The Pollution–Climate Feedback System

The IPCC measures the push each agent gives to the climate as effective radiative forcing. By 2019, carbon dioxide contributed 2.16 W/m² relative to 1750, methane 0.54, nitrous oxide 0.21, halogenated gases 0.41 and ozone 0.47, with a wide range of 0.24–0.71 for ozone. Against these, aerosols pulled the other way: −0.22 W/m² by scattering and absorbing sunlight and −0.84 W/m² by brightening clouds, the most uncertain term of all, with a 5–95% range of −1.45 to −0.25. The net human-caused forcing was 2.72 W/m², and most of the uncertainty in that total comes from aerosols.

Forcing by agent hides a second layer: what was actually emitted. The IPCC's attribution by emitted substance shows that methane emissions added 0.61°C of warming by 2019, 0.43°C directly, 0.14°C through the ozone they create and a little through stratospheric water vapour. Nitrogen oxides, a classic urban pollutant, warmed by making ozone but cooled more by shortening methane's lifetime, netting −0.15°C. Volatile organic compounds and carbon monoxide did the reverse, lengthening methane's life and making ozone, for +0.22°C. Sulphur dioxide offset 0.43°C; black carbon added 0.09°C, including its darkening of snow.

The heatmap below makes the interactions visible. Each row is a substance people emit; each column is a physical route by which it changes temperature. Off-diagonal cells are the couplings: methane acting through ozone, nitrogen oxides acting through methane, black carbon acting through clouds and snow. They are the reason that air-quality policy is also climate policy, whether intended or not, and the reason that some cuts — methane, black carbon, carbon monoxide and volatile organics — help both at once while others, chiefly sulphur dioxide, trade one benefit against another.

Aerosol forcing grew with industrial coal burning and reached its strongest, about −1.4 to −1.5 W/m², between 1980 and the early 1990s. At that point it cancelled roughly two-thirds of the forcing from greenhouse gases and ozone. Since then, clean-air laws in Europe and North America, followed by China's campaign against coal smoke, have cut world sulphur dioxide emissions from a peak of 141 million tonnes in 1979 to 73 million in 2022. The European Union's emissions fell from 35.5 to 2.1 million tonnes and the United States' from 29.5 to 2.0; China's fell from 38.1 million tonnes in 2006 to 10.4. India's have risen, and since 2021 exceed China's.

The result is visible in the forcing record: aerosols offset 28% of the greenhouse and ozone forcing in 2019, down from 67% in 1980. These cuts saved very large numbers of lives, and the health case for continuing them is overwhelming. But they also removed cooling. The clearest recent example is shipping: in 2020 the International Maritime Organization cut the sulphur allowed in ship fuel from 3.5% to 0.5%. Published estimates of the resulting forcing range from about 0.07 W/m², confined to shipping lanes, to 0.2 W/m² over the global ocean, with several climate models agreeing on roughly 0.1–0.14 W/m² and a warming of about 0.04–0.05°C over the following decades.

The mask matters for the near term because aerosols last days to weeks while carbon dioxide lasts centuries. Once emissions stop, the cooling vanishes almost at once and the underlying warming shows through. The IPCC's assessment is that other human drivers, principally aerosols, cooled the planet by 0.0–0.8°C between 1850–1900 and 2010–2019. The width of that range is one of the main reasons that near-term warming is uncertain. The response is not to keep polluting but to cut, at the same time, the pollutants that warm.

The map below links nine parts of the system. Arrows show the direction of influence; their colour shows its character. Rust arrows add warming, blue dashed arrows are the aerosol mask, magenta arrows are damage to health, and moss and violet arrows run through ecosystems and toxic metals. Each link carries the evidence behind it. The point of the map is that the parts form loops, and loops either amplify a change or damp it.

Four reinforcing loops stand out. The ozone penalty: warming speeds the chemistry that makes ground-level ozone, which is itself a greenhouse gas; global ozone exposure has risen since 1990, attributed to emissions and rising temperatures. The fire loop: hotter, drier conditions fuel wildfires that release carbon and smoke. The ozone–sink loop: ozone damages leaves, cutting crop yields and the land's uptake of carbon dioxide, so that more stays in the air. The permafrost loop: thaw releases carbon and could mobilise the mercury stored in frozen soils.

Two further patterns matter for policy. The mask is a trade-off: cutting sulphate pollution saves lives but removes cooling. The methane–ozone win is the opposite: methane makes ozone, so cutting methane reduces warming, ozone deaths and crop losses together. Many of these loops are slow or uncertain in size — the permafrost-mercury pathway in particular rests on modelled estimates — but their direction is well established, and together they mean that pollution and climate damage compound rather than simply add.

The State of Global Air 2025 report attributes 7.9 million deaths in 2023 to air pollution, about one in eight deaths worldwide and the second-largest risk factor after high blood pressure. Outdoor fine particles accounted for 4.9 million, household air pollution from cooking and heating with solid fuels for 2.8 million, and ozone for 470,000. Ninety per cent of the deaths were in low- and middle-income countries; India and China each had more than two million. The 2025 Lancet Countdown attributes 2.5 million deaths a year to air pollution from burning fossil fuels, a subset of the total that overlaps with the outdoor estimate.

The climate side of the system adds its own burden. Heat-related deaths averaged 546,000 a year in 2012–2021, according to the Lancet Countdown, and smoke from wildfires caused a record 154,000 deaths in 2024. Exposure to ozone is rising: global population-weighted ozone reached 49.8 ppb in 2020, highest in South Asia, and 42% of the world's people live where it exceeds the World Health Organization's least stringent interim target. State of Global Air attributes the rise to emissions and to rising temperatures, the so-called ozone–climate penalty.

Wildfire smoke shows why the loops need careful reading. Global burned area recorded by the Global Wildfire Information System has not risen since 2012; savanna and grassland fires have been declining for decades as farming spreads, a trend Andela and colleagues documented in 2017. Yet smoke deaths have grown: the Finnish Meteorological Institute, which computes the Lancet indicator, finds the increase driven mainly by more people living where smoke reaches them, while fire activity has intensified in particular regions such as western Canada and the United States. Heat, smoke and ozone all land hardest on people who are already exposed and least able to protect themselves.

Ground-level ozone enters leaves through their pores and damages the cells that photosynthesise. Studies summarised in the State of Global Air 2025 attribute yield losses of about 7.2% for wheat, 6.7% for soybean, 3.6% for maize and 2.6% for rice to ambient ozone, and project a loss of 16.8 million tonnes of European wheat in 2050 from rising ozone. The same damage reduces the land's uptake of carbon dioxide. Sitch and colleagues showed in 2007 that the indirect forcing from ozone's effect on plants could exceed the direct forcing of the ozone itself, closing a loop from air pollution back to greenhouse warming.

Methane links ecosystems and climate in both directions. Agriculture, waste and fossil-fuel extraction are the main human sources, but wetlands are the largest natural one, and they emit more when warm and wet. Atmospheric methane reached 1,936 ppb in 2025, 2.65 times its pre-industrial level. Its annual growth surged to records of 12.2, 16.1 and 16.1 ppb in 2020, 2021 and 2022, against an average of 0.4 ppb a year in 2000–2006, before easing to 6.6 ppb in 2025. Scientists are still debating the causes; rising wetland emissions in the tropics and a temporary fall in the atmosphere's capacity to destroy methane are leading explanations.

Frozen soils add a toxic dimension. Schuster and colleagues estimated in 2018 that soils in the northern permafrost region hold about 1,656 thousand tonnes of mercury, nearly twice as much as all other soils, the ocean and the atmosphere combined, based on cores from Alaska scaled up with soil-carbon maps. For comparison, human activity emitted about 2,220 tonnes of mercury to the air in 2015, 38% of it from artisanal and small-scale gold mining. How much permafrost mercury thaw will release, and how much will reach food webs as toxic methylmercury, remains uncertain, but the store is several hundred times annual human emissions.

The IPCC's emulator splits future warming into its drivers. In the very low emissions scenario, SSP1-1.9, the aerosol contribution shrinks from about −0.44°C in 2020 to −0.17°C in 2050 as the air is cleaned, adding about 0.27°C. Falling methane and other greenhouse gases offset part of it. In the high scenario, SSP3-7.0, aerosol cooling stays roughly constant because pollution control is weak, while methane's warming grows. The IPCC's summary judgement is that, across its scenarios, the combined changes in methane, aerosols and ozone precursors lead to net warming in the near and long term, and that strong, rapid methane cuts would limit the warming from declining aerosol pollution and improve air quality.

A pair of experimental scenarios isolates the choice. Both start from SSP3-7.0 and apply strong air-quality controls; one leaves methane unchanged, the other also cuts methane. Relative to 2020, warming from short-lived forcers by 2050 is about 0.18°C in SSP3-7.0 itself, 0.26°C with clean air alone — because the aerosol mask is removed while methane keeps rising — and 0.02°C when methane is cut too. The difference between the two clean-air variants, about 0.24°C by 2050, is the value of pairing the policies.

These are median results from a simple climate emulator and carry wide ranges, chiefly because aerosol forcing is uncertain: in SSP1-1.9 the 5–95% range of the aerosol contribution by 2050 runs from 0.08°C to 0.52°C. They also omit the health side. The clean-air-only variant still prevents a large number of deaths; the point is that it should not travel alone. In the long run, carbon dioxide dominates every scenario, which is why the shared source — fossil combustion — has to shrink regardless.

Seen as one system, the main policy levers sort into three groups. Double wins cut warming and toxic exposure together: reducing methane from oil and gas, coal mines, landfills and rice; reducing black carbon from diesel engines, brick kilns and household stoves; and reducing carbon monoxide and volatile organics. The Global Methane Assessment estimated that a 45% cut in human-caused methane this decade would avoid nearly 0.3°C of warming by the 2040s, 255,000 premature deaths, 775,000 asthma-related hospital visits and 26 million tonnes of crop losses a year. Phasing down HFCs under the Kigali Amendment is a climate-only win, estimated by UNEP to avoid up to 0.4°C by 2100.

Trade-offs are concentrated in sulphur. Scrubbing sulphur from coal plants and ship fuel saves lives but removes cooling, as the shipping rule showed. The right response is to pair them with double wins, not to delay them. Cutting nitrogen oxides is mixed for climate — it reduces ozone but lets methane live longer — and clearly positive for health and crops. Root fixes shrink the shared source: replacing fossil combustion with clean electricity removes carbon dioxide, sulphur, nitrogen oxides, particles and mercury at once. Its near-term climate effect is muted by unmasking, but it is the only lever that ends the problem.

The calculator lets readers mix the IPCC's component pathways — aerosols, methane, ozone and HFCs — and see the short-lived-forcer warming to 2100. It shows the asymmetry directly: choosing strong aerosol controls alone pushes the line up; adding strong methane and ozone pathways brings it down. Because the IPCC emulated each component separately, the combinations are approximate. The matrix beneath summarises the direction of each lever across the system, drawing on the evidence in this report.

Forcing by agent (effective radiative forcing, 1750–2019, best estimate with 5–95% range) and the attribution of 1750–2019 warming to emitted substances and forcing routes are taken from the public data repository of IPCC AR6 Working Group I Chapter 7, which produced Figures 6.12 and 7.8 and Figure SPM.4. The attribution uses a two-layer emulator constrained to the assessed forcing and climate sensitivity; values are medians. The "VOCs + CO" row combines non-methane volatile organic compounds and carbon monoxide. Future contributions by forcer (relative to 1850–1900) are the median emulator results for SPM.4; short-lived-forcer contributions relative to 2020, including the SSP3-7.0 low near-term-climate-forcer variants, come from the Chapter 6 output.

Sulphur dioxide is from the Community Emissions Data System via Our World in Data; the European Union is the sum of its 27 members, and the rest of world includes international shipping and unallocated emissions. Methane is NOAA's global annual mean (2025 preliminary); the 1750 value of 729 ppb is from AR6 Table 7.5. Burned area is from the Global Wildfire Information System via Our World in Data, 2012–2024. Health numbers come from State of Global Air 2025 (GBD 2023), the 2025 Lancet Countdown, and the Global Methane Assessment; crop losses from Tai et al. (2021) as reported in State of Global Air; shipping forcing from four peer-reviewed studies; mercury from Schuster et al. (2018) and UNEP's Global Mercury Assessment 2018.

The feedback map is qualitative: arrows show established directions of influence, and the numbers attached are the best available estimates rather than loop gains. Death estimates come from different years and methods and should not be summed; the fossil-fuel estimate overlaps with the outdoor-particle estimate. The calculator adds median component contributions from different scenarios and is approximate. Related reports in this series: S-curves against the clock, the interdependencies of twenty-first-century risk and human exposure across the polycrisis century.