We can safely experiment with reflecting sunlight away from Earth. Here’s how | Dakota Gruener and Daniele Visioni

TThe world is warming rapidly – ​​and our options for avoiding catastrophic damage are narrowing. 2024 was the first full year to exceed the 19th century average by more than 1.5°C. Emissions continue to rise, and fossil fuel consumption is expected to reach a new peak in 2025. Permanent carbon removal technologies – often cited as a solution – only remove tens of thousands of tons per year, almost nothing compared to the 5 to 10 billion tons needed. Reducing emissions and increasing carbon removal remain essential. But they may not be enough.

As suffering increases and ecosystems collapse, more and more people will ask: can we do anything to prevent this damage? The idea of ​​reflecting a small fraction of incoming sunlight to reduce warming is not a new idea. In 1965, Lyndon B. Johnson’s scientific advisors proposed it as only a way to cool the planet. The Earth already reflects about 30% of incident sunlight; increasing this fraction slightly – say to 31% – could strengthen the planet’s natural heat shield. But how?

In 1991, Mount Pinatubo erupted and sent about 15 million tons of sulfur dioxide into the stratosphere, cooling the planet by about 0.5°C. This eruption became a natural experiment and inspired the idea of ​​stratospheric aerosol injection (SAI). Models suggest that SAI could offset 1°C of warming with around 12 million tonnes of SO₂ per year – far less than we currently unintentionally emit from industrial processes, but with a much greater cooling effect.

Let’s be clear: SAI is not a substitute for reducing emissions. If it is deployed and then suddenly stopped, the planet will experience a rapid rebound in warming. Poorly designed or poorly coordinated interventions could alter precipitation patterns catastrophically. But that’s exactly why the research is needed – not to greenlight deployment, but to understand whether SAI could ever be used safely, effectively and in the public interest.

Some argue that the risks of misuse mean it shouldn’t even be studied. We disagree. Careful and open research can determine whether a well-governed approach could reduce harm, particularly for the most vulnerable. It can also surface risks and failure modes early, making unwise proposals less likely to gain traction. In this sense, research acts as a safeguard and not as a slippery slope.

But how do you know if something is safe or too risky? We don’t need to reinvent the wheel. Medicine solved the “too risky to test” dilemma 60 years ago by codifying clinical trials in stages. A similarly structured, staged program for SAI can safely provide the evidence that policymakers will eventually need.

Right now, we’re stuck in the “preclinical” phase, or phase zero: lab work and computer models. These are excellent tools – they have successfully predicted the risks of increased emissions – but we cannot build confidence in their predictions without verifying that they correctly capture the key processes of the ISC. How do aerosols form, evolve and disperse in the stratosphere? How do they interact with the environment? These are key factors for any solid assessment. What would clinical trial-like phases for SAI look like?

The first phase would involve releasing a tiny amount of SO₂ – about 10 tonnes of SO₂ (a fraction of what many coal-fired power plants emit in a day) – at appropriate altitudes and carefully measuring its change using a suite of instruments: aircraft, ground and satellite. This amount would be far too small to affect the climate but would allow researchers to study how aerosols form and behave – which remains one of the biggest scientific uncertainties in the field. Comparing these observations to model predictions would quickly test key assumptions and help identify areas where current projections are robust – and where they need to be refined.

A potential phase two experiment could be 10 or 100 times larger – still orders of magnitude smaller than a “small” volcanic eruption like Mount Ruang, which injected around 300,000 tonnes at once in 2024 and had yet to no measurable impact on the global climate. This would allow researchers to study how aerosols mix and distribute. How fast do particles travel? How do they interact with the stratospheric circulation? Are our models capturing this correctly? Otherwise, what are we missing? Are we observing something totally unexpected? The observation capabilities required for these tests would also be essential to detect unauthorized deployment.

Once researchers around the world had a chance to review the data and draw their own conclusions, the evidence could be put to the test: Are governments interested in moving forward with something that begins to resemble a deployment? If so, the research would move to a phase three – similar to a phase four trial in post-licensure medicine – involving deliberate mild cooling, perhaps around 0.1°C over five years, under constant observation and strict monitoring. Such a slow (and reversible) rollout, if coupled with a strong governance framework, would be the opposite of an irregular or reckless rollout.

The world may never need to reflect sunlight. But if this is the case, the only way to make a responsible decision regarding its use will be to produce concrete evidence, in a transparent manner, Before a crisis forces our hand. This means putting tools, rules and monitoring mechanisms in place now, not later.

We see the UK’s Advanced Research and Invention Agency (Aria) program as an important first step in this direction. In the laboratory that one of us leads, a new project funded by Aria is developing the theoretical bases for determining the minimum scale at which an outdoor experiment could significantly reduce key uncertainties – an essential basis for any future research to be conducted safely and transparently. And at reflective, the organization that one of us leads, we work to support open science, careful coordination, and strong public accountability across the field.

Outdoor research is not a slippery slope to deployment. This is how we ensure that any future decisions – whether to move forward, reject the idea altogether, or refine it – are based on facts and not fear or wishful thinking. Done well, small-scale experiments can reduce both scientific uncertainties and political risks. The real danger is not asking the question. You have to wait too long to know the answer.

• Dakota Gruener is CEO of Reflective, a nonprofit climate initiative that accelerates the rate at which sunlight reflects, and Dr. Daniele Visioni is an assistant professor of Earth and atmospheric sciences at Cornell University.