The ozone layer has become one of the great environmental success stories of the past half century. After the world banned the chlorofluorocarbons and other ozone-depleting chemicals responsible for the Antarctic ozone hole, satellite instruments began detecting the first tentative signs of healing in the upper stratosphere. But a sweeping new analysis of tropical ozone, published in Atmospheric Chemistry and Physics, delivers a sobering caveat to that narrative: the satellite records scientists rely on to verify recovery disagree with one another so profoundly that they currently cannot be used to judge whether computer models are projecting the future of the ozone layer correctly.
The study, led by Sean M. Davis of the NOAA Chemical Sciences Laboratory in Boulder, Colorado, brought together an international team of more than forty researchers to tackle a deceptively simple question. Over the tropics, the belt between 30 degrees south and 30 degrees north that covers half the planet’s surface and is home to roughly 60 percent of humanity, how much has ozone actually changed since the year 2000, and do the world’s chemistry-climate models agree with what the satellites have seen? The answer, it turns out, depends heavily on which dataset you trust, and that dependence has consequences for how confidently we can forecast ozone’s trajectory through the rest of this century.
To disentangle the picture, the team divided the atmosphere into four vertical layers: the troposphere, which holds about 13 percent of the total ozone column; the lower stratosphere, contributing roughly 28 percent; the middle stratosphere, the largest share at about 41 percent; and the upper stratosphere, which supplies the remaining 18 percent. This layering matters because different physical processes dominate each region. In the lower stratosphere, ozone is governed mainly by transport, as the greenhouse-gas-driven acceleration of the Brewer-Dobson circulation pushes air upward through the tropical tropopause and thins the ozone layer there. In the upper stratosphere, chemistry rules: declining chlorine from the Montreal Protocol and cooling temperatures from rising carbon dioxide both push ozone upward, producing robust recovery. The troposphere adds its own complexity, because ozone there is manufactured from precursor emissions such as nitrogen oxides, which are rising across the developing tropics even as they fall over the midlatitudes of the Northern Hemisphere.
The statistical engine of the study was a dynamical linear model, a flexible regression technique that separates the slow background trend from the noise of natural variability, including the El Niño Southern Oscillation, the quasi-biennial oscillation of equatorial winds, the 11-year solar cycle, and stratospheric aerosol from volcanic eruptions. Applying this method to more than a dozen merged satellite datasets and reanalyses, and to two generations of coordinated chemistry-climate model simulations known as CCMI-1 and CCMI-2022, the researchers computed probability distributions of ozone change over the 2000 to 2021 period and measured how much those distributions overlapped.
The headline numbers look encouraging at first glance. Both model generations simulate an increase in tropical total column ozone since 2000, and the newer CCMI-2022 ensemble, with a mean change of +2.5 Dobson units, agrees slightly better with the observational estimate of +3.3 Dobson units than the older CCMI-1 ensemble does at +1.6 Dobson units. But that apparent agreement conceals systematic differences lurking beneath the surface. In the troposphere, the newer models show a substantial increase of about +1.5 Dobson units since 2000, driven largely by stronger rises in precursor emissions, whereas the older generation shows almost no change at +0.3 Dobson units. That single difference explains most of the gap between the two model generations in total column ozone.
More troubling still is the behavior of the observations themselves. In the upper stratosphere, observational estimates of the 2000 to 2021 change range from +0.6 Dobson units in the GOZCARDS dataset to +2.6 Dobson units in the SAGE-OSIRIS-SAGE composite, with as little as 2 percent overlap between the two distributions. In the middle stratosphere, the spread is equally stark, with one dataset showing essentially no change and another showing an increase of 2.5 Dobson units. Even datasets built largely from the same Aura Microwave Limb Sounder instrument, such as SWOOSH and GOZCARDS, diverge, apparently because of differences in how the pre-2004 data from earlier instruments were processed and merged. In the troposphere, the disagreement is arguably the worst of all: one composite shows zero change while another shows an increase of 2.4 Dobson units, with only 1 percent overlap between them.
Why does this matter so much? Because the models exhibit a tantalizing property that, in principle, could turn observations into a predictive tool. The researchers found that across the older CCMI-1 ensemble, models with faster ozone recovery during the first two decades of the twenty-first century also tend to project stronger ozone growth through the rest of the century, a correlation that holds in every layer of the stratosphere and is especially strong in the middle stratosphere. This kind of relationship, known as an emergent constraint, allows scientists to combine a measured present-day trend with the model-derived correlation to narrow the range of possible futures. Before the satellites can play that role, however, two conditions must hold: the correlation must be robust across model generations, and the observational trend estimates must be precise enough to be useful.
Both conditions fail. The emergent relationship that is so clear in the older model ensemble largely disappears in the newer CCMI-2022 simulations, surviving only as a borderline correlation in the middle stratosphere. The reasons remain uncertain, but the authors point to differences in greenhouse gas forcing between the two experimental designs, improved but changed representations of the quasi-biennial oscillation, and documented differences in stratospheric transport between model generations as plausible culprits. Meanwhile, when the researchers propagated the spread of observational trends through the model-derived relationships, the resulting uncertainty in the observationally informed end-of-century ozone projection was as large as, or larger than, the spread of the raw model projections themselves. Depending on which satellite record is treated as truth, one could conclude that the models either systematically overestimate or systematically underestimate future tropical ozone recovery.
Not everything in the analysis is mired in ambiguity. The lower stratosphere, the region where earlier studies first flagged an unexpected decline, shows the most consistent story. Nearly all observational datasets and both model ensembles agree that tropical lower stratospheric ozone has decreased since 2000, with multi-model means of −0.6 Dobson units for CCMI-1 and −1.1 Dobson units for CCMI-2022, and the observations trending somewhat more negative on average than the models. This decline is broadly consistent with the expected greenhouse-gas-driven speeding up of the shallow branch of the Brewer-Dobson circulation, although questions remain about the roles of tropopause height changes and very short-lived halogen compounds. Looking further ahead, the models robustly project continued lower stratospheric ozone decline through 2100, averaging about −5 Dobson units, even as the upper stratosphere gains roughly +6 Dobson units from chemical recovery.
The ultimate message of the study is a call to action for the observational community. The spread among existing satellite trend estimates, particularly in the troposphere and the middle and upper stratosphere, is wide enough to swamp the signal scientists are trying to measure, and reconciling those differences, possibly by hunting for instrumental drifts in the source records, is now a critical priority. Stable, trend-quality measurements from current and future satellite platforms, combined with a better understanding of why successive model generations behave differently, will be essential before observations can serve as the yardstick against which projections of the ozone layer’s future are judged. Until then, the recovery of the ozone layer over the tropics, the region where billions of people and much of Earth’s biodiversity depend on it, remains a story whose next chapter cannot yet be read with confidence.
Subject of Research: Uncertainties in observed tropical stratospheric and tropospheric ozone trends and their implications for constraining model projections of future ozone change
Article Title: Uncertainties in recent tropical stratospheric and tropospheric ozone changes limit the use of observational constraints for assessing future ozone change
Article References: Uncertainties in recent tropical stratospheric and tropospheric ozone changes limit the use of observational constraints for assessing future ozone change. (n.d.). https://doi.org/10.5194/acp-26-13817-2026
Image Credits: AI Generated
DOI: 10.5194/acp-26-13817-2026
Keywords: ozone layer, tropical ozone, stratosphere, troposphere, satellite observations, chemistry-climate models, Montreal Protocol, Brewer-Dobson circulation, ozone recovery, emergent constraint, CCMI, Dobson units
News Source: Bethany Barker. (October 10, 2026). Satellite Records of Tropical Ozone Disagree Too Much to Predict Future Recovery, Study Finds. Scienmag.
