Nobody Can Spot a Secret Geoengineering Test Right Now. A New Study Says a Satellite Could.

Picture a company or a country quietly releasing sulphur dioxide from high-altitude balloons to test whether it can cool the planet. No announcement, no permit, no oversight. A new study in Atmospheric Measurement Techniques says the world currently has no way to catch them doing it, but proposes a solution: an infrared satellite instrument that could detect and measure such a test within hours to a few days of the first release.

The technique is stratospheric aerosol injection, or SAI, the most widely discussed form of solar geoengineering. The idea copies nature: volcanoes inject sulphur dioxide into the stratosphere, where it becomes tiny sulphate particles that scatter sunlight and cool the planet. Mount Pinatubo’s 1991 eruption injected an estimated 15 to 20 teragrams of sulphur dioxide and temporarily lowered global mean surface temperatures by more than one degree Celsius. SAI would do the same thing on purpose, continuously, to offset warming from greenhouse gases.

The new research, led by Pasquale Sellitto of the Laboratoire Interuniversitaire des Systèmes Atmosphériques in France with colleagues from Germany, Spain, the Netherlands, Belgium, and the United States, starts from an uncomfortable premise. There is no global framework regulating SAI, despite repeated calls for one from the United Nations Environment Programme, the European Parliament, and national academies. And according to estimates from the EU-funded Co-Create project, a small-scale outdoor experiment could be carried out today with existing technology for between 10 and 100 million US dollars, using stratospheric balloons to release anywhere from a few kilograms to a few hundred tonnes of sulphur dioxide.

The blind spot

Current satellites were built to watch volcanoes, not secret experiments. A typical moderate volcanic eruption injects a few teragrams of sulphur dioxide, that is millions of tonnes, which is tens of thousands to billions of times more than a plausible early SAI test. Instruments tuned to volcanic plumes simply cannot resolve something that small.

The study audits the existing fleet and the picture is not reassuring. Solar occultation instruments like SAGE III/ISS deliver excellent precision but their coverage is sparse; they could only distinguish an SAI perturbation at injection rates of 1 to 2 teragrams of sulphur per year, far above anything an exploratory test would release. Limb-scattering instruments such as OMPS-LP operate in visible wavelengths, so they cannot chemically identify sulphate aerosol at all. Space LiDARs like EarthCARE’s ATLID can detect faint aerosol layers but cannot determine what they are made of. Nadir infrared sounders like IASI can identify sulphate but give almost no information about altitude. Meanwhile, the infrared occultation instrument ACE-FTS is expected to be retired within a few years, and several other middle-atmosphere missions are nearing the end of their lives. One group of atmospheric scientists has warned of an “imminent data desert” in stratospheric monitoring. Against this backdrop, an early geoengineering test would be, for all practical purposes, invisible.

How the researchers tested detection

The team turned to a mission concept called CAIRT, the Changing-Atmosphere Infra-Red Tomography explorer, which was one of four candidates shortlisted in the European Space Agency’s Earth Explorer 11 competition. CAIRT was ultimately not selected, but its design is unusual: the first limb sounder in space using imaging Fourier-transform thermal infrared technology, observing the atmosphere edge from 5 to 115 kilometres altitude with spectral resolution of 0.25 wavenumbers or better across the mid-infrared range. That design delivers three-dimensional profiles of trace gases and aerosols with vertical resolution of 1 to 3 kilometres.

To evaluate whether such an instrument could spot a clandestine test, the researchers built what they call a pseudo-reality from state-of-the-art climate simulations. They ran the Community Earth System Model (version 2) with the Whole Atmosphere Community Climate Model (version 6) under the Geoengineering Model Intercomparison Project, using the Geo SSP5-34-OS 2.0 scenario in which sulphur dioxide is injected at about 20 to 25 kilometres altitude to hold warming to 2 degrees Celsius. In the first simulated deployment year, 2034, the injection rates are deliberately tiny: about 0.035 teragrams per year at 15 degrees north, roughly 100 tonnes per day, and about 0.005 teragrams per year at 15 degrees south, roughly 15 tonnes per day. Those are the kinds of rates a near-term unilateral test might actually use.

From that simulation the team generated synthetic observations of what a CAIRT-like instrument would measure, including realistic noise derived from the mission’s feasibility studies. Detection exploits the chemicals’ spectral fingerprints: sulphur dioxide near 8.7 and 7.3 micrometres, sulphate aerosol near 12.5 and 8.3 micrometres. Retrieval uncertainties came out at 0.15 to 0.25 parts per billion for sulphur dioxide and 1.0 to 1.5 parts per trillion for sulphate aerosol, remarkably close to background stratospheric levels.

The world is currently blind to the earliest and most plausible forms of geoengineering. This study offers the first proven-in-simulation path to seeing them.

What a CAIRT-like instrument would actually see

The results were precise. In the simulated February 2034 scene, the synthetic observations clearly detected both injection sites, pinned their altitude at 25 to 30 kilometres, and quantified the sulphur dioxide mass intercepted at 3.1 tonnes against a true value of 3.4 tonnes at the northern site, and 0.4 tonnes against 0.5 tonnes at the southern one, biases of roughly 10 and 25 percent. Even the weaker southern injection, cumulatively only a few hundred tonnes by the end of February, stood well above the detection limit, and the plume’s descent, zonal spreading, and poleward drift were all faithfully captured.

The technique also tracked the aftermath through the simulated year. It followed the conversion of sulphur dioxide into sulphate aerosol as a new particle layer formed at the injection longitudes, spread to encircle the globe, and dispersed toward higher latitudes in both hemispheres. Differences between the synthetic observations and the model truth generally stayed within 10 percent inside the plume. The instrument even reproduced the self-lofting of the aerosol layer, in which particles absorb solar radiation, heat the surrounding air, and lift the layer to higher altitudes over time, a phenomenon observed in reality after the 2019 Raikoke eruption.

Why this matters for governance

Detection is not governance, and the authors do not pretend otherwise. Knowing that someone is injecting sulphur dioxide is not the same as having the legal authority to stop them, and no such authority exists at the global level. SAI remains deeply controversial for good reasons: modelling suggests side effects including shifts in regional precipitation, altered temperature gradients between hemispheres and between the poles and the equator, and delayed recovery of the stratospheric ozone layer. A sudden halt to a large deployment could trigger a rapid rebound warming known as termination shock.

Arms control treaties work because parties can confirm compliance; climate interventions cannot be governed if they cannot be seen.

The open question is whether anything like CAIRT will fly. The mission lost the Earth Explorer 11 selection, and its scientific case is now being made in part through studies like this one. If a CAIRT-like mid-infrared limb sounder does not launch in the coming decades, the blind spot persists and the middle atmosphere keeps drifting toward the data desert. The surveillance gap is quantified, the fix is designed, and the fix has no launch date.