarios, budget, and timeline assumptions is given in the Methodology.
How to use the Roadmap
The roadmap shows a timeline of research activities organized by phase, with activities grouped into four research tracks: impacts, aerosols, engineering, and observations. These can also be grouped by duration or cost. Funding determines how much research runs in parallel and whether activities can be compressed, but never bypasses a gate. Each research activity can be expanded to show its scientific justification, scope of work, and estimated budget and timeline, and each gate opens to the body of evidence required to reach it and the questions its decision rests on.
Acknowledgements & feedback
We want to acknowledge the very many conversations and individuals that helped shape these research activities. For their detailed advice and information, we particularly thank Marta Abalos, Phil Ansell, Bianca Baier, Alexandre Baron, Ewa Bednarz, David Brown, Tahlia Crabtree, Sebastian Eastham, Jim Franke, Greg Frost, Colleen Golja, Hamish Gordon, Cameron Homeyer, Todd Hutchinson, David Keith, Frank Keutsch, JF Lamarque, Chris Lennard, Dana Moreno, Liz Moyer, Dan Murphy, Jessica Neu, Jeff Pierce, Prateek Ranjan, Karen Rosenlof, Kate Smith, Hongwei Sun, Colm Sweeney, Simone Tilmes, Ed Young, and Peter Zoogman.
If you see something we've missed or that you disagree with, use the feedback link on any page. We intend to improve the roadmap over time, and input from the research community will be vital for doing so.
Contact: roadmap@reflective.org