Editing IPCC:AR6/WGI/Chapter-6 (section)

== 6.3 Evolution of Atmospheric SLCF Abundances ==

<div id="h1-4-siblings" class="h1-siblings"></div>

This section assesses the evolution of atmospheric abundance <sup>[[#footnote-004|1]]</sup> of SLCFs since AR5 based on observations and modelling, our knowledge of SLCF burden and distribution, and our understanding of the trends over longer time scales. In addition to emissions (Section 6.2), atmospheric chemistry (gas and heterogeneous chemistry), deposition (including wet and dry removal), and transport processes play a major role in determining the atmospheric distribution, budget and lifetime of SLCFs. The distribution and lifetime of SLCFs are further influenced by the modulation of chemical and physical processes in response to a changing climate. Therefore, the time evolution of atmospheric abundance of SLCFs is characterized by many complex non-linear interactions occurring at varying temporal and spatial scales. For this Assessment, global-scale, long-term measurements are employed only for a few gaseous SLCFs while for most short-lived species regional-scale observations and global models are relied upon.

<div id="box-6.1" class="h2-container box-container"></div>

'''Box 6.1 | Atmospheric Abundance of SLCFs: From Process-level Studies to Global Chemistry–Climate Models'''

<div id="h2-12-siblings" class="h2-siblings"></div>

Changes in the atmospheric distribution of SLCFs determine their radiative forcing, and climate and air-quality impacts. This box provides an overview of how process-level understanding of the distribution and evolution of chemical compounds is derived and where uncertainties come from.

Process-level understanding of tropospheric gas and aerosol chemistry developed through laboratory and simulation chamber experiments, as well as quantum chemical theory, is used to generate chemical mechanisms. Atmospheric simulation chambers are designed to identify the chemical pathways and quantify reaction kinetics in isolation from atmospheric transport, deposition and emission processes. Ideally the chemical regimes studied are representative for ambient atmospheric complexity and concentrations (e.g., McFiggans et al. , 2019). Recently, quantum chemical theory has advanced to a level that it can provide kinetic and product information in a parameter range not possible with laboratory experiments (Vereecken et al. , 2015) . Iterative and interlinked use of simulation chamber and quantum chemical theory has led to improved knowledge of chemical mechanisms (Peeters et al. , 2009, 2014; Nguyen et al. , 2010; Fuchs et al. , 2013). For application in chemistry–climate models (CCMs), the chemical mechanisms need to be computationally efficient, requiring simplifications. Such simplifications include reduced hydrocarbon representations, the application of lumping techniques (one compound or a chemical structure representing a family of compounds, for example, as done for parametrizing SOA formation) and/or the implementation of artificial operators representing key steps of the chemistry ( [[#Emmerson--2009|Emmerson and Evans, 2009]] ; [[#Xia--2009|Xia et al., 2009]] ; [[#Stockwell--2020|Stockwell et al., 2020]] ). Additionally, aerosol microphysical processes (nucleation, coagulation, condensation, evaporation and sedimentation) that determine the evolution of aerosol number concentrations and size particle distribution are represented in parametrized forms in global models with varying levels of complexity ( [[#Mann--2014|Mann et al., 2014]] ).

A wide range of in situ and remotely sensed observations are used to characterize atmospheric chemical composition. Measurements made routinely as part of long-term monitoring programmes are particularly useful for assessing long-term trends and variability, and spatial distributions (Sections 2.2, 6.3 and 7.3.3), while intensive field campaigns provide a more comprehensive view of atmospheric composition at a specific location for a limited time, facilitating an improved process-level understanding. Retrieval of atmospheric concentrations from satellites, in particular, has been tremendously useful for providing global continuous coverage, although the retrievals themselves depend on prior information of atmospheric composition usually derived from models. Over the last decade or so, observations of atmospheric concentrations have been combined with information from global chemistry–climate models to produce global assimilation and forecasting systems with the purpose of producing chemical reanalysis or improving model inputs (i.e., emissions or boundary conditions) and forecasts ( [[#Miyazaki--2015|Miyazaki et al., 2015]] ; [[#Randles--2017|Randles et al., 2017]] ; [[#Inness--2019|Inness et al., 2019]] ).

Global three-dimensional CCMs (Box 6.1, Figure 1) represent the full coupling of chemistry with climate physics (e.g., Morgenstern et al. , 2017) with different levels of complexity (e.g., interactive aerosols with or without tropospheric and/or stratospheric chemistry). Methane concentrations are typically prescribed or constrained to observations while emissions of other SLCFs (or their precursors) are either prescribed or calculated interactively in the current generation of CCMs (Collins et al. , 2017) . CCMs, now part of Earth system models (ESMs), are applied extensively to simulate the distribution and evolution of chemical compounds on a variety of spatial and temporal scales to improve current knowledge, make future projections and investigate global scale chemistry–climate interactions and feedbacks ( [[IPCC:Wg1:Chapter:Chapter-3#3.8.2.2|Section 3.8.2.2]] ). CCMs are also used to interpret observations to disentangle the processes that drive observed variability and trends. Some aspects of air quality, such as diurnal peaks or local threshold violations, strong gradient in chemical regimes and coupling between processes cannot be captured by relatively coarse spatial resolution (&gt;50 km) global CCMs ( [[#Markakis--2014|Markakis et al., 2014]] ) and necessitate subsequent downscaling modelling exercises.

The skill of CCMs is typically assessed by their ability to reproduce observed abundance, trends and variability of chemical compounds. However, uncertainty remains large because of observation limitations (errors and uncertainties, spatial and temporal coverage),

[[File:3db269acc6280b8188bdc035ebb9b542 IPCC_AR6_WGI_Box_6_1_Figure_1.png]]

Box 6.1, Figure 1 |

'''Knowledge exchange between laboratory/theoretical studies, observations and global climate–chemistry models (CCMs) to inform our understanding of short-lived climate forcers (SLCFs).'''
model parametrizations (e.g., chemical mechanisms, photolysis schemes, parametrizations for mixing and convective transport, and deposition), model input parameters (e.g., reaction rate constants, emissions) and an incomplete understanding of the physical and chemical processes that determine SLCF distributions ( [[#Brasseur--2017|Brasseur and Jacob, 2017]] ; [[#Young--2018|Young et al., 2018]] ). CCMs can therefore not capture every aspect of atmospheric chemical composition, but are expected to represent, as faithfully as possible, the sensitivity of chemical compounds to their drivers (e.g., anthropogenic emissions). Models are evaluated in multiple ways to identify their strengths and weaknesses in explaining the evolution of SLCF abundances. For example, CCM simulations are performed in the nudged or offline meteorology mode, that is, driven by observed or reanalysed meteorology rather than in the free-running mode, for consistent comparison of modelled chemical composition with observations for a specific time period ( [[#Dameris--2013|Dameris and Jöckel, 2013]] ). However, caution is exercised as nudging can alter the model climate resulting in unintentional impacts on the simulated atmospheric physics and/or chemistry ( [[#Orbe--2018|Orbe et al., 2018]] ; [[#Chrysanthou--2019|Chrysanthou et al., 2019]] ). Chemical mechanisms implemented in CCMs are evaluated and intercompared to assess their skill in capturing relevant chemistry features (e.g., [[#Brown-Steiner--2018|Brown-Steiner et al., 2018]] ). The multi-model ensemble approach, employed for evaluating climate models, has been particularly useful for characterizing errors in CCM simulations of SLCFs related to structural uncertainty and internal variability (Naik et al. , 2013; Shindell et al. , 2013; Young et al. , 2013; Turnock et al. , 2020) . However, as discussed in Box 4.1, this approach is unable to capture the full uncertainty range.

This assessment draws upon results from single-model studies and recent multi-model intercomparisons (e.g., AeroCom, CCMI), in particular those endorsed by CMIP6 (see Table 1.3), which then allows for the full consideration of robustness and uncertainty due to model structures and processes. Based on the collective information provided in this body of literature, the CMIP6 multi-model ensemble is largely fit-for-purpose of evaluating the influence of SLCFs on radiative forcing, climate and non-CO <sub>2</sub> biogeochemical feedbacks. Additionally, CMIP6 models are fit for capturing the global air pollution response to changes in emissions and meteorology, but have difficulty in simulating the mean state ( [[#Turnock--2020|Turnock et al., 2020]] ). The set of CMIP6 simulations has been used to update the relations between emissions and surface temperature at the heart of the emulators (Cross-Chapter Box 7.1) and update emissions metrics ( [[IPCC:Wg1:Chapter:Chapter-7#7.6|Section 7.6]] ). Emulators and emissions metrics are used in this chapter (Sections 6.6 and 6.7) to assess more specifically the effect of the individual SLCFs for each sector and region, which would be of prohibitive computing cost with CCMs. CCMs are also used to build global source-receptor models which use relations between surface concentrations and emissions. Such a model is used to assess the impact of various mitigation policies on air quality (Sections 6.5 and 6.7).

<div id="6.3.1" class="h2-container"></div>

<span id="methane-ch-4"></span>
=== 6.3.1 Methane (CH <sub>4</sub> ) ===

<div id="h2-13-siblings" class="h2-siblings"></div>

The global mean surface mixing ratio of methane has increased by 156% since 1750 ( [[IPCC:Wg1:Chapter:Chapter-2#2.2.3.4|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 ( [[IPCC:Wg1:Chapter:Chapter-2#2.2.3.3.2|Section 2.2.3.3.2]] ) largely driven by anthropogenic activities as assessed in [[IPCC:Wg1:Chapter:Chapter-5|Chapter 5]] ( [[IPCC:Wg1:Chapter:Chapter-5#5.2.2|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--2001|Prather et al., 2001]] ). The AR5 based its assessment of methane lifetime on [[#Prather--2012|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, <sup>[[#footnote-003|2]]</sup> tropospheric halogen loss and soil uptake were assessed to be 150 ± 50 years, 200 ± 100 years, and 120 ± 24 years, respectively ( [[#Myhre--2013|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--2019|]] [[#Zhao--2019|]] [[#Zhao--2019|Zhao et al., 2019]] ; [[#Stevenson--2020|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--2017|Prather and Holmes, 2017]] ; [[#Naus--2019|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 <sup>–1</sup> from BU chemistry–climate models in the Chemistry–Climate Modelling Initiative (CCMI) to 514 [474–529] Tg yr <sup>–1</sup> from TD inverse modelling ( [[IPCC:Wg1:Chapter:Chapter-5#5.2.2|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--2020|Saunois et al., 2020]] ; [[#Zhao--2020a|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--2019|]] [[#Zhao--2019|]] [[#Zhao--2019|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.

<div id="_idContainer014" class="_idGenObjectStyleOverride-1"></div>

'''Table 6.2 |''' '''Methane lifetime due to chemical losses, soil uptake and total atmospheric lifetime based on CMIP6 multi-model analysis, and bottom-up and top-down methane budget estimates in Table 5.2.''' Bottom-up and top-down methane lifetimes are calculated using the central estimates of the respective sinks for the mean 2008–2017 period in Table 5.2 together with the mean 2008–2017 global methane concentration of 1815 ppb (see Annex III) converted to methane burden using a fill-factor of 2.75 Tg/ppb from [[#Prather--2012|Prather et al. (2012)]] . Values in parentheses show the minimum and maximum range.

{| class="wikitable"
|-
| Study

| Total Chemical Lifetime (years)

| Soil Lifetime

(years)

| Total Atmospheric Lifetime

(years)

| Number of Models/ Inversions

|-
| [[#Stevenson--2020|Stevenson et al. (2020)]] <sup>a</sup>

| 8.3 (8.1–8.6) <sup>b</sup>

| 160

| 8.0 (7.7–8.2)

| 3 CMIP6 ESMs

|-
| Bottom-up (based on Table 5.2)

| 8.3 (6.2–9.8)

| 166 (102–453)

| 8.0 (6.3–10.0)

| 7 CCMI CCMs/CTMs

|-
| Top-down (based on Table 5.2)

| 9.7 (9.4–10.5)

| 135 (116–185)

| 9.1 (8.7–10.0)

| 7 inversion systems

|-
| AR6 assessed value <sup>c</sup>

| 9.7 ± 1.1

| 135 ± 44

| 9.1 ± 0.9

| Based on top-down with uncertainty estimate from AR5

|}

<sup>a</sup> Mean over 2005–2014

<sup>b</sup> Does not include lifetime due to tropospheric halogen loss

<sup>c</sup> Uncertainties indicate ±1 standard deviation

The methane perturbation lifetime ( τ <sub>pert</sub> ) 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|Prather, 1994]] ; [[#Fuglestvedt--1996|Fuglestvedt et al., 1996]] ; [[#Holmes--2013|Holmes et al., 2013]] ; [[#Holmes--2018|Holmes, 2018]] ). Since perturbation lifetime relates changes in emissions to changes in burden, it is used to determine the emissions metrics assessed in [[IPCC:Wg1:Chapter:Chapter-7|Chapter 7]] ( [[IPCC:Wg1:Chapter:Chapter-7#7.6|Section 7.6]] ). The perturbation lifetime is related to the atmospheric lifetime as τ <sub>pert</sub> = f * τ <sub>total</sub> where f is the feedback factor and is calculated as f = 1/(1-s), where s = δ (ln τ <sub>total</sub> )/ δ (ln[CH <sub>4</sub> ]) ( [[#Prather--2001|Prather et al., 2001]] ). Since there are no observational constraints for either τ <sub>pert</sub> 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--2013|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|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--2021b|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 [[IPCC:Wg1:Chapter:Chapter-7#7.6|Section 7.6]] .

<div id="6.3.2" class="h2-container"></div>

<span id="ozone-o-3"></span>
=== 6.3.2 Ozone (O <sub>3</sub> ) ===

<div id="h2-14-siblings" class="h2-siblings"></div>

<div id="6.3.2.1" class="h3-container"></div>

<span id="tropospheric-ozone"></span>
==== 6.3.2.1 Tropospheric Ozone ====

<div id="h3-7-siblings" class="h3-siblings"></div>

About 10% of the total atmospheric ozone column resides in the troposphere. The ozone forcing on climate strongly depends on its vertical and latitudinal distribution in the troposphere. The lifetime of ozone in the troposphere ranges from a few hours in polluted urban regions to up to few months in the upper troposphere. Observed tropospheric ozone concentrations range from less than 10 ppb over the tropical Pacific Ocean to as much as 100 ppb in the upper troposphere and more than 100 ppb downwind of major ozone precursor emissions regions. An ensemble of five CMIP6 models including whole atmospheric chemistry and interactive ozone has been shown to simulate consistently the present-day ozone distribution (north to south and latitudinal gradients) and its seasonal variability when compared with observations from sondes, background surface stations and satellite products ( [[#Griffiths--2021|Griffiths et al., 2021]] ). The biases, whose magnitude is similar to AR5, are lower than 15% against climatological seasonal cycles from ozonesondes with an overestimate in the Northern Hemisphere and an underestimate in the Southern Hemisphere ( [[#Griffiths--2021|Griffiths et al., 2021]] ). The CMIP6 multi-model ensemble estimate of the global mean lifetime of ozone for present-day conditions is 25.5 ± 2.2 days ( [[#Griffiths--2021|Griffiths et al., 2021]] ), which is within the range of previous multi-model estimates ( [[#Stevenson--2006|Stevenson et al., 2006]] ; [[#Young--2013|Young et al., 2013]] ), indicating a ''high level of confidence'' .

The AR5 assessed the tropospheric ozone burden to be 337 ± 23 Tg for the year 2000 based on the ACCMIP ensemble of model simulations ( [[#Myhre--2013|Myhre et al., 2013]] ). Multiple satellite products, ozonesondes and CCMs are used to estimate tropospheric ozone burden (Table 6.3). Satellite products provide lower-bound values as they exclude regions under polar night conditions ( [[#Gaudel--2018|Gaudel et al., 2018]] ). The tropospheric ozone burden values from multi-model exercises are within the range of the observational estimates despite different definitions of the tropopause for multi-model estimates which can lead to differences of about 10% on the ozone-burden model estimates ( [[#Griffiths--2021|Griffiths et al., 2021]] ). Weighted by their number of members, CMIP6 and CCMI multi-model estimates and observational estimates of tropospheric ozone burden in about the year 2010, lead to an assessment of the tropospheric ozone burden of 347 ± 28 Tg for 2010.

<div id="_idContainer015" class="_idGenObjectStyleOverride-1"></div>

'''Table 6.3 |''' '''Global tropospheric ozone budget terms and burden based on multi-model estimates and observations for present conditions.''' All uncertainties quoted as '''±''' 1 standard deviation. Values of tropospheric ozone burden with asterisk indicate average over the latitudinal zone 60 <sup>°</sup> N–60 <sup>°</sup> S. STE = stratospheric–tropospheric exchange.

{| class="wikitable"
|-
| '''Period'''

| '''Burden'''

'''(Tg)'''

| '''Production'''

'''Tg y''' '''r''' <sup>–1</sup>

| '''Loss'''

'''Tg y''' '''r''' <sup>–1</sup>

| '''Deposition'''

'''Tg y''' '''r''' <sup>–1</sup>

| '''STE'''

'''Tg y''' '''r''' <sup>–1</sup>

| '''Number of Models/Reference'''

|-
| colspan="7"| Models

|-
| ~2000

time slice (1995–2004)

| 347 ± 30

| 4510 ± 566

| 3948 ± 379

| 846 ± 44

| 284 ± 193

| rowspan="2"| CMIP6 <sup>a</sup> (5 Earth system models for burden and 4 models for budget terms)

( [[#Griffiths--2021|Griffiths et al., 2021]] )

|-
| ~2010

time slice (2005–2014)

| 356 ± 31

| 4708 ± 589

| 4122 ± 399

| 863 ± 40

| 277 ± 201

|-
| ~2000

| 341 ± 31

(309 ± 31)*

|
| rowspan="2"| CCMI <sup>b</sup> (9 models) ( [[#Archibald--2020|Archibald et al., 2020]] )

|-
| 2010

| 345 ± 30

(314 ± 29)*

|
|-
| '''~''' 2000

| 340 ± 34

| 4937 ± 656

| 4442 ± 570

| 996 ± 203

| 535 ± 161

| TOAR <sup>c</sup> (based on 32–49 models participating in inter-model comparisons and single-model studies)

( [[#Young--2018|Young et al., 2018]] )

|-
| colspan="7"| Observations

|-
| 2010–2014

| 338 ± 6

|
| TOST <sup>d</sup> , IASI <sup>e</sup> -FORLI, and IASI-SOFRID

( [[#Gaudel--2018|Gaudel et al., 2018]] )

|-
| 2010–2014

| 302 ± 12*

|
| TOST, IASI-FORLI, IASI-SOFRID, OMI <sup>f</sup> /MLS, OMI-SAO and OMI-RAL

( [[#Gaudel--2018|Gaudel et al., 2018]] )

|}

<sup>a</sup> CMIP6: Coupled Model Intercomparison Project Phase 6; <sup>b</sup> CCMI: Chemistry–Climate Model Initiative; <sup>c</sup> toAR Tropospheric Ozone Assessment Report; <sup>d</sup> toST Trajectory-mapped Ozonesonde dataset for the Stratosphere and Troposphere; <sup>e</sup> IASI Infrared Atmospheric Sounding Interferometer; <sup>f</sup> OMI Ozone Monitoring Instrument.

The tropospheric ozone budget is controlled by chemical production and loss, by stratospheric–tropospheric exchange (STE), and by deposition at the Earth’s surface, whose magnitude are calculated by CCMs (Table 6.3). Despite The high agreement of the model ensemble mean with observational estimates in the present-day tropospheric ozone burden, the values of individual budget terms can vary widely across models in CMIP6, consistent with previous model intercomparison experiments ( [[#Young--2018|Young et al., 2018]] ). Furthermore, single-model studies have shown that the halogen chemistry, which is typically neglected from model chemistry schemes in CCMs, may have a notable impact on the ozone budget, as halogens, particularly of marine origin, take part in efficient ozone-loss catalytic cycles in the troposphere (Saiz-Lopez et al. , 2012; Sarwar et al. , 2015; Sherwen et al. , 2016) .

Because of the heterogeneous distribution of ozone, limited observations or proxies do not provide accurate information about the global pre-industrial abundance, posing a challenge to the estimation of the historical evolution of tropospheric ozone. Therefore, global CCMs complemented by observations are relied upon for estimating the long-term changes in tropospheric ozone. The AR5 concluded that anthropogenic changes in ozone precursor emissions are unequivocally responsible for the increase in tropospheric ozone between 1850 and the present ( [[#Myhre--2013|Myhre et al., 2013]] ). Based on limited isotopic evidence, [[IPCC:Wg1:Chapter:Chapter-2|Chapter 2]] assesses that the global tropospheric ozone increased by less than 40% between 1850 and 2005 ( ''low confidence'' ) ( [[IPCC:Wg1:Chapter:Chapter-2#2.2.5.3|Section 2.2.5.3]] ). The CMIP6 models are in line with this increase of tropospheric ozone with an ensemble-mean value of 109 ± 25 Tg (model range) from 1850–1859 to 2005–2014 (Figure 6.4). This increase is higher than the AR5 value of 100 ± 25 Tg from 1850–2010 due to higher ozone precursor emissions in CMIP6. However, the AR5 and CMIP6 values are close when considering the reported uncertainties. The uncertainties are equivalent in CMIP6 and AR5 despite enhanced inclusion of coupled processes in the CMIP6 ESMs (e.g., biogenic NMVOC emissions or interactive stratospheric ozone chemistry).

<div id="_idContainer017" class="_idGenObjectStyleOverride-1"></div>

[[File:5358ba242352e968d18a4ead22513723 IPCC_AR6_WGI_Figure_6_4.png]]

'''Figure 6.4 |''' '''Time evolution of global annual mean tropospheric ozone burden (in Tg) from 1850 to 2100.''' Multi-model means for CMIP6 historical experiment (1850–2014) from UKESM1-LL-0, CESM2-WACCM, MRI-ESM2-0, GISS-E2.1-G and GFDL-ESM4 and for ScenarioMIP SSP3-7.0 experiment (2015–2100) are represented with their inter-model standard deviation (±1 standard deviation, shaded areas). Observation-based global tropospheric ozone burden estimate (from Table 6.3) is for 2010–2014. Tropospheric Ozone Assessment Report (TOAR) multi-model mean value (from Table 6.3) is for 2000 with a ±1 standard deviation error-bar. Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP) multi-model means are for 1850, 1930, 1980 and 2000 time slices with ±1 standard deviation error-bars. The troposphere is masked by the tropopause pressure calculated in each model using the WMO thermal tropopause definition. Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Since the mid-20th century, the CMIP6 model ensemble shows a higher global trend (Figure 6.4). Since the mid-1990s, the trends are better documented by observations ( [[IPCC:Wg1:Chapter:Chapter-2#2.2.5.3|Section 2.2.5.3]] ) and indicate spatial heterogeneity. In particular, in situ observations at remote surface sites and in the lower free troposphere indicate positive trends that are far more common than negative trends, especially in the northern tropics and across Southern and Eastern Asia (Figure 6.5). The CMIP6 ensemble and observations largely agree on the magnitude of the global positive trend since 1997 (0.82 ± 0.13 Tg yr <sup>–1</sup> in the model ensemble; 0.70 ± 0.15 Tg yr <sup>–1</sup> in the ozonesonde dataset; 0.83 ± 0.85 Tg yr <sup>–1</sup> in the satellite ensemble) and qualitatively reproduce positive trends in the Southern Hemisphere ( [[#Griffiths--2021|Griffiths et al., 2021]] ). More analyses are needed for evaluation in other parts of the world to assess the skill of the recent ensemble based on CMIP6 emissions.

<div id="_idContainer019" class="Basic-Text-Frame"></div>

[[File:ce65b712100397d7960e48942eac57b6 IPCC_AR6_WGI_Figure_6_5.png]]

'''Figure 6.5 |''' '''Decadal tropospheric ozone trends''' '''since 1994.''' Trends are shown at 28 remote and regionally representative surface sites ( [[#Cooper--2020|Cooper et al., 2020]] ) and in 11 regions of the lower free troposphere (650 hPa, about 3.5 km) as measured by In-Service Aircraft for a Global Observing System (IAGOS) above Europe, north-eastern USA, south-eastern USA, western North America, north-east China, South East Asia, southern India, the Persian Gulf, Malaysia/Indonesia, the Gulf of Guinea and northern South America ( [[#Gaudel--2020|Gaudel et al., 2020]] ). High-elevation surface sites are &gt;1500 m above sea level. All trends end with the most recently available year but begin in 1995 or 1994. The sites and datasets are the same as those used in Figure 2.8, further details on data sources and processing are available in the [[IPCC:Wg1:Chapter:Chapter-2|Chapter 2]] data table (Table 2.SM.1).

In summary, there is ''high confidence'' in the estimated present-day (about 2010) global tropospheric ozone burden based on an ensemble of models and observational estimates (347 ± 28 Tg), but there is ''medium confidence'' among the individual models for their estimates of the tropospheric ozone-related budget terms.

Evidence from successive multi-model intercomparisons and the limited isotopic evidence agree on the magnitude of the increase of the tropospheric ozone burden from 1850 to the present day in response to anthropogenic changes in ozone precursor emissions corroborating AR5 findings. This increase is assessed to be 109 ± 25 Tg ( ''medium confidence'' ). The CMIP6 model ensemble shows a constant global increase since the mid-20th century whose rate is consistent with that derived from observations since the mid-1990s.

<div id="6.3.2.2" class="h3-container"></div>

<span id="stratospheric-ozone"></span>
==== 6.3.2.2 Stratospheric Ozone ====

<div id="h3-8-siblings" class="h3-siblings"></div>

Ninety percent of the total column ozone (TCO) resides in the stratosphere. The chemical lifetime of ozone in the stratosphere ranges from less than a day in the upper stratosphere to several months in the lower stratosphere ( [[#Bekki--2009|Bekki and Lefevre, 2009]] ). Global stratospheric ozone trends based on observations are assessed in [[IPCC:Wg1:Chapter:Chapter-2|Chapter 2]] ( [[IPCC:Wg1:Chapter:Chapter-2#2.2.5.2|Section 2.2.5.2]] ). The CMIP6 model ensemble shows that global TCO has slightly changed from 1850–1960 ( [[#Keeble--2021|Keeble et al., 2021]] ). The rapid decline in the 1970s and 1980s due to halogenated ozone-depleting substances (ODSs, as assessed in [[IPCC:Wg1:Chapter:Chapter-2#2.2.5.2|Section 2.2.5.2]] from observations) until the end of the 1990s, followed by a slight increase since then, is captured by the models ( [[#Keeble--2021|Keeble et al., 2021]] ). Overall, the observed climatology patterns and annual cycle amplitudes are well represented in the CMIP6 ensemble mean. The CMIP6 ensemble overestimates the observed TCO values by up to 6% (10–20 Dobson Units (DU)) globally in the NH and SH mid-latitudes, and in the tropics, but the trend in these regions is well captured between 1960 and 2014. However, there is poor agreement between the individual CMIP6 models in the pre-industrial period and throughout the historical period, with model TCO values spread across a range of about 60 DU. The global stratospheric ozone column decreased by 14.3 ± 8.7 DU from 1850–2014 ( [[#Keeble--2021|Keeble et al., 2021]] ).

Model simulations attribute about half of the observed upper-stratospheric ozone increase after 2000 to the decline of ODS since the late 1990s while the other half of the ozone increase is attributed to the slowing of gas-phase ozone destruction cycles due to cooling of the upper stratosphere by increasing GHGs ( [[#Aschmann--2014|Aschmann et al., 2014]] ; [[#Oberländer-Hayn--2015|Oberländer-Hayn et al., 2015]] ).

In summary, global stratospheric ozone column has decreased from pre-industrial period to present day in response to the ODS-induced ozone rapid decline in the 1970s and 1980s, followed by slow, and still incomplete, recovery. There is ''medium confidence'' that global stratospheric ozone column has changed by 14.3 ± 8.7 DU between 1850 and 2014.

<div id="6.3.3" class="h2-container"></div>

<span id="precursor-gases"></span>
=== 6.3.3 Precursor Gases ===

<div id="h2-15-siblings" class="h2-siblings"></div>

<div id="6.3.3.1" class="h3-container"></div>

<span id="nitrogen-oxides-no-x"></span>
==== 6.3.3.1 Nitrogen Oxides (NO <sub>x</sub> ) ====

<div id="h3-9-siblings" class="h3-siblings"></div>

The distribution of tropospheric NO <sub>x</sub> is highly variable in space and time owing to its short lifetime coupled with highly heterogeneous emission and sink patterns. NO <sub>x</sub> undergoes chemical processing, including the formation of nitric acid (HNO <sub>3</sub> ), nitrate (NO <sup>–</sup> <sub>3</sub> ), and organic nitrates (e.g., alkyl nitrate and peroxyacyl nitrate), atmospheric transport, and deposition. Despite challenges in retrieving quantitative information from satellite observations ( [[#Duncan--2014|Duncan et al., 2014]] ; [[#Lin--2015|Lin et al., 2015]] ; [[#Lorente--2017|Lorente et al., 2017]] ; [[#Silvern--2018|Silvern et al., 2018]] ), improved accuracy and resolution of satellite-derived tropospheric NO <sub>2</sub> columns over the past two decades have advanced understanding of the global distribution, long-term trends and source attribution of NO <sub>x</sub> . Long-term average tropospheric NO <sub>2</sub> column based on multiple satellite-borne instruments (Figure 6.6a) reveals the highest NO <sub>2</sub> levels over the most populated, urbanized and industrialized regions of the world corresponding to high NO <sub>x</sub> emissions source regions ( [[#Krotkov--2016|Krotkov et al., 2016]] ; [[#Georgoulias--2019|Georgoulias et al., 2019]] ). Enhanced but highly variable NO <sub>2</sub> columns are also associated with biomass-burning regions as well as areas influenced by lightning activity ( [[#Miyazaki--2014|Miyazaki et al., 2014]] ; [[#Tanimoto--2015|Tanimoto et al., 2015]] ).

<div id="_idContainer021" class="Basic-Text-Frame"></div>

[[File:54fa0416445da6c9329b02a4769f09b2 IPCC_AR6_WGI_Figure_6_6.png]]

'''Figure 6.6 |''' '''Long-term climatological mean (a) and time evolution (b) of tropospheric nitrogen dioxide (NO''' <sub>2</sub> ''') vertical column density.''' Data are from the merged GOME/SCIAMACHY/GOME-2 (TM4NO2A version 2.3) dataset for the period 1996–2016 ( [[#Georgoulias--2019|Georgoulias et al., 2019]] ). Time evolution of NO <sub>2</sub> column shown in panel (b) is normalized to the fitted 1996 levels for the 10 regions shown as boxes in panel (a). Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

Observational constraints derived from the isotopic composition of atmospheric nitrate inferred from ice cores provide evidence of increasing anthropogenic NO <sub>x</sub> sources since pre-industrial times ( [[#Hastings--2009|Hastings et al., 2009]] ; [[#Geng--2014|Geng et al., 2014]] ). Global NO <sub>x</sub> emissions trends in bottom-up inventories (Section 6.2.1) as well as model simulations of nitrogen deposition ( [[#Lamarque--2013a|Lamarque et al., 2013a]] ) are in qualitative agreement with these observational constraints. CMIP6 ESMs exhibit stable NO <sub>x4</sub> burden over the first half of the 20th century and then a sharp increase driven by a factor of three increase in emissions, however, the magnitude of this increase remains uncertain due to poor observational constraints on pre-industrial concentrations of NO <sub>x</sub> <sub></sub> ( [[#Griffiths--2021|Griffiths et al., 2021]] ).

The AR5 reported NO <sub>2</sub> decreases by 30–50% in Europe and North America, and increases by more than a factor of two in Asia, over the 1996–2011 period based on satellite observations ( [[#Hartmann--2013|Hartmann et al., 2013]] ). Extension of this analysis covering the time period up to 2015 reveals that NO <sub>2</sub> has continued to decline over the USA, Western Europe and Japan ( [[#Schneider--2015|Schneider et al., 2015]] ; [[#Duncan--2016|Duncan et al., 2016]] ; [[#Krotkov--2016|Krotkov et al., 2016]] ) because of effective fossil fuel NO <sub>x</sub> emissions controls (Section 6.2), although this rate of decline has slowed down post-2011 ( [[#Jiang--2018|Jiang et al., 2018]] ). Satellite observations also reveal a 32% decline in NO <sub>2</sub> column over China after peaking in 2011 (Figure 6.6b), consistent with declining NO <sub>x</sub> emissions (Section 6.2) due to the implementation of emissions-control strategies (de Foy et al. , 2016; Irie et al. , 2016; F. Liu et al. , 2016) . Over Southern Asia, tropospheric NO <sub>2</sub> levels have grown rapidly with increases of 50% during 2005–2015, largely driven by hotspot areas in India experiencing rapid expansion of the power sector ( [[#Duncan--2016|Duncan et al., 2016]] ; [[#Krotkov--2016|Krotkov et al., 2016]] ). Further analysis indicates that many parts of India have also undergone a reversal in NO <sub>2</sub> trends since 2011 that has been attributed to a combination of factors, including a slowdown in economic growth, implementation of cleaner technologies, non-linear NO <sub>x</sub> chemistry, and meteorological variability ( [[#Georgoulias--2019|Georgoulias et al., 2019]] ). Satellite data reveals spatially heterogeneous NO <sub>2</sub> trends over the Middle East with an overall increase over 2005–2010 and a decrease over large parts of the region after 2011–2012. The reasons for trend reversal within individual areas are diverse, including warfare, imposed sanctions, and air-quality controls ( [[#Lelieveld--2015a|Lelieveld et al., 2015a]] ; [[#Georgoulias--2019|Georgoulias et al., 2019]] ). Satellite-derived tropospheric NO <sub>2</sub> levels over Africa and Latin America do not show a clear trend; both increasing and decreasing trends are observed over large agglomerations in these regions since the early 2000s ( [[#Schneider--2015|Schneider et al., 2015]] ; [[#Duncan--2016|Duncan et al., 2016]] ).

In summary, global tropospheric NO <sub>x</sub> abundance has increased from 1850–2015 ( ''high confidence'' ). Satellite observations of tropospheric NO <sub>x</sub> indicate strong regional variations in trends over 2005–2015. There is ''high confidence'' that NO <sub>2</sub> has declined over the USA and Western Europe since the mid-1990s and increased over China until 2011. NO <sub>2</sub> trends have reversed (declining) over China beginning in 2012 and NO <sub>2</sub> has increased over Southern Asia by 50% since 2005 ( ''medium confidence'' ).

<div id="_idContainer022" class="_idGenObjectStyleOverride-1"></div>

'''Table 6.4 |''' '''Summary of the global CO trends based on model estimates and observations.'''

{| class="wikitable"
|-
! '''Analysis Period'''

! '''Trends: Regions'''

! '''Reference/Methodology'''

|-
| colspan="3"| '''Global/Hemispheric'''

|-
| 2003–2015

| –0.86% yr <sup>–1</sup>

| [[#Flemming--2017|Flemming et al. (2017)]] Model assimilating MOPITT

|-
| 2002–2013

| –1.4% yr <sup>–1</sup>

| [[#Gaubert--2017|Gaubert et al. (2017)]]

Model assimilating MOPITT

|-
| 2002–2018

| –0.50 ± 0.3% yr <sup>–1</sup> : 60°N–60°S (MOPITT)

–0.56 ± 0.3% yr <sup>–1</sup> ; <sup>–</sup> 0.61 ± 0.2% yr <sup>–1</sup> : 0°–60°N

–0.35 ± –0.3% yr <sup>–1</sup> ; -0.33±0.3% yr <sup>–1</sup> : 0°–60°S

| [[#Buchholz--2021|Buchholz et al. (2021)]]

Satellite Observations

MOPITT; AIRS

|-
| 2000–2017

| –0.32 ± 0.05% yr <sup>–1</sup>

| [[#Zheng--2019|Zheng et al. (2019)]]

Satellite Observations

MOPITT

|-
| 2003–2014

| around –2.5 to 0.5 ppb yr <sup>–1</sup> : Northern Hemisphere around –0.5 to 0 ppb yr <sup>–1</sup> : Southern Hemisphere

| [[#Flemming--2017|Flemming et al. (2017)]] NOAA Carbon Cycle Cooperative Global Air Sampling Network

|-
| 2001–2013

| –2.19 to –0.80 ppb yr <sup>–1</sup> : Northern Hemisphere (Upper Troposphere/Tropopause Layer)

| [[#Cohen--2018|Cohen et al. (2018)]] IAGOS Airborne

|-
| colspan="3"| '''Pacific/Tropics'''

|-
| 2004–2013

(Spring Mean)

| –2.9 ± 2.6 ppb yr <sup>–1</sup> : Mauna Loa (19.54°N, 155.58°W)

| [[#Gratz--2015|Gratz et al. (2015)]] Ground-based

|-
| 2004–2013

(Spring Mean)

| –2.6 ± 1.8 ppb yr <sup>–1</sup> : Sand Island Midway (28.21°N, 177.38°W)

| [[#Gratz--2015|Gratz et al. (2015)]] Ground-based

|-
| colspan="3"| '''Europe'''

|-
| 1996–2006

| –0.45 ± 0.16% yr <sup>–1</sup> : Jungfraujoch (46.6°N, 8.0°E)

–1.00 ± 0.24% yr <sup>–1</sup> : Zugspitze (47.4°N, 11.0°E)

–0.62 ± 0.19% yr <sup>–1</sup> : Harestua (60.2°N, 10.8°E)

0.61 ± 0.16% yr <sup>–1</sup> : Kiruna (67.8°N, 20.4°E)

| [[#Angelbratt--2011|Angelbratt et al. (2011)]] Ground-based

|-
| 2001–2011

May to Sep

| –3.1 ± 0.30 ppb yr <sup>–1</sup> : Pico Mt. Obs (38.47°N, 28.40°W) –1.4 ± 0.20 ppb yr <sup>–1</sup> : Mace Head, Ireland

| [[#Kumar--2013|Kumar et al. (2013)]] Ground-based

|-
| 2002–2018

| –-0.89 ± 0.1% yr <sup>–1</sup> : Europe (45°N–55°N, 0°E–15°E)

| [[#Buchholz--2021|Buchholz et al. (2021)]]

Satellite Observations

MOPITT

|-
| colspan="3"| '''North America'''

|-
| 2001–2010

| –2.5 ppb yr <sup>–1</sup> : Thompson Farm (43.11°N, 70.95°W)

–2.3 ppb yr <sup>–1</sup> : Mt. Washington (44.27°N, 71.30°W)

+2.8 ppb yr <sup>–1</sup> : Castle Springs (43.75°N, 71.35°W)

–3.5 ppb yr <sup>–1</sup> : Pack Monadnock (42.86°N, 71.88°W)

–2.8 ppb yr <sup>–1</sup> : Whiteface Mountain (44.40°N, 73.90°W)

–4.3 ppb yr <sup>–1</sup> : Pinnacle State Park (42.09°N, 77.21°W)

| [[#Zhou--2017|Zhou et al. (2017)]] Ground-based

|-
| 2004–2013

(Spring Mean)

| –3.2 ± 2.9 ppb yr <sup>–1</sup> : Mt. Bachelor Observatory

| [[#Gratz--2015|Gratz et al. (2015)]] Ground-based

|-
| 2004–2012

(Spring Mean)

| –2.8 ± 1.8 ppb yr <sup>–1</sup> : Shemya Island (55.21°N, 162.72°W)

| [[#Gratz--2015|Gratz et al. (2015)]] Ground-based

|-
| 2002–2018

| –0.85 ± 0.1%yr <sup>–1</sup> : Eastern USA (35°N–40°N, –95°E–75°E)

| [[#Buchholz--2021|Buchholz et al. (2021)]]

Satellite Observations

MOPITT

|-
| colspan="3"| '''Asia'''

|-
| 2005–2018

| –0.46 ± 0.14% yr <sup>–1</sup> : Eastern Asia

| [[#Zheng--2018a|Zheng et al. (2018a)]]

WDCGG Ground-based

|-
| 2005–2018

| –0.41 ± 0.09% yr <sup>–1</sup> : Eastern Asia

| [[#Zheng--2018a|Zheng et al. (2018a)]]

MOPITT

|-
| 2002–2018

| –1.18 ± 0.3% yr <sup>–1</sup> : (Northeast China 30°E–40°E, 110°E–123°E) –0.28 ± 0.2% yr <sup>–1</sup> : (North India 20°N–30°N, 70°E–95°E)

| [[#Buchholz--2021|Buchholz et al. (2021)]]

Satellite Observations

MOPITT

|}

<div id="6.3.3.2" class="h3-container"></div>

<span id="carbon-monoxide-co"></span>
==== 6.3.3.2 Carbon Monoxide (CO) ====

<div id="h3-10-siblings" class="h3-siblings"></div>

About half of the atmospheric CO burden is due to its direct emissions and the remainder is due to the atmospheric oxidation of methane and NMVOCs. Reaction with OH is the primary sink of CO with a smaller contribution from dry deposition.

Since AR5, advances in satellite retrievals (e.g., Worden et al. , 2013; Warner et al. , 2014; Buchholz et al. , 2021) , ground-based column observations (e.g., Zeng et al. , 2012; Té et al. , 2016) , airborne platforms (e.g., [[#Cohen--2018|Cohen et al., 2018]] ; [[#Petetin--2018|Petetin et al., 2018]] ), surface measurement networks (e.g., Andrews et al. , 2014; Schultz et al. , 2015; Prinn et al. , 2018; Pétron et al. , 2019) and assimilation products (e.g., [[#Deeter--2017|Deeter et al., 2017]] ; [[#Flemming--2017|Flemming et al., 2017]] ; [[#Zheng--2019|Zheng et al., 2019]] ) have resulted in better characterization of the present-day atmospheric CO distribution. Typical annual mean surface CO concentrations range from around 120 ppb in the Northern Hemisphere to around 40 ppb in the Southern Hemisphere ( [[#Pétron--2019|Pétron et al., 2019]] ). The sub-regional patterns in CO reflect the distribution of emissions sources. Seasonal hotspots are linked to areas of biomass burning in tropical South America, equatorial Africa, South East Asia and Australia. A study using data assimilation techniques estimates a global mean CO burden of 356 ± 27 Tg over the 2002–2013 period ( [[#Gaubert--2017|Gaubert et al., 2017]] ).

Global models generally capture the global spatial distribution of the observed CO concentrations but have regional biases of up to 50% (e.g., [[#Emmons--2020|Emmons et al., 2020]] ; [[#Horowitz--2020|Horowitz et al., 2020]] ). Despite updated emissions datasets, the global multi-model and single-model simulations persistently underestimate observed CO concentrations at northern high and mid-latitudes as well as in the Southern Hemisphere, but with smaller biases compared with that in the Northern Hemisphere (Naik et al. , 2013; Stein et al. , 2014; Monks et al. , 2015; Strode et al. , 2015). Models are biased high in the tropics, particularly over highly polluted areas in India and Eastern Asia ( [[#Strode--2016|Strode et al., 2016]] ; [[#Yarragunta--2017|Yarragunta et al., 2017]] ).

Estimates of global CO burden simulated by global models generally fall within the range of that derived from data assimilation techniques, though the spread across the models is large (Naik et al. , 2013; Stein et al. , 2014; Zeng et al. , 2015; Myriokefalitakis et al. , 2016) . There is a large diversity in model-simulated CO budget driven by uncertainties in CO sources and sinks, particularly those related to in situ production from NMVOCs and loss due to reaction with OH ( [[#Stein--2014|Stein et al., 2014]] ; [[#Zeng--2015|Zeng et al., 2015]] ; [[#Myriokefalitakis--2016|Myriokefalitakis et al., 2016]] ). Global CO budget analysis from a multi-model ensemble for more recent years, including results from the CMIP6 model runs, are not yet available.

Reconstructions of CO concentrations based on limited ice-core samples in the Northern Hemisphere high latitudes suggest CO mole fractions of about 145 ppb in the 1950s, which rose by 10–15 ppb in the mid- 1970s, and then declined by about 30–130 ppb by 2008 ( [[#Petrenko--2013|Petrenko et al., 2013]] ). The negative trends since the 1990s are often attributed to emissions regulations from road transportation in North America and Europe. Due to limited observations prior to the satellite era, long-term global CO trends are based on estimates from models. An increase of global CO burden of about 50% for the year 2000 relative to 1850 is found in CMIP6 ( [[#Griffiths--2021|Griffiths et al., 2021]] ).

The AR5 reported a global CO decline of about 1% yr <sup>–1</sup> based on satellite data from 2002–2010, but biases in instruments rendered ''low confidence'' in this trend. The AR5 also indicated a small CO decrease from in situ networks but did not provide quantitative estimates. New analysis of CO trends performed since AR5 and based on different observational platforms and assimilation products show a decline globally and over most regions during the last one to two decades with varying amplitudes partly depending on the period of analysis (Table 6.4). Inversion-based analysis attributes the global CO decline during the past two decades to decreases in anthropogenic and biomass-burning CO emissions despite probable increase in atmospheric CO chemical production (Gaubert et al. , 2017; Jiang et al. , 2017; Zheng et al. , 2019). Furthermore, [[#Buchholz--2021|Buchholz et al. (2021)]] report a slowdown in global CO decline in 2010–2018 compared to 2002–2010, although the magnitude and sign of this change in the trend varies regionally. Global models prescribed with emissions inventories developed prior to the CMIP6 inventory capture the declining observed CO trends over North America and Europe but not over Eastern Asia ( [[#Strode--2016|Strode et al., 2016]] ). CMIP6 models driven by CMIP6 emissions simulate a negative trend in global CO burden over the 1990–2020 period ( [[#Griffiths--2021|Griffiths et al., 2021]] ), however the simulated trends have not yet been evaluated against observations.

In summary, our understanding of present-day global CO distribution has increased since AR5 with newer and improved observations and reanalysis. There is ''high confidence'' that global CO burden is declining since 2000. Evidence from observational CO reanalysis suggests this decline is driven by reductions in anthropogenic CO emissions, however this is yet to be corroborated by global ESM studies with the most recent emissions inventories.

<div id="6.3.3.3Non-Methane" class="h3-container"></div>

<span id="non-methane-volatile-organic-compounds-nmvocs"></span>
==== 6.3.3.3 Non-Methane Volatile Organic Compounds (NMVOCs) ====

<div id="h3-11-siblings" class="h3-siblings"></div>

NMVOCs encompass thousands of compounds with lifetimes from hours to days to months, and abundances and chemical composition highly variable with respect to space and time. Although the biogenic source (Section 6.2.2) dominates the global NMVOC budget, anthropogenic activities are the main driver of long-term trends in the abundance of many compounds.

Information on the global distribution of individual NMVOCs is scarce, except for the less reactive compounds having lifetimes of several days to months. Based on measurements from polar firn air samples and ground-based networks, AR5 reported that the abundances of the predominantly anthropogenic light alkanes (C <sub>2</sub> -C <sub>5</sub> ) increased until 1980 and declined afterwards. The decline was attributed to air-quality emissions controls and to fugitive emissions decreases following the collapse of the Soviet Union ( [[#Simpson--2012|Simpson et al., 2012]] ). Since AR5, scarce ground-based measurements have shown that the decline in C <sub>2</sub> -C <sub>3</sub> alkanes ended around 2008 and their abundances are since growing again, which is primarily attributed to increasing North American emissions (Section 6.2.1). Furthermore, since AR5 the evolution of ethane levels during the past millennium was made accessible by analysis of ice-core samples ( [[#Nicewonger--2016|Nicewonger et al., 2016]] ). The large observed interpolar ratio of ethane in pre-industrial times (3.9) corroborates a large geologic source of ethane previously put forward by ( [[#Etiope--2009|Etiope and Ciccioli, 2009]] ), and narrows down its likely global magnitude ( [[#Nicewonger--2018|Nicewonger et al., 2018]] ) ( ''low to medium confidence'' ). The incorporation of geologic emissions in CCMs is not yet systematic though a one-model study has shown improved agreement of the results with observations ( [[#Dalsøren--2018|Dalsøren et al., 2018]] ).

Formaldehyde (HCHO) is a short-lived, high-yield product of NMVOC oxidation, and formaldehyde column data from satellite instruments can therefore inform on trends in anthropogenic NMVOC abundances over very industrialized regions. The AR5 reported significant positive trends in formaldehyde between 1997 and 2009 over northeastern China (4% yr <sup>–1</sup> ) and negative trends over northeastern US cities. Since AR5, there is ''robust evidence'' and ''high agreement'' of an upward trend of HCHO over eastern China, though large regional disparities exist in the trends ( [[#De%20Smedt--2015|De Smedt et al., 2015]] ; [[#Shen--2019|Shen et al., 2019]] ) with a possible negligible or decreasing trend over Beijing and the Pearl River Delta. In other world regions, in particular North America, there is ''limited'' to ''medium evidence'' for significant changes in the HCHO columns, except in regions where the trend is particularly strong (e.g., the Houston area: –2.2% yr <sup>–1</sup> over 2005–2014) and the Alberta oil sands (+3.8% yr <sup>–1</sup> ; <sup></sup> [[#Zhu--2017|Zhu et al., 2017]] ). Over the northeastern USA, even the sign of the trend differs between studies (De Smedt et al. , 2015; Zhu et al. , 2017) for reasons that are unclear.

In summary, after a decline between 1980 and 2008, abundances of light NMVOCs have increased again over the Northern Hemisphere due to the extraction of oil and gas in North America ( ''high confidence'' ). Trends in satellite HCHO observations, used as a proxy of anthropogenic NMVOC over industrialized areas, show a significant positive trend over eastern China ( ''high confidence'' ) but also indicate large regional disparities in the magnitude of the trends over China and even in their signs over North America.

<div id="6.3.3.4" class="h3-container"></div>

<span id="ammonia-nh-3"></span>
==== 6.3.3.4 Ammonia (NH <sub>3</sub> ) ====

<div id="h3-12-siblings" class="h3-siblings"></div>

Ammonia is the most abundant alkaline gas in the atmosphere. Its present-day source is dominated by livestock and crop production (Section 6.2). Ammonia reacts with nitric acid and sulphuric acid to produce ammonium sulphate and ammonium nitrate, which contribute to the aerosol burden (Section 6.3.5.2), promotes aerosol nucleation by stabilizing sulphuric acid clusters ( [[#Kirkby--2011|Kirkby et al., 2011]] ), and contributes to nitrogen deposition (Section 6.4.4; [[#Sheppard--2011|Sheppard et al., 2011]] ; [[#Flechard--2020|Flechard et al., 2020]] ). Trends in NH <sub>3</sub> were not assessed in AR5.

Considerable expansion of satellite (Clarisse et al. , 2009; [[#Shephard--2015|Shephard and Cady-Pereira, 2015]] ; Warner et al. , 2016) and ground-based observations (Miller et al. , 2014; Y. Li et al. , 2016; Pan et al. , 2018) has improved our understanding of the spatial distribution and seasonal to interannual variability of ammonia, and advanced its representations in models (e.g., [[#Zhu--2015|Zhu et al., 2015]] ). Regionally, peak NH <sub>3</sub> concentrations are observed over large agricultural (e.g., northern India, the USA Midwest and Central Valley) and biomass-burning regions, in good qualitative agreement with emissions inventories (Van Damme et al. , 2015, 2018) . However, several large agricultural and industrial hotspots have been found to be missing or greatly underestimated in emissions inventories (Van Damme et al. , 2018) . NH <sub>3</sub> exhibits a strong vertical gradient, with a maximum in the boundary layer ( [[#Schiferl--2016|Schiferl et al., 2016]] ), and can be transported into the upper troposphere and lower stratosphere (UTLS), particularly in the Asian Monsoon region, as indicated by observations ( [[#Froyd--2009|Froyd et al., 2009]] ; [[#Höpfner--2016|Höpfner et al., 2016]] , 2019) and theoretical considerations (Ge et al. , 2018) . There is a large range in the present-day NH <sub>34</sub> burden (from 0.04–0.7 TgN) simulated by CCMs, highlighting deficiencies in the process-level representation of NH <sub>3</sub> in current global models ( [[#Bian--2017|Bian et al., 2017]] ). The underestimate of surface NH <sub>3</sub> concentrations ( [[#Bian--2017|Bian et al., 2017]] ) further highlights such deficiencies and the limitations in comparing site-specific observations with relatively coarse-resolution models.

Observations show that NH <sub>3</sub> concentration has been increasing in recent decades in the USA (Butler et al. , 2016; Warner et al. , 2016; Yu et al. , 2018) , Western Europe (van Zanten et al. , 2017; Warner et al. , 2017; Wichink Kruit et al. , 2017; Tang et al. , 2018) , and China ( [[#Warner--2017|Warner et al., 2017]] ; M. [[#Liu--2018|]] [[#Liu--2018|Liu et al., 2018]] ). This trend has been attributed to a combination of increasing ammonia emissions ( [[#Sutton--2013|Sutton et al., 2013]] ; [[#Fowler--2015|Fowler et al., 2015]] ) and decreases in the chemical reaction of NH <sub>3</sub> with nitric and sulphuric acids associated with reductions in SO <sub>2</sub> and NO <sub>x</sub> emissions whose rate depends on the region ( [[#Warner--2017|Warner et al., 2017]] ; [[#Yao--2019|Yao and Zhang, 2019]] ). Over longer time scales, CCMs simulate an increase of the NH <sub>34</sub> burden by a factor of two to seven since pre-industrial conditions ( [[#Xu--2012|Xu and Penner, 2012]] ; [[#Hauglustaine--2014|Hauglustaine et al., 2014]] ).

In summary, progress has been made in the understanding of the spatio-temporal distribution of ammonia, though representation of NH <sub>3</sub> remains rather unsatisfactory due to process-level uncertainties. Evidence from observations and models suggests that ammonia concentrations have been increasing over recent decades due to emissions and chemistry. There is ''high confidence'' that the global NH <sub>34</sub> burden has increased considerably from the pre-industrial period to the present day, although the magnitude of the increase remains uncertain.

<div id="6.3.3.5" class="h3-container"></div>

<span id="sulphur-dioxide-so-2"></span>
==== 6.3.3.5 Sulphur Dioxide (SO <sub>2</sub> ) ====

<div id="h3-13-siblings" class="h3-siblings"></div>

The AR5 did not assess trends in SO <sub>2</sub> concentrations. Trends in SO <sub>2</sub> abundances are consistent with the overall anthropogenic emissions changes as presented in Section 6.2 and Figure 6.18. Long-term surface-based in situ observations in North America and Europe show reductions of more than 80% since the measurements began around 1980 (Table 6.5). Europe had the largest reductions in the first part of the period while the highest reduction came later in North America. Observed trends are qualitatively reproduced by global and regional models over North America and Europe during the period 1990–2015 for which emissions changes are well quantified (Table 6.5; [[#Aas--2019|Aas et al., 2019]] ).

<div id="_idContainer023" class="_idGenObjectStyleOverride-1"></div>

'''Table 6.5 |''' '''Summary of changes or trends in atmospheric abundance of sulphur dioxide (SO''' <sub>2</sub> ''') and sulphate (SO''' <sub>4</sub> <sup>2–</sup> ''') aerosols based on in situ and satellite observations.'''

{| class="wikitable"
|-
| '''Analysis Period'''

| '''Trends in SO''' <sub>2</sub>

| '''Trends in Particulate SO''' <sub>4</sub> <sup>2–</sup>

| '''Reference'''

|-
| colspan="4"| Global Models/Assimilated Models

|-
| 1990–2000

| –8.54 ± 1.40% yr <sup>–1</sup> (EU, 43 sites)

–2.63 ± 0.30% yr <sup>–1</sup> (NA, 53 sites)

| –5.23 ± 1.17% yr <sup>–1</sup> (EU, 41 sites)

–1.94 ± 0.43% yr <sup>—</sup> (NA 101 sites)

| [[#Aas--2019|Aas et al. (2019)]]

|-
| 2000–2015

| –0.41 ± 0.92% yr <sup>–1</sup> (EA, 19 sites)

–4.86 ± 1.31% yr <sup>–1</sup> (EU, 47 sites)

–4.40 ± 0.93% yr <sup>–1</sup> (NA, 77 sites)

| 0.02 ± 0.91% yr <sup>–1</sup> (EA, 13 sites)

–3.26 ± 0.85% yr <sup>–1</sup> (EU, 36 sites)

–3.18 ± 0.66% yr <sup>–1</sup> (NA, 218 sites)

| [[#Aas--2019|Aas et al. (2019)]]

|-
| colspan="4"| Ground-based In Situ Observations

|-
| 1980–1990

| –5.03 ± 2.04% yr <sup>–1</sup> (EU, 20 sites)

–2.5% yr <sup>–1</sup> (US)

| –2.56 ± 3.10% yr <sup>–1</sup> (EU, 16 sites)

–1.80 ± 4.09% yr <sup>–1</sup> (US SO <sub>4</sub> <sup>2–</sup> in precipitation, 78 sites)

| [[#Aas--2019|Aas et al. (2019)]]

[https://www.epa.gov/air-trends/sulfur-dioxide-trends US EPA] <sup>a</sup>

|-
| 1990–2000

| –7.56 ± 1.81% yr <sup>–1</sup> (EU, 43 sites)

–3.27 ± 1.69% yr <sup>–1</sup> (NA, 53 sites)

| –5.16 ± 2.11% yr <sup>–1</sup> (EU, 41 sites)

–2.08 ± 1.44% yr <sup>–1</sup> (NA, 101 sites)

| [[#Aas--2019|Aas et al. (2019)]]

|-
| 2000–2015

| –0.14 ± 5.32% yr <sup>–1</sup> (EA, 19 sites)

–3.89 ± 2.16% yr <sup>–1</sup> (EU, 47 sites)

–4.69 ± 1.35% yr <sup>–1</sup> (NA, 77 sites)

| 2.68 ± 9.41% yr <sup>–1</sup> (EA, 13 sites)

–2.67 ± 2.03% yr <sup>–1</sup> (EU, 36 sites)

–3.15 ± 1.30% yr <sup>–1</sup> (NA, 218 sites)

| [[#Aas--2019|Aas et al. (2019)]]

|-
| colspan="4"| Change Based on Satellite Observations

|-
| 2005–2015

| ca –80% (Eastern US)

|
| [[#Krotkov--2016|Krotkov et al. (2016)]]

|-
| 2005–2015

| ca –60% (Eastern EU)

|
| [[#Krotkov--2016|Krotkov et al. (2016)]]

|-
| 2005–2015

| 200 ± 50% (India)

|
| [[#Krotkov--2016|Krotkov et al. (2016)]]

|-
| 2005 (and 2012) –2015

| ca –50% (The North China Plain)

|
| [[#Krotkov--2016|Krotkov et al. (2016)]]

|}

<sup>a</sup> https://www.epa.gov/air-trends/sulfur-dioxide-trends

In situ observations over other parts of the world are scattered. However, the limited in situ observations in Eastern Asia indicate an increase in atmospheric SO <sub>2</sub> up to around 2005 and then a decline ( [[#Aas--2019|Aas et al., 2019]] ). This is confirmed by satellite observations ( [[#Krotkov--2016|Krotkov et al., 2016]] ), which further reveal a rapid decline in SO <sub>2</sub> since around 2012 or 2013 ( [[#Krotkov--2016|Krotkov et al., 2016]] ; [[#Zheng--2018b|Zheng et al., 2018b]] ). In India, on the other hand, SO <sub>2</sub> levels have doubled between 2005 and 2015 ( [[#Krotkov--2016|Krotkov et al., 2016]] ).

In summary, surface and satellite observations indicate strong regional variations in trends of atmospheric SO <sub>2</sub> abundance. The SO <sub>2</sub> concentrations in North America and Europe have declined over 1980–2015 with slightly stronger reductions in North America (70 ± 20%) than in Europe (58 ± 32%) over 2000–2015, though Europe had larger reductions than the US in the prior decade (1990–2000). In Asia, the SO <sub>2</sub> trends are more scattered, though there is ''medium confidence'' that there was a strong increase up to around 2005, followed by a steep decline in China, while over India, the concentrations are increasing steadily.

<div id="6.3.4" class="h2-container"></div>

<span id="short-lived-halogenated-species"></span>
=== 6.3.4 Short-lived Halogenated Species ===

<div id="h2-16-siblings" class="h2-siblings"></div>

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 [[IPCC:Wg1:Chapter:Annex-iii|Annex III]] and discussed in [[IPCC:Wg1:Chapter:Chapter-2|Chapter 2]] ( [[IPCC:Wg1:Chapter:Chapter-2#2.2.4|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 origin, while chlorinated species have significant additional anthropogenic sources ( [[#Carpenter--2014|Carpenter et al., 2014]] ; [[#Hossaini--2015|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--2018|Engel et al., 2018]] ). This increase is mostly due to dichloromethane (CH <sub>2</sub> cl <sub>2</sub> ), a species that has predominantly anthropogenic sources reflected by three-times higher concentrations in the Northern Hemisphere than in the Southern Hemisphere ( [[#Hossaini--2017|Hossaini et al., 2017]] ). The upward dichloromethane trend is corroborated by upper-tropospheric aircraft data over the period 1998–2014 ( [[#Elvidge--2015b|Elvidge et al., 2015b]] ; [[#Oram--2017|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--2018|Engel et al., 2018]] ).

<div id="6.3.5" class="h2-container"></div>

<span id="aerosols"></span>
=== 6.3.5 Aerosols ===

<div id="h2-17-siblings" class="h2-siblings"></div>

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 [[IPCC:Wg1:Chapter:Chapter-2#2.2.6|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 <sub>2.5</sub> , dominant aerosol type and aerosol composition, consistent with the findings in AR5.

<div id="_idContainer025" class="Basic-Text-Frame"></div>

[[File:179ed8251722ef95f81759ee7c0bd069 IPCC_AR6_WGI_Figure_6_7.png]]

'''Figure 6.7 |''' '''Distribution of PM''' <sub>2.5</sub> <sub></sub> '''composition mass concentration (in μg m''' <sup>–3</sup> ''') for the major PM''' <sub>2.5</sub> '''aerosol 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 <sub>2.5</sub> 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--2014|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--2015|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--2004|Celis et al., 2004]] ; [[#Feng--2006|Feng et al., 2006]] ; [[#Bourotte--2007|Bourotte et al., 2007]] ; [[#Fuzzi--2007|Fuzzi et al., 2007]] ; [[#Mariani--2007|Mariani and de Mello, 2007]] ; [[#Molina--2007|Molina et al., 2007]] , 2010; [[#Favez--2008|Favez et al., 2008]] ; [[#Mkoma--2008|Mkoma, 2008]] ; [[#Aggarwal--2009|Aggarwal and Kawamura, 2009]] ; [[#Mkoma--2009|Mkoma et al., 2009]] ; [[#de%20Souza--2010|de Souza et al., 2010]] ; [[#Li--2010|Li et al., 2010]] ; [[#Martin--2010|Martin et al., 2010]] ; [[#Radhi--2010|Radhi et al., 2010]] ; [[#Weinstein--2010|Weinstein et al., 2010]] ; [[#Batmunkh--2011|Batmunkh et al., 2011]] ; [[#Gioda--2011|Gioda et al., 2011]] ; [[#Pathak--2011|Pathak et al., 2011]] ; F. [[#Zhang--2012|]] [[#Zhang--2012|Zhang et al., 2012]] ; [[#Cho--2013|Cho and Park, 2013]] ; [[#Zhao--2013|Zhao et al., 2013]] ; [[#Wang--2019|Wang et al., 2019]] ; [[#Kuzu--2020|Kuzu et al., 2020]] ). Further details on data sources and processing are available in the chapter data table (Table 6.SM.3).

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 ( [[IPCC:Wg1:Chapter:Chapter-7#7.3.3.1|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--2021|Gliß et al., 2021]] ), although different remote-sensing estimates obtain different estimates of global mean AOD. [[#Gliß--2021|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--2020|Cherian and Quaas, 2020]] ; [[#Gliß--2021|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--2020|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 SO <sub>2</sub> 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 SO <sub>2</sub> 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 SO <sub>2</sub> emissions over Eastern Asia since 2011 is underestimated in the CMIP6 emissions database ( [[#Hoesly--2018|Hoesly et al., 2018]] ), indicating that the observed AOD change over Eastern Asia may not be captured accurately by CMIP6 models ( [[#Wang--2021|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--2020|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).

<div id="_idContainer027" class="Basic-Text-Frame"></div>

[[File:2de691d569e4677992259ed244abf7c1 IPCC_AR6_WGI_Figure_6_8.png]]

'''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).

<div id="6.3.5.1" class="h3-container"></div>

<span id="sulphate-so-4-2"></span>
==== 6.3.5.1 Sulphate (SO <sub>4</sub> <sup>2–</sup> ) ====

<div id="h3-14-siblings" class="h3-siblings"></div>

Sulphate aerosols (or sulphate-containing aerosols) are emitted directly or formed in the atmosphere by gas- and aqueous-phase oxidation of precursor sulphur gases, including SO <sub>2</sub> , DMS and carbonyl sulphide (OCS), emitted from anthropogenic and natural sources (Section 6.2). Sulphate aerosols influence climate forcing directly by either scattering solar radiation or absorbing longwave radiation, and indirectly by influencing cloud micro- and macrophysical properties and precipitation ( [[#Boucher--2013|Boucher et al., 2013]] ; [[#Myhre--2013|Myhre et al., 2013]] ). Additionally, sulphate aerosols and sulphate deposition have a large impact on air quality and ecosystems ( [[#Reis--2012|Reis et al., 2012]] ). The majority of sulphate particles are formed in the troposphere, however, SO <sub>2</sub> and other longer-lived natural precursors, such as OCS, transported into the stratosphere, contribute to the background stratospheric aerosol layer ( [[#Kremser--2016|Kremser et al., 2016]] ). SO <sub>2</sub> emissions from volcanic eruptions are a significant source of stratospheric sulphate loading (see [[IPCC:Wg1:Chapter:Chapter-2|Chapter 2]] for reconstruction of stratospheric aerosol optical depth and [[IPCC:Wg1:Chapter:Chapter-7|Chapter 7]] for radiative forcing of volcanic aerosols). Furthermore, studies suggest sulphate contributions from anthropogenic SO <sub>2</sub> emissions transported into the stratosphere could have a consequent impact on radiative forcing ( [[#Myhre--2004|Myhre et al., 2004]] ; [[#Yu--2016|Yu et al., 2016]] ). However, there is significant uncertainty in the relative importance of this stratospheric sulphate source ( [[#Kremser--2016|Kremser et al., 2016]] ).

Process understanding of sulphate production pathways from SO <sub>2</sub> emissions has seen some progress since AR5. More specifically, many global climate models now have a more complete description of chemical reactions such that oxidant levels (including ozone) are better described, include a pH-dependence of SO <sub>2</sub> oxidation (e.g., [[#Kirkevåg--2018|Kirkevåg et al., 2018]] ; [[#Bauer--2020|Bauer et al., 2020]] ) , and implement explicit descriptions of ammonium and nitrate aerosol components, which may influence the partitioning of sulphate (Bian et al. , 2017; Lund et al. , 2018a) . The pH influences the heterogeneous chemistry as well as the physical properties of the aerosols, and this topic has been a subject of growing interest since AR5 ( [[#Cheng--2016|Cheng et al., 2016]] ; [[#Freedman--2019|Freedman et al., 2019]] ; [[#Nenes--2020|Nenes et al., 2020]] ). Increases in cloudwater pH have been shown to significantly increase the radiative forcing due to sulphate aerosols ( [[#Turnock--2019|Turnock et al., 2019]] ).

Sulphate is removed from the atmosphere by dry deposition and wet scavenging, and these processes depend on the characteristics of the Earth’s surface, and the intensity, frequency and amount of precipitation ( [[#Boucher--2013|Boucher et al., 2013]] ). Even though there have been some improvements since AR5, representation of atmospheric transport and of wet scavenging and related cloud processes remains a key source of uncertainty in the simulated aerosol distribution and lifetime, with further consequences for the sulphate forcing estimates ( [[#Kristiansen--2016|Kristiansen et al., 2016]] ; [[#Lund--2018a|Lund et al., 2018a]] ). There are also still relatively large uncertainties in the emission height used in models affecting the simulated aerosol distribution (Yang et al. , 2019a) .

Based on long-term surface-based in situ observations, AR5 reported a strong decline in sulphate aerosols in Europe and the USA over 1990–2009, with the largest decreases occurring before 2000 in Europe and post-2000 in the USA. Since AR5, atmospheric measurements in conjunction with model results have provided insights into the spatial and temporal distribution of sulphate and sulphur deposition ( [[#Vet--2014|Vet et al., 2014]] ; [[#Tan--2018|Tan et al., 2018]] ; [[#Aas--2019|Aas et al., 2019]] ). The in situ observations in North America and Europe reveal substantial reduction since the measurements started around 1980, though the trends have not been linear through this period (Table 6.5). Several regional studies agree with these trend estimates for Europe (Banzhaf et al. , 2015; Theobald et al. , 2019) and North America ( [[#Sickles%20II--2015|Sickles II and Shadwick, 2015]] ; Paulot et al. , 2016) . Further, the concentrations of primary emitted SO <sub>2</sub> (Section 6.3.3.5) show greater decreases than secondary sulphate aerosols over these regions due to a combination of higher oxidation rate (hence more SO <sub>2</sub> converted to SO <sub>4</sub> <sup>2–</sup> ) and increased dry deposition rate of SO <sub>2</sub> (Fowler et al. , 2009; Banzhaf et al. , 2015) . In situ observations over other parts of the world are scattered (Figure 6.7), and the lack of observations makes it too uncertain to quantify regional representative trends ( [[#Hammer--2018|Hammer et al., 2018]] ). However, limited in situ observations in Eastern Asia indicate an increase in atmospheric sulphate up to around 2005 and then a decline ( [[#Aas--2019|Aas et al., 2019]] ), which is confirmed by satellite observations of SO <sub>2</sub> (Section 6.3.3.5). In India, on the other hand, satellite observations indicate a rapid increase in the SO <sub>2</sub> levels ( [[#Krotkov--2016|Krotkov et al., 2016]] ), and long-term measurements of sulphate in precipitation in India further provide evidence of an increasing trend from 1980–2010 ( [[#Bhaskar--2017|Bhaskar and Rao, 2017]] ; [[#Aas--2019|Aas et al., 2019]] ). Further improvements in global trend assessments are expected with new integrated reanalysis products from the Earth-system data assimilation projects ( [[#Randles--2017|Randles et al., 2017]] ; [[#Inness--2019|Inness et al., 2019]] ).

Indirect evidence of decadal trends in the atmospheric loading of sulphur are provided by Alpine ice cores, mainly influenced by European sources ( [[#Engardt--2017|Engardt et al., 2017]] ), and ice cores from Svalbard ( [[#Samyn--2012|Samyn et al., 2012]] ) and Greenland ( [[#Patris--2002|Patris et al., 2002]] ; [[#Iizuka--2018|Iizuka et al., 2018]] ) influenced by sources in Europe and North America. These show similar patterns with a weak increase from the end of the 19th century up to around 1950, followed by a steep increase up to around 1980, and then a significant decrease over the next two decades (see also [[IPCC:Wg1:Chapter:Chapter-2#2.2.6|Section 2.2.6]] ). This general trend is consistent with the emissions of SO <sub>2</sub> in North America and Europe (Figures 6.18 and 6.19; Hoesly et al. , 2018) .

Global and regional models qualitatively reproduce observed trends over North America and Europe for the period 1990–2015 for which emissions changes are generally well quantified ( [[#Aas--2019|Aas et al., 2019]] ; [[#Mortier--2020|Mortier et al., 2020]] ), building confidence in the relationship between emissions, concentration, deposition and radiative forcing derived from these models. However, the models seem to systematically underestimate sulphate ( [[#Bian--2017|Bian et al., 2017]] ; [[#Lund--2018a|Lund et al., 2018a]] ) and AOD ( [[#Lund--2018a|Lund et al., 2018a]] ; [[#Gliß--2021|Gliß et al., 2021]] ), and there are quite large differences in the models’ distribution of the concentration fields of sulphate driven by differences in the representation of photochemical production and sinks of aerosols. One global model study also highlighted biases in simulated sulphate trends over the 2001–2015 period over eastern China due to uncertainties in the CEDS anthropogenic SO <sub>2</sub> emissions trends (Paulot et al. , 2018a) .

In summary, there is ''high confidence'' that the global tropospheric sulphate burden increased from 1850 to around 2005, but there are large regional differences in the magnitude. Sulphate aerosol concentrations in North America and Europe have declined over 1980–2015 with slightly stronger reductions in North America (47 ± 20%) than in Europe (40 ± 30%) over 2000–2015, though Europe had larger reductions in the prior decade (1990–2000; 52 ± 21% and 21 ± 14% respectively for Europe and North America). In Asia, the trends are more scattered, though there is ''medium confidence'' that there was a strong increase up to around 2005, followed by a steep decline in China, while over India, the concentrations are increasing steadily.

<div id="6.3.5.2" class="h3-container"></div>

<span id="ammonium-nh-4-and-nitrate-aerosols-no-3"></span>
==== 6.3.5.2 Ammonium (NH <sub>4</sub> <sup>+</sup> ) and Nitrate Aerosols (NO <sub>3</sub> <sup>–</sup> ) ====

<div id="h3-15-siblings" class="h3-siblings"></div>

Ammonium sulphate and ammonium nitrate aerosols are formed when NH <sub>3</sub> reacts with nitric acid (HNO <sub>3</sub> ) and sulphuric acid (H <sub>2</sub> SO <sub>4</sub> ), produced in the atmosphere by the oxidation of NO <sub>x</sub> and SO <sub>2</sub> respectively. Ammonium nitrate is formed only after H <sub>2</sub> SO <sub>4</sub> is fully neutralized. NH <sub>4</sub> <sup>+</sup> and NO <sub>3</sub> <sup>–</sup> aerosols produced via these gas-to-particle reactions are a major fraction of fine-mode particles (with diameter &lt;1µm) affecting air quality and climate. Coarse-mode nitrate, formed by the heterogeneous reaction of nitric acid with dust and sea salt, dominates the overall global nitrate burden, but has little radiative effect ( [[#Hauglustaine--2014|Hauglustaine et al., 2014]] ; [[#Bian--2017|Bian et al., 2017]] ). Trends in ammonium (NH <sub>4</sub> <sup>+</sup> ) and nitrate (NO <sub>3</sub> <sup>–</sup> ) were not assessed in AR5.

Global model present-day estimates of the global NH <sub>4</sub> <sup>+</sup> burden range from 0.1–0.6 TgN ( [[#Bian--2017|Bian et al., 2017]] ). Models generally simulate surface NH <sub>4</sub> <sup>+</sup> concentrations better than surface NH <sub>3</sub> concentrations ( [[#Bian--2017|Bian et al., 2017]] ), which reflects its thermodynamic control by SO <sub>4</sub> <sup>2–</sup> rather than NH <sub>3</sub> ( [[#Shi--2017|Shi et al., 2017]] ). The concomitant increases of NH <sub>3</sub> , SO <sub>2</sub> , and NO <sub>x</sub> emissions (see Section 6.2) have led to a factor of three to nine increase in the simulated NH <sub>4</sub> <sup>+</sup> burden from 1850–2000 ( [[#Hauglustaine--2014|Hauglustaine et al., 2014]] ; [[#Lund--2018a|Lund et al., 2018a]] ), driven primarily by ammonium sulphate (70–90%). The increases in the NH <sub>3</sub> and NH <sub>4</sub> <sup>+</sup> burdens are indirectly supported by the observed increase of NH <sub>4</sub> <sup>+</sup> concentration in ice cores in mid- to high latitudes (Kang et al. , 2002; Kekonen et al. , 2005; Lamarque et al. , 2013; Iizuka et al. , 2018) .

Ammonium nitrate is semi-volatile, which results in complex spatial and temporal patterns in its concentrations ( [[#Putaud--2010|Putaud et al., 2010]] ; [[#Hand--2012|Hand et al., 2012]] a; H. [[#Zhang--2012|]] [[#Zhang--2012|Zhang et al., 2012]] ), reflecting variations in its precursors, NH <sub>3</sub> and HNO <sub>3</sub> , as well as SO <sub>4</sub> <sup>2–</sup> , non-volatile cations, temperature and relative humidity (Nenes et al. , 2020) . High relative humidity and low temperature as well as elevated fine particulate matter loading (Huang et al. , 2014; Petit et al. , 2015; H. Li et al. , 2016; Sandrini et al. , 2016) favour nitrate production. Measurements reveal the high contribution of NO <sub>3</sub> <sup>–</sup> to surface PM <sub>2.5</sub> (&gt;30%) in regions with elevated regional NO <sub>x</sub> and NH <sub>3</sub> emissions, such as the Paris area ( [[#Beekmann--2015|Beekmann et al., 2015]] ; [[#Zhang--2019|Zhang et al., 2019]] ), northern Italy ( [[#Masiol--2015|Masiol et al., 2015]] ; [[#Ricciardelli--2017|Ricciardelli et al., 2017]] ), Salt Lake City ( [[#Kuprov--2014|Kuprov et al., 2014]] ; [[#Franchin--2018|Franchin et al., 2018]] ), the North China Plains ( [[#Guo--2014|Guo et al., 2014]] ; [[#Chen--2016|Chen et al., 2016]] ) and New Delhi ( [[#Pant--2015|Pant et al., 2015]] ). Recent observations also show that ammonium nitrate contributes to the Asian tropopause aerosol layer ( [[#Vernier--2018|Vernier et al., 2018]] ; [[#Höpfner--2019|Höpfner et al., 2019]] ). Model diversity in simulating the present-day global fine-mode NO <sub>3</sub> <sup>–</sup> burden is large with two multi-model intercomparison studies reporting estimates in the range of 0.14–1.88 Tg and 0.08–0.93 Tg respectively ( [[#Bian--2017|Bian et al., 2017]] ; [[#Gliß--2021|Gliß et al., 2021]] ). Models differ in their estimates of the global tropospheric nitrate burden by up to a factor of 13 with differences remaining nearly the same across the CMIP5 and CMIP6 generation of models ( [[#Bian--2017|Bian et al., 2017]] ; [[#Gliß--2021|Gliß et al., 2021]] ). While regional patterns in the concentration of fine-mode NO <sub>3</sub> <sup>–</sup> are qualitatively captured by models, the simulation of fine-mode NO <sub>3</sub> <sup>–</sup> is generally worse than that of NH <sub>4</sub> <sup>+</sup> or SO <sub>4</sub> <sup>2–</sup> ( [[#Bian--2017|Bian et al., 2017]] ). This can be partly attributed to the semi-volatile nature of ammonium nitrate and biases in the simulation of its precursors ( [[#Heald--2014|Heald et al., 2014]] ; [[#Paulot--2016|Paulot et al., 2016]] ), including the sub-grid scale heterogeneity in NO <sub>x</sub> and NH <sub>3</sub> emissions ( [[#Zakoura--2018|Zakoura and Pandis, 2018]] ).

Models indicate that the burden of fine-mode NO <sub>3</sub> <sup>–</sup> has increased by a factor of two to five from 1850–2000 ( [[#Xu--2012|Xu and Penner, 2012]] ; [[#Hauglustaine--2014|Hauglustaine et al., 2014]] ; [[#Lund--2018a|Lund et al., 2018a]] ), an increase that has accelerated between 2001 and 2015 ( [[#Lund--2018a|Lund et al., 2018a]] ; [[#Paulot--2018b|Paulot et al., 2018b]] ). The sensitivity of NO <sub>3</sub> <sup>–</sup> to changes in NH <sub>3</sub> , SO <sub>4</sub> <sup>2–</sup> , and HNO <sub>3</sub> is determined primarily by aerosol pH, temperature, and aerosol liquid water ( [[#Guo--2016|Guo et al., 2016]] , 2018; [[#Weber--2016|Weber et al., 2016]] ; [[#Nenes--2020|Nenes et al., 2020]] ). In regions where aerosol pH is high, changes in NO <sub>3</sub> <sup>–</sup> follow changes in NO <sub>x</sub> emissions, consistent with the observed increase of ammonium nitrate in northern China from 2000–2015 ( [[#Wen--2018|Wen et al., 2018]] ) and its decrease in the US Central Valley ( [[#Pusede--2016|Pusede et al., 2016]] ). In contrast, the decrease in SO <sub>2</sub> emissions in the south-east USA has caused little change in NO <sub>3</sub> <sup>–</sup> <sub></sub> from 1998–2014 as nitric acid largely remains in the gas phase due to highly acidic aerosols ( [[#Weber--2016|Weber et al., 2016]] ; [[#Guo--2018|Guo et al., 2018]] ).

In summary, there is ''high confidence'' that the NH <sub>4</sub> <sup>+</sup> and NO <sub>3</sub> <sup>–</sup> burdens have increased from the pre-industrial period to the present day, although the magnitude of the increase is uncertain especially for NO <sub>3</sub> <sup>–</sup> . The sensitivity of NH <sub>4</sub> <sup>+</sup> and NO <sub>3</sub> <sup>–</sup> to changes in NH <sub>3</sub> , H <sub>2</sub> SO <sub>4</sub> and HNO <sub>3</sub> is well understood theoretically. However, it remains challenging to represent in models, in part because of uncertainties in the simulation of aerosol pH, and only a minority of ESMs consider nitrate aerosols in CMIP6.

<div id="6.3.5.3" class="h3-container"></div>

<span id="carbonaceous-aerosols"></span>
==== 6.3.5.3 Carbonaceous Aerosols ====

<div id="h3-16-siblings" class="h3-siblings"></div>

Carbonaceous aerosols are black carbon (BC) <sup>[[#footnote-002|3]]</sup> , which is soot made almost purely of carbon, and organic aerosols <sup>[[#footnote-001|4]]</sup> (OA), which also contain hydrogen and oxygen and can be of both primary (POA) or secondary (SOA) origin. BC and a fraction of OA called brown carbon (BrC) absorb solar radiation. The various components of carbonaceous aerosols have different optical properties, so the knowledge of their partition, mixing, coating and ageing is essential to assess their climate effect ( [[IPCC:Wg1:Chapter:Chapter-7#7.3.3.1.2|Section 7.3.3.1.2]] ).

Carbonaceous aerosols receive attention in the scientific and policy arena due to their radiative forcing, and their sizeable contribution to PM in an air-quality context (Rogelj et al. , 2014b; Harmsen et al. , 2015; Shindell et al. , 2016; Haines et al. , 2017; Myhre et al. , 2017) . BC exerts a positive forcing, but the forcing from carbonaceous aerosol as a whole is negative ( [[#Bond--2013|Bond et al., 2013]] ; [[#Thornhill--2021b|Thornhill et al., 2021b]] ). On average, carbonaceous aerosols account for 50–70% of PM with a diameter lower than 1 µm in polluted and pristine areas (Zhang et al. , 2007; Carslaw et al. , 2010; Andreae et al. , 2015; Monteiro dos Santos et al. , 2016; Chen et al. , 2017) .

An extensive review on BC ( [[#Bond--2013|Bond et al., 2013]] ) discussed limitations in inferring its atmospheric abundance and highlighted inconsistencies between different terminology and related measurement techniques ( [[#Petzold--2013|Petzold et al., 2013]] ; [[#Sharma--2017|Sharma et al., 2017]] ). Due to a lack of global observations, AR5 only reported declining total carbonaceous aerosol trends from the USA and a declining BC trend from the Arctic based on data available up to 2008. Since AR5, the number of observation sites has grown worldwide (Figure 6.7) but datasets suitable for global trend analyses remain limited ( [[#Reddington--2017|Reddington et al., 2017]] ; [[#Laj--2020|Laj et al., 2020]] ). Locally, studies based on observations from rural and background sites have reported decreasing surface carbonaceous aerosol trends in the Arctic, Europe, the USA, Japan and India (Table 6.6). Increases in carbonaceous aerosol concentrations in some rural sites of the western USA have been associated with wildfires ( [[#Hand--2013|Hand et al., 2013]] ; [[#Malm--2017|Malm et al., 2017]] ). Long-term OA observations are scarce, so their trends outside of the USA are difficult to assess. Ice-core analysis has provided insight into carbonaceous aerosol trends predating the satellite and observation era over the Northern Hemisphere ( [[IPCC:Wg1:Chapter:Chapter-2#2.2.6|Section 2.2.6]] , Figure 2.9b).

<div id="_idContainer028" class="_idGenObjectStyleOverride-1"></div>

'''Table 6.6 |''' '''Summary of the regional carbonaceous aerosol trends at background observation sites.'''

{| class="wikitable"
|-
| Species

| Analysis Period

| Change/Trends

| References

|-
| rowspan="6"| BC

| 1990–2009

| Arctic Sites (Alert, Barrow, Ny-Alesund)

−2% yr <sup>−1</sup>

| [[#Sharma--2013|Sharma et al. (2013)]]

|-
| 1970–2010

| Finland (Kevo remote site)

−1.8% yr <sup>–1</sup>

| [[#Dutkiewicz--2014|Dutkiewicz et al. (2014)]]

|-
| 2005–2014

| Germany (rural site)

−2% yr <sup>–1</sup>

| [[#Kutzner--2018|Kutzner et al. (2018)]]

|-
| 2009–2016

| United Kingdom (Harwell rural site)

−8% yr <sup>–1</sup>

| [[#Singh--2018|Singh et al. (2018)]]

|-
| 2009–2019

| Japan (Fukue Island)

−5.8 ± 1.5% yr <sup>–1</sup>

| [[#Kanaya--2020|Kanaya et al. (2020)]]

|-
| 2009–2015

| India (Darjeeling mountain site)

−5% yr <sup>–1</sup>

| [[#Sarkar--2019|Sarkar et al. (2019)]]

|-
| OA

| 2001–2015

| USA (IMPROVE sites east of 100°W)

−2% yr <sup>–1</sup>

| [[#Malm--2017|Malm et al. (2017)]]

|-
| rowspan="2"| Total Carbon (EC + OC)

| 1990–2010

| USA (IMPROVE sites)

Western USA: −4 to −5% yr <sup>–1</sup>

Eastern USA: −1 to −2% yr <sup>–1</sup>

| [[#Hand--2013|Hand et al. (2013)]]

|-
| 2002–2010

| Spain (Montseny rural site)

−5% yr <sup>–1</sup>

| [[#Querol--2013|Querol et al. (2013)]]

|}

Knowledge of carbonaceous aerosol atmospheric abundance continues to rely on global models due to a lack of global-scale observations. For BC, models agree within a factor of two with measured surface mass concentrations in Europe and North America, but underestimate concentrations at the Arctic surface by one to two orders of magnitude, especially in winter and spring ( [[#Lee--2013|Lee et al., 2013]] ; [[#Lund--2018a|Lund et al., 2018a]] ). For OA, AeroCom models underestimate surface mass concentrations by a factor of two over urban areas, as their low horizontal resolution prevents them from resolving local pollution peaks ( [[#Tsigaridis--2014|Tsigaridis et al., 2014]] ; [[#Lund--2018a|Lund et al., 2018a]] ). Models agree within a factor of two with OA surface concentrations measured at remote sites, where surface concentrations are more spatially uniform ( [[#Tsigaridis--2014|Tsigaridis et al., 2014]] ).

BC and OA lifetimes are estimated to be 5.5 days ± 35% and 6.0 days ± 29% (median ± 1 standard deviation), respectively, based on an ensemble of 14 models ( [[#Gliß--2021|Gliß et al., 2021]] ). Disagreement in simulated lifetime leads to horizontal and vertical variations in predicted carbonaceous aerosol concentrations, with implications for radiative forcing ( [[#Samset--2013|Samset et al., 2013]] ; [[#Lund--2018b|Lund et al., 2018b]] ). Airborne campaigns have provided valuable vertical-profile measurements of carbonaceous aerosol concentrations (Schwarz et al. , 2013; Freney et al. , 2018; Hodgson et al. , 2018; Schulz et al. , 2019; D. Zhao et al. , 2019; Morgan et al. , 2020). Compared to those measurements, models tend to transport BC too high in the atmosphere, suggesting that lifetimes are not larger than 5.5 days ( [[#Samset--2013|Samset et al., 2013]] ; [[#Lund--2018b|Lund et al., 2018b]] ). Newly developed size-dependent wet-scavenging parametrization for BC ( [[#Taylor--2014|Taylor et al., 2014]] ; [[#Schroder--2015|Schroder et al., 2015]] ; [[#Ohata--2016|Ohata et al., 2016]] ; G. [[#Zhang--2017|]] [[#Zhang--2017|Zhang et al., 2017]] ; [[#Ding--2019|Ding et al., 2019]] ; [[#Moteki--2019|Moteki et al., 2019]] ; [[#Motos--2019|Motos et al., 2019]] ) may lead to decreased BC lifetimes and improve agreement with observed vertical profiles.

Simulated BC burdens show a large spread among models ( [[#Gliß--2021|Gliß et al., 2021]] ), despite using harmonised primary emissions, because of differences in BC removal efficiency linked to different treatment of ageing and mixing, particularly in strong source regions. The multi-model median BC burden for the year 2010 from [[#Gliß--2021|Gliß et al. (2021)]] , based on 14 AeroCom models, is 0.131 ± 0.047 Tg (median ± 1 standard deviation). The range encompasses values reported by independent single-model estimates ( Huang et al. , 2013; Lee et al. , 2013; Sharma et al. , 2013; Q. Wang et al. , 2014; Tilmes et al. , 2019) .

Simulated OA burdens also show a large spread among global models, with [[#Gliß--2021|Gliß et al. (2021)]] reporting a multi-model median of 1.91 ± 0.65 Tg for the year 2010. The large spread reflects the wide range in the complexity of the OA parametrizations, particularly for SOA formation, as well as in the primary OA emissions ( [[#Tsigaridis--2014|Tsigaridis et al., 2014]] ; [[#Gliß--2021|Gliß et al., 2021]] ). The uncertainties are particularly large in model estimates of SOA production rates, which vary between 10 and 143 Tg yr <sup>–1</sup> ( [[#Tsigaridis--2014|Tsigaridis et al., 2014]] ; [[#Hodzic--2016|Hodzic et al., 2016]] ; [[#Tilmes--2019|Tilmes et al., 2019]] ). While the level of complexity in the representation of OA in global models has increased since AR5 ( [[#Shrivastava--2017|Shrivastava et al., 2017]] ; [[#Hodzic--2020|Hodzic et al., 2020]] ), limitations in process-level understanding of the formation, ageing and removal of organic compounds lead to uncertainties in the global model predictions of global OA burden and distribution as well as the relative contribution of POA and SOA to OA. [[#Jo--2016|Jo et al. (2016)]] estimated that BrC contributes about 20% of total OA burden. That would give BrC a burden similar to that of BC ( ''low confidence'' ), enhancing the overall forcing exerted by carbonaceous aerosol absorption ( [[#Zhang--2020|Zhang et al., 2020]] ).

In summary, the lack of global-scale observations of carbonaceous aerosols, the complexity of processes influencing them, and the large spread in their simulated global budget and burdens means that there is only ''low confidence'' in the quantification of the present-day atmospheric distribution of individual components of carbonaceous aerosols. Global trends in carbonaceous aerosols cannot be characterized due to limited observations, but sites representative of background conditions have reported multi-year declines in BC over several regions of the Northern Hemisphere.

<div id="6.3.6" class="h2-container"></div>

<span id="implications-of-slcf-abundances-for-atmospheric-oxidizing-capacity"></span>
=== 6.3.6 Implications of SLCF Abundances for Atmospheric Oxidizing Capacity ===

<div id="h2-18-siblings" class="h2-siblings"></div>

The atmospheric oxidising capacity is determined primarily by tropospheric hydroxyl (OH) radical and to a smaller extent by NO <sub>3</sub> radical, ozone, hydrogen peroxide (H <sub>2</sub> O <sub>2</sub> ) 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 <sub>x</sub> leads to the production of tropospheric ozone. OH also contributes to the formation of aerosols from oxidation of SO <sub>2</sub> 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 <sub>x</sub> emissions control the secondary OH production, while in pristine air it occurs via other mechanisms involving, in particular, isoprene ( [[#Lelieveld--2016|Lelieveld et al., 2016]] ; [[#Wennberg--2018|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 <sub>x</sub> 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 <sub>x</sub> 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 <sub>x</sub> , 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 <sub>x</sub> emissions and deep convection (Turner et al. , 2018) , and fire emissions ( [[#Rowlinson--2019|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--2018|Prinn et al., 2018]] and references therein) or derived from global atmospheric chemistry models ( [[#Lelieveld--2016|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--2013|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--2017|Rigby et al., 2017]] ; [[#Turner--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--2019|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--2020|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--2017|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 <sub>x</sub> and decreasing CO emissions ( [[#Stevenson--2020|Stevenson et al., 2020]] ).

<div id="_idContainer029" class="_idGenObjectStyleOverride-1"></div>

'''Table 6.7 |''' '''Summary of global OH trends and interannual variability from studies post 2010.'''

{| class="wikitable"
|-
| '''Reference'''

| '''Time Period'''

| '''OH Trends and IAV'''

| '''Approach'''

|-
| colspan="4"| Inversion and Empirical Methods Based on Observations

|-
| [[#Montzka--2011|Montzka et al. (2011)]]

| 1998–2007

| 2.3 ± 1.3% (IAV)

| 3D inversion

|-
| [[#Ciais--2013|Ciais et al. (2013)]]

| 2000s

| within ±5% (IAV)

| AR5 based on inversions

|-
| [[#McNorton--2016|McNorton et al. (2016)]]

| 1993–2011

| ±2.3% (IAV)

| Box-model inversion

|-
| [[#Rigby--2017|Rigby et al. (2017)]]

| 1980–2014

| 10% increase from the late 1990s–2004; 10% decrease from 2004–2014

| Box-model inversion

|-
| [[#Turner--2017|Turner et al. (2017)]]

| 1983–2015

| about 7% increase in 1991–2001; 7% decrease in 2003–2016

| Box-model inversion

|-
| [[#Nicely--2018|Nicely et al. (2018)]]

| 1980–2015

| 1.6% (IAV)

| Empirical reconstruction

|-
| [[#McNorton--2018|McNorton et al. (2018)]]

| 2003–2015

| 1.8 ± 0.4% decrease

| 3D inversion

|-
| [[#Naus--2019|Naus et al. (2019)]]

| 1994–2015

| 3.8 ± 3.2% increase

| Box-model inversion

|-
| [[#Patra--2021|Patra et al. (2021)]]

| 1996–2015

| 2–3% IAV, no trend

| 3D inversion

|-
| colspan="4"| Global CTMs, CCMs and ESMs

|-
| [[#John--2012|John et al. (2012)]]

| 1860–2005

1980–2000

| 6% decrease

About 3% increase

| CCM

|-
| [[#Holmes--2013|Holmes et al. (2013)]]

| 1997–2009

| 0.7–1.1% (IAV)

| Multi-model CTMs

|-
| [[#Ciais--2013|Ciais et al. (2013)]]

| 2000s

| within ±3% (IAV)

| AR5 based on CCMs

|-
| [[#Murray--2013|Murray et al. (2013)]]

| 1998–2006

| increasing trend

| 3D CTM

|-
| [[#Naik--2013|Naik et al. (2013)]]

| 1980–2000

1850–2000

| 3.5 ± 2.2 % increase

−0.6 ± 8.8%

| Multi-model CCMs/CTMs

|-
| [[#Murray--2014|Murray et al. (2014)]]

| 1770s–1990s

| 5.3% increase

| CCM

|-
| [[#Dalsøren--2016|Dalsøren et al. (2016)]]

| 1970–2012

| 8% increase

| 3D CTM

|-
| [[#Gaubert--2017|Gaubert et al. (2017)]]

| 2002–2013

| 7% increase

| CCM with assimilated satellite CO observations

|-
| [[#Zhao--2019|]] [[#Zhao--2019|]] [[#Zhao--2019|Zhao et al. (2019)]]

| 1960–2010

1980–2000

| 1.9 ± 1.2 % (IAV)

4.6 ± 2.4 % increase

| Multi-model CCMs/CTMs

|-
| [[#Stevenson--2020|Stevenson et al. (2020)]]

| 1980–2014

1850–1980

| 9% increase

no trend

| Multi-model ESMs

|}

<div id="_idContainer031" class="Basic-Text-Frame"></div>

[[File:297c58cb2efbea12401a3db5425cecee IPCC_AR6_WGI_Figure_6_9.png]]

'''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).

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--2015|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 <sub>x</sub> and BVOCs, and model representation of isoprene oxidation chemistry ( [[#Achakulwisut--2015|Achakulwisut et al., 2015]] ; [[#Hopcroft--2017|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--2013|Naik et al., 2013]] ; [[#Voulgarakis--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--2017|Nicely et al., 2017]] ; [[#Wild--2020|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--2020|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 <sub>x</sub> 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.

<div id="6.4" class="h1-container"></div>

<span id="slcf-radiative-forcing-and-climate-effects"></span>