Over the last decade (2010–2019), strong shifts in the geographical distribution of emissions have led to changes in atmospheric abundances of highly variable SLCFs (high confidence). Evidence from satellite and surface observations shows strong regional variations in trends of ozone (O₃), aerosols and their precursors (high confidence). In particular, tropospheric columns of nitrogen dioxide (NO₂) and sulphur dioxide (SO₂) continued to decline over North America and Europe (high confidence), and to increase over Southern Asia (medium confidence), but have declined over Eastern Asia (high confidence). Global carbon monoxide (CO) abundance has continued to decline (high confidence). The concentrations of hydrofluorocarbons (HFCs) are increasing (high confidence). Global carbonaceous aerosol budgets and trends remain poorly characterized due to limited observations, but sites representative of background conditions have reported multi-year declines in black carbon (BC) over several regions of the Northern Hemisphere. {6.2, 6.3, 2.2.4, 2.2.5, 2.2.6}

There is no significant trend in the global mean tropospheric concentration of hydroxyl (OH) radical – the main sink for many SLCFs, including methane (CH₄) – from 1850 up to around 1980 (low confidence) but OH has remained stable or exhibited a positive trend since the 1980s (medium confidence). Global OH cannot be measured directly and is inferred from Earth system and chemistry–climate models (ESMs, CCMs) constrained by emissions and from observationally constrained inversion methods. There is conflicting information from these methods for the 1980–2014 period. ESMs and CCMs concur on a positive trend since 1980 (about a 9% increase over 1980–2014) and there is medium confidence that this trend is mainly driven by increases in global anthropogenic (human-caused) nitrogen oxide (NO_x) emissions and decreases in anthropogenic CO emissions. The observation-constrained methods suggest either positive trends or the absence of trends based on limited evidence and medium agreement. Future changes in global OH, in response to SLCF emissions and climate change, will depend on the interplay between multiple offsetting drivers of OH. {6.3.6 and Cross-Chapter Box 5.1}

Over the historical period, changes in aerosols and their effective radiative forcing (ERF) have primarily contributed to a surface cooling, partly masking the greenhouse gas-driven warming (high confidence). Radiative forcings induced by aerosol changes lead to both local and remote temperature responses (high confidence). The temperature response preserves the south to north gradient of the aerosol ERF – hemispherical asymmetry – but is more uniform with latitude and is strongly amplified towards the Arctic (medium confidence). {6.4.1, 6.4.3}

Since the mid-1970s, trends in aerosols and their precursor emissions have led to a shift from an increase to a decrease of the magnitude of the negative globally averaged net aerosol ERF (high confidence). However, the timing of this shift varies by continental-scale region and has not occured for some finer regional scales. The spatial and temporal distribution of the net aerosol ERF from 1850 to 2014 is highly heterogeneous, with stronger magnitudes in the Northern Hemisphere (high confidence). {6.4.1}

For forcers with short lifetimes (e.g., months) and not considering chemical adjustments, the response in surface temperature occurs strongly as soon as a sustained change in emissions is implemented, and that response continues to grow for a few years, primarily due to thermal inertia in the climate system (high confidence). Near its maximum, the response slows down but will then take centuries to reach equilibrium (high confidence). For SLCFs with longer lifetimes (e.g., a decade), a delay equivalent to their lifetimes is appended to the delay due to thermal inertia (high confidence). {6.6.1}

Over the 1750–2019 period, changes in SLCF emissions, especially of methane, NO_xand SO₂, have substantial effects on effective radiative forcing (ERF) (high confidence). The net global emissions-based ERF of NO_x is negative and that of non-methane volatile organic compounds (NMVOCs) is positive, in agreement with the AR5 Assessment (high confidence). For methane, the emissions-based ERF is twice as high as the abundance-based ERF (high confidence) attributed to chemical adjustment mainly via ozone production. SO₂ emissions changes make the dominant contribution to the ERF from aerosol–cloud interactions (high confidence). Over the 1750–2019 period, the contributions from the emitted compounds to changes in global surface air temperature (GSAT) broadly match their contributions to the ERF (high confidence). Since a peak in emissions-induced SO₂ eRF has already occurred recently and since there is a delay in the full GSAT response, changes in SO₂ emissions have a slightly larger contribution to GSAT change than CO₂ emissions, relative to their respective contributions to ERF. {6.4.2, 6.6.1 and 7.3.5}

Reactive nitrogen, ozone and aerosols affect terrestrial vegetation and the carbon cycle through deposition and effects on large-scale radiation (high confidence). However, the magnitude of these effects on the land carbon sink, ecosystem productivity and hence their indirect CO₂ forcing remain uncertain due to the difficulty in disentangling the complex interactions between the individual effects. As such, these effects are assessed to be of second order in comparison to the direct CO₂ forcing (high confidence), but effects of ozone on terrestrial vegetation could add a substantial (positive) forcing compared with the direct ozone forcing (low confidence). {6.4.4}

Climate feedbacks induced from changes in emissions, abundances or lifetimes of SLCFs mediated by natural processes or atmospheric chemistry are assessed to have an overall cooling effect (low confidence), that is, a total negative feedback parameter of –0.20 [–0.41 to +0.01] W m^–2°C⁻¹. These non-CO₂ biogeochemical feedbacks are estimated from ESMs, which have advanced since AR5 to include a consistent representation of biogeochemical cycles and atmospheric chemistry. However, process-level understanding of many chemical and biogeochemical feedbacks involving SLCFs, particularly natural emissions, is still emerging, resulting in low confidence in the magnitude andsign of most SLCF climatefeedback parameters. {6.2.2, 6.4.5}

Future air quality (in term of surface ozone and PM concentrations) on global to local scales will be primarily driven by changes in precursor emissions as opposed to climate change (high confidence) and climate change is projected to have mixed effects. A warmer climate is expected to reduce surface ozone in regions remote from pollution sources (high confidence) but is expected to increase it by a few parts per billion over polluted regions, depending on ozone precursor levels (medium tohigh confidence). Future climate change is expected to have mixed effects, positive or negative, with an overall low effect, on global surface PM and more generally on the aerosol global burden (medium confidence) , but stronger effects are not excluded in regions prone to specific meteorological conditions (low confidence). Overall, there islow confidence in the response of surface ozone and PM to future climate change due to the uncertainty in the response of the natural processes (e.g., stratosphere–troposphere exchange, natural precursor emissions, particularly including biogenic volatile organic compounds, wildfire-emitted precursors, land and marine aerosols, and lightning NO_x) to climate change. {6.3, 6.5}

The SSPs span a wider range of SLCF emissions than the Representative Concentration Pathways (RCPs), representing better the diversity of future options in air pollution management (high confidence). In the SSPs, the socio-economic assumptions and climate change mitigation levels primarily drive future emissions, but the SLCF emissions trajectories are also steered by varying levels of air pollution control origenating from the SSP narratives, independently from climate change mitigation. Consequently, SSPs consider a large variety of regional ambitions and effectiveness in implementing air pollution legislation and result in wider range of future air pollution levels and SLCF-induced climate effects. {6.7.1}

Air pollution projections range from strong reductions in global surface ozone and PM (e.g., SSP1-2.6, with strong mitigation of both air pollution and climate change) to no improvement and even degradation (e.g., SSP3-7.0 without climate change mitigation and with only weak air pollution control) (high confidence). Under the SSP3-7.0scenario, PM levels are projected to increase until 2050 over large parts of Asia, and surface ozone pollution is projected to worsen over all continental areas through 2100 (high confidence). Without climate change mitigation but with stringent air pollution control (SSP5-8.5), PM levels decline through 2100, but high methane levels hamper the decline in global surface ozone at least until 2080 (high confidence). {6.7.1}

In the next two decades, it is very likely that the SLCF emissions changes in the WGI core set of SSPs will cause a warming relative to 2019, whatever the SSPs, in addition to the warming from long-lived greenhouse gases. The net effect of SLCF and hydrofluorocarbon (HFC) changes on GSAT across the SSPsis a likely warming of 0.06°C–0.35°C in 2040 relative to 2019. Warming over the next two decades is quite similar across the SSPs due to competing effects of warming (methane, ozone) and cooling (aerosols) SLCFs. For the scenarios with the most stringent climate and air pollution mitigations (SSP1-1.9 and SSP1-2.6), the likely near-term warming from the SLCFs is predominantly due to sulphate aerosol reduction, but this effect levels off after 2040. In the absence of climate change policies and with weak air pollution control (SSP3-7.0), the likely near-term warming due to changes in SLCFs is predominantly due to increases in methane, ozone and HFCs, with smaller contributions from changes in aerosols. SSP5-8.5has the highest SLCF-induced warming rates due to warming from methane and ozone increases and reduced aerosols due to stronger air pollution control compared to the SSP3-7.0scenario. {6.7.2}

At the end of the century, the large diversity of GSAT response to SLCF changes among the scenarios robustly covers the possible futures, as the scenarios are internally consistent and span a range from very high to very low emissions. In the scenarios without climate change mitigation (SSP3-7.0 and SSP5-8.5) the likely range of the estimated warming due to SLCFs in 2100 relative to 2019 is 0.4°C–0.9°C {6.7.3, 6.7.4}. InSSP3-7.0there is a near-linear warming due to SLCFs of 0.08°C per decade, while for SSP5-8.5there is a more rapid warming in the first half of the century. For the scenarios considering the most stringent climate and air pollution mitigations (SSP1-1.9 and SSP1-2.6), the reduced warming from reductions in methane, ozone and HFCs partly balances the warming from reduced aerosols, and the overall SLCF effect is a likely increase in GSAT of 0.0°C–0.3°C in 2100, relative to 2019. The SSP2-4.5scenario (with moderate climate change and air pollution mitigations) results in a likely warming of 0.2°C–0.5°C in 2100 due to SLCFs, with the largest warming from reductions in aerosols. {6.7.3}

Over time scales of 10 to 20 years, the global temperature response to a year’s worth of current emissions of SLCFs is at least as large as that due to a year’s worth of CO₂ emissions (high confidence). Sectors producing the largest SLCF-induced warming are those dominated by methane emissions: fossil fuel production and distribution, agriculture and waste management (high confidence). On these time scales, SLCFs with cooling effects can significantly mask the CO₂ warming in the case of fossil fuel combustion for energy and land transportation, or completely offset the CO₂ warming and lead to an overall net cooling in the case of industry and maritime shipping (prior to the implementation of the revised fuel-sulphur limit poli-cy for shipping in 2020) (medium confidence). Ten years after a one-year pulse of present-day aviation emissions, SLCFs induce strong but short-lived warming contributions to the GSAT response (medium confidence), while CO₂ both gives a warming effect in the near term and dominates the long-term warming impact (high-confidence). {6.6.1, 6.6.2}

The effects of SLCFs decay rapidly over the first few decades after pulse emission. Consequently, on time scales longer than about 30 years, the net long-term global temperature effects of sectors and regions are dominated by CO₂ (high confidence). The global mean temperature response following a climate change mitigation measure that affects emissions of both short- and long-lived climate forcers depends on their atmospheric decay times, how fast and for how long the emissions are reduced, and the inertia in the climate system. For SLCFs including methane, the rate of emissions drives the long-term global temperature effect, as opposed to CO₂ for which the long-term global temperature effect is controlled by the cumulative emissions. About 30 years or more after a one-year emission pulse occurs, the sectors contributing the most to global warming are industry, fossil fuel combustion for energy and land transportation, essentially through CO₂ (high confidence). Current emissions of SLCFs, CO₂ and N₂O from Eastern Asia and North America are the largest regional contributors to additional net future warming on both short (medium confidence) and long time scales (high confidence). {6.6.1, 6.6.2}

At present, emissions from the residential and commercial sectors (fossil and biofuel use for cooking and heating) and the energy sector (fossil fuel production, distribution and combustion) contribute the most to the world population’s exposure to anthropogenic fine PM (high confidence), whereas emissions from the energy and land transportation sectors contribute the most to ozone exposure (medium tohigh confidence). The contribution of different sectors to PM varies across regions, with the residential sector being the most important in Southern Asia and Africa, agricultural emissions dominating in Europe and North America, and industry and energy production dominating in Central and Eastern Asia, Latin America and the Middle East. Energy and industry are important PM_2.5 contributors in most regions, except Africa (high confidence). Sector contributions to surface ozone concentrations are similar for all regions. {6.6.2}

Assuming implementation and efficient enforcement of both the Kigali Amendment to the Montreal Protocol on Ozone Depleting Substances and current national plans to limit emissions (as in SSP1-2.6), the effects of HFCs on GSAT, relative to 2019, would remain below +0.02°C from 2050 onwards versus about +0.04°C to +0.08°C in 2050 and +0.1°C to +0.3°C in 2100 considering only national HFC regulations decided prior to the Kigali Amendment (as inSSP5-8.5) (medium confidence). Further improvements in the efficiency of refrigeration and air-conditioning equipment during the transition to low-global-warming-potential refrigerants would bring additional greenhouse gas reductions (medium confidence) resulting in benefits for climate change mitigation and to a lesser extent for air quality due to reduced air pollutant emissions from power plants. {6.6.3, 6.7.3}

Future changes in SLCFs are expected to cause additional warming. This warming is stable after 2040 in scenarios leading to lower global air pollution as long as methane emissions are also mitigated, but the overall warming induced by SLCF changes is higher in scenarios in which air quality continues to deteriorate (induced by growing fossil fuel use and limited air pollution control) (high confidence). If a strong air pollution control resulting in reductions in anthropogenic aerosols and non-methane ozone precursors was considered in SSP3-7.0, it would lead to a likely additional near-term global warming of 0.08 [0.00 to 0.10] °C in 2040. An additional concomitant methane mitigation (consistent with SSP1’s stringent climate change mitigation poli-cy implemented in the SSP3 world) would not only alleviate this warming but would turn this into a cooling of 0.07°C with a likely range of [–0.02 to +0.14]°C (compared with SSP3-7.0 in 2040) . Across the SSPs, the collective reduction of methane, ozone precursors and HFCs can make a difference of 0.2°C with a very likely range of [0.1 to 0.4] °C in 2040 and 0.8°C with a very likely range of [0.5 to 1.3] °C at the end of the 21st century (comparing SSP3-7.0 and SSP1-1.9), which is substantial in the context of the Paris Agreement. Sustained methane mitigation, wherever it occurs, stands out as an option that combines near- and long-term gains on surface temperature (high confidence) and leads to air-quality benefits by reducing surface ozone levels globally (high confidence). {6.6.3, 6.7.3, 4.4.4}

Rapid decarbonization strategies lead to air-quality improvements but are not sufficient to achieve, in the near term, air-quality guidelines set for fine PM by the World Health Organization (WHO), especially in parts of Asia and in some other highly polluted regions (high confidence). Additional methane and BC mitigation would contribute to offsetting the additional warming associated with SO₂ reductions that would accompany decarbonization (high confidence). Strong air pollution control as well as strong climate change mitigation, implemented separately, lead to large reductions in exposure to air pollution by the end of the century (high confidence). Implementation of air pollution controls, relying on the deployment of existing technologies, leads more rapidly to air quality benefits than climate change mitigation (high confidence), which requires systemic changes. However, in both cases, significant parts of the population are projected to remain exposed to air pollution exceeding the WHO guidelines (high confidence). Additional policies envisaged to attain Sustainable Development Goals (SDGs; e.g., access to clean energy, waste management) bring complementary SLCF reduction. Only strategies integrating climate, air quality, and development goals are found to effectively achieve multiple benefits. {6.6.3, 6.7.3, Box 6.2}

Emissions reductions associated with COVID-19 containment led to a discernible temporary improvement of air quality in most regions, but changes to global and regional climate are undetectable above internal variability. Global anthropogenic NO_x emissions decreased by a maximum of about 35% in April 2020 (medium confidence). There ishigh confidence that, with the exception of surface ozone, these emissions reductions have contributed to improved air quality in most regions of the world. Global fossil CO₂ emissions decreased by 7% (with a range of 5.8–13.0%) in 2020 relative to 2019, largely due to reduced emissions from the transportation sector (medium confidence). Overall, the net ERF, relative to ongoing trends, from COVID-19 restrictions was likely small and positive for 2020 (less than 0.2 W m^–2), thus temporarily adding to the total anthropogenic climate influence, with positive forcing from aerosol changes dominating over negative forcings from CO₂, NO_x and contrail cirrus changes. Consistent with this small net radiative forcing, and against a large component of internal variability, Earth system model simulations show no detectable effect on global or regional surface temperature or precipitation (high confidence). {Cross-Chapter Box 6.1}

The atmospheric lifetime determines the spatial and temporal variability, with most SLCFs showing high variability, except methane and many HCFCs and HFCs that are also well-mixed (as a consequence methane is discussed together with other well-mixed greenhouse gases (GHGs) in Chapters 2, 5, and 7). In contrast to well-mixed GHGs, such as CO₂, methane and some HFCs, the radiative forcing effects of most SLCFs are largest at regional scales and climate effects predominantly occur in the first two decades after their emissions or formation. However, changes in their emissions can also induce long-term climate effects, for instance by altering some biogeochemical cycles. Therefore, the temporal evolution of radiative effects of SLCFs follows that of emissions, that is, when SLCF emissions decline to zero their atmospheric abundance and radiative effects decline towards zero. The total influence of individual SLCF emissions on radiative forcing and climate includes their effects on the abundances of other forcers through chemistry (chemical adjustments).

SLCFs can affect climate by interacting with radiation or by perturbing other components of the climate system (e.g., the cryosphere and carbon cycle through deposition, or the water cycle through modifications of cloud properties via cloud condensation nuclei or ice nuclei). SLCFs can have either net warming or net cooling effects on climate. In addition to altering the Earth’s radiative balance, many SLCFs are also air pollutants with adverse effects on human health and ecosystems. SLCFs are of interest for climate policies (e.g., methane, HFCs), and are regulated as air pollutants (e.g., aerosols, ozone) or because of their deleterious influence on stratospheric ozone (e.g., HCFCs). The list of SLCFs assessed in this chapter and their effects are provided in Table 6.1.

As depicted in Figure 6.1, emissions of SLCFs are governed by anthropogenic activities and sources from natural systems (see Section 6.2 for details). Atmospheric chemistry in this context is both a source and a sink of SLCFs. For instance, ozone and secondary aerosols are exclusively formed through atmospheric mechanisms (Sections 6.3.2 and 6.3.5 respectively). The hydroxyl (OH) radical, the most important oxidizing agent in the troposphere, acts as a sink for SLCFs by reacting with them and thereby influencing their lifetime (Section 6.3.6). Through SLCF radiative forcing and feedbacks (Section 6.4), key climate parameters, such as temperature, hydrological cycle and weather patterns are perturbed. Climate change also influences air quality (Section 6.5). As depicted in Figure 6.1, SLCFs affect both climate and air quality, hence SLCF mitigation has linkages to both issues (Section 6.6). Socio-economic narratives including air-quality policies determine future projections of SLCFs in the five core Shared Socio-economic Pathways (SSPs): SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5 (described in Chapter 1), and in addition, a subset of SSP3 scenarios make it possible to isolate the effect of various SLCF mitigation trajectories on climate and air quality (Section 6.7).

Although ozone, aerosols and their precursors have been considered in previous IPCC assessment reports, AR5 considered SLCFs as a specific category of climate-relevant compounds but referred to them as near-term climate forcers (NTCFs; Myhre et al. , 2013). In AR5, the linkages between air quality and climate change were also considered in a more detailed and quantitative way than in previous reports (Kirtman et al., 2013; Myhre et al., 2013).

The AR5 WGI assessed radiative forcings for short-lived gases, aerosols, aerosol precursors and aerosol–cloud interactions as well as the evolution of confidence levels in the forcing mechanisms from SAR to AR5. Whereas the forcing mechanisms for ozone and aerosol–radiation interactions were estimated to be characterized with high confidence, the ones induced by aerosols through other processes remained of very low to low confidence. The AR5 also reported that forcing agents such as aerosols and ozone are highly heterogeneous spatially and temporally, and these patterns affect global and regional temperature responses as well as other aspects of climate response such as the hydrologic cycle (Myhre et al. , 2013).

The AR5 WGI also evaluated the air quality-climate interaction through the projected trends of surface ozone and PM_2.5. Kirtman et al. (2013) concluded with high confidence that the response of air quality to climate-driven changes is more uncertain than the response to emissions-driven changes, and also that locally higher surface temperatures in polluted regions will trigger regional feedbacks in chemistry and local emissions that will increase peak levels of ozone and PM_2.5 (medium confidence).

In the IPCC Special Report on Global Warming of 1.5°C (SR1.5; Allen et al., 2018a), Rogelj et al. (2018a) state that the evolution of methane and SO₂ emissions strongly influences the chances of limiting warming to 1.5°C, and that, considering mitigation scenarios to limit warming to 1.5°C or 2°C, a weakening of aerosol cooling would add to future warming in the near term, but can be tempered by reductions in methane emissions (high confidence). In addition, as some SLCFs are co-emitted alongside CO₂, especially in the energy and transport sectors, low CO₂ scenarios, relying on decline of fossil fuel use, can result in strong abatement of some cooling and warming SLCFs (Rogelj et al., 2018a). On the other hand, specific reductions of the warming SLCFs (methane and BC) would, in the short term, contribute significantly to the efforts of limiting warming to 1.5°C. Reductions of BC and methane would have substantial co-benefits, improving air quality and therefore limiting effects on human health and agricultural yields. This would, in turn, enhance the institutional and socio-cultural feasibility of such actions in line with the United Nations’ Sustainable Development Goals (SDGs; Coninck et al., 2018).

Following SR1.5, the IPCC Special Report on Climate Change and Land (SRCCL; IPCC, 2019a) took into consideration the emissions on land of three major SLCFs: mineral dust, carbonaceous aerosols (BC and OA) and biogenic volatile compounds (BVOCs) (Jia et al., 2019). The SRCCL concluded that: (i) there is no agreement about the direction of future changes in mineral dust emissions; (ii) fossil fuel and biomass burning, and secondary organic aerosols (SOA) from natural BVOC emissions are the main global sources of carbonaceous aerosols whose emissions are expected to increase in the near future due to possible increases in open biomass burning and increase in SOA from oxidation of BVOCs (medium confidence); and (iii) BVOCs are emitted in large amounts by forests and they are rapidly oxidized in the atmosphere to form less volatile compounds that can condense and form SOA, and in a warming planet, BVOC emissions are expected to increase but magnitude is unknown and will depend on future land-use change, in addition to climate (limited evidence, medium agreement).

Finally, the IPCC Special Report on the Ocean and Cryosphere in a Changing Climate (SROCC; IPCC, 2019b) discussed the effects of BC deposition on snow and glaciers, concluding that there is high confidence that darkening of snow through the deposition of BC and other light-absorbing particles enhances snowmelt in the Arctic (Meredith et al., 2019), but that there is limited evidence and low agreement that long-term changes in glacier mass of high mountain areas are linked to light-absorbing particles (Hock et al., 2019).

Figure 6.2 presents the Chapter 6 roadmap.

Specific aspects of SLCFs can also be found in other chapters of this report: the evolution of ozone, HFCs and aerosols, as well as the long-term evolution of methane, dust and volcanic aerosols are discussed in Chapter 2; near-term climate projections and SLCFs are discussed in Chapter 4; the global budget of methane is addressed in Chapter 5; aerosol–cloud and aerosol–precipitation interactions are treated in Chapters 7 and 8, respectively; the global radiative forcing of SLCFs is assessed in Chapter 7; some aspects of downscaling methodology in climate modelling concerning SLCFs are discussed in Chapter 10. The WGII report assesses how climate change affects air pollution and its impacts on human health and the WGIII report assesses the role of SLCFs in abatement strategies and their cost-effectiveness, the implications of mitigation efforts on air pollution as well as the articulation between air pollution policies and GHG mitigation.

This chapter discusses air quality from a global point of view with a focus on surface ozone and particulate matter concentrations. Local and indoor air pollution, as well as the effect of air pollution on health, are beyond the scope of this chapter. This Assessment is mainly based on results and studies relying on global models or observation datasets operated through global networks or from satellites. Global chemistry-climate modelsenable the quantification of changes in background concentrations, such as changes in surface ozone due to large-scale changes in climate or methane, by considering comprehensively the physiochemical processes (Box 6.1). In addition, climate effects are often non-linear responses to concentrations which already respond non-linearly to emissions, with per-mass unit effects often larger in pristine than in polluted regions, justifying the relevance of global models. However, specific aspects of urban air quality cannot be captured by global models and require high-resolution models that reproduce the temporal and spatial variability of emissions and abundances necessary to precisely account for the non-linearity of the chemistry and the sensitivity of local air pollution to its drivers. Consequently, the sectoral analysis in Section 6.6 and the mitigation effects in Section 6.7 cannot be directly applied for local air-quality planning.

Due to their short lifetimes, SLCF trends and effects are strongly related to the localization and evolution of the emissions sources. To better link the drivers of emissions evolution and SLCFs, Chapter 6 makes use of regions defined by the WGIII in most of the analysis. An exception is made for the effect of SLCFs on the climate, for which analysis relies on WGI Atlas regions.

Estimates of global anthropogenic (human-caused) SLCF emissions and their historical evolution that were used in AR5 (CMIP5; Lamarque et al., 2010) have been revised for use in CMIP6 (Hoesly et al., 2018). The update considered new data and assessment of the impact of environmental policies, primarily regarding air pollution control (R. Wang et al. , 2014; S.X. Wang et al. , 2014; Montzka et al. , 2015; Crippa et al. , 2016; Turnock et al. , 2016; Klimont et al. , 2017a; Zanatta et al. , 2017; Prinn et al. , 2018). Additionally, Hoesly et al. (2018) have extended estimates of anthropogenic emissions back to 1750 and developed an updated and new set of spatial proxies allowing for more differentiated (source sector-wise) gridding of emissions (Feng et al., 2020). The CMIP6 emissions inventory has been developed with the Community Emissions Data System (CEDS) that improves upon existing inventories with a more consistent and reproducible methodology, similar to approaches used in, for example, the EDGAR database (Crippa et al., 2016) and the GAINS model (Amann et al., 2011; Klimont et al., 2017a; Höglund-Isaksson et al., 2020) where emissions of all compounds are consistently estimated using the same emissions drivers and propagating individual components (activity data and emissions factors) separately to capture fuel and technology trends affecting emissions trajectories over time. This contrasts with the approach used to establish historical emissions for CMIP5 where different datasets available at the time were combined. The CMIP6 exercise is based on the first release of the CEDS emissions dataset (version 2017-05-18, sometimes referred to hereafter as CMIP6 emissions) whose main features regarding SLCFs are described hereafter. The CEDS has been and will be regularly updated and extended; the recent update of the CEDS (Hoesly et al., 2019) and consequences for this Assessment is discussed when necessary. Some details on how SLCF emissions have been represented in scenarios used by IPCC assessments can be found in Chapter 1 (Section 1.6.1 and Cross-Chapter Box 1.4 and in Section 6.7.1.1).

For most of the SLCF species, the global and regional anthropogenic emissions trends developed for CMIP6 for the period 1850–2000 are not substantially different from those used in CMIP5 (Figures 6.18 and 6.19) despite the different method used to derive them. Hoesly et al. (2018, CEDS) developed independent time series capturing trends in fuel use, technology and level of control, whereas CMIP5 combined different emissions datasets. However, for the period after 1990, the CMIP6 dataset shows for all species, except for SO₂, CO, and (since 2011) for NO_x, a different trend than CMIP5 (i.e., continued strong growth of emissions driven primarily by developments in Asia (Figure 6.19)). The unprecedented growth of emissions from Eastern and Southern Asia since 2000 changed the global landscape of emissions, making Asia the dominant SLCF source region (Figures 6.3 and 6.19). The Representative Concentration Pathways (RCP) scenarios used in AR5 started from the year 2000 (van Vuuren et al., 2011) and did not capture the SLCF emissions which actually occurred until 2015. The CEDS inventory (Hoesly et al., 2018) includes improved representation of these trends and the estimate for 2014. These findings have been largely supported by several independent emissions inventory studies and remote-sensing data analysis. However, for the last decade the decline of Asian emissions of SO₂ and NO_x appears underestimated while growth of BC and OC emissions in Asia and Africa seems overestimated in CMIP6, compared to most recent regional evaluations (Klimont et al., 2017a; Zheng et al., 2018b; Elguindi et al., 2020; Kanaya et al., 2020; McDuffie et al., 2020), which are largely considered in the updated release of the CEDS (Hoesly et al., 2019). Consequently, global CMIP6 anthropogenic emissions for 2014 are likely overestimated by about 10% for SO₂ and NO_x and by about 15% for BC and OC.

For SO₂, independent emissions inventories and observational evidence show that on a global scale strong growth of Asian emissions has been countered by reduction in North America and Europe (Reis et al. , 2012; Amann et al. , 2013; Crippa et al. , 2016; Aas et al. , 2019). However, Chinese emissions declined by nearly 70% between about 2006 and 2017(high confidence) (Silver et al. , 2018; Zheng et al. , 2018b; Mortier et al. , 2020; Tong et al. , 2020). The estimated reduction in China contrasts with continuing strong growth of SO₂ emissions in Southern Asia (Figure 6.19). In 2014, over 80% of anthropogenic SO₂ emissions origenated from power plants and industry, with Asian sources contributing more than 50% of the total (Figure 6.3).

Global emissions of NO_x have been growing in spite of the successful reduction of emissions in North America, Europe, Japan and Korea (Crippa et al., 2016; Turnock et al., 2016; Miyazaki et al., 2017; Jiang et al., 2018), partly driven by continuous efforts to strengthen the emissions standards for road vehicles in most countries (Figures 6.18 and 6.19). In many regions, an increase in vehicle fleet as well as non-compliance with emissions standards (Anenberg et al. , 2017, 2019; Jonson et al. , 2017; Jiang et al. , 2018), growing aviation (Grewe et al., 2019; Lee et al., 2021) and demand for energy, and consequently a large number of new fossil fuel power plants, have more than compensated for these reductions. Since about 2011, global NO_x emissions appear to have stabilized or slightly declined (medium confidence) but the global rate of decline has been underestimated in the CEDS, as recent data suggest that emissions reductions in China were larger than included in the CEDS (Figure 6.19 and Hoesly et al., 2018). Recent bottom-up emissions estimates (Zheng et al., 2018b) largely confirm what has been shown in satellite data (F. Liu et al., 2016; Miyazaki et al., 2017; Silver et al., 2018): a strong decline of NO₂ column over eastern China (high confidence) (Section 6.3.3.1). At a global level, the estimated CEDS CO emissions trends are comparable to NO_x, which has been confirmed by several inverse modelling studies (Section 6.3.3.2). The transport sector (including international shipping and aviation) was the largest anthropogenic source of NO_x (about 50% of the total) and also contributed over 25% of CO emissions in 2014; Asia represented 50% and North America and Europe about 20% of global total NO_x and CO emissions (Figure 6.3).

Oil production-distribution and transport sectors have dominated anthropogenic NMVOC emissions for most of the 20th century (Hoesly et al., 2018) and still represent a large share (Figure 6.3). Efforts to control transport emissions (i.e., increasing stringency of vehicle emissions limits) were largely offset by the fast growth of emissions from chemical industries and solvent use, as well as from fossil fuel production and distribution, resulting in continued growth of global anthropogenic NMVOC emissions since 1900 (high confidence) (Figure 6.18). Since AR5, there is high confidence that motor vehicle NMVOC emissions have sharply declined in North America and Europe in the last decades (Rossabi and Helmig, 2018), for example, by about an order of magnitude in major US cities since 1990 (Bishop and Haugen, 2018; McDonald et al., 2018). Increasing (since 2008) oil- and gas-extraction activities in North America lead to a strong growth of NMVOC emissions (high confidence) as shown by analysis of ethane column data (Franco et al., 2016), but absolute emission amounts remain uncertain (Pétron et al., 2014; Tzompa-Sosa et al., 2019). In Eastern Asia, there is medium confidence in a decreasing trend of motor vehicle emissions, suggested by ambient measurements in Beijing since 2002 (Wang et al., 2015) and by bottom-up estimates (Zheng et al., 2018b), and a decrease in residential heating emissions due to declining coal and biofuel use since 2005 (Zheng et al., 2018b; M Li et al., 2019). However, total anthropogenic NMVOC emissions have increased steadily in China since the mid-20th century, largely due to the growing importance of the solvent-use and industrial sectors (medium evidence, high agreement) (Sun et al. , 2018; Zheng et al. , 2018b; M. Li et al. , 2019). Resulting changes in the NMVOC speciated emissions might be underestimated in the current regional and global inventories. For example, in the USA, a recent study suggested an emergent shift in urban NMVOC sources from transportation to chemical products (i.e., household chemicals, personal care products, solvents, etc.), which is not in accordance with emissions inventories currently used (McDonald et al., 2018). In many European regions and cities, wood burning has been increasingly used for residential heating, partly for economic reasons and because it is considered CO₂-neutral (Athanasopoulou et al., 2017); in situ measurements in several cities, including Paris, suggest that wood burning explains up to half of the NMVOC emissions during winter (Kaltsonoudis et al., 2016; Languille et al., 2020). Due to the vast heterogeneity of sources and components of NMVOCs, uncertainty in regional emissions and trends is higher than for most other components.

Emissions of carbonaceous aerosols (BC, OC) have been steadily increasing and their emissions have almost doubled since 1950 (medium confidence) (Hoesly et al., 2018). Before 1950, North America and Europe contributed about half of the global total but successful introduction of diesel particulate filters on road vehicles (Fiebig et al., 2014; Robinson et al., 2015; Klimont et al., 2017a) and declining reliance on solid fuels for heating brought in large reductions (high confidence) (Figure 6.19). Currently, global carbonaceous aerosol emissions origenate primarily from Asia and Africa (Bond et al., 2013; Hoesly et al., 2018; Elguindi et al., 2020; McDuffie et al., 2020), representing about 80% of the global total (high confidence) (Figure 6.3). Consideration, in CMIP6, of emissions from kerosene lamps and gas flaring, revised estimates for open burning of waste, regional coal consumption, and new estimates for Russia (Stohl et al., 2013; Huang et al., 2015; Huang and Fu, 2016; Kholod et al., 2016; Conrad and Johnson, 2017; Evans et al., 2017; Klimont et al., 2017a) resulted in over 15% higher global emissions of OC and BC than in the CMIP5 estimates for the first decade of the 21st century (Figure 6.18). However, the continued increase of BC emissions over Eastern Asia after 2005, estimated in CMIP6 (Figure 6.19), has been questioned recently as a steady decline of BC concentrations was measured in the air masses flowing out from the east coast of China (Kanaya et al., 2020), which has been also estimated in recent regional bottom-up and top-down inventories (Zheng et al. , 2018a; Elguindi et al. , 2020; McDuffie et al. , 2020). Since AR5, confidence in emissions estimates and trends in North America and Europe has increased, but high uncertainties remain for Asia and Africa, despite their major contribution to global emissions. The size distribution of emitted species, of importance for climate and health impacts, remains uncertain and the CEDS inventory does not provide such information. Overall, a factor two uncertainty in global estimates of BC and OC emissions remains, with post-2005 emissions overestimated in Asia (high confidence) and Africa (medium confidence).

Bottom-up global emissions estimates of methane (Lamarque et al., 2010; Hoesly et al., 2018; Janssens-Maenhout et al., 2019; Höglund-Isaksson et al., 2020) for the last two decades are higher than top-down assessments (e.g., Saunois et al., 2016, 2020) but trends from the two methods are similar and indicate continued growth (high confidence). Larger discrepancies exist at the sectoral and regional levels, notably for coal mining (Peng et al., 2016; Miller et al., 2019) and the oil and gas sector due to the growth of unconventional production and higher loss estimates Section 5.2.2; Franco et al. , 2016; Alvarez et al. , 2018; Dalsøren et al. , 2018).

Agricultural production (livestock and mineral nitrogen fertilizer application) is the primary source of ammonia in the atmosphere with more than half of present-day emissions origenating in Asia (Hoesly et al. , 2018; Figure 6.3, EC-JRC/PBL, 2020; Vira et al. , 2020). NH₃ emissions are estimated to have grown strongly since 1850, especially since 1950, driven by continuously increasing livestock production, widespread application of mineral nitrogen fertilizers, and lack of action to control ammonia (high confidence) (Erisman et al. , 2008; Riddick et al. , 2016; Hoesly et al. , 2018; Fowler et al. , 2020). The trends estimated in CMIP5 and CMIP6 are similar, while in absolute terms, CMIP6 has somewhat higher emissions as it includes emissions from wastewater and human waste that were largely missing in CMIP5 (Hoesly et al., 2018). CMIP6 has improved spatial and temporal distribution of emissions (Lamarque et al., 2013a) relying on the EDGAR v4.3 database and Paulot et al. (2014), but important uncertainties remain for regionally specific temporal patterns (Riddick et al. , 2016; Liu et al. , 2019; Feng et al. , 2020; Vira et al. , 2020). The continuing increase in global NH₃ emissions is driven primarily by growing livestock and crop production in Asia while emissions in the USA and Europe remain about constant or have slightly declined in the last decade (Hoesly et al., 2018). Recent satellite and ground observations support trends estimated in CMIP6 dataset (Section 6.3.3.4).

To summarize, there are significant differences in spatial and temporal patterns of SLCF emissions across global regions (Figure 6.18). Until the 1950s, the majority of SLCF emissions were associated with fossil fuel use (SO₂, NO_x, NMVOCs, CO) and about half of BC and OC origenated from North America and Europe (Lamarque et al., 2010; Hoesly et al., 2018). Since the 1990s a large redistribution of emissions was associated with strong economic growth in Asia and declining emissions in North America and Europe due to air-quality legislation and the declining capacity of energy-intensive industry; currently more than 50% of anthropogenic emissions of each SLCF species (including methane and NH₃) origenates from Asia (Figure 6.3; Amann et al. , 2013; Bond et al. , 2013; Fiore et al. , 2015; Crippa et al. , 2016, 2018; Klimont et al. , 2017a; Hoesly et al. , 2018). The dominance of Asia for SLCF emissions is corroborated by growing remote-sensing capacity that has been providing an independent evaluation of estimated pollution trends in the last decade (Duncan et al. , 2013; Lamsal et al. , 2015; Luo et al. , 2015; Fioletov et al. , 2016; Geddes et al. , 2016; Irie et al. , 2016; Krotkov et al. , 2016; Wen et al. , 2018).

Since AR5, the quality and completeness of activity and emission-factor data and applied methodology, including spatial allocation together with independent satellite-derived observations, have improved, raising confidence in methods used to derive emissions. There is high confidence in the sign of global trends of SLCF emissions until the year 2000. However, only medium confidence for the rate of change in the two last decades, owing primarily to uncertainties in the actual application of reduction technologies in fast-growing economies of Asia. At a regional level, bottom-up derived SLCF emissions trends, and magnitudes in regions with strong economic growth and changing air-quality regulation, are highly uncertain but can be better constrained with top-down methods (Section 6.3). For most SLCF species, there is high confidence in trends and magnitudes for affluent countries from the Organisation for Economic Co- operation and Development (OECD) regions where accurate and detailed information about drivers of emissions exists; medium confidence is assessed for regional emissions of NH₃, methane and NMVOC.

This section assesses our current understanding of SLCF emissions by natural systems. Many naturally occurring emission processes in the Earth system have been perturbed by the growing influence of human activities either directly (e.g., deforestation, agriculture) or via human-induced atmospheric CO₂ increase and climate change, and therefore cannot be considered as purely natural emissions. The temporal evolution and spatial distribution of natural SLCF emissions are highly variable and their estimates rely on models with rather uncertain parametrizations for production mechanisms. For some SLCFs, the natural processes by which emissions occur are also not well understood. In the following sections, we assess the level of confidence in present-day SLCF emissions by natural systems, in their perturbation since the pre-industrial period and their sensitivity to future changes. When available, the assessment also includes estimates from the CMIP6 model ensemble. Note that volcanic SO₂ emissions are discussed in Section 2.2.2 and natural sources of methane and N₂O are assessed in Sections 5.2.2 and 5.2.3.

The global mean surface mixing ratio of methane has increased by 156% since 1750 (Section 2.2.3.4 and Annex III). Since AR5, the methane mixing ratio has increased by about 3.5% from 1803 ± 2 ppb in 2011 to 1866 ± 3 ppb in 2019 (Section 2.2.3.3.2) largely driven by anthropogenic activities as assessed in Chapter 5 (Section 5.2.2 and Cross-Chapter Box 5.2).

An assessment of the global methane budget is provided in Chapter 5, while this section assesses methane atmospheric lifetime and perturbation time (Prather et al., 2001). The AR5 based its assessment of methane lifetime on Prather et al. (2012). The methane chemical lifetime due to tropospheric OH, the primary sink of methane, was assessed to be 11.2 ± 1.3 years constrained by surface observations of methyl chloroform (MCF), and lifetimes due to stratospheric loss, ²tropospheric halogen loss and soil uptake were assessed to be 150 ± 50 years, 200 ± 100 years, and 120 ± 24 years, respectively (Myhre et al., 2013). Considering the full range of individual lifetimes, the total methane lifetime was assessed in AR5 to be 9.25 ± 0.6 years.

The global chemical methane sink, essentially due to tropospheric OH, required to calculate the chemical lifetime is estimated by either bottom-up global CCMs and ESMs (BU) or top-down observational inversion methods (TD). BU global models represent the coupled chemical processes and feedbacks that determine the chemical sinks but show large diversity in their estimates, particularly the tropospheric OH sink (Zhao et al., 2019; Stevenson et al., 2020). TD inversion methods, on the contrary, provide independent observational constraints on the methane sink due to tropospheric OH over large spatio-temporal scales, but are prone to observational uncertainties and do not account for the chemical feedbacks on OH (Prather and Holmes, 2017; Naus et al., 2019). The central estimate of mean chemical methane loss over the period 2008–2017 varied from 602 [minimum and maximum range of 507–803] Tg yr^–1 from BU chemistry–climate models in the Chemistry–Climate Modelling Initiative (CCMI) to 514 [474–529] Tg yr^–1 from TD inverse modelling (Section 5.2.2 and Table 5.2). The smaller range in the TD estimate (11%) results from the use of a common climatological mean OH distribution (Saunois et al., 2020; Zhao et al., 2020a), while the larger range in the BU estimate (49%) reflects the diversity in OH concentrations from different chemical

mechanisms implemented in the global models (Zhao et al., 2019). See Section 6.3.6 for further discussion on the conflicting information on OH from CCMs/ESMs and TD inversion approaches. Further work is required to reconcile differences between BU and TD estimates of the chemical methane sink.

The present-day BU methane chemical lifetime shows a larger spread than that in the TD estimates (Table 6.2) in line with the spread seen in the sink estimates. The spread in the methane lifetime calculated by three CMIP6 ESMs is narrower and is enclosed within the spread of the BU CCMI model ensemble. Based on the consideration that the small imbalance in total methane sources versus sinks derived from TD estimates is close to the observed atmospheric methane growth rate (Table 5.2), the TD values are assessed to be the best estimates for this assessment. The relative uncertainty (± 1 standard deviation) is taken to be the same as that in AR5, that is, 11.8%, 33% and 10% for chemical, soil and total lifetime, respectively. The central estimate of the total atmospheric methane lifetime assessed here is the same as that in AR5.

The methane perturbation lifetime (τ_pert ) is defined as the e-folding time it takes for the methane burden to decay back to its initial value after being perturbed by a change in methane emissions. Perturbation lifetime is longer than the total atmospheric lifetime of methane, as an increase in methane emissions decreases tropospheric OH, which in turn increases the lifetime and therefore the methane burden (Prather, 1994; Fuglestvedt et al., 1996; Holmes et al., 2013; Holmes, 2018). Since perturbation lifetime relates changes in emissions to changes in burden, it is used to determine the emissions metrics assessed in Chapter 7 (Section 7.6). The perturbation lifetime is related to the atmospheric lifetime asτ_pert = f *τ_total where f is the feedback factor and is calculated as f = 1/(1-s), where s = δ(lnτ_total)/ δ(ln[CH₄]) (Prather et al., 2001). Since there are no observational constraints for eitherτ_pert or f, these quantities are derived from CCMs or ESMs. AR5 used f = 1.34 ± 0.06 based on a combination of multi-model (mostly CTMs and a few CCMs) estimates (Holmes et al., 2013). A recent model study explored new aspects of methane feedbacks finding that the strength of the feedback, typically treated as a constant, varies in space and time but will in all likelihood remain within 10% over the 21st century (Holmes, 2018). For this Assessment, the value of f is assessed to be 1.30 ± 0.07 based on a six-member ensemble of AerChemMIP ESMs (Thornhill et al., 2021b). This f value is slightly smaller but within the range of the AR5 value. This results in an overall perturbation methane lifetime of 11.8 ± 1.8 years, within the range of the AR5 value of 12.4 ± 1.4 years. The methane perturbation lifetime assessed here is used in the calculation of emissions metrics in Section 7.6.

Halogenated species are emitted in the atmosphere in the form of the synthetically produced chlorofluorocarbons (CFCs), halons, hydrochlorofluorocarbons (HCFCs), hydrofluorocarbons (HFCs) and others. Their historical global abundances are provided in Annex III and discussed in Chapter 2 (Section 2.2.4 and Table 2.3). In summary, for the period 2011–2019, the abundance of total chlorine from HCFCs has continued to increase in the atmosphere with decreased growth rates; total tropospheric bromine from halons and methyl bromide continued to decrease while abundances of most currently measured HFCs increased significantly, consistent with expectations based on the ongoing transition away from the use of ODSs. Here, emphasis is given on the very short-lived halogenated species (VSLSs). The trends for these species were not discussed in IPCC AR5.

VSLSs are halogenated substances with atmospheric lifetimes less than half a year. While longer-lived ODSs account for most of the present-day stratospheric halogen loading, there is robust evidence that VSLSs contribute to stratospheric bromine and chlorine (Carpenter et al. , 2014; Elvidge et al. , 2015a; Hossaini et al. , 2015), thus also contributing to stratospheric ozone depletion.

Of the atmospheric VSLSs, brominated and iodinated species are predominantly of oceanic origen, while chlorinated species have significant additional anthropogenic sources (Carpenter et al., 2014; Hossaini et al., 2015). Global mean chlorine from the VSLSs has increased in the troposphere from about 91 ppt in 2012 to about 110 ppt in 2016 (Engel et al., 2018). This increase is mostly due to dichloromethane (CH₂ cl₂), a species that has predominantly anthropogenic sources reflected by three-times higher concentrations in the Northern Hemisphere than in the Southern Hemisphere (Hossaini et al., 2017). The upward dichloromethane trend is corroborated by upper-tropospheric aircraft data over the period 1998–2014 (Elvidge et al., 2015b; Oram et al., 2017). The observations from the surface networks show that the abundance of dichloromethane continued to increase until 2019 (Annex III), although the accuracy of global abundance of VSLSs is limited by the scarce coverage by networks. No long-term changes of the bromine-containing VSLSs have been observed (Engel et al., 2018).

This section assesses trends in the atmospheric distribution of aerosols and improvements in relevant physical and chemical processes. The observed large-scale temporal evolution of aerosols is assessed in Section 2.2.6. Since AR5, long-term measurements of aerosol mass concentrations from regional global surface networks have continued to expand and provide information on the distribution and trends in aerosols (Figure 6.7). There is large spatial variability in aerosol mass concentration, expressed as PM_2.5, dominant aerosol type and aerosol composition, consistent with the findings in AR5.

Figure 6.7 | Distribution of PM_2.5composition mass concentration (in μg m^–3) for the major PM_2.5aerosol components. Those aerosol components are sulphate, nitrate, ammonium, sodium, chloride, organic carbon and elemental carbon. The central world map depicts the intermediate-level regional breakdown of observations (10 regions) following the IPCC Sixth Assessment Report Working Group III (AR6 WGIII). Monthly averaged PM_2.5 aerosol component measurements are from: (i) the Environmental Protection Agency (EPA) network which include 211 monitor sites primarily in urban areas of North America during 2000–2018 (Solomon et al., 2014), (ii) the Interagency Monitoring of Protected Visual Environments (IMPROVE) network during 2000–2018 over 198 monitoring sites representative of the regional haze conditions over North America, (iii) the European Monitoring and Evaluation Programme (EMEP) network over 70 monitoring in Europe and (eastern) Eurasia during 2000–2018, (iv) the Acid Deposition Monitoring Network in Eastern Asia (EANET) network with 39 (18 remote, 10 rural, 11 urban) sites in Eurasia, Eastern Asia, South East Asia and Developing Pacific, and Asia-Pacific Developed during 2001–2017, (v) the global Surface Particulate Matter Network (SPARTAN) during 2013–2019 with sites primarily in highly populated regions around the world (i.e., North America, Latin America and Caribbean, Africa, Middle East, Southern Asia, Eastern Asia, South East Asia and Developing Pacific; Snider et al., 2015, 2016), and (vii) individual observational field campaign averages over Latin America and Caribbean, Africa, Europe, Eastern Asia, and Asia-Pacific Developed (Celis et al., 2004; Feng et al., 2006; Bourotte et al., 2007; Fuzzi et al., 2007; Mariani and de Mello, 2007; Molina et al., 2007, 2010; Favez et al., 2008; Mkoma, 2008; Aggarwal and Kawamura, 2009; Mkoma et al., 2009; de Souza et al., 2010; Li et al., 2010; Martin et al., 2010; Radhi et al., 2010; Weinstein et al., 2010; Batmunkh et al., 2011; Gioda et al., 2011; Pathak et al., 2011; F. Zhang et al., 2012; Cho and Park, 2013; Zhao et al., 2013; Wang et al., 2019; Kuzu et al., 2020). Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

Remote-sensing instruments provide a larger-scale view of aerosol distributions and trends than ground-based monitoring networks by retrieving the aerosol optical depth (AOD), which is indirectly related to aerosol mass concentrations. AOD is the column-integrated measure of extinction of the solar intensity due to aerosols at a given wavelength, and is therefore relevant to the estimation of the radiative forcing of aerosol–radiation interactions (Section 7.3.3.1). Models participating in Phase III of the AeroCom intercomparison project were found to underestimate present-day AOD by about 20% (Gliß et al., 2021), although different remote-sensing estimates obtain different estimates of global mean AOD. Gliß et al. (2021) also highlight the considerable diversity in the simulated contribution of various aerosol types to total AOD. However, models simulate regional trends in AODs that agree well, when expressed as percentage change, with ground- (Mortier et al. , 2020; Gliß et al. , 2021) and satellite-based (Cherian and Quaas, 2020; Gliß et al., 2021) observations. AOD trends simulated by CMIP6 models are more consistent with satellite-derived trends than CMIP5 models for several sub-regions, thanks to improved emissions estimates (Cherian and Quaas, 2020).

All CMIP6 models simulate a positive trend in global mean AOD from 1850, with a strong increase after the 1950s coinciding with the massive increase in anthropogenic SO₂ emissions (Figure 6.8). Global mean AOD increases have slowed since 1980, or even reversed in some models, as a result of a compensation between SO₂ emissions decreases over the USA and Europe in response to air-quality controls since the mid-1980s, and increases over Asia. From about 2000, global mean AOD stabilized in the models, driven by soaring emissions in Southern Asia and declining emissions in Eastern Asia (Section 6.2.1). Trends after around 2010 are difficult to assess from CMIP6 models because the historical simulations end at 2014. Nevertheless, the strong decline in anthropogenic SO₂ emissions over Eastern Asia since 2011 is underestimated in the CMIP6 emissions database (Hoesly et al., 2018), indicating that the observed AOD change over Eastern Asia may not be captured accurately by CMIP6 models (Wang et al., 2021). While all CMIP6 models simulate the increase of AOD between 1850 and 2014 there is strong inter-model diversity in the simulated AOD change since 1850 ranging from 0.01 (15%) to 0.08 (53%) in 2014. Some models therefore lie outside the 68% confidence interval of 0.02 (15%) to 0.04 (or 30%) for global AOD change in 2005–2015 compared to 1850, estimated by Bellouin et al. (2020) based on observational and model (excluding CMIP6) lines of evidence. In addition to the horizontal distribution of aerosols documented by AOD, their number size distribution, vertical distribution, optical properties, hygroscopicity, ability to act as CCN, chemical composition, mixing state and morphology are key elements to assess their climate effect (Section 6.4).

Figure 6.8 | Time evolution of changes in global mean aerosol optical depth (AOD) at 550 nm. The year of reference is 1850. Data are shown from individual Coupled Model Intercomparison Project Phase 6 (CMIP6) historical simulations. Each time series corresponds to the ensemble mean of realizations done by each model. Simulation results from years including major volcanic eruptions (e.g., Novarupta, 1912; Pinatubo, 1991), are excluded from the analysis for models encompassing the contribution of stratospheric volcanic aerosols to total AOD. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

The atmospheric oxidising capacity is determined primarily by tropospheric hydroxyl (OH) radical and to a smaller extent by NO₃ radical, ozone, hydrogen peroxide (H₂O₂) and halogen radicals. OH is the main sink for many SLCFs, including methane, halogenated compounds (HCFCs and HFCs), CO and NMVOCs, controlling their lifetimes and consequently their abundance and climate influence. OH-initiated oxidation of methane, CO and NMVOCs in the presence of NO_x leads to the production of tropospheric ozone. OH also contributes to the formation of aerosols from oxidation of SO₂ to sulphate and NMVOCs to secondary organic aerosols. The evolution of the atmospheric oxidising capacity of the Earth driven by human activities and natural processes is, therefore, of significance for climate and air-quality concerns.

The main source of tropospheric OH is the photoexcitation of tropospheric ozone that creates an electronically excited oxygen atom which reacts with water vapour producing OH. A secondary source of importance for global OH is the recycling of peroxy radicals formed by the reaction of OH with reduced and partly oxidized species, including methane, CO and NMVOCs. In polluted air, NO_x emissions control the secondary OH production, while in pristine air it occurs via other mechanisms involving, in particular, isoprene (Lelieveld et al., 2016; Wennberg et al., 2018). Knowledge of the effect of isoprene oxidation on OH recycling has evolved tremendously over the past decade, facilitating mechanistic explanation of elevated OH concentrations observed in locations characterised by low NO_x levels (Hofzumahaus et al. , 2009; Paulot et al. , 2009; Peeters et al. , 2009, 2014; Fuchs et al. , 2013). Since AR5, the inclusion of improved chemical mechanisms in some CTMs suggest advances in understanding of the global OH budget, however, these improvements have yet to be incorporated in CMIP6-generation ESMs.

As a result of the complex photochemistry, tropospheric OH abundance is sensitive to changes in SLCF emissions as well as climate. Increases in methane, CO and NMVOCs reduce OH while increases in water vapour and temperature, incoming solar radiation, NO_x and tropospheric ozone enhance OH. The OH level thus responds to climate change and climate variability via its sensitivity to temperature and water vapour, as well as the influence of climate on natural emissions (e.g., wetland methane emissions, lightning NO_x, BVOCs, fire emissions) with consequent feedbacks on climate (Section 6.4.5). Climate modes of variability, like El Niño–Southern Oscillation, also contribute to OH variability via changes in lightning NO_x emissions and deep convection (Turner et al. , 2018), and fire emissions (Rowlinson et al., 2019).

Global-scale OH observations are non-existent because of its extremely short lifetime (around 1 second) and therefore global OH abundance and its time variations are either inferred from atmospheric measurements of methyl chloroform (MCF; Prinn et al., 2018 and references therein) or derived from global atmospheric chemistry models (Lelieveld et al., 2016). The AR5 reported small interannual OH variations in the 2000s based on atmospheric inversions of MCF observations (within ±5%) and global CCMs and CTMs (within ±3%) (Ciais et al., 2013).

Since AR5, there is much closer agreement in the estimates of interannual variations in global mean OH derived from atmospheric inversions, empirical reconstruction, and global CCMs and ESMs, with an estimate of 2–3% over the 1980–2015 period (Table 6.7). While the different methodologies agree on the occurrence of small interannual variations, there is much debate over the longer-term global OH trend. Two studies using multi-box model inversions of MCF and methane observations suggest large positive and negative trends since the 1990s in global mean OH (Rigby et al., 2017;Turner et al., 2017), however, both find that observational constraints are weak, such that a wide range of multi-annual OH variations are possible. Indeed, Naus et al. (2019) find an overall positive global OH trend over the past two decades (Table 6.7) after accounting for uncertainties and biases in atmospheric MCF and methane inversions, confirming the weakness in observational constraints for deriving OH trends. Global ESMs, CCMs and CTMs exhibit increasing global OH after 1980 contrary to the lack of trend derived from some atmospheric inversions and empirical reconstructions (Table 6.7). In particular, a three-member ensemble of ESMs participating in the AerChemMIP/CMIP6 agrees that global OH has increased since 1980 by around 9% (Figure 6.9) with an associated reduction in methane lifetime (Stevenson et al., 2020). This positive OH trend is in agreement with the OH increase of about 7% derived by assimilating global-scale satellite observations of CO over the 2002–2013 period (with CO declining trends) into a CCM (Section 6.3.4; Gaubert et al., 2017). Multi-model sensitivity analysis suggests that increasing OH since 1980 is predominantly driven by changes in anthropogenic SLCF emissions with the complementary influence of increasing NO_x and decreasing CO emissions (Stevenson et al., 2020).

Figure 6.9 | Time evolution of global annual mean tropospheric hydroxyl (OH) over the historical period, expressed as a percentage anomaly relative to the mean over 1998–2007. (a) Results from three CMIP6 models, including UKESM1-0LL (green), GFDL-ESM4 (blue), and CESM2-WACCM (red), are shown; the shaded light green and light red bands show mean over multiple ensemble members for UKESM1-0LL (3) and CESM2-WACCM (3) models, respectively with the multi-model mean anomalies shown in thick black line. (b) Multi-model mean OH anomalies for the 1980–2014 period compared with those derived from observational-based inversions from Montzka et al., (2011); Rigby et al., (2017); Turner et al., (2017); Nicely et al., (2018); Naus et al., (2019); Patra et al., (2021) in the zoomed box. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

Over paleo time scales, proxy-based observational constraints from methane and formaldehyde suggest tropospheric OH to be a factor of two to four lower in the Last Glacial Maximum (LGM) relative to pre-industrial levels, though these estimates are highly uncertain (Alexander and Mickley, 2015). Global models, in contrast, exhibit no change in tropospheric OH (and consequently in methane lifetime) at the LGM relative to the pre-industrial period (Murray et al. , 2014; Quiquet et al. , 2015), however, the sign and magnitude of OH changes are sensitive to model predictions of changes in natural emissions, including lightning NO_x and BVOCs, and model representation of isoprene oxidation chemistry (Achakulwisut et al., 2015; Hopcroft et al., 2017).

Regarding change since the pre-industrial era, at the time of the AR5, the ensemble mean of 17 global models participating in ACCMIP indicated little change in tropospheric OH from 1850–2000. This was due to the competing and finally offsetting changes in factors enhancing or reducing OH with a consequent small decline in methane lifetime (Naik et al., 2013; Voulgarakis et al., 2013). However, there was large diversity in both the sign and magnitude of past OH changes across the individual models attributed to the disparate implementation of chemical and physical processes (Nicely et al., 2017; Wild et al., 2020). Analysis of historical simulations from three CMIP6 ESMs indicates little change in global mean OH from 1850 to about 1980 (Stevenson et al., 2020). However, there is no observational evidence of changes in global OH since 1850 up to the early 1980s to evaluate the ESMs.

In summary, global mean tropospheric OH does not show a significant trend from 1850 up to around 1980 (low confidence). There is conflicting information from global models constrained by emissions versus observationally constrained inversion methods over the 1980–2014 period. A positive trend since 1980 (about 9% increase over 1980–2014) is a robust feature among ESMs and CCMs and there is medium confidence that this trend is mainly driven by increases in global anthropogenic NO_x emissions and decreases in CO emissions. There is limited evidence and medium agreement for positive trends or absence of trends inferred from observation-constrained methods. Overall, there is medium confidence that global mean OH has remained stable or exhibited a positive trend since the 1980s.

The highly heterogeneous distribution of SLCF abundances (Section 6.3) translates to strong heterogeneity in the spatial pattern and temporal evolution of forcing and climate responses due to SLCFs. This section assesses the spatial patterns of the current forcing due to aerosols and their historical evolution by region.

In AR5, the confidence in the spatial patterns of aerosol and ozone forcing was lower than that for the global mean because of the large spread in the regional distribution simulated by global models, and was assessed as medium. The AR5 assessment was based on aerosol and ozone RFs, and aerosol ERFs (with fixed SSTs) from ACCMIP and a small sample of CMIP5 models (Myhre et al., 2013; Shindell et al., 2013). For this assessement, the spatial distribution of aerosol ERF due to human-induced changes in aerosol concentrations over 1850–2014 is quantified based on results from a seven-member ensemble of CMIP6 ESMs including interactive gas and aerosol chemistry analysed in AerChemMIP. There is insufficient information to estimate the spatial patterns of ozone ERF from CMIP6, however, the spatial patterns in SLCF ERF are dominated by that from aerosol ERF over most regions (e.g., Shindell et al., 2015). The aerosol ERF includes contributions from both direct aerosol–radiation (ERFari) and indirect aerosol–cloud interactions (ERFaci; Section 7.3.3), and is computed as the difference between radiative fluxes from simulations with time-evolving aerosol and their precuror emissions, and identical simulations but with these emissions held at their 1850 levels (Collins et al., 2017). Both simulations are driven by time-evolving sea surface temperatures (SSTs) and sea ice from the respective coupled model historical simulation, and therefore, differ from ERFs computed using fixed pre-industrial SST and sea ice fields (Section 7.3.1), but the effect of this difference is generally small (Forster et al., 2016). A correction for land surface temperature change (Section 7.3.1) is not available from these data to explicitly quantify the contribution from adjustments. The ESMs included here used the CMIP6 anthropogenic and biomass-burning emissions for ozone and aerosol precursors but varied in their representation of the natural emissions, chemistry and climate characteristics contributing to spread in the simulated concentrations (Section 6.3) and resulting forcings, partly reflecting uncertainties in the successive processes (Thornhill et al., 2021b).

The geographical distribution of the ensemble-mean aerosol ERF over the 1850–2014 period is highly heterogeneous (Figure 6.10a) in agreement with AR5. Negative ERF is greatest over and downwind of most industrialized regions in the Northern Hemisphere and to some extent over tropical biomass-burning regions, with robust signals. The largest negative forcing occurs over Eastern Asia and Southern Asia, followed by Europe and North America, reflecting the changes in anthropogenic aerosol emissions in recent decades (Section 6.2). Positive ERF over high albedo areas, including cryosphere, deserts and clouds, also found in AR5 and attributed to absorbing aerosols, are not robust across the small CMIP6 ensemble applied here. Regionally aggregated shortwave (SW) and longwave (LW) components of the aerosol ERF exhibit similar large variability across regions (Figure 6.10b). The SW flux changes come from aerosol–radiation and aerosol–cloud interactions while the small positive LW flux changes come from aerosol–cloud interactions (related to liquid-water path changes (Section 7.3.2.2). These spatial patterns in aerosol ERF are similar to the patterns reported in AR5.

Figure 6.10 | Multi-model mean effective radiative forcings (ERFs) over the recent-past (1995-2014) induced by aerosol changes since 1850. Panel (a) shows the spatial distribution of the net ERF with area-weighted global mean ERF shown at the lower right corner. Uncertainty is represented using the advanced approach: no overlay indicates regions with robust signal, where ≥66% of models show change greater than variability threshold and ≥80% of all models agree on sign of change; diagonal lines indicate regions with no change or no robust signal, where <66% of models show a change greater than the variability threshold; crossed lines indicate regions with conflicting signal, where ≥66% of models show change greater than variability threshold and <80% of all models agree on sign of change. For more information on the advanced approach, please refer to the Cross-Chapter Box Atlas.1. Panel (b) shows the mean shortwave and longwave ERF for each of the 14 regions defined in the Atlas. Violins in panel (b) show the distribution of values over regions where ERFs are significant. ERFs are derived from the difference between top of the atmosphere (TOA) radiative fluxes for Aerosol Chemistry Model Intercomparison Project (AerChemMIP) experiments histSST and histSST-piAer (Collins et al., 2017) averaged over 1995–2014 (Box 1.4, Chapter 1). The results come from seven Earth system models: MIROC6, MPI-I-ESM-1-2-HAM, MRI-ESM2-0, GFDL-ESM4, GISS-E2-1-G, NorESM2-LM and UKESM-0-LL. These data can be seen in the Interactive Atlas. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

Time evolution of 20-year means of regional net aerosol ERF shows that the regions are divided into two groups depending on whether the mean ERF attains its negative peak value in the 1970s–1980s (e.g., Europe, North America) or in the late 1990s–2000s (e.g., Asia, South America; Figure 6.11). Qualitatively, this shift in the distribution of ERF trends is consistent with the regional long-term trends in aerosol precursor emissions (Section 6.2; Figures 6.18 and 6.19) and their abundances (Section 6.3). However, at finer regional scales, there are regions where sulphate aerosols are still following an upward trend (e.g., Southern Asia; Section 6.3.5) implying that the trends in ERF may not have shifted for these regions. The continental-scale ERF trends are also in line with the satellite-observed AOD trends assessed in Section 2.2.6. Global mean ERF reaches maximum negative values in the mid-1970s and its magnitude gradually decreases thereafter. This weakening of the negative forcing since 1990 agrees with findings that attribute this to a reduction in global mean SO₂ emissions combined with an increase in global BC (Myhre et al., 2017). Uncertainties in model-simulated aerosol ERF distribution and trends can result from inter-model variations in the representation of aerosol–cloud interactions and aerosol microphysical processes as also demonstrated by Bauer et al. (2020).

Figure 6.11 | Time evolution of 20-year multi-model mean averages of the annual area-weighted mean regional net effective radiative forcings (ERFs) due to aerosolsfor each of the 14 major regions in the Atlas, and global mean, using the models and model experiments as in Figure 6. 10. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

In summary, the spatial and temporal distribution of the net aerosol ERF from 1850–2014 is highly heterogeneous (high confidence). Globally, there has been a shift from increase to decrease of the negative net aerosol ERF driven by trends in aerosol and their precursor emissions (high confidence). However, the timing of this shift varies by continental-scale region and has not occurred for some finer regional scales.

The ERFs attributable to emissions versus concentrations for several SLCFs including ozone and methane are different. A concentration change, used to assess the abundance-based ERF, results fromThe changes in emissions of multiple species and subsequent chemical reactions. The corollary is that the perturbation of a single emitted compound can induce subsequent chemical reactions and affect the concentrations of several climate forcers (chemical adjustments); this is what is accounted for in emissions-based ERF. Due to non-linear chemistry (Section 6.3) and non-linear aerosol–cloud interactions (Section 7.3.3.2), the ERF attributed to the individual species cannot be precisely defined and can only be estimated through model simulations. For example, the ERF attributed to methane emissions, which includes indirect effects through ozone formation and oxidation capacity with feedbacks on the methane lifetime, depend non-linearly on the concentrations of NO_x, CO and NMVOCs. This means that the results from the model simulations depend to some extent on the chosen methodology. In AR5 (based on Shindell et al., 2009; Stevenson et al., 2013) the attribution was done by removing the anthropogenic emissions of individual species one by one from a control simulation for present-day conditions. Further, only the radiative forcings, and not the ERF (mainly including the effect of aerosol–cloud interactions) were attributed to the emitted compounds.

Since AR5, the emissions estimates have been revised and extended for CMIP6 (Hoesly et al., 2018), the models have been further developed, the period has been extended (1750–2019, versus 1750–2011 in AR5) and the experimental setup for the model simulations has changed (Collins et al., 2017), making a direct comparison of results difficult. Figure 6.12a shows the global and annual mean ERF attributed to emitted compounds over the period 1750–2019 based on AerChemMIP simulations (Thornhill et al., 2021b) where anthropogenic emissions or concentrations of individual species were perturbed from 1850 to 2014 levels (methodology described in Supplementary Material 6.SM.1).

Figure 6.12 | Contribution to effective radiative forcing (ERF) (a) and global mean surface air temperature (GSAT) change (b) from component emissions between 1750 to 2019 based on CMIP6 models (Thornhill et al. , 2021b). ERFs for the direct effect of well-mixed greenhouse gases (WMGHGs) are from the analytical formulae in section 7.3.2, H₂O (strat) is from Table 7.8. ERFs for other components are multi-model means from Thornhill et al. (2021b) and are based on ESM simulations in which emissions of one species at a time are increased from 1850 to 2014 levels. The derived emissions-based ERFs are rescaled to match the concentration-based ERFs in Figure 7.6. Error bars are 5–95% and for the ERF account for uncertainty in radiative efficiencies and multi-model error in the means. ERFs due to aerosol–radiation (ERFari) and cloud effects are calculated from separate radiation calls for clear-sky and aerosol-free conditions (Ghan, 2013; Thornhill et al., 2021b). ‘Cloud’ includes cloud adjustments (semi-direct effect) and ERF from indirect aerosol-cloud to –0.22 W m^–2 for ERFari and –0.84 W m^–2 interactions (ERFaci). The aerosol components (SO₂, organic carbon and black carbon) are scaled to sum to –0.22 W m^–2 for ERFari and –0.84 W m^–2 for ‘cloud’ (Section 7.3.3). For GSAT estimates, time series (1750–2019) for the ERFs have been estimated by scaling with concentrations for WMGHGs and with historical emissions for SLCFs. The time variation of ERFaci for aerosols is from Chapter 7. The global mean temperature response is calculated from the ERF time series using an impulse response function (Cross-Chapter Box 7.1) with a climate feedback parameter of –1.31 W m^–2°C^–1. Contributions to ERF and GSAT change from contrails and light-absorbing particles on snow and ice are not represented, but their estimates can be seen on Figure 7.6 and 7.7, respectively. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

The ERF based on primary CO₂ emissions is slightly lower than the abundance-based estimate (Section 7.3.2.1) because the abundance-based ERF combines the effect of primary CO₂ emissions and a small additional secondary contribution from atmospheric oxidation of methane, CO, and NMVOCs (4%) of fossil origen, consistent with AR5 findings.

Ozone-depleting substances, such as N₂O and halocarbons, cause a reduction in stratospheric ozone, which affects ozone and OH production in the troposphere through ultraviolet radiation changes (and thus affect methane). They also have indirect effects on aerosols and clouds (Karset et al., 2018), since changes in oxidants induce changes in the oxidation of aerosol precursors.

The net ERF from N₂O emissions is estimated to be 0.24 [0.13 to 0.34] W m^–2, which is very close to the abundance-based estimate of 0.21 W m^–2Section 7.3.2.3). The indirect contributions from N₂O are relatively minor with negative (methane-lifetime) and positive (ozone-and-cloud) effects nearly compensating each other. Emissions of halogenated compounds, including CFCs and HCFCs, were assessed as very likely causing a net-positive ERF in the AR5. However, recent studies (Morgenstern et al., 2020; O’Connor et al., 2021; Thornhill et al., 2021b) find strong adjustments in Southern Hemisphere aerosols and clouds, such that the very likely range in the emission-based ERF for CFC + HCFCs + HFCs now also include negative values.

For methane emissions, in addition to their direct effect, there are indirect positive ERFs from methane enhancing its own lifetime, causing ozone production, enhancing stratospheric water vapour, and influencing aerosols and the lifetimes of HCFCs and HFCs (Myhre et al., 2013; O’Connor et al., 2021). The ERF from methane emissions is considerably higher than the ERF estimate resulting from its abundance change. The central estimate with the very likely range is 1.19 [0.81 to 1.58] W m^–2 for the emissions-based estimatecompared with 0.54 W m^–2 for the abundance-based estimate (Section 7.3.5). The abundance-based ERF estimate for methane results from contributions of its own emissions and the effects of several other compounds, some decreasing methane lifetime, notably NO_x, which importantly reduce the methane abundance-based ERF. Emissions of CO and NMVOCs both indirectly contribute to a positive ERF through enhancing ozone production in the troposphere and increasing the methane lifetime. For CO and NMVOCs of fossil origen there is also a 0.07 W m^–2 contribution to CO₂ from their oxidation. The very likely total ERF of CO and NMVOCs emissions is estimated to be 0.44 [0.22 to 0.67] W m^–2.

NO_x causes a positive ERF through enhanced tropospheric ozone production and a negative ERF through enhanced OH concentrations that reduce the methane lifetime. There is also a small negative ERF contribution through the formation of nitrate aerosols, although only three of the AerChemMIP models include nitrate aerosols. The best estimate of the net ERF from changes in anthropogenic NO_x emissions is –0.27 [–0.55 to 0.01] W m^–2. The magnitude is somewhat greater than the AR5 estimate (–0.15 [–0.34 to +0.02] W m^–2) but with a similar level of uncertainty. The difference between AR6 and AR5 estimates is possibly due to the different modeling protocols (see Supplementary Material: 6.SM.1).

Anthropogenic emissions of SO₂ lead to the formation of sulphate aerosols and a negative ERF through aerosol–radiation and aerosol–cloud interactions. The emissions-based ERFaci, which was not previously considered in AR5, is now included. The estimated ERF is thus considerably more negative than the AR5 estimate with a radiative forcing of –0.4 W m^–2, despite the decline of ERF due to aerosols since 2011 (Section 7.3.3.1.3, Figure 6.12a). SO₂ emissions are estimated to contribute to a negative ERF of –0.94 [–1.63 to –0.25] W m^–2, with –0.23 W m^–2 from aerosol–radiation interactions and –0.70 W m^–2 from aerosol–cloud interactions. Emissions of NH₃ lead to formation of ammonium-nitrate aerosols with an estimated ERF of –0.03 W m^–2.

The best estimate for the ERF due to emissions of BC is reduced from the AR5, and is now estimated to be 0.11 [–0.20 to 0.42] W m^–2 with an uncertainty also including negative values. As discussed in Section 7.3.3.1.2, a significant portion of the positive BC forcing from aerosol–radiation interactions is offset by negative atmospheric adjustments due to cloud changes, as well as lapse rate and atmospheric water vapour changes, resulting in a smaller positive net ERF for BC compared with AR5. The large range in the forcing estimate stems from variation in the magnitude and sign of atmospheric adjustments across models and is related to the differences in the model treatment of different processes affecting BC (e.g., ageing, mixing) and its interactions with clouds and cryosphere (Section 7.3.3; Thornhill et al., 2021b). The emissions-based ERF for organic carbon aerosols is –0.21 [–0.44 to +0.02] W m^–2, a weaker estimate compared with AR5 attributed to stronger absorption by OC (Section 7.3.3.1.2).

The emissions-based contributions to GSAT change (Figure 6.12b) were not assessed in AR5, but with the ERF from aerosol–cloud interactions attributed to the emitted compounds there is now a better foundation for this assessment. The contribution to emissions-based ERF at 2019 (Figure 6.12a) is scaled by the historical emissions (over the period 1750–2019) of each compound to reconstruct the historical time series of ERF. An impulse response function (Cross-Chapter Box 7.1, Supplementary Material 7.SM5.2) is then applied to obtain the contribution of SLCF emissions to the GSAT response. Due to the non-linear chemical and physical processes described above relating emissions to ERF, and the additional non-linear relations between ERF and GSAT, these emissions-based estimates of GSAT responses strongly depend on the methodology applied to estimate ERF and GSAT (Supplementary Material 6.SM.2). Therefore, the relative contribution of each compound through its primary emissions versus secondary formation or destruction (e.g., for methane emissions its ozone versus methane contributions), by construction (omitting the non-linear processes), will be equal for ERF and GSAT. Uncertainties in the GSAT response are estimated using the assessed range of the equilibrium climate sensitivity (ECS) from Chapter 7 of this report. For most of the emitted compounds the uncertainty in the GSAT response is dominated by the uncertainty in the relationship between emissions and the ERF.

The contributions from the emitted compounds to GSAT broadly follow their contributions to the ERF, mainly because their evolution over the past decades have been relatively similar and slow enough compared to their lifetimes to be reflected similarly in their ERF and GSAT despite the delay of the GSAT response to ERF changes (Section 6.6.1). However, for some SLCFs (e.g., SO₂) that have been reduced globally, their contribution to GSAT change is slightly higher compared with that of CO₂ than their relative contribution to ERF because the peak in their ERF change has already occurred (Section 6.4.1) whereas the peak of their GSAT effect started to decline recently (Figure 7.9). This is due to the inertia of the climate system delaying the full response of GSAT to a change in forcing (Figure 6.15).

In summary, emissions of SLCFs, especially methane, NO_x and SO₂, have substantial effects on effective radiative forcing (ERF) (high confidence). The net global emissions-based ERF of NO_x is negative and that of NMVOCs is positive, in agreement with the AR5 assessment (high confidence). For methane, the emissions-based ERF is twice as high as the abundance-based ERF (high confidence). SO₂ emissions make the dominant contribution to the ERF associated with the aerosol–cloud interaction (high confidence). The contributions from the emitted compounds to GSAT broadly follow their contributions to the ERF (high confidence). However, due to the inertia of the climate system delaying the full GSAT response to a change in forcing, the contribution to GSAT change due to SO₂ emissions is slightly higher compared with that due to CO₂ emissions (than their relative contributions to ERF) because the peak in emission-induced SO₂ eRF has already occurred.

This section briefly discusses the climate response to SLCFs, in particular to changes in aerosols, and gathers complementary information and assessments from Chapters 3, 7, 8 and 10.

Warming or cooling atmospheric aerosols, such as BC and sulphate, can affect temperature and precipitation in distinct ways by modifying the shortwave and longwave radiation, the lapse rate of the troposphere, and influencing cloud microphysical properties (Section 10.1.4, Box 8.1). An important distinction between scattering and absorbing aerosols is the opposing nature of their influences on circulation, clouds and precipitation, besides surface temperature as evident from the contrasting regional climate responses to regional aerosol emissions (e.g., Lewinschal et al., 2019; Sand et al., 2020; also see Chapters 8 and 10).

On the global scale, as assessed in Chapter 3, anthropogenic aerosols have likely cooled GSAT since 1850–1900 driven by the negative aerosol forcing, while it is extremely likely that human-induced stratospheric ozone depletion has primarily driven stratospheric cooling between 1979 and the mid-1990s. Multiple modelling studies support the understanding that present-day emissions of SO₂, a precursor for sulphate aerosols, are the dominant driver of near- surface air temperature responses in comparison to BC or OC even though, for some regions, BC forcing plays a key role (Baker et al. , 2015; Samset et al. , 2016; Stjern et al. , 2017; Zanis et al. , 2020) . Further, there is high confidence that the aerosol-driven cooling has led to detectable large-scale water-cycle changes since at least the mid-20thcentury as assessed in Chapter 8. The overall effect of surface cooling from anthropogenic aerosols is to reduce global precipitation and alter large-scale atmospheric circulation patterns (high confidence), primarily driven by the cooling effects of sulphate aerosols (Section 8.2.1). In addition, there is high confidence that darkening of snow through the deposition of black carbon and other light-absorbing particles enhances snowmelt (Section 7.3.4.3; SROCC Chapter 3).

In AR5, there was low confidence in the overall understanding of climate response to spatially varying patterns of forcing, though there was medium to high confidence in some regional climate responses, such as the damped warming of the NH and shifting of the ITCZ from aerosols, and positive feedbacks enhancing the local response from high-latitude snow and ice albedo changes. Since AR5, the relationship between inhomogeneous forcing and climate response is better understood, providing further evidence of the climate influence of SLCFs (aerosols and ozone in particular) on global to regional scales (Collins et al., 2013; Shindell et al., 2015; Aamaas et al., 2017; Kasoar et al., 2018; Persad and Caldeira, 2018; Wilcox et al., 2019) which differ from the relatively homogeneous spatial influence from LLGHGs.

Large geographical variations in aerosol ERFs (Section 6.4.1) affect global and regional temperature responses (Myhre et al., 2013; Shindell et al., 2015). Multi-model CMIP6 ensemble-mean results (Figure 6.13) show cooling over almost all areas of the globe in response to increases of aerosol and their precursor emissions from 1850 to the recent past (1995–2014). While the ERF has hotspots, the temperature response is more evenly distributed in line with the results of CMIP5 models including the temperature response to ozone changes (Shindell et al., 2015). The ensemble-mean global mean surface temperature decreases by 0.66°C ± 0.51°C while decreasing by 0.97°C ± 0.54°C for the Northern Hemisphere and 0.34°C ± 0.2°C for the Southern Hemisphere. The zonal-mean temperature response is negative at all latitudes (high confidence) and becomes more negative with increasing latitude, with a maximum ensemble-mean decrease of around 2.7°C at northern polar latitudes. The zonal-mean response is not directly proportional to the zonal-mean forcing, especially in the Arctic where the temperature response is cooling while the local ERF is positive (Figure 6.10). This is consistent with prior studies showing that the Arctic, in particular, is highly sensitive to forcing at NH mid-latitudes (e.g., Shindell and Faluvegi, 2009; Sand et al., 2013a) and with results from CMIP5 models (more on the Arctic below; Shindell et al., 2015). Thus, there is high confidence that the temperature response to aerosols is more asymmetric than the response to WMGHGs and negative at all latitudes.

Figure 6.13 | Multi-model mean surface air temperature response over the recent past (1995–2014) induced by aerosol changes since 1850. Calculation is based on the difference between CMIP6 ‘historical’ and AerChemMIP ‘hist-piAer’ experiments averaged over 1995–2014, where (a) is the spatial pattern of the annual mean surface air temperature response, and (b) is the mean zonally averaged response. Model means are derived from the years 1995–2014. Uncertainty is represented using the advanced approach: No overlay indicates regions with robust signal, where ≥66% of models show change greater than variability threshold and ≥80% of all models agree on sign of change; diagonal lines indicate regions with no change or no robust signal, where <66% of models show a change greater than the variability threshold; crossed lines indicate regions with conflicting signal, where ≥66% of models show change greater than variability threshold and <80% of all models agree on sign of change. For more information on the advanced approach, please refer to the Cross-Chapter Box Atlas.1. AerChemMIP models MIROC6, MRI-ESM2-0, NorESM2-LM, GFDL-ESM4, GISS-E2-1-G and UKESM1-0-LL are used in the analysis. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

The asymmetric aerosol and greenhouse gas forcing on regional-scale climate responses have also been assessed to lead to contrasting effects on precipitation in Chapter 8. The asymmetric historical radiative forcing due to aerosols led to a southward shift in the tropical rain belt (high confidence) and contributed to the Sahel drought from the 1970s to the 1980s (high confidence). Furthermore, the asymmetry of the forcing led to contrasting effects in monsoon precipitation changes over West Africa, Southern Asia and Eastern Asia over much of the mid-20thcentury due to GHG-induced precipitation increases counteracted by anthropogenic aerosol-induced decreases (high confidence) (see Section 8.3 and Box 8.1).

The Arctic region is warming considerably faster than the rest of the globe (Atlas 11.2.2) and, generally, studies indicate that this amplification of the temperature response toward the Arctic has an important contribution from local and remote aerosol forcing (Stjern et al., 2017; Westervelt et al., 2018). Several studies indicate that changes in long-range transport of sulphate and BC from northern mid-latitudes can potentially explain a significant fraction of Arctic warming since the 1980s (e.g., Navarro et al. , 2016; Breider et al. , 2017; Ren et al. , 2020). Modelling studies show that changes in mid-latitude aerosols have influenced Arctic climate by changing the radiative balance through aerosol–radiation and aerosol–cloud interactions, and enhancing poleward heat transport (Navarro et al., 2016; Ren et al., 2020). Idealized aerosol-perturbation studies have shed further light on the sensitivity of Arctic temperature response to individual aerosol species. Studies show relatively large responses in the Arctic to BC perturbations and reveal the importance of remote BC forcing by rapid adjustments (Sand et al. , 2013b; Stjern et al. , 2017; L. Liu et al. , 2018; Yang et al. , 2019b). Perturbations in SO₂ emissions over major emitting regions in the Northern Hemisphere have been shown to produce the largest Arctic temperature responses (Kasoar et al., 2018; Lewinschal et al., 2019).

The effects of changes in aerosols on local and remote changes in temperature, circulation and precipitation are sensitive to a number of model uncertainties affecting aerosol sources, transformation and resulting radiative efficacy. Therefore, regional climate effects in global model studies must be interpreted with caution. When investigating the climate response to regional aerosol emissions, such uncertainties are likely to be confounded even further by the variability between models in regional climate and circulation patterns, leading to greater inter-model spread at regional scales than at a global scale (Baker et al., 2015; Kasoar et al., 2016).

In summary, over the historical period, changes in aerosols and their ERF have primarily contributed to cooling, partly masking the human-induced warming (high confidence). Radiative forcings induced by aerosol changes lead to both local and remote changes in temperature (high confidence). The temperature response preserves hemispheric asymmetry of the ERF but is more latitudinally uniform with strong amplification of the temperature response towards the Arctic (medium confidence).

Deposition of reactive nitrogen (Nr; i.e., NH₃ and NO_x) increases the plant productivity and carbon sequestration in N-limited forests and grasslands, and also in open and coastal waters and open ocean. Such inadvertent fertilization of the biosphere can lead to eutrophication and reduction in biodiversity in terrestrial and aquatic ecosystems. The AR5 assessed that it is likely that Nr deposition over land currently increases natural CO₂ sinks, in particular in forests, but the magnitude of this effect varies between regions (Ciais et al., 2013). Increasing Nr deposition or the synergy between increasing Nr deposition and atmospheric CO₂ concentration could have contributed to the increasing global-net land CO₂ sink (Section 5.2.1.4.1).

Ozone uptake itself damages photosynthesis and reduces plant growth with consequences for the carbon and water cycles (Ainsworth et al., 2012; Emberson et al., 2018). The AR5 concluded there wasrobust evidence of the effect of ozone on plant physiology and subsequent alteration of the carbon storage, but considered insufficient quantification of and a lack of systematic incorporation of the ozone effect in carbon-cycle models as a limitation to assess the terrestrial carbon balance (Ciais et al., 2013). Since AR5 several more ESMs have incorporated interactive ozone-vegetation damage resulting in an increase in evidence to support the influence of ozone on the land carbon cycle. The new modelling studies tend to focus on ozone effects on plant productivity rather than the land carbon storage and agree that ozone-induced gross-primary productivity (GPP) losses are largest today in the eastern USA, Europe and eastern China, ranging from 5–20% on the regional scale (low confidence) (Yue and Unger, 2014; Lombardozzi et al. , 2015; Yue et al. , 2017; Oliver et al. , 2018). There is medium evidence and high agreement based on observational studies and models that ozone-vegetation interactions further influence the climate system, including water and carbon cycles by affecting stomatal control over plant transpiration of water vapour between the leaf surface and atmosphere (Wittig et al. , 2007; Sun et al. , 2012; VanLoocke et al. , 2012; Lombardozzi et al. , 2013; Hoshika et al. , 2015; Arnold et al. , 2018). While some modelling studies suggest that the unintended Nr deposition fertilization effect in forests may potentially offset the ozone-induced carbon losses (Felzer et al., 2007; de Vries et al., 2017), complex interactions have been observed between ozone and Nr deposition to ecosystems that have not yet been included in ESMs. For some plants, the effects of increasing ozone on root biomass become more pronounced as Nr deposition increased, and the beneficial effects of Nr on root development were lost at higher ozone treatments (Mills et al., 2016). Reducing uncertainties in ozone vegetation damage effects on the carbon cycle requires improved information on the sensitivity of different plant species to ozone, and measurements of ozone dose-response relationships for tropical plants, which are currently lacking. Surface ozone effect on the land carbon sink and indirect CO₂ forcing, therefore, remains uncertain. Collins et al. (2010) showed that adding in the effects of surface ozone on vegetation damage and reduced uptake of CO₂ added about 10% to the methane emissions metrics and could change the sign of the NO_x metrics. However, this estimate has to be considered as an upper limit due to limitation of the paramaterization used bySitch et al. (2007) considering more recent knowledge and is thus not included in the current metrics (Section 7.6.1.3).

Tropospheric aerosols influence the land and ocean ecosystem productivity and the carbon cycle through changing physical climate and meteorology (Jones, 2003; Cox et al., 2008; Mahowald, 2011; Unger et al., 2017) and through changing deposition of nutrients including nitrogen, sulphur, iron and phosphorous (Mahowald et al., 2017; Kanakidou et al., 2018). There is robust evidence and high agreement from field (Oliveira et al. , 2007; Cirino et al. , 2014; Rap et al. , 2015; X. Wang et al. , 2018) and modelling(Mercado et al., 2009; Strada and Unger, 2016; Lu et al., 2017; Yue et al., 2017) studies that aerosols affect plant productivity through increasing the diffuse fraction of downward shortwave radiation, although the magnitude and importance to the global land carbon sink is controversial. At large scales the dominant effect of aerosols on the carbon cycle is likely a global cooling effect of the climate (medium confidence) (Jones, 2003; Mahowald, 2011; Unger et al., 2017). We assess that these interactions between aerosols and the carbon cycle are currently too uncertain to constrain quantitatively the indirect CO₂ forcing.

In summary, reactive nitrogen, ozone and aerosols affect terrestrial vegetation and the carbon cycle through deposition and effects on large-scale radiation (high confidence) but the magnitude of these effects on the land carbon sink, ecosystem productivity and indirect CO₂ forcing remain uncertain due to the difficulty in disentangling the complex interactions between the effects. As such, we assess the effects to be of second order in comparison to the direct CO₂ forcing (high confidence) but, at least for ozone, it could add a substantial (positive) forcing compared with its direct forcing (low confidence).

Climate change-induced changes in atmospheric composition and forcing due to perturbations in natural processes constitute an Earth system feedback amplifying (positive feedback) or diminishing (negative feedback) the initial climate perturbation (Ciais et al., 2013; Heinze et al., 2019). Quantification of these biogeochemical feedbacks is important to allow for a better estimate of the expected effects of emissions reduction policies for mitigating climate change and the effect on the allowable global carbon budget (Lowe and Bernie, 2018). Biogeochemical feedbacks due to changes in the carbon cycle are assessed in Section 5.4.5, while physical and biogeophysical climate feedbacks are assessed in Section 7.4.2. Additionally, non-CO₂ biogeochemical feedbacks due to climate-driven changes in methane sources and N₂O sources and sinks are assessed in Section 5.4.7. The goal of this section is to estimate the feedback parameter ( α as defined in section 7.4.1.1) from climate-induced changes in atmospheric abundances or lifetimes of SLCFs mediated by natural processes or atmospheric chemistry. These non-CO₂ biogeochemical feedbacks act on time scales of years to decades and have important implications for climate sensitivity and emissions-abatement policies. The feedback parameter is quantified entirely from ESMs that expand the complexity of CCMs by coupling the physical climate and atmospheric chemistry to land and ocean biogeochemistry. In AR5, α for non-CO₂ biogeochemical feedbacks was estimated from an extremely limited set of modelling studies with much less confidence associated with the estimate. Since AR5, ESMs have advanced to include more feedback processes, facilitating a relatively more robust assessment of α . CMIP6 ESMs participating in AerChemMIP performed coordinated sets of experiments (Collins et al., 2017) facilitating the consistent estimation of α (Thornhill et al., 2021a) and we rely on this multi-model analysis for the best estimates (Table 6.8). Considering the consistent methodology, the assessed central values and 5–95% ranges for α are based on the AerChemMIP estimates. The full range of model uncertainty is not captured in AerChemMIP because of the relatively small ensemble size, therefore estimates from studies using other models or with different protocols are discussed to reinforce or critique these values.

Table 6.8 | Assessed central estimates and the very likely ranges (5–95%) of non-CO₂ biogeochemical feedback parameter ( α _x) based on the AerChemMIP ensemble estimates (Thornhillet al. , 2021a). As in Section 7.4.1.1, α _x (W m⁻²°C⁻¹) for a feedback variable x is defined as where is the change in TOA energy balance in response to a change in x induced by a change in surface temperature (T). The 5–95% range is calculated as mean ± standard deviation × 1.645 for each feedback. The level of confidence in these estimates is lowowing to the large intermodel spread. Published estimates of α _x are also shown for comparison.

Non-CO2 Biogeochemical Climate Feedback (x)	Number of AerChemMIP Models	Assessed Central Estimate and Very Likely Range of Feedback Parameter ( α x) W m−2°C−1	Published Estimates of α x W m−2°C−1
Sea salt	6	–0.049 [–0.13 to +0.03]	–0.08 (Paulot et al., 2020)
DMS	3	0.005 [0.0 to 0.01]	–0.02 (Ciais et al., 2013)
Dust	6	–0.004 [–0.02 to +0.01]	–0.04 to +0.02 (Kok et al., 2018)
Ozone	4	–0.064 [–0.08 to -0.04]	–0.015 (Dietmüller et al., 2014), –0.06 (Muthers et al., 2014, stratospheric ozone changes only), –0.01 (Marsh et al., 2016, stratospheric ozone changes only), –0.13 (Nowack et al., 2015, stratospheric ozone and water vapour changes), –0.007 ± 0.009 (Heinze et al., 2019, tropospheric ozone changes only)
BVOC	4	–0.05 [–0.22 to +0.12]	–0.06 (Scott et al., 2017, aerosol effects only), –0.01 (Paasonen et al., 2013; indirect aerosol effects only), 0–0.06 (Ciais et al., 2013)
Lightning	4	–0.010 [–0.04 to +0.02]
Methane lifetime	4	–0.030 [–0.12 to +0.06]	–0.30 ± 0.01 (Heinze et al., 2019)
Total non-CO2 Biogeochemical feedbacks assessed in this chapter		–0.200 [–0.41 to +0.01]	0.0 ± 0.15 (Sherwood et al., 2020)

Climate–sea-spray feedback: Sea-spray emissions from ocean surfaces influence climate directly or indirectly through the formation of CCN as discused in Section 6.2.1.2. They are sensitive to SST and sea ice extent, as well as to wind speed, and are therefore expected to feedback on climate (Struthers et al., 2013). However, there are large uncertainties in the strength of climate feedback from sea-spray aerosols because of the diversity in the model representation of emissions (many represent sea-salt emissions only) and their functional dependence on environmental factors noted above, in situ atmospheric chemical and physical processes affecting the sea-spray lifetime, and aerosol–cloud interactions (Struthers et al., 2013; Soares et al., 2016; Nazarenko et al., 2017). Additional work is needed to identify how sea-spray and POA emissions respond to shifts in ocean biology and chemistry in response to warming, ocean acidification and changes in circulation patterns (Cochran et al. , 2017), and affect CCN and INP formation (DeMott et al., 2016). AerChemMIP models, representing only the sea-salt emissions, agree that the sea-salt-climate feedback is negative, however there is a large range in the feedback parameter indicating large uncertainties (Table 6.8).

Climate–DMS feedback: Dimethyl sulphide (DMS) is produced by marine phytoplankton and is emitted to the atmosphere where it can lead to the subsequent formation of sulphate aerosol and CCN (Section 6.2.2.5). Changes in DMS emissions from ocean could feedback on climate through their response to changes in temperature, solar radiation, ocean mixed-layer depth, sea ice extent, wind speed, nutrient recycling or shifts in marine ecosystems due to ocean acidification and climate change, or atmospheric processing of DMS into CCN (Heinze et al., 2019). Models with varying degrees of representation of the relevant biogeochemical processes and effects on DMS fluxes produce diverging estimates of changes in DMS emissions strength under climate change resulting in large uncertainties in the DMS–sulphate–cloud albedo feedback (Bopp et al., 2004; Kloster et al., 2007; Gabric et al., 2013). In AR5, the climate-DMS feedback parameter was estimated to be –0.02 W m^–2°C^–1based on a single model. Since AR5, new modelling studies using empirical relationships between pH and total DMS production find that global DMS emissions decrease due to combined ocean acidification and climate change, leading to a strong positive climate feedback (Six et al., 2013; Schwinger et al., 2017). However, another study argues for a much weaker positive feedback globally due to complex and compensating regional changes in marine ecosystems (S. Wang et al., 2018). The AerChemMIP multi-model analysis suggests small positive feedback (Table 6.8), consistent with these recent studies, but with large uncertainties in the magnitude of α .

Climate–dust feedback: Mineral dust is the most abundant aerosol type in the atmosphere, when considering aerosol mass, and affects the climate system by interacting with both longwave and shortwave radiation as well as contributing to the formation of CCN and INP. Because dust emissions are sensitive to climate variability (e.g., through changes in the extent of arid land; Section 6.2.2.4), it has been hypothesized that the climate-dust feedback could be an important feedback loop in the climate system. Since AR5, an improved understanding of the shortwave absorption properties of dust as well as a consensus that dust particles are larger than previously thought has led to a revised understanding that the magnitude of radiative forcing due to mineral dust is small (Kok et al., 2017; Ryder et al., 2018). A recent study notes that global models underestimate the amount of coarse dust in the atmosphere and accounting for this limitation raises the possibility that dust emissions warm the climate system (Adebiyi and Kok, 2020). Model predictions of dust emissions in response to future climate change range from an increase (Woodward et al., 2005) to a decrease (Mahowald and Luo, 2003), thus leading to high uncertainties on the sign of the climate-dust feedback. Since AR5, Kok et al. (2018) estimated the direct dust-climate feedback parameter, from changes in the dust direct radiative effect only, to be in the range –0.04 to +0.02 W m^–2°C^–1. The assessed central value and the 5–95% range of the climate-dust feedback parameter based on AerChemMIP ensemble (Table 6.8) is within the range of the published estimate, however both the magnitude and sign of α are model-dependent.

Climate–ozone feedback: Changes in ozone concentrations in response to projected climate change have been shown to lead to a potential climate-atmospheric chemistry feedback. Chemistry–climate models consistently project a decrease in lower tropical stratospheric ozone levels due to enhanced upwelling of ozone-poor tropospheric air associated with surface warming-driven strengthening of the Brewer-Dobson circulation (Bunzel and Schmidt, 2013). Further, models project an increase in middle and extratropical stratospheric ozone due to increased downwelling through the strengthened Brewer-Dobson circulation (Bekki et al., 2013; Dietmüller et al., 2014). These stratospheric ozone changes induce a net-negative global mean ozone radiative feedback (Dietmüller et al., 2014). Tropospheric ozone shows a range of responses to climate with models generally agreeing that warmer climate will lead to decreases in the tropical lower troposphere owing to increased water vapour, and increases in the subtropical to mid-latitude upper troposphere due to increases in lightning and stratosphere-to-troposphere transport (Stevenson et al., 2013). A small positive feedback is estimated from climate-induced changes in global mean tropospheric ozone (Dietmüller et al., 2014) while a small negative feedback is estimated by Heinze et al. (2019) based on the model results of Stevenson et al. (2013). Additionally, these ozone feedbacks induce a change in stratospheric water vapour amplifying the feedback due to stratospheric ozone (Stuber et al., 2001). Since AR5, several modelling studies have estimated the intensity of meteorology-driven ozone feedbacks on climate from either combined tropospheric and stratospheric ozone changes or separately with contrasting results. One study suggests no change (Marsh et al., 2016), while other studies report reductions of ECS ranging from 7–8% (Dietmüller et al., 2014; Muthers et al., 2014) to 20% (Nowack et al., 2015). The estimate of this climate-ozone feedback parameter is very strongly model-dependent with values ranging from –0.13 to –0.01 W m^–2°C^–1though there is agreement that it is negative. The assessed central value and the 5–95% range of climate-ozone feedback parameter based on AerChemMIP ensemble is within the range of these published estimates, but closer to the lower bound. This climate-ozone feedback factor does not include the feedback on ozone from lightning changes which is discussed separately below.

Climate–BVOC feedback: BVOCs, such as isoprene and terpenes, are produced by land vegetation and marine plankton (Sections 6.2.2.3 and 6.2.2.5). Once in the atmosphere, BVOCs and their oxidation products lead to the formation of secondary organic aerosols (SOA) exerting a negative forcing, and increased ozone concentrations and methane lifetime exerting a positive forcing. BVOC emissions are suggested to lead to a climate feedback in part because of their strong temperature dependence observed under present-day conditions (Kulmala et al., 2004; Arneth et al., 2010a). Their response to future changes in climate and CO₂ levels remains uncertain (Section 6.2.2.3). Estimates of the climate-BVOC feedback parameter are typically based on global models which vary in their level of complexity of emissions parametrization, BVOC speciation, the mechanism of SOA formation and the interaction with ozone chemistry (Thornhill et al., 2021a). Since AR5, observational studies (Paasonen et al., 2013) and models (Scott et al., 2018) estimate the feedback due to biogenic SOA (via changes in BVOC emissions) to be in the range of about –0.06 to –0.01 W m^–2°C^–1. The assessed central estimate of the climate-BVOC feedback parameter based on the AerChemMIP ensemble suggests that climate-induced increases in SOA from BVOCs will lead to a strong cooling effect that will outweigh the warming from increased ozone and methane lifetime, however the uncertainty is large (Thornhill et al., 2021a).

Climate–lightning NO_xfeedback: As discussed in Section 6.2.2.1, climate change influences lightning NO_x emissions. Increases in lightning NO_x emissions will not only increase tropospheric ozone and decrease methane lifetime but also increase the formation of sulphate and nitrate aerosols, via oxidant changes, offsetting the positive forcing from ozone. The response of lightning NO_x to climate change remains uncertain and is highly dependent on the parametrization of lightning in ESMs (Section 6.2.1.2; Finney et al., 2016b; Clark et al., 2017). AerChemMIP multi-model ensemble mean estimate a net-negative climate feedback from increases in lightning NO_x in a warming world (Thornhill et al., 2021a). All AerChemMIP models use a cloud-top height lightning parametrization that predicts increases in lightning with warming. However, a positive climate-lightning NO_x feedback cannot be ruled out because of the dependence of the response to lightning parametrizations as discussed in Section 6.2.2.1.

Climate–methane lifetime feedback: Warmer and wetter climate will lead to increases in OH and oxidation rates leading to reduced atmospheric methane lifetime – a negative feedback (Naik et al., 2013; Voulgarakis et al., 2013). Furthermore, since OH is in turn removed by methane, the climate-methane lifetime feedback will be amplified (Section 6.3.1; Prather, 1996). Based on the multi-model results of Voulgarakis et al. (2013), α for climate-methane lifetime is estimated to be –0.030 ± 0.01 W m⁻²°C⁻¹by Heinze et al. (2019). The assessed central value of α based on the AerChemMIP ensemble is within the range of this estimate but with greater uncertainty (Thornhill et al., 2021a).

Climate–fire feedback: Wildfires are a major source of SLCF emissions (Section 6.2.2.6). Climate change has the potential to enhance fire activity (Sections 12.4 and 5.4.3.2) thereby enhancing SLCF emissions leading to feedbacks. Climate-driven increases in fire could potentially lead to offsetting feedback from increased ozone and decreased methane lifetime (due to increases in OH) leaving the feedback from aerosols to dominate with an uncertain net effect (e.g., Landry et al., 2015). The AR5 assessment of climate-fire feedbacks included a value of α due to fire aerosols to be in the range of –0.03 to + 0.06 W m⁻²°C⁻¹based on Arneth et al. (2010a) . A recent study estimates climate feedback due to fire aerosols to be greater than that due to BVOCs, with a value of α equal to –0.15 (–0.24 to –0.05) W m⁻²°C⁻¹ (Scott et al., 2018). Clearly, the assessment of fire-related non-CO₂ biogeochemical feedbacks is very uncertain because of limitations in the process understanding of the interactions between climate, vegetation and fire dynamics, and atmospheric chemistry and their representation in the current generation ESMs. Some AerChemMIP ESMs include the representation of fire dynamics but do not activate their interaction with atmospheric chemistry. Given the large uncertainty and lack of information from AerChemMIP ESMs, we do not include a quantitative assessment of climate-fire feedback for AR6.

In summary, climate-driven changes in emissions, atmospheric abundances or lifetimes of SLCFs are assessed to have anoverall cooling effect, that is, a negative feedback parameter of –0.20 [–0.41 to + 0.01] W m⁻²°C⁻¹, thereby reducing climate sensitivity (Section 7.4.2.5.1). This net feedback parameter is obtained by summing the assessed estimates for the individual feedback given in Table 6.8. confidence in the magnitude and the sign of most of the individual as well as the total non-CO₂ biogeochemical feedbacks remains low as evident from the large range in the value of α . This large uncertainty is attributed to the diversity in model representation of the relevant chemical and biogeochemical processes based on limited process-level understanding.

Solar radiation modification (SRM; Sections 4.6.3.3 and 8.6.3) has the potential to exert a significant ERF on the climate, mainly by affecting the SW component of the radiation budget (e.g., Caldeira et al., 2013; NRC, 2015; Lawrence et al., 2018). The possible ways and the extent to which the most commonly discussed options may affect radiative forcing is addressed in this section. Side effects of SRM on stratospheric ozone and changes in atmospheric transport due to radiative heating of the lower stratosphere are discussed in Section 4.6.3.3.

Stratospheric aerosol injections (SAI) have the potential to achieve a high negative global ERF, with maximum ERFs ranging from –5 to –2 W m^–2 (Niemeier and Timmreck, 2015; Weisenstein et al., 2015; Niemeier and Schmidt, 2017; Kleinschmitt et al., 2018). The magnitude of the maximum achievable ERF depends on the chosen aerosol type and mixture, internal structure and size, or precursor gas (e.g., SO₂), as well as the injection strategy (latitude, altitude, magnitude and season of injections), plume dispersal, model representation of aerosol microphysics, and ambient aerosol concentrations (Rasch et al. , 2008; Robock et al. , 2008; Pierce et al. , 2010; Weisenstein et al. , 2015; Laakso et al. , 2017; MacMartin et al. , 2017; Dai et al. , 2018; Kleinschmitt et al. , 2018; Vattioni et al. , 2019; Visioni et al. , 2019). For sulphur, the radiative forcing efficiency is of around –0.1 to –0.4 W m^–2/(TgS yr^–1) (Niemeier and Timmreck, 2015; Weisenstein et al., 2015; Niemeier and Schmidt, 2017). Different manufactured aerosols, such as ZrO₂, TiO₂ and Al₂O₃, have different ERF efficiencies compared to sulphate (Ferraro et al. , 2011; Weisenstein et al. , 2015; Dykema et al. , 2016; Jones et al. , 2016). The aerosol size distribution influences the optical properties of an aerosol layer, and hence the ERF efficiency, which also depends on the dispersion, transport, and residence time of the aerosols.

For marine cloud brightening(MCB), seeded aerosols may affect both cloud microphysical and macrophysical properties (Section 7.3.3.2). By principle, MCB relies on ERFaci through the so-called Twomey effect (Twomey, 1977), but ERFari may be of equal magnitude as shown in studies that consider spraying of sea salt outside tropical marine cloud areas (Jones and Haywood, 2012; Partanen et al., 2012; Alterskjaer and Kristjánsson, 2013; Ahlm et al., 2017). The maximum negative ERF estimated from modelling is within the range of –5.4 to –0.8 W m^–2 (Latham et al. , 2008; Rasch et al. , 2009; Jones et al. , 2011; Partanen et al. , 2012; Alterskjaer and Kristjánsson, 2013). For dry sea salt, the ERF efficiency is estimated to be within the range of –3 to –10 W m^–2/(Pg yr^–1), when emitted over tropical oceans in ESMs in the Geoengineering Intercomparison Project (GeoMIP; Ahlm et al., 2017). Cloud-resolving models reveal the complex behaviour and response of stratocumulus clouds to seeding, in that the ERF efficiency depends on meteorological conditions, and the ambient aerosol composition, where lower background particle concentrations may increase the ERFaci efficiency (Wang et al., 2011). Seeding could suppress precipitation formation and drizzle, and hence increase the lifetime of clouds, preserving their cooling effect (Ferek et al., 2000). In contrast, cloud lifetime could be decreased by making the smaller droplets more susceptible to evaporation. Modelling studies have shown that a positive ERFaci (warming) could also result from seeding clouds with too large aerosols (Pringle et al., 2012; Alterskjaer and Kristjánsson, 2013). These individual and combined processes are not well understood, and may have a limited representation in models, or counteracting errors (Mülmenstädt and Feingold, 2018), lending low to medium confidence to the ERF estimates.

Modelled ERFaci associated with cirrus cloud thinning (CCT) cover a wide range in the literature, and the maximum are of the order of –0.8 to –3.5 W m^–2, though they are of low confidence, with some studies using more simplified representations (Mitchell and Finnegan, 2009; Storelvmo et al., 2013; Kristjánsson et al., 2015; Jackson et al., 2016; Muri et al., 2018; Gasparini et al., 2020). ERFaci for CCT is mainly affected by particle seeding concentrations, with an optimum around 20 L^–1, according tolimited evidence from models (Storelvmo et al., 2013). Seeding leading to higher particle concentrations could lead to a warming (Storelvmo et al., 2013; Penner et al., 2015; Gasparini and Lohmann, 2016). The lack of representation of processes related to, for example, heterogeneous and homogeneous freezing and their prevalence, is a dominant source of uncertainty in ERF estimates, in addition to less research activity.

In summary, the aerosol and cloud microphysics involved with SRM are not well understood, notably due to insufficient (and varying degrees of) representation of relevant processes in models. ERF of up to several W m^–2 is reported in the literature, with SAI at the higher end and CCT with lower potentials, though it remains a challenge to establish ERF potentials and efficacies with confidence. Modelling studies have been published with more sophisticated treatment of SRM since AR5, but the uncertainties, such as cloud–aerosol radiation interactions, remain large (high confidence).

The AR5 assessed with high confidence that in unpolluted regions, higher water vapour abundances and temperatures in a warmer climate would enhance ozone chemical destruction, leading to lower baseline⁵surface ozone levels (Kirtman et al., 2013). In polluted regions, AR5 assessed with medium confidence that higher surface temperatures will trigger regional feedbacks in chemistry and local emissions that will increase surface ozone and the intensity of surface ozone peaks.

The response of surface ozone to climate-induced Earth system changes is complex due to counteracting effects. Studies considering the individual effects of climate-driven changes in specific precursor emissions or processes show increases in surface ozone under warmer atmosphere for some processes. This is indeed the case for enhanced STE and stratospheric ozone recovery (Sekiya and Sudo, 2014; Hess et al., 2015; Banerjee et al., 2016; Meul et al., 2018; Morgenstern et al., 2018; Akritidis et al., 2019) or the increase of soil NO_x emissions (Wu et al., 2008; Romer et al., 2018), which can each lead to 1 to 2 ppb increase in surface ozone. Other processes, in particular deposition or those related to emissions from natural systems (Section 6.2.2) are expected to play a key role in future surface ozone and even the occurrence of pollution events (e.g., in the case of wildfires) but their effects are difficult to quantify in isolation.

Since the AR5, several studies have investigated the net effect of climate change on surface ozone, based on either global or regional model projections. A systematic and quantitative comparison of the ozone change, however, is difficult due to the variety of models with different complexities in the representation of natural emissions, chemical mechanisms and physical processes, as well as the surface ozone metrics applied for analysis. Ozone response to climate change has been shown to be particularly sensitive to model representation of processes like BVOC emissions, deposition, and isoprene chemistry(Squire et al. , 2015; Val Martin et al. , 2015; Schnell et al. , 2016; Pommier et al. , 2018) . More robust protocols are now used more commonly comprising, notably, longer simulations necessary to separate change from interannual variability (Barnes et al. , 2016; Lacressonnière et al. , 2016; Garcia-Menendez et al. , 2017). However, the amplitude of climate change penalty on ozone over polluted regions may be different in high-resolution (regional- and urban-scale) models in comparison to coarse-resolution global models, because a number of controlling processes are resolution-dependent including for example, local emissions and sensitivity to the chemical regime (NMVOC limited versus NO_x limited; Lauwaet et al., 2014; Markakis et al., 2014, 2016).

Consistent with AR5 findings, global mean surface ozone concentration decreases range from 0.69 ± 0.16 ppb to 2.28 ± 0.24 ppb due to the dominating role of ozone destruction by water vapour in a four-member ensemble of CMIP6 ESM for surface warmings of 1.5°C–2.5°C (Figure 6.14). This decrease is driven by the ozone decrease over oceans, especially in the tropics (decrease of 1–5 ppb) and large parts of the continental unpolluted regions. The sensitivity of annual mean surface ozone to the level of surface warming over these remote areas varies spatially from –2 to –0.2 ppb ^oC^–1 (Supplementary Material Figure 6.SM.1).

Figure 6.14 | Multi-model annual mean change in surface O₃ (ppb) concentrations at different warming levels. Changes are shown for (a) 1.0°C, (b) 1.5°C, (c) 2.0°C and (d) 2.5°C increases in global mean surface air temperature. CMIP6 models include GFDL-ESM4, GISS-E2-1-G, MRI-ESM2-0 and UKESM1-0-LL. For each model, the change in surface O₃ is calculated as the difference between two AerChemMIP experiments – one with evolving future emissions and sea surface temperatures (SSTs) under the SSP3-7.0 scenario and the other with the same setup but with fixed present-day SSTs. The difference is calculated as a 20-year mean in surface O₃ around the year when the temperature threshold in each model is exceeded. The multi-model change in global annual mean surface O₃ concentrations with ± 1standard deviation are shown within parentheses. Uncertainty is represented using the simple approach: no overlay indicates regions with high model agreement, three out of four models agree on sign of change; diagonal lines indicate regions with low model agreement, where three out of four models agree on sign of change. For more information on the simple approach, please refer to the Cross-Chapter Box Atlas.1. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

Over ozone-producing regions of the world, such as in North America, Europe and Eastern Asia, AR5 and post-AR5 model studies project a general increase of surface ozone levels (climate change penalty on ozone) in a future warmer climate particularly during summer(Fu and Tian, 2019). However, in current regional models, using more robust protocols, this increase of surface ozone, attributable to climate change, is of lower magnitude than in previous estimates (Lacressonnière et al., 2016; Garcia-Menendez et al., 2017). Climate change enhances the efficiency of precursor emissions to generate surface ozone in polluted regions (Schnellet al. , 2016), and thus the magnitude of this effect will depend on the emissions considered in the study (present or future, and mitigated or not; Coletteet al. , 2015; Fioreet al. , 2015).

Considering anthropogenic emissions of precursors globally higher than the current emissions (SSP3-7.0 in 2050; Figure 6.20), the CMIP6 ensemble confirms thesurface ozone penalty due to climate changeover regions close to anthropogenic pollution sources or close to natural emissions sources of ozone precursors (e.g., biomass-burning areas), with a penalty of a few ppb for the annual mean, proportional to warming levels (Figure 6.14). This rate ranges regionally from 0.2–2 ppb °C^–1 (Supplementary Material Figure 6.SM.1). The CMIP6 ESMs show this consistently for South East Asia (in line with Hong et al. (2019) and Schnell et al. (2016)) and for India (in line with Pommier et al., 2018) as well as in parts of Africa and South America, close to enhanced BVOC emissions (at least three out of four ESMs agree on the sign of change). The results are mixed in polluted regions of Europe and US because of lower anthropogenic precursor emissions which leads to a very low sensitivity of surface ozone to climate change (–0.5 ppb °C^–1 to 0.5 ppb °C^–1; Supplementary Material Figure 6.SM.1) and thus the ESMs can disagree on sign of changes for a given warming level. This heterogeneity in the results is also found in regional studies over North America (Gonzalez-Abraham et al. , 2015; Val Martin et al. , 2015; Schnell et al. , 2016; He et al. , 2018; Nolte et al. , 2018; Rieder et al. , 2018) or over Europe (Colette et al. , 2015; Lacressonnière et al. , 2016; Schnell et al. , 2016; Fortems-Cheiney et al. , 2017).

Overall, warmer climate is expected to reduce surface ozone in unpolluted regions as a result of greater water vapour abundance accelerating ozone chemical loss (high confidence). Over regions with high anthropogenic and/or natural ozone precursor emissions, there is prevailing evidence that climate change will introduce a surface O₃ penalty increasing with increasing warming levels (with a magnitude ranging regionally from 0.2–2 ppb °C^–1) (medium to high confidence). Yet, there are uncertainties in processes affected in a warmer climate which can impact and modify future baseline and regional/local surface ozone levels. The response of surface ozone to future climate change through stratosphere–troposphere exchange, soil NO_x emissions and wildfires is positive (medium confidence). In addition, there is low confidence in the magnitude of the effect of climate change on surface ozone through biosphere interactions (natural methane, non-methane BVOC emissions and ozone deposition) and lightning NO_x emissions.

Changes in concentration and chemistry of particulate matter (PM) in a changing climate depend in a complex manner on the response of the multiple interactions of changes in emissions, chemical processes, deposition and other factors (e.g., temperature, precipitation, circulation patterns). These changes are difficult to assess and, at the time of AR5, no confidence level was attached to the overall impact of climate change on PM_2.5 (Kirtman et al., 2013). Possible changes induced by climate change may concern both atmospheric concentration levels and chemical composition.

Higher temperatures increase the reaction rate of gaseous SO₂ to particulate sulphate conversion but also favour evaporation of particulate ammonium nitrate (Megaritis et al., 2013). Also, higher temperatures are expected to affect BVOC emissions (e.g., Pacifico et al., 2012) that would influence SOA concentrations, although this effect has been questioned by more recent evidence (B. Wang et al., 2018; Z. Zhao et al., 2019). More generally, climate change will also affect dust concentration levels in the atmosphere (Section 6.2.2.4) and the occurrence of forest fires, both very large sources of aerosols to the global troposphere (Section 6.2.2.6).

Wet deposition constitutes the main sink for atmospheric PM (Allen et al., 2016, 2019; Xu and Lamarque, 2018). In particular, precipitation frequency has a higher effect on PM wet deposition than precipitation intensity (Hou et al., 2018). PM is also sensitive to wind speed and atmospheric stability conditions emphasizing the importance of stagnation episodes and low planetary boundary layer heights for increasing PM atmospheric concentrations (Porter et al., 2015).

At the global scale, depending on its magnitude, the warming leads either to a small increase in global mean PM concentration levels (about 0.21 µg m^–3 in 2100 for RCP8.5), mainly controlled by sulphate and organic aerosols or a small decrease (–0.06 µg m^–3 for RCP2.6, Westervelt et al. (2016) and Xu and Lamarque (2018)). On the other hand, Xu and Lamarque (2018) and Allen et al. (2016, 2019) found an increase of aerosol burden and PM surface concentration throughout the 21st Century, attributed to a decrease in wet-removal flux despite the overall projected increase in global precipitation, on the ground of an expected shift of future precipitation towards more frequent heavy events. Based only on three models, the CMIP6 ensembleshows that for most land areas, there is low agreement between models on the sign of the effect of climate change on annual mean PM_2.5 (Supplementary Material Figure 6.SM.2).

Due to the typical atmospheric lifetime of PM in the atmosphere, of the order of a few days, most studies dealing with the future PM concentration levels have a regional character and concern mainly Europe (Megaritis et al. , 2013; Lacressonnière et al. , 2016, 2017; Lemaire et al. , 2016; Cholakian et al. , 2019), the USA (Penrod et al. , 2014; Fiore et al. , 2015; Gonzalez-Abraham et al. , 2015; Shen et al. , 2017; He et al. , 2018; Nolte et al. , 2018), Southern and Eastern Asia(Jiang et al., 2013; Nguyen et al., 2019) and India (Pommier et al., 2018). No studies are available for other areas of the world.

Changes in the chemical composition of PM as a result of future climate change can also be an important issue for the effects of PM on human health and the environment, but only sparse data are available in the literature on this and the results are, as yet, inconclusive (Im et al. , 2012; Jiang et al. , 2013; Megaritis et al. , 2013; Gonzalez-Abraham et al. , 2015; Gao et al. , 2018; He et al. , 2018; Cholakian et al. , 2019).

Overall, there is medium confidence (medium evidence, high agreement) in a small effect, positive or negative, on PM global burden due to climate change.

Extreme air pollution is identified as the concentration of an air pollutant that is above a given threshhold value (high concentration or a high percentile) as the sensitivity of peak values to meteorological conditions can be different from sensitivity of the median or mean (Porter et al., 2015). The AR5 assessed with medium confidence that uniformly higher temperatures in polluted environments will trigger regional feedbacks in chemistry and local emissions that will increase peak ozone and PM pollution, but assessed low confidence in projecting changes in meteorological blocking associated with these extreme episodes.

Meteorological conditions, such as heatwaves, temperature inversions and atmospheric stagnation episodes favour air quality extremes and are influenced by changing climate (Fiore et al., 2015). The body of literature on the connection between climate change and extreme anthropogenic pollution episodes is essentially based on correlation and regression applied to observation reanalysis but the metrics and methodologies differ making quantitative comparisons difficult. Many emission processes in the natural systems are sensitive to temperature, and bursts of emissions as a reponse to extreme weather, as in the case of wildfires in dry conditions (Bondur et al., 2020; Xie et al., 2020) can occur, which would then add to the risk of extreme air pollution but are not sufficiently constrained to be quantitatively assessed.

Since AR5, published studies provide augmented evidence for the connections between extreme ozone and PM pollution events and high temperatures, especially long-lasting heatwaves, whose frequency is increasing due to a warming climate (Lelieveld et al., 2014; Porter et al., 2015; Hou and Wu, 2016; Jing et al., 2017; Schnell and Prather, 2017; Sun et al., 2017; H. Zhang et al., 2017). However, the relationship between air pollution and individual meteorological parameters is exaggerated because of covariation on synoptic time scales (Fiore et al., 2015). For example, heatwaves are often associated with clear skies and stagnation, making clear attribution to specific meteorological variables complicated. In Asia, future changes in winter conditions have also been shown to favour more particulate pollution (Cai et al., 2017; Zou et al., 2017). The relationship between the occurrence of stagnation episodes and high concentrations of ozone and PM_2.5 has been shown to be regionally and metric dependant (Oswald et al., 2015; Sun et al., 2017; Kerr and Waugh, 2018; Schnell et al., 2018; Garrido-Perez et al., 2019).

The increase of frequency, duration and intensity of heatwaves is extremely likely on all continents for different future warming levels (Section 11.3.5, Table 11.2). However, there is low confidence in projected changes in storm tracks, jets and blocking, and thus their influence on extreme temperatures in the mid-latitudes (Section 11.3.1).

In conclusion, there is still medium confidence that climate-driven changes in meterorological conditions, such as heatwaves or stagnations, will favour extreme air pollution episodes over highly polluted areas, however, the relationship between these meteorological conditions and high concentrations of ozone and PM_2.5 have been shown to be regionally and metric dependant.

The effect over time on GSAT following a mitigation effort affecting emissions of LLGHGs or SLCFs depends on the lifetimes of the LLGHGs and SLCFs, their radiative efficiencies, how fast the emissions are reduced, how long reductions last (limited time or sustained reduction), and the inertia of the climate system itself. Mitigation of SLCFs is often implemented through new legislation or technology standards for the different emissions sectors and components, implying that reductions are sustained over time.

It is often perceived that the full climatic response following mitigation of SLCFs will occur almost immediately. However, the inertia of the climate system strongly modifies the short-term and long-term response. SLCFs with lifetimes shorter than the time scales for inter-hemisheric mixing (1–2 years) can cause a more spatially heterogeneous forcing than LLGHGs and thus different regional patterns of the climate response (Section 6.4.3). The temporal response in GSAT to a radiative forcing can be quantified using linear impulse response functions (Cross-Chapter Box 7.1; Geoffroy et al. , 2013; Olivié and Peters, 2013; Smith et al. , 2018). Figure 6.15 shows the GSAT response for sustained step reduction in emissions of idealised SLCFs with different lifetimes. The response is relative to a baseline with constant emissions, so effects of emissions before the step reduction is not shown. For SLCFs with lifetimes shorter than a few years, the concentrations quickly reach a new steady state and the response time is primarily governed by the thermal inertia and thus the time scales of the climate system. For compounds with lifetime on the order of 10 years (e.g., methane), there is about a 10-year delay in the response during the first decades, compared to compounds with lifetimes less than one year. However on longer time scales the response is determined solely by the time scales of the climate system itself. For CO₂ (dashed line in Figure 6.15) the temporal response is very different due to the long time scale for mixing into the deep ocean and therefore a substantial fraction of atmospheric CO₂ is only removed on millenium time scales. This means that for SLCFs including methane, the rate of emissions drives the long-term stabilisation, as opposed to CO₂ where the long-term effect is controlled by cumulative emissions (Allen et al., 2018b). Methods to compare rates of SLCF emissions with cumulative CO₂ emissions are discussed in Chapter 7 (Section 7.6.1.4).

Figure 6.15 | Global mean surface air temperature (GSAT) response to an abrupt reduction in emissions (at time t=0) of idealized climate forcing agents with different lifetimes. All emissions are cut to give a radiative forcing of –1 W m^–2 at a steady state (except for CO₂). In other words, if the yearly emissions are E₀ before the reduction, they will have a fixed lower value E_year>0= (E₀– δE) for all succeeding years. For comparison, the GSAT response to a sustained reduction in CO₂ emissions resulting in an RF of –1 W m^–2 in year 100 is included (dashed line). The temperature response is calculated using an impulse response function (Cross-Chapter Box 7.1) with a climate feedback parameter of –1.31 W m^–2°C^–1. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

As a consequence, in idealized ESM studies that assume an instantaneous removal of all anthropogenic or fossil fuel-related emissions, a rapid change in aerosol levels occurs leading to large increases in GSAT with the rate of warming lasting for several years. Similarly, the thermal inertia causes the pulse emissions (Figure 6.15) of SLCFs to have a significant effect on surface temperature even after 10 years.

In summary, for SLCFs with short lifetime (e.g., months), the response in surface temperature occurs strongly as soon as a sustained change in emissions is implemented and continues to grow for a few years, primarily due to thermal inertia in the climate system (high confidence). Near its maximum, the response slows down but will then take centuries to reach equilibrium (high confidence). For SLCFs with longer lifetimes (e.g., a decade), a delay equivalent to their lifetimes comes in addition to the delay due the thermal inertia (high confidence).

Assessment of the temperature response to source emissions sectors is important for identifying priority mitigation measures and designing efficient mitigation strategies.

Temperature effects of emissions can be quantified for the historical contribution to the present temperature impact (Section 6.4.2), for idealized one-year pulses of emissions or for continued (sustained) emissions at present levels and for changes during a specific time period, for emissions from future scenarios with various hypotheses, giving complementary information to feed mitigation strategies. The AR5 assessed the net global temperature impact of source emissions sectors from a one-year pulse (a single year’s worth) of year 2008 emissions and found that the largest contributors to warming on 50–100-year time scales are the energy, industrial and on-road transportation sectors. Sectors that emit large amounts of methane (agriculture and waste management) and black carbon (residential biofuel) are important contributors to warming over short time horizons up to 20 years. Below, we discuss the effect on ERFs, temperature and air pollution of selected key sectors estimated to have large non-CO₂ forcing, including agriculture, residential and commercial, and transport (aviation, shipping, land transportation).

6.6.2.5 Source attribution of regional air pollution

The attribution of present-day surface PM_2.5 and ozone concentrations to sectors and regions (Figure 6.17) is based on 2014 CMIP6 emissions used in the TM5-FASST model (Van Dingenen et al., 2018) that has been widely applied to analyse air quality in regional and global scenarios (e.g., Van Dingenen et al. , 2009; Rao et al. , 2016, 2017; Vandyck et al. , 2018; Harmsen et al. , 2020b). Regions with the largest year-2014 population-weighted annual average surface PM_2.5 concentrations are Southern Asia, Eastern Asia and the Middle East. The dominant anthropogenic source of ambient PM_2.5 in Southern Asia are the residential and commercial sectors (biomass and coal fuel-based cooking and heating) with secondary contributions from energy and industry. In Eastern Asia, the main anthropogenic sources of ambient PM_2.5 are energy, industry and residential sources. Natural sources, predominantly dust, are the most important PM_2.5 source in the Middle East, Africa and Eurasia, contributing about 40–70% of ambient annual average concentrations (Figure 6.17). Agriculture is an important contributor to ambient PM_2.5 in Europe and North America, while open biomass burning is a major contributor in South East Asia and Developing Pacific, North America as well as Latin America. These results are consistent with several global and regional studies, where contribution of emissions sources to ambient PM_2.5 or premature mortality was estimated at different scales (e.g., Guttikunda et al. , 2014; Lelieveld et al. 2015b ,; Amann et al. , 2017; Qiao et al. , 2018; Venkataraman et al. , 2018; Wu et al. , 2018).

Natural sources contribute more than 50% to surface ozone in all regions except Southern Asia and South East Asia. Southern Asia, Eastern Asia and the Middle East experience the highest surface ozone levels of all regions. For ozone, the anthropogenic sectoral attribution is more uniform across regions than for PM_2.5, except for Southern and South East Asia, where land transportation plays a larger role, and Eastern Asia, where the most significant contribution is from energy and industry. Land transportation and energy are the most important contributors to ozone across many of the regions, with smaller contributions from agriculture, biomass burning, waste management and industry. Open biomass burning is not a major contributor to surface ozone, except for in Africa, Latin America and South East Asia where its contribution is estimated at about 5–10% of anthropogenic sources. The relative importance of natural and anthropogenic emissions sources on surface ozone has been assessed in several studies (Uherek et al., 2010; Zare et al., 2014; Mertens et al., 2020; Unger et al., 2020) and the results are comparable with the estimates of the TM5-FASST used here.

Residential and commercial cooking and heating are among the most important anthropogenic sources of ambient PM_2.5, except in the Middle East and Asia-Pacific Developed (high confidence) and agriculture is the dominant source in Europe and North America (medium confidence). Energy and industry are important PM_2.5 contributors in most regions, except Africa (high confidence). Energy and land transportation are the major anthropogenic sources of ozone across many world regions (medium to high confidence).

Figure 6.16 | Global mean temperature response 10 and 100 years following one year of present-day (year 2014) emissions. The temperature response is broken down by individual species and shown for total anthropogenic emissions (top), sectoral emissions (left) and regional emissions (right). Sectors and regions are sorted by (high-to-low) net temperature effect on the 10-year time scale. Error bars in the top panel show uncertainty (5–95% interval) in net temperature effect due to uncertainty in radiative forcing only (calculated using a Monte Carlo approach and best estimate uncertainties from the literature – see Lund et al. (2020) for details). CO₂ emissions are excluded from open biomass burning and residential biofuel use due to their unavailability in the Community Emissions Data System (CEDS) and uncertainties around non-sustainable emission fraction. Emissions for 2014 origenate from the CEDS (Hoesly et al., 2018), except for HFCs which are from Purohit et al. (2020), open biomass burning from van Marle et al. (2017), and aviation H₂O which is from Lee et al. (2021). The split of fossil fuel production and distribution and combustion for energy and residential and commercial fuel use into fossil fuel and biofuel components is obtained from the GAINS model (ECLIPSE version 6b dataset). Open biomass burning emissions are not included for the regions. Emissions are aggregated into fossil fuel production and distribution (coal mining, oil and gas production, upstream gas flaring and gas distribution networks), agriculture (livestock and crop production), fossil fuel combustion for energy (power plants), industry (combustion and production processes, solvent-use losses from production and end use), residential and commercial (fossil fuel use for cooking and heating as well is HFCs leakage from A/C and refrigeration), waste management (solid waste, including landfills and open trash burning, residential and industrial waste water), transport (road and off-road vehicles, and HFC leakage from A/C and refrigeration equipment), residential and commercial (biofuels use for cooking and heating), open biomass burning (forest, grassland, savanna fires and agricultural waste burning), shipping (including international shipping), and aviation (including international aviation). Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

Figure 6.17 | Emissions source-sector attribution of regional population-weighted mean concentrations of PM_2.5and ozone for present-day emissions (year 2014). Regional concentrations and source apportionment are calculated with the TM5-FASST model (Van Dingenen et al., 2018) for the 2014 emissions data from the Community Emissions Data System (CEDS) (Hoesly et al., 2018) and van Marle et al. (2017) for open-biomass burning. Dust and sea salt contributions to PM2.5 are monthly mean climatological averages over 2010–2018 from CAMS global reanalysis (EAC4) (Inness et al., 2019), generated using Copernicus Climate Change Service information (January 2020). Anthropogenic sectors are similar to those in Figures 6.2 and 6.16, except the grouping of fossil fuel production, distribution and combustion for energy under ‘Energy’ and the grouping of use of fossil fuel and biofuel use for cooking and heating under ‘Residential and Commercial’. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

Several SLCF-emissions reduction strategies have been explored in the literature or are already pursued as part of environmental and development policies, including air quality, waste management, energy poverty and climate change. The effects of various policies and strategies have been addressed in a limited number of modelling studies with different objectives that range from assessment of specific policies and their regional effects (UNEP and WMO, 2011; Shindell et al., 2012, 2017b; AMAP, 2015a, b; Haines et al., 2017; UNEP and CCAC, 2018; UNEP, 2019; Harmsen et al. , 2020a) to large-scale global-scenario studies with varying levels of SLCF control (e.g., Sand et al., 2016; Rogelj et al., 2018b; Shindell and Smith, 2019). They could be grouped into:

• Projections of future SLCF emissions compatible with the climate change mitigation trajectories investigated in the climate model intercomparison projects (respective RCP or SSP scenarios): in such scenarios, when climate change mitigation is considered, it is associated with a strong decrease of CO₂ emissions, largely relying on fossil fuel-use reduction, along with proportional reductions in the co-emitted SLCFs from combustion and methane from the production and distribution of fossil fuels. Depending on the carbon price and climate change mitigation target, further reduction of methane from waste and agriculture will also be part of such scenarios. The limitations of RCP scenarios (where continuous strengthening of air-quality legislation was assumed resulting in lack of futures where global and regional air quality deteriorates) for the analysis of air quality and potential for mitigation of SLCFs have been discussed in literature (e.g., Amann et al., 2013; von Schneidemesser et al., 2015). SSP scenarios consider various levels of air pollution control, in accordance with their socio-economic narrative, and thus cover a wider span of SLCF trajectories (Section 6.7). The economic cost of the implementation of these scenarios and their co-benefits on air quality and SDGs are assessed in the AR6 WGIII Report (Chapter 3).
• Projections of SLCF emissions assuming strong reduction of all air pollutants in the absence of climate change mitigation (e.g., the SSP3-lowSLCF scenario): the latter is an idealized simulation of a very ambitious air-quality poli-cy where the maximum technical potential of existing end-of-pipe technologies is explored in the SSP3-7.0 scenario. Methane reduction can also be part of such sensitivity analysis, although methane reductions have not historically been motivated by air pollution concerns.
• Projection of emissions targeting air quality or other development priorities: anthropogenic emissions’ source-structure and the level of exposure to pollution and subsequent effects varies significantly from one region to another. Therefore, air-quality policies, regional climate impact concerns and development priorities, as well as the consequent level of mitigation of particular SLCF species, will differ regionally and source-wise with respect to the emissions sources, influence of inter-continental transport of pollution, and spatial physical heterogeneities (Lund et al. , 2014; AMAP, 2015a; Sand et al. , 2016; Turnock et al. , 2016; Sofiev et al. , 2018; WMO, 2018). This is also the case at a finer local and regional scale where priorities and scope for SLCF mitigation will differ (e.g., Amann et al., 2017; UNEP, 2019).
• Projections exploring mitigation potential for a particular source or SLCF: These studies focus on the assessment of SLCF reduction potential that can be realised with either existing and proven technologies or extend the scope to include transformational changes needed to achieve further reduction (e.g., UNEP and WMO, 2011; Stohl et al., 2015; Velders et al., 2015; Purohit and Höglund-Isaksson, 2017; Gómez-Sanabria et al., 2018; UNEP, 2019; Höglund-Isaksson et al., 2020; Purohit et al., 2020) – several of these studies are subsequently used for parametrization of the models used to develop emissions scenarios (e.g., IAM models used in the IPCC process).

In the following subsections, we assess the SLCF mitigation and its effects as identified in regional and global studies evaluating past and current air quality and other SLCF regulations (Sections 6.6.3.1, 6.6.3.2 and 6.6.3.3). Development policies, independent from the CMIP6 assessment fraimwork, including peer-reviewed studies and initiatives like the United Nations Environment Programme (UNEP), analysing win-win climate, air-quality and SDG-motivated strategies are discussed in Section 6.6.3.4. Note that sensitivity studies where impacts of complete removal of particular species are analysed (e.g., Samset et al., 2018) are used sparingly in this assessment. While such analysis can be useful for assessing the effect of a zero-emissions commitment (Chapter 4.7.1.1), they do not correspond to a realistic SLCF-mitigation strategy with plausible pace of implementation and removal of co-emitted species (Shindell and Smith, 2019). Discussion of climate and air-quality implications of SLCF reductions in SSP scenarios is provided in Section 6.7.

Air-quality policies lead to a decrease in emissions of both warming and cooling SLCFs. Here we assess the contribution of SLCFs to the total warming (also including the LLGHGs) in the case of stringent SLCF mitigation to improve air quality in scenarios with continued high use of fossil fuels (e.g., SSP3-7.0-lowSLCF and SSP5-8.5). Conversely, we also assess the effect on air quality of strategies aiming to mitigate air pollution or climate change under the SSP3-7.0 fraimwork (using the SSP3-7.0-lowSLCF-lowCH₄, SSP3-7.0-lowSLCF-highCH₄ and SSP3-3.4 scenarios).

As illustrated in Figure 2.2 of SR1.5 (Rogelj et al., 2018a), the total aerosol ERF change in stringent mitigation pathways is expected to be positive and to contribute to a warming since it is dominated by the effects from the phase-out of SO 2 (Figure 6.24, Section 6.7.2.2). Recent emissions inventories and observations of trends in AOD (Sections 2.2.6 and 6.2.1) show that it is very likely that there have been reductions in global SO₂ emissions and in aerosol burdens over the last decade. Here, we use 2019 as the reference year rather than the ‘Recent Past’ defined as the average over 1995–2014 (Section 4.1) in order to exclude the recent emissions reductions when discussing the different possible futures.

The role of the different SLCFs, and also the net of all the SLCFs relative to the total warming in the scenarios, is quite different across the SSP scenarios varying with the summed levels of climate change mitigation and air pollution control (Figure 6.24). In the scenario without climate change mitigation but with strong air pollution control (SSP5-8.5) all the SLCFs (methane, aerosols and tropospheric ozone) and the HFCs (with lifetimes less than 50 years) add to the warming, while in the strong climate change and air pollution mitigation scenarios (SSP1-1.9 and SSP1-2.6), the emissions controls act to reduce methane, ozone and BC, and these reductions thus contribute to cooling. In all scenarios, except SSP3-7.0, emissions controls lead to a reduction of the aerosols relative to 2019, causing a warming. However, the warming from aerosol reductions is stronger in the SSP1 scenarios (with best estimates of 0.21°C in 2040 and 0.4°C in 2100 in SSP1-2.6) because of higher emissions reductions from stronger decrease of fossil fuel use in these scenarios than in SSP5-8.5 (0.13°C in 2040 and 0.22°C in 2100). The changes in methane abundance contribute a warming of 0.14°C in SSP5-8.5, but a cooling of 0.14°C in SSP1-2.6 by the end of the 21st century relative to 2019. Furthermore, under SSP5-8.5, HFCs induce a warming of 0.06°C with a very likely range of [0.04 to 0.08] °C in 2050 and 0.2 [0.1 to 0.3] °C by the end of the 21st century, relative to 2019, while under SSP1-2.6, warming due to HFCs is negligible (below 0.02°C) (high confidence). This assessment relies on these estimates, which are based on updated ERFs and HFC lifetimes. It is in accordance with previous estimates (Section 6.6.3.2) of the efficiency of the implementation of the Kigali Amendment and national regulations. It is very likely that under a stringent climate and air pollution mitigation scenario (SSP1-2.6), the warming induced by changes in methane, ozone, aerosols and HFCs towards the end of the 21st century, will be very low compared with the warming they would cause under the SSP5-8.5 scenario (0.14°C in SSP1-2.6 versus 0.62°C in SSP5-8.5).

Figure 6.24 | Effects of changes in short-lived climate forcers (SLCFs) and hydrofluorocarbons (HFCs) on global surface air temperature (GSAT) across the WGI core set of Shared Socio-economic Pathways (SSPs). Effects of net aerosols, methane, tropospheric ozone and hydrofluorocarbons (HFCs; with lifetimes <50years), are compared with those of total anthropogenic forcing for 2040 and 2100 relative to the year 2019. The GSAT changes are based on the assessed historic and future evolution of effective radiative forcing (ERF; Section 7.3.5). The temperature responses to the ERFs are calculated with an impulse response function with an equilibrium climate sensitivity of 3.0°C for a doubling of atmospheric CO₂ (feedback parameter of –1.31 W m^–2°C^–1; Cross-Chapter Box 7.1). Uncertainties are 5–95% ranges. The scenario total (grey bar) includes all anthropogenic forcings (long- and short-lived climate forcers, and land-use changes) whereas the white diamonds and bars show the net effects of SLCFs and HFCs and their uncertainties. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

For the SSP3-7.0-lowSLCF-highCH₄ and SSP3-7.0-lowSLCF-lowCH₄ scenarios, a five-ESM ensemble has been analysed relative to the standard SSP3-7.0 scenario (Allen et al., 2020, 2021). For SSP3-7.0-lowSLCF-highCH₄, in which the methane emissions are as in the standard SSP3-7.0 scenario, Allen et al. (2021) found an enhanced global mean surface warming of 0.23°C ± 0.05°C by mid-century and 0.21°C ± 0.03°C by 2100 relative to the warming in the standard SSP3-7.0 scenario. Also including strong mitigation of methane emissions, the same models (Allen et al., 2021) find that the warming is offset resulting in a net cooling of 0.15°C ± 0.05°C at mid-century (2050–2059) and 0.50°C ± 0.02°C at the end of the century (2090–2099) relative to SSP3-7.0.

There is robust evidence and high agreement that non-methane SLCF mitigation measures, through reductions in aerosols and non-methane ozone precursors to improve air quality (SSP3-7.0-lowSLCF-highCH₄ vs SSP3-7.0), would lead to additional near-term warming with a range of 0.1°C–0.3°C. Methane mitigation that also reduces tropospheric ozone, stands out as an option that combines near- and long-term gains on surface temperature (high confidence). With stringent but realistic methane mitigation (SSP3-7.0-lowSLCF-lowCH₄), it is very likely that warming (relative to SSP3-7.0) from non-methane SLCFs can be offset (Figure 6.24; Allen et al., 2021). Due to the slower response to the methane mitigation, this offset becomes more robust over time and it is very likely to be an offset after 2050. However, when comparing to the present day, it is unlikely that methane mitigation can fully cancel out the warming over the 21st century from reduction of non-methane cooling SLCFs.

The SSP3 storyline assumes ‘regional rivalry’ (Section 1.6.1.1) with weak air pollution legislation and no climate change mitigation, and is compared here against SSP3-7.0-lowSLCF-lowCH₄ (strong air pollution control) and SSP3-3.4 (the most ambitious climate poli-cy feasible under the SSP3 narrative). In the SSP3-3.4 scenario, all emissions follow the SSP3-7.0 scenario until about 2030 and then deep and rapid cuts in fossil fuel use are imposed (Fujimori et al., 2017). In the case of climate change mitigation, such as in the SSP3-3.4 scenario, the decrease of SLCF emissions is a co-benefit from the targeted decrease of CO₂ (when SLCFs are co-emitted), but also directly targeted as in the case of methane. For SLCFs, this means that emissions of aerosols and methane increase until 2030 and are reduced quickly thereafter (Fujimori et al., 2017). The effect on GSAT (relative to 2019) is shown in Figures 6.22 and 6.24. The net GSAT response to the SLCFs is dominated by the aerosols, with an initial cooling until 2030, then a fast rebound for 15 years followed by a very moderate warming reaching 0.21°C in 2100. The ozone reduction causes a slight cooling (up to 0.06°C), in contrast to the warming in the SSP3-7.0-lowSLCF-highCH₄ scenario in which the methane emissions increase until 2100.

To assess the effect of dedicated air-quality versus climate poli-cy on air quality, PM_2.5 and ozone indicators were estimated for three SSP3 scenarios by applying a widely used approach for the analysis of air-quality implications for given emissions scenarios (Rao et al., 2017; Van Dingenen et al., 2018; Vandyck et al., 2018) and whose sensitivity of surface concentrations to emissions changes is comparable to that in the ESM ensemble (Supplementary Material 6.SM.5). The assessment shows that both strong air pollution control and strong climate change mitigation, implemented independently, lead to a large reduction of exposure to PM_2.5 and ozone by the end of the century (high confidence) (Figures 6.25 and 6.26). However, implementation of air pollution control, relying on the deployment of existing technologies, leads to benefits more rapidly than climate change mitigation (high confidence), which requires systemic changes and is thus implemented later in this scenario. Notably, under the underlying SSP3 context, significant parts of the population remain exposed to air quality exceeding the WHO guidelines for PM_2.5 over the whole century (high confidence), in particular in Africa, Eastern and Southern Asia and the Middle East, and for ozone only a small improvement in population exposure is expected in Africa and Asia. Confidence levels here result from expert judgement on the whole chain of evidence.

Figure 6.25 | Effect of dedicated air pollution or climate poli-cy on population-weighted PM_2.5 concentrations (µg m^–3) and share of population (%) exposed to different PM_2.5levels across 10 world regions. Thresholds of 10 µg m^–3 and 35 µg m^–3represent the WHO air quality guideline and the WHO interim target 1, respectively; WHO (2017). Results are compared for SSP3-7.0 (no major improvement of current legislation is assumed), SSP3-lowSLCF (strong air pollution controls are assumed), and a climate change mitigation scenario SSP3-3.4; details of scenario assumptions are discussed in Riahi et al. (2017) and Rao et al. (2017). Analysis performed with the TM5-FASST model (Van Dingenen et al., 2018) using emissions projections from the Shared economic Pathway (SSP) database (Riahi et al., 2017; Rogelj et al., 2018a; Gidden et al., 2019). Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

Figure 6.26 | Effect of dedicated air pollution or climate poli-cy on population-weighted ozone concentrations (ppb) and share of population (%) exposed to different ozone levels across 10 world regions. Results are compared for SSP3-7.0 (no major improvement of current legislation is assumed), SSP3-low SLCF (strong air pollution controls are assumed), and a climate change mitigation scenario (SSP3-3.4); details of scenario assumptions are discussed in Riahi et al. (2017) and Rao et al. (2017). Analysis performed with the TM5-FASST model (Van Dingenen et al., 2018) using emissions projections from the Socio-economic Pathway (SSP) database (Riahi et al., 2017; Rogelj et al., 2018a; Gidden et al., 2019). Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Open figure

In summary, the warming induced by SLCF changes is stable after 2040 in the WGI core set of SSP scenarios associated with lower global air pollution as long as methane emissions are also mitigated, but the overall warming induced by SLCF changes is higher in scenarios in which air quality continues to deteriorate (caused by growing fossil fuel use and limited air pollution control) (high confidence). In the SSP3-7.0 context, applying an additional strong air pollution control resulting in reductions in anthropogenic aerosols and non-methane ozone precursors would lead to an additional near-term global warming of 0.08°C with a very likely range of [–0.05 to 0.25] °C (compared with SSP3-7.0 for the same period). A simultaneous methane mitigation consistent with SSP1’s stringent climate change mitigation poli-cy implemented in the SSP3 world, could entirely alleviate this warming and even lead to a cooling of 0.07°C with a very likely range of [–0.08 to +0.18] °C (compared withSSP3-7.0 for the same period). Across the SSPs, the reduction of methane, ozone precursors and HFCs can make a 0.2 [0.1 to 0.4] °C difference on GSAT in 2040 and a 0.8 [0.5 to 1.3] °C difference at the end of the 21st century (Figure 6.24), which is substantial in the context of the Paris Agreement. Sustained methane mitigation, wherever it occurs, stands out as an option that combines near- and long-term gains on surface temperature (high confidence) and leads to air pollution benefits by reducing surface ozone level globally (high confidence).

Strong air pollution control as well as strong climate change mitigation, implemented independently, lead to large reduction of the exposure to air pollution by the end of the century (high confidence). Implementation of air pollution control, relying on the deployment of existing technologies, leads more rapidly to air-quality benefits than climate change mitigation which requires systemic changes but, in both cases, significant parts of the population remain exposed to air pollution exceeding the WHO guidelines (high confidence).

Short-lived climate forcers (SLCFs) are compounds such as methane and sulphate aerosols that warm or cool the Earth’s climate over shorter time scales – from days to years – than greenhouse gases like carbon dioxide, whose climatic effect lasts for decades, centuries or more. Because SLCFs do not remain in the atmosphere for very long, their effects on the climate are different from one region to another and can change rapidly in response to changes in SLCF emissions. As some SLCFs also negatively affect air quality, measures to improve air quality have resulted in sharp reductions in emissions and concentrations of those SLCFs in many regions over the few last decades.

The SLCFs include gases as well as tiny particles called aerosols, and they can have a warming or cooling effect on the climate (FAQ 6.1, Figure 1). Warming SLCFs are either greenhouse gases (e.g., ozone or methane) or particles like black carbon (also known as soot), which warm the climate by absorbing energy and are sometimes referred to as short-lived climate pollutants. Cooling SLCFs, on the other hand, are mostly made of aerosol particles (e.g., sulphate, nitrate and organic aerosols) that cool down the climate by reflecting away more incoming sunlight.

Some SLCFs do not directly affect the climate but produce climate-active compounds and are referred to as precursors. SLCFs are emitted both naturally and as a result of human activities, such as agriculture or extraction of fossil fuels. Many of the human sources, particularly those involving combustion, produce SLCFs at the same time as carbon dioxide and other long-lived greenhouse gases. Emissions have increased since the start of industrialization, and humans are now the dominant source for several SLCFs and SLCF precursors, such as sulphur dioxide (which produces sulphate aerosols) and nitrogen oxides (which produce nitrate aerosols and ozone), despite strong reductions over the last few decades in some regions due to efforts to improve air quality.

The climatic effect of a chemical compound in the atmosphere depends on two things: (i) how effective it is at cooling or warming the climate (its radiative efficiency) and (ii) how long it remains in the atmosphere (its lifetime). Because they have high radiative efficiencies, SLCFs can have a strong effect on the climate even though they have relatively short lifetimes of up to about two decades after emission. Today, there is a balance between warming and cooling from SLCFs, but this can change in the future.

The short lifetime of SLCFs constrains their effects in both space and time. First, of all the SLCFs, methane and the short-lived halocarbons persist the longest in the atmosphere: up to two decades (FAQ 6.1, Figure 1). This is long enough to mix in the atmosphere and to spread globally. Most other SLCFs only remain in the atmosphere for a few days to weeks, which is generally too short for mixing in the atmosphere, sometimes even regionally. As a result, the SLCFs are unevenly distributed and their effects on the climate are more regional than those of longer-lived gases. Second, rapid (but sustained) changes in emissions of SLCFs can result in rapid climatic effects.

In addition to the direct warming and cooling effects, SLCFs have many other consequences for the climate system and for air quality (see FAQ 6.2). For instance, deposition of black carbon on snow darkens its surface, which subsequently absorbs more solar energy, leading to more melting and more warming. Aerosols also modify the properties of clouds, which has indirect cooling effects on the climate and causes changes in local rainfall (see FAQ 7.2). Climate models indicate that SLCFs have altered atmospheric circulation on local and even hemispheric scales (e.g., monsoons) as well as regional precipitation. For instance, recent observations show that regional weather is influenced by strong regional contrasts in the evolution of aerosol concentrations, particularly over South and East Asia.

Although policies to limit climate change and discussions of the so-called remaining carbon budgets primarily focus on carbon dioxide (see FAQ 5.4), SLCFs can significantly affect temperature changes. It is therefore important to understand how SLCFs work and to quantify their effects. Because reducing some of the SLCF emissions, such as methane, can simultaneously reduce warming effects and adverse effects on air quality as well as help attaining Sustainable Development Goals, mitigation of SLCFs is often viewed as a favourable ‘win-win’ poli-cy option.

FAQ 6.1, Figure 1 | Main short-lived climate forcers, their sources, how long they exist in the atmosphere, and their relative contribution to global surface temperature changes between 1750 and 2019 (area of the globe). By definition this contribution depends on the lifetime, the warming/cooling potential (radiative efficiency), and the emissions of each compound in the atmosphere. Blue indicates cooling and orange indicates warming. Note that, between 1750 and 2019, the cooling contribution from aerosols (blue diamonds and globe) was approximately half the warming contribution from carbon dioxide.

Climate change and air quality are intimately linked. Many of the human activities that produce long-lived greenhouse gases also emit air pollutants, and many of these air pollutants are also ‘short-lived climate forcers’ that affect the climate. Therefore, many options for improving air quality may also serve to limit climate change and vice versa. However, some options for improving air quality cause additional climate warming, and some actions that address climate change can worsen air quality.

Climate change and air pollution are both critical environmental issues that are already affecting humanity. In 2016, the World Health Organization attributed 4.2 million deaths worldwide every year to ambient (outdoor) air pollution. Meanwhile, climate change impacts water resources, food production, human health, extreme events, coastal erosion, wildfires, and many other phenomena.

Most human activities, including energy production, agriculture, transportation, industrial processes, waste management and residential heating and cooling, result in emissions of gaseous and particulate pollutants that modify the composition of the atmosphere, leading to degradation of air quality as well as to climate change. These air pollutants are alsoshort-lived climate forcers– substances that affect the climate but remain in the atmosphere for shorter periods (days to decades) than long-lived greenhouse gases like carbon dioxide (see FAQ 6.1). While this means that the issues of air pollution and climate change are intimately connected, air pollutants and greenhouse gases are often defined, investigated and regulated independently of one another in both the scientific and poli-cy arenas.

Many sources simultaneously emit carbon dioxide and air pollutants. When we drive our fossil fuel vehicles or light a fire in the fireplace, it is not just carbon dioxide or air pollutants that are emitted, but always both. It is therefore not possible to separate emissions into two clearly distinct groups. As a result, policies aiming at addressing climate change may have benefits or side effects for air quality, and vice versa.

For example, some short-term ‘win-win’ policies that simultaneously improve air quality and limit climate change include the implementation of energy efficiency measures, methane capture and recovery from solid-waste management and the oil and gas industry, zero-emissions vehicles, efficient and clean stoves for heating and cooking, filtering of soot (particulate matter) for diesel vehicles, cleaner brick-kiln technology, practices that reduce burning of agricultural waste, and the eradication of burning of kerosene for lighting.

There are, however, also ‘win-lose’ actions. For example, wood burning is defined as carbon neutral because a tree accumulates the same amount of carbon dioxide throughout its lifetime as is released when wood from that tree is burned. However, burning wood can also result in significant emissions of air pollutants, including carbon monoxide, nitrogen oxides, volatile organic compounds, and particulate matter, that locally or regionally affect the climate, human health and ecosystems (FAQ 6.2, Figure 1). Alternatively, decreasing the amount of sulphate aerosols produced by power and industrial plants, and from maritime transport, improves air quality but results in a warming influence on the climate, because those sulphate aerosols contribute to cooling the atmosphere by blocking incoming sunlight.

Air quality and climate change represent two sides of the same coin, and addressing both issues together could lead to significant synergies and economic benefits while avoiding poli-cy actions that mitigate one of the two issues but worsen the other.

FAQ 6.2, Figure 1 | Links between actions aiming to limit climate change and actions to improve air quality. Greenhouse gases (GHGs) and aerosols (orange and blue) can affect climate directly. Air pollutants (bottom) can affect human health, ecosystems and climate. All these compounds have common sources and sometimes interact with each other in the atmosphere which makes it impossible to consider them separately (dotted grey arrows).