Volume 11, Issue 12 e2023EF003697
Research Article
Open Access

Inequalities in Air Pollution Exposure and Attributable Mortality in a Low Carbon Future

C. L. Reddington

Corresponding Author

C. L. Reddington

Institute for Climate and Atmospheric Science (ICAS), School of Earth and Environment, University of Leeds, Woodhouse, UK

Correspondence to:

C. L. Reddington,

[email protected]

Search for more papers by this author
S. T. Turnock

S. T. Turnock

Met Office Hadley Centre, Exeter, UK

University of Leeds Met Office Strategic (LUMOS) Research Group, University of Leeds, Woodhouse, UK

Search for more papers by this author
L. Conibear

L. Conibear

The Tomorrow Companies Inc., Boston, MA, USA

Search for more papers by this author
P. M. Forster

P. M. Forster

Priestley Centre for Climate Futures, University of Leeds, Leeds, UK

Search for more papers by this author
J. A. Lowe

J. A. Lowe

Met Office Hadley Centre, Exeter, UK

Priestley International Centre for Climate, University of Leeds, Leeds, UK

Search for more papers by this author
L. Berrang Ford

L. Berrang Ford

Priestley International Centre for Climate, University of Leeds, Leeds, UK

Search for more papers by this author
C. Weaver

C. Weaver

Institute for Climate and Atmospheric Science (ICAS), School of Earth and Environment, University of Leeds, Woodhouse, UK

Priestley International Centre for Climate, University of Leeds, Leeds, UK

Search for more papers by this author
B. van Bavel

B. van Bavel

Priestley International Centre for Climate, University of Leeds, Leeds, UK

Search for more papers by this author
H. Dong

H. Dong

School of Environmental Science and Engineering, Shanghai Jiao Tong University, Shanghai, China

Search for more papers by this author
M. R. Alizadeh

M. R. Alizadeh

Department of Bioresource Engineering, McGill University, Montreal, QC, Canada

Search for more papers by this author
S. R. Arnold

S. R. Arnold

Institute for Climate and Atmospheric Science (ICAS), School of Earth and Environment, University of Leeds, Woodhouse, UK

Search for more papers by this author
First published: 15 December 2023

Abstract

Understanding the costs and benefits of climate change mitigation and adaptation options is crucial to justify and prioritize future decarbonization pathways to achieve net zero. Here, we quantified the co-benefits of decarbonization for air quality and public health under scenarios that aim to limit end-of-century warming to 2°C and 1.5°C. We estimated the mortality burden attributable to ambient PM2.5 exposure using population attributable fractions of relative risk, incorporating projected changes in population demographics. We found that implementation of decarbonization scenarios could produce substantial global reductions in population exposure to PM2.5 pollution and associated premature mortality, with maximum health benefits achieved in Asia around mid-century. The stringent 1.5ºC-compliant decarbonization scenario (SSP1-1.9) could reduce the PM2.5-attributable mortality burden by 29% in 2050 relative to a middle-of-the-road scenario (SSP2-4.5), averting around 2.9 M annual deaths worldwide. While all income groups were found to benefit from improved air quality through a combination of decarbonization and air pollution controls, the smallest health benefits are experienced by the low-income population. The disparity in PM2.5 exposure across income groups is projected to reduce by 2100, but a 30% disparity between high- and low-income groups persists even in the strongest mitigation scenario. Further, without additional and targeted air quality measures, low- and lower-middle-income populations (predominantly in Africa and Asia) will continue to experience PM2.5 exposures that are over three times the World Health Organization Air Quality Guideline.

Key Points

  • Decarbonization has the potential to generate substantial health co-benefits by averting millions of premature deaths associated with PM2.5 exposure across all income groups

  • The low-income population is predicted to experience the smallest health benefits of decarbonization and continue to be exposed to PM2.5 concentrations that are over three times that of the World Health Organization Air Quality Guideline

  • Under a decarbonization future pathway, the global socioeconomic disparity in PM2.5 exposure reduces but persists at around 30% by the end of the century

Plain Language Summary

Implementation of decarbonization strategies to mitigate future climate change can provide additional benefits or “co-benefits” through improved air quality and public health. Quantifying these benefits and how they manifest across different world regions and income groups is essential to incentivize climate action. In this work we have quantified the air pollution health co-benefits for three different possible future scenarios: one “middle-of-the-road” scenario and two decarbonization scenarios. We found that by following a future decarbonization pathway instead of a “middle-of-the-road” pathway, can generate substantial air quality and public health benefits worldwide, particularly in Asia around 2050. While all income groups were found to benefit from improved air quality through decarbonization, the smallest health benefits are experienced by the low-income population. Inequalities in air pollution exposure between the lower-income and high-income groups were found to reduce rapidly under a decarbonization pathway, but persist through to 2100 even under the strongest mitigation. Further, without additional and targeted air quality measures, low- and lower-middle-income populations (predominantly in Africa and Asia) will continue to experience air pollution levels that exceed the World Health Organization Air Quality Guideline.

1 Introduction

The Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment report confirmed the need for rapid reductions in both carbon dioxide emissions and in emissions of other greenhouse gases (GHGs) such as methane by 2030 (Riahi et al., 2022). These reductions toward a mid-century net zero target are the minimum necessary to satisfy the Paris Agreement temperature goals. At the same time actual policies are falling short (UNEP, 2022). Climate change mitigation can be incentivized by the realization of co-benefits, such as improved health, wealth, air quality, water availability, and access to nature (IPCC, 2022). These benefits and where they fall are poorly quantified, leading to difficulties designing co-benefits into specific policies and interventions. Here we detail the quantification of one possible co-benefit of future decarbonization: air quality health effects.

Long-term exposure to ambient fine particulate matter (PM2.5) is associated with a range of negative health outcomes including cardiovascular diseases, respiratory diseases, lung cancer, and subsequent premature mortality (Chen & Hoek, 2020; Liu et al., 2021; Park et al., 2021; Yu et al., 2021; Yuan et al., 2019; Zhu et al., 2021). At present, exposure to ambient PM2.5 pollution is the largest environmental risk factor for disease and premature death globally (GBD 2019 Risk Factor Collaborators, 2020). Previous studies have demonstrated that emissions from the combustion of fossil fuels (coal, oil, and natural gas) are a major contributor to the global premature mortality burden attributable to ambient PM2.5 exposure (Lelieveld et al., 2019; McDuffie et al., 2021; Vohra et al., 2021). Emissions from residential combustion of coal or solid biofuel are a major contributor to regional ambient PM2.5-attributable premature mortality, particularly across South and East Asia (Reddington et al., 2019) and West Africa (Gordon et al., 2023). Over recent decades, emission control efforts have delivered notable reductions in PM2.5 exposure across some regions, such as North America (Butt et al., 2017) and Europe (Turnock et al., 2016), and more recently in China (e.g., Conibear, Reddington, Silver, Chen, Arnold, et al., 2022; Silver, Conibear, et al., 2020; Silver, He, et al., 2020). Despite the reductions across these regions, ambient PM2.5 exposure has been increasing globally, with increases mainly occurring in countries with a low to middle socioeconomic status for example, countries in South Asia, Southeast Asia, North Africa, West Africa, and the Middle East (GBD 2019 Risk Factor Collaborators, 2020). The increases in PM2.5 exposure in these regions are likely to be linked to changes in anthropogenic air pollutant emissions (e.g., Koplitz et al., 2017; Osipov et al., 2022; Shi et al., 2020; Xu et al., 2019), although other factors may play a role.

Ambient PM2.5 pollution exposure is often greater in populations with a lower socioeconomic status compared to those with a high socioeconomic status (Fairburn et al., 2019; Hajat et al., 2015; Miao et al., 2015), with low- and middle-income countries in Asia and Africa experiencing some of the highest PM2.5 concentrations globally (Shaddick et al., 2020; WHO, 2022). Sub-national PM2.5 exposure inequalities are also observed in countries with a high level of income and overall health inequality such as the United States (Colmer et al., 2020; Jbaily et al., 2022; Y. Wang, Zhou, et al., 2017). These inequalities can be partly explained by the non-linear relationship between PM2.5 exposure and socioeconomic development. Ambient PM2.5 concentrations tend to increase with industrialization and per-capita gross domestic product (GDP), and then subsequently decrease as air quality control measures are introduced with increasing resources and awareness of the health implications (Lim et al., 2020; S. Wang, Zhou, et al., 2017). Furthermore, higher-income countries/regions have in some cases “outsourced” their manufacturing (and associated air pollutant emissions) to lower-income countries/regions with less stringent air pollution controls (Xia et al., 2018; Zhang et al., 2017), which can exacerbate the disparities (Nansai et al., 2020). Additional drivers of inequality in ambient air pollution exposure arise from polluting activities that are predominantly undertaken by poorer communities (Rao et al., 2021; Reddington et al., 2021).

Inequities in PM2.5 exposure can be compounded by other socioeconomic factors that increase the vulnerability and disease susceptibility of a population, such as poor healthcare and nutrition (O'Neill et al., 2003) and population ageing (Conibear, Butt, Knote, Spracklen, & Arnold, 2018; Rafaj et al., 2021). Lower-income countries tend to suffer from reduced access to healthcare (O'Neill et al., 2003), while high-income countries tend to have older (more vulnerable) populations than low- and middle-income countries (United Nations, 2019). Overall, despite current differences in population vulnerability, 92% of the 2019 ambient PM2.5-attributable mortality burden was in low- and middle-income countries (GBD Collaborative Network, 2020). Even accounting for differences in population size, the ambient PM2.5-attributable mortality rates in middle-income countries were two to three times greater than in high-income countries (GBD Collaborative Network, 2020). Rapid increases in population age in the least developed countries may increase this disparity, with two-thirds of the global population aged 60 years and over expected to live in lower- and middle-income countries by 2050 (United Nations, 2019). The extent to which socioeconomic disparities in ambient PM2.5 pollution exposure and health impacts will continue in the future, as low- and middle-income populations develop economically and address their air quality problems, has not yet been examined.

Improvements in air quality and public health can be achieved by implementing climate mitigation strategies that involve reductions of both GHG emissions and co-emitted air pollutants (Amann et al., 2020; Chowdhury et al., 2018; Fujimori et al., 2020; Hamilton et al., 2021; Shindell et al., 2018; Silva et al., 2016; West et al., 2013). The estimated global economic value of these air quality and health co-benefits could potentially offset the costs of climate change policy implementation and GHG reductions (Aleluia Reis et al., 2022; Markandya et al., 2018; Sampedro et al., 2020; Scovronick et al., 2019; Vandyck et al., 2018; West et al., 2013). However, it is important to note that proposed climate mitigation/net-zero measures can have a range of impacts on air quality (Royal Society, 2021). A number of specific measures that would likely yield the largest co-benefits for climate and air quality have been highlighted in recent reports (Royal Society, 2021; UNEP, 2019; UNESCAP, 2023). Overall, the greatest health benefits are likely to result from implementation of climate mitigation policies in combination with stringent air pollution control measures (Amann et al., 2020; Conibear, Reddington, Silver, Arnold, et al., 2022; Harmsen et al., 2020; Likhvar et al., 2015; Partanen et al., 2018; Rafaj et al., 2021; Rao et al., 2016; Shim et al., 2021).

The new Shared Socioeconomic Pathways (SSPs) (O'Neill et al., 2014) combine a range of potential future climate policies with varying degrees of air pollution control (Rao et al., 2016). Recent studies have assessed the impacts of the SSPs on global air quality (Allen et al., 2020; Shim et al., 2021; Turnock et al., 2020) and public health in Africa (Shindell et al., 2022), China (Wang et al., 2022) and globally (Turnock et al., 2023; Yang et al., 2023), demonstrating that strong mitigation of both climate and air pollutants in the SSPs could yield large reductions in PM2.5 concentrations across all world regions. Turnock et al. (2023) showed that there are potential penalties to the future air pollution health burden in some world regions due to a warming climate that could offset benefits from reductions in air pollutant emissions, highlighting the importance of mitigating both climate and anthropogenic air pollution sources simultaneously.

Here, we examine and quantify the air quality and health co-benefits of future decarbonisation pathways that were designed with the aim of meeting the Paris Agreement temperature targets of 2°C and 1.5°C by the end of the century. This is the first multi-model quantification of future global PM2.5-attributable health impacts of the 2°C- and 1.5°C-compliant SSP1 scenarios using the current generation of Earth system models, which account for changes in both emissions and climate and simulate non-linear impacts of climate change on PM2.5 concentrations. We examine how PM2.5 exposure and associated health outcomes under different decarbonisation scenarios vary with socioeconomic status, and we make the first quantification of future socioeconomic disparities in PM2.5-exposure and health.

2 Data and Methods

Here we briefly describe the emission scenarios, models, and health impact assessment methodology used. The methods are described in further detail in the supplementary material (see Section S1 in Supporting Information S1). Our results are reported for six continental regions (shown in Figure S1 in Supporting Information S1) and for four socioeconomic groups (see Section 2.4).

2.1 Future Baseline and Decarbonization Scenarios

We used existing model data from experiments conducted as part of the Coupled Model Intercomparison Project Phase 6 (CMIP6; Eyring et al., 2016) by the latest generation of Earth system and climate models. The CMIP6 model simulations were driven by prescribed GHG concentrations based on future scenarios that combine a particular climate mitigation target (in terms of an anthropogenic radiative forcing reached by 2100) and the range of emission mitigation measures necessary to achieve it, within the social, economic, and environmental developments of the individual SSP (O'Neill et al., 20142016; Riahi et al., 2017). We selected three future scenarios (SSP2-4.5, SSP1-2.6, and SSP1-1.9) used in the future experiments conducted as part of the Scenario Model Intercomparison Project (ScenarioMIP; O'Neill et al., 2016), a sub-MIP of CMIP6. The scenario selected to be our future baseline, SSP2-4.5, combines the “Middle-of-the-Road” socioeconomic development of SSP2, with a medium radiative forcing target of 4.5 W m−2 by 2100. The mitigation (decarbonization) scenarios used in this study combine the “Sustainable development” pathway of SSP1 with the lower end of the range of future forcing pathways aimed to limit warming to either well below 2°C (with a radiative forcing target of 2.6 W m−2; SSP1-2.6) or below 1.5°C (with a radiative forcing target of 1.9 W m−2; SSP1-1.9) by 2100.

The underlying SSP storylines include varying degrees of air pollution emission controls, with the implementation and strength of the controls linked to socioeconomic development (Rao et al., 2017). The SSP associated with our future baseline scenario SSP2-4.5 includes “medium strength” air pollution controls, which means that the implementation of air pollution controls is assumed to continue along current national trajectories with some “catch-up” assumed for lower-income countries, where emission controls are achieved at lower income levels than when higher-income countries began controls (Rao et al., 2017). Over the 21st century, it is assumed that air pollution concentration targets become more ambitious, and the enforcement of these targets becomes increasingly effective (Rao et al., 2017). The combination of medium strength air pollution controls and climate mitigation in SSP2-4.5 results in an eventual decrease in global air pollutant emissions (Gidden et al., 2019). The SSP associated with the decarbonization scenarios SSP1-1.9 and SSP1-2.6, includes “strong” air pollution controls, which means that ambitious and stringent air pollution controls are assumed to be implemented rapidly across high-income countries, with successful achievement of air pollutant targets that go beyond current legislation in the medium to long term (Rao et al., 2017). Lower-income countries are assumed to catch-up relatively quickly with high-income countries (Rao et al., 2017). The combination of strong air pollution controls and stringent climate mitigation in SSP1-1.9 and SSP1-2.6 results in relatively rapid reductions in global air pollutant emissions (Gidden et al., 2019).

2.2 CMIP6 Model Simulations

To investigate future changes in PM2.5 pollution between 2015 and 2100, we calculated global surface distributions of PM2.5 concentrations using data from five CMIP6 models (see Table S1 in Supporting Information S1) with data available for the SSP2-4.5, SSP1-2.6, and SSP1-1.9 experiments. Simulated surface PM2.5 concentrations were calculated at the native model grid and then re-gridded to a consistent horizontal grid, before generating multi-model means for 5-year time intervals between 2015 and 2100. To improve the representation of real-world ambient PM2.5 concentrations for the health impact assessment, the present-day CMIP6-simulated PM2.5 data were corrected to observation-based estimates of PM2.5 concentrations from van Donkelaar et al. (2021). The steps involved in processing and observationally correcting the data are described in Section S1.1 and shown in Figure S2 in Supporting Information S1. The sensitivity of including additional aerosol components in the calculation of future changes in multi-model mean PM2.5 concentrations is explored in Section S2 in Supporting Information S1.

2.3 Health Impact Assessment

We performed an air pollution health impact assessment to estimate the future premature mortality burden attributable to long-term exposure to ambient PM2.5 concentrations under the different model scenarios, using population attributable fractions of relative risk following Conibear, Reddington, Silver, Arnold, et al. (2022). The relative risk for a specific PM2.5 exposure and population age group was estimated using the Global Exposure Mortality Model (GEMM) (Burnett et al., 2018). Long-term PM2.5 exposure was calculated as the population-weighted 5-year mean PM2.5 concentrations from the observationally corrected CMIP6 multi-model mean data (at 0.125° × 0.125° resolution). We used the GEMM for non-accidental mortality (non-communicable disease (NCD) plus lower respiratory infections (LRI)) for adults over 25 years of age, with age-specific risk function parameters for each 5-year age group between 25 and 80+ years (Burnett et al., 2018). The uncertainty range in our PM2.5-attributable premature mortality estimates was calculated based on the derived uncertainty intervals at the 95% confidence level from the GEMM exposure-outcome association (Burnett et al., 2018). Henceforth in the paper we refer to the PM2.5-attributable premature mortality burden as the PM2.5-attributable mortality burden. Futher details on the health impact assessment calculation can be found in Section S1.2 in Supporting Information S1.

We chose to use the GEMM in our health impact assessment framework as it is based only on cohort studies of exposure to ambient PM2.5 concentrations and has been used in several recent studies to quantify future PM2.5-attributable mortality burdens under different scenarios (Conibear, Reddington, Silver, Arnold, et al., 2022; Shindell et al., 2022; Turnock et al., 2023; Wang et al., 2022; Yang et al., 2023). For completeness, the results calculated in this study using the GEMM are compared with those calculated using the recent function from Weichenthal et al. (2022) (combined with the Fusion function from Burnett et al. (2022)) in Section S3 in Supporting Information S1.

For each country, current and future cause-specific (NCD and LRI) baseline mortality rates and population age structure were taken from International Futures (IFs) for adults aged 25–80 years in 5-year age intervals and for 80 years plus (Frederick S. Pardee Center for International Futures, 2021). Current and future global gridded population count at a resolution of 0.125° × 0.125° was taken from Jones and O'Neill (20162020). Future changes in global population count follow the SSP2 pathway (Jones & O'Neill, 20162020; Figure S3 in Supporting Information S1) to be consistent with our baseline scenario (SSP2-4.5). Future changes in baseline mortality rates and population age structure follow the IFs “Base Case” (or “Current Path”) scenario (Turner et al., 2017), which is closest to SSP2 of all the SSPs in terms of storyline (Hughes & Narayan, 2021), but is slightly more pessimistic in terms of projections of some outcomes (Burgess et al., 2023; Hughes, 2019). We assumed that the current and future population (count and age structure) and baseline mortality rates were kept the same for all three scenarios (SSP1-1.9, SSP1-2.6, and SSP2-4.5) in order to isolate the effects of the differences in projected air pollutant emissions and simulated PM2.5 concentrations on future PM2.5 exposure and attributable mortality.

2.4 Income Groups

We grouped the global population into four socioeconomic groups (low-, lower-middle-, upper-middle-, and high-income; Figure S4 in Supporting Information S1) based on population-weighted per-capita GDP, following the method of Alizadeh et al. (2022). We advance the method of Alizadeh et al. (2022) by using future projected GDP data for the years of 2020–2100 in 10-year intervals, rather than fixed values for 2015. We used global gridded, spatially downscaled GDP data at the 1/12-degree grid scale from Murakami et al. (2021), which accounts for sub-national (population-level) variability in GDP. We selected GDP data that develop in line with economic development of the SSP2 pathway, with moderate economic growth projected for existing major cities, to be consistent with our baseline scenario (SSP2-4.5) and the population and health data sets (Section 2.3). We assumed that the global SSP2 GDP data stays fixed for each 10-year period, so for the intermediate 5-year time intervals (i.e., 2015–2019, 2025–2029, 2035–2039,… etc.) we used the GDP data from the starting year of the following decade (i.e., 2020, 2030, 2040,… etc.).

To calculate GDP per capita, the gridded GDP data set was regridded to match the grid resolution of the population data (0.125° × 0.125°) and divided by the SSP2 population count for the corresponding time period. As in Alizadeh et al. (2022), the weight for each grid cell was calculated by normalizing its population count by the total global population count. The boundaries of the income groups were calculated for each 10-year time period as the population-weighted 25th, 50th,, and 75th quantiles of the per-capita GDP distribution, which meant that the population count was similar across the income groups for each time period. In simple terms, the low-income population group (referred to as a “region” in our results following Alizadeh et al. (2022)) represents the population with the lowest-quartile per-capita GDP globally.

We note that the identification of the socioeconomic status of a region or population is complex (Hajat et al., 2021) and the measure used here to classify income groups is one of many different measures used in the air pollution health literature. Our estimated population-weighted country-level income classifications for 2020 are comparable to those assigned by the World Bank based on national values of gross national income per capita in 2020 (World Bank, 2022).

3 Results

3.1 Future Air Pollution Exposure

Global PM2.5 exposure is projected to reduce considerably during the 21st century under both the baseline scenario (SSP2-4.5) and the decarbonization scenarios (SSP1-2.6, SSP1-1.9). Under SSP2-4.5, global PM2.5 exposure reduces from 28.5 μg m−3 in 2015–2019 to 16.9 μg m−3 in 2095–2099. Relative to SSP2-4.5, the decarbonization scenarios consistently produce greater reductions in global PM2.5 exposure: by 22% (SSP1-2.6) and 26% (SSP1-1.9) on average across the 21st century. Several factors are responsible for the model-simulated changes in global PM2.5 concentrations, including future changes in anthropogenic emissions, natural emissions, and climate. However, the predicted reductions in PM2.5 in the decarbonization scenarios are likely to be driven mainly by the projected reductions in primary anthropogenic emissions of organic carbon (Figure S5 in Supporting Information S1), black carbon, and sulfur dioxide, resulting from the implementation of stringent air pollution controls and climate change mitigation policies. The CMIP6 models include changes in natural aerosol, such as mineral dust and biogenic secondary organic aerosol (SOA), in response to changes in climate, which likely drive the diversity in model estimates of PM2.5 exposure in some regions (e.g., Shindell et al., 2022). We note that the projected future changes in PM2.5 exposure shown in our results do not include simulated future changes in nitrate and ammonium aerosol concentrations, which likely results in an underestimation in the overall simulated changes in PM2.5 exposure in each scenario (see Section S2 in Supporting Information S1).

The magnitude, timing, and rate of reductions in PM2.5 exposure vary strongly between the different scenarios and between different regions of the world. Figure 1 shows the variation in simulated PM2.5 exposure, calculated as population-weighted 5-year mean PM2.5 concentration, in six continental regions (Figure S1 in Supporting Information S1) under the three scenarios between 2015 and 2100. In the Americas and Asia, the decarbonization scenarios produce consistent reductions in PM2.5 exposures relative to the baseline across the 21st century, with average reductions between 2015 and 2100 of 13%–16% in South & Central America, 20%–24% in North America, and 26%–32% in Asia. In Europe, PM2.5 exposure is similar in all three scenarios up to around 2030, after which we see the additional air quality benefits from decarbonization. Hence, over the course of the 21st century overall, the decarbonization scenarios produce average reductions of 16% and 19% relative to the baseline across Europe. In Africa, the decarbonization scenarios lead to an air quality penalty toward the end of the century, where the baseline PM2.5 exposure decreases beyond the levels predicted by SSP1-2.6 and SSP1-1.9 (as a result of the projected changes in anthropogenic aerosol emissions; Figure S5 in Supporting Information S1). However, the decarbonization scenarios predict strong reductions in PM2.5 exposure relative to the baseline during mid-century, leading to overall average reductions of 13%–15%. In Oceania, present-day PM2.5 exposures are relatively low and are predicted to reduce by small amounts toward 2100 under the decarbonization scenarios, relative to the baseline (by 5% on average). Under SSP1-1.9, PM2.5 exposure in Oceania is projected to increase by ∼1 μg m−3 between 2035 and 2045, which is driven by carbonaceous aerosol emissions from forest burning (see Figures S5f and S6 in Supporting Information S1; Gidden et al., 2019).

Details are in the caption following the image

Variation in predicted PM2.5 exposures between 2015 and 2100 under the baseline scenario, SSP2-4.5, and two decarbonization scenarios, SSP1-2.6 and SSP1-1.9, in six continental regions. PM2.5 exposure was calculated as the population-weighted 5-year mean PM2.5 concentration in each region from 2015 to 2019 to 2095–2099 (plotted as 2020 to 2100). PM2.5 concentrations are from the observation-corrected multi-model mean CMIP6 data. The multi-model diversity in predicted PM2.5 exposures is not shown in order to be able distinguish the multi-model mean values under each scenario more clearly.

3.2 Future Air Pollution Exposure by Income Region

Relative to present day, PM2.5 exposure is predicted to reduce across all four income regions (low, lower-middle, upper-middle, and high) by the end of the century, under the three future scenarios (Figure 2). The predicted reduction in PM2.5 exposure from present-day levels is largest in the lower-middle-income region (49%–54%) and smallest in the high-income region (22%–30%). However, populations in the high-income region are consistently exposed to the lowest PM2.5 concentrations of all four socioeconomic groups across the century. The highest PM2.5 exposures are experienced by populations in the low- and lower-middle-income regions.

Details are in the caption following the image

Variation in predicted PM2.5 exposures in four income regions (low, lower-middle, upper-middle, and high) between 2015 and 2100 under the (a) baseline scenario, SSP2-4.5, and two decarbonization scenarios, (b) SSP1-2.6 and (c) SSP1-1.9. The dashed line shows the “PM2.5 disparity”: the difference between the income regions with the greatest and lowest PM2.5 exposures in each 5-year interval. PM2.5 exposure was calculated as the population-weighted 5-year mean PM2.5 concentration in each income region from 2015 to 2019 to 2095–2099 (plotted as 2020 to 2100). PM2.5 concentrations are from the observation-corrected multi-model mean CMIP6 data. The income regions are calculated based on the population-weighted per-capita gross domestic product (GDP). The projected population count and GDP data, and thus the regions within each socioeconomic group, vary with time in 10-year intervals.

In the baseline scenario (SSP2-4.5) the PM2.5 exposures in the low and lower-middle income regions are predicted to increase initially toward 2040, and then decrease toward 2100. In the decarbonization scenarios (SSP1-2.6, SSP1-1.9), all income regions experience relatively rapid reductions in PM2.5 exposure up to around 2040. The predicted reductions in PM2.5 exposure in the low-income region in SSP1-2.6 and SSP1-1.9 are not as strong as for the lower-middle-income region. Thus, during the latter half of the century the low-income region experiences the highest PM2.5 exposure of all the socioeconomic groups. These projected changes in PM2.5 exposure across the income groups remain consistent with the inclusion of future changes in nitrate and ammonium aerosol in the PM2.5 calculation (see Section S2, Figures S7 and S8 in Supporting Information S1).

Populations in Asia make up largest share of the total lower-middle- and upper-middle-income populations across the century (51%–74%; Figure S9 in Supporting Information S1), hence the magnitude and temporal pattern of PM2.5 exposure for these income regions in Figure 2 are comparable to Figure 1e. While populations in Asia also make up a large proportion of the low-income region population (38%–48%), populations in Africa make up the largest proportion (39%–53%), which explains why the reduction in PM2.5 exposure is weaker beyond ∼2030 than for the lower-middle-income region under the decarbonization scenarios (Figures 2b and 2c), resembling Figure 1d. The regional contribution to the high-income population is more mixed, with large contributions from Asia (45%–54%), North America (14%–22%), and Europe (12%–24%) (Figure S9 in Supporting Information S1). Toward the end of the century, there is an increasing contribution of populations in Africa (with relatively high PM2.5 exposure) to the high-income population (up to 19%), which is why there is a small increase in PM2.5 exposure beyond 2070 for this income region in Figures 2b and 2c.

The global socioeconomic disparity in PM2.5 exposure is predicted to reduce by the end of the century, but remain considerable, under all three scenarios (Figure 2). Under SSP2-4.5, the difference in the 2015–2019 mean PM2.5 exposure between the high-income region and the income region experiencing the greatest exposure (lower-middle) is 18.6 μg m−3 (51%), increasing up to 25.5 μg m−3 (63%) around 2035, and then reducing to 5.1 μg m−3 (27%) by the end of the century. Under the decarbonization scenarios, the PM2.5-exposure disparity continually decreases from present-day; going from 17.6 μg m−3 (51%) in 2015–2019 between the high- and lower-middle-income regions to 5.4 μg m−3 (31%) in 2095–2099 between the high- and low-income regions under SSP1-1.9. Overall, these results demonstrate that a range of future anthropogenic emission pathways could act to reduce the global inequality in PM2.5 exposure by the end of the century. However, immediate reduction in the global PM2.5 exposure inequalities in the near term, is only achieved under a decarbonization scenario.

3.3 Future Compliance With the WHO Air Quality Guideline

Across the 21st century, the decarbonization scenarios consistently result in greater proportions of the global population moving into compliance with the WHO Air Quality Guideline (AQG) for PM2.5 of 5 μg m−3 annual mean concentration (WHO, 2021), when compared to the baseline scenario. Figure 3a shows the fraction of the worlds' population exposed to ambient PM2.5 concentrations within the AQG as predicted by the three scenarios for selected years. In 2095–2099, the decarbonization scenarios produce a 52% (SSP1-2.6) and 61% (SSP1-1.9) increase in the population exposed to AQG-compliant PM2.5 concentrations, relative to the baseline scenario. This suggests that by following a decarbonization pathway, an additional 0.45–0.53 billion people could have a significantly reduced risk to acute and chronic health effects associated with PM2.5 pollution by the end of the century. However, it is important to note that even with the strongest air pollution controls, as implemented in SSP1-1.9 and SS1-2.6, a large fraction of the worlds' population (∼85%) remains exposed to concentrations above the AQG at the end of the 21st century.

Details are in the caption following the image

(a) Proportion of the global population exposed to PM2.5 concentrations in compliance with the World Health Organization Air Quality Guideline (AQG) for PM2.5 (5 μg m−3) as predicted by the baseline scenario (SSP2-4.5) and the decarbonization scenarios (SSP1-2.6 and SSP1-1.9). (b) The relative contribution of six continental regions to the AQG-compliant population as predicted by scenarios SSP2-4.5 and SSP1-1.9. (c) The relative contribution of the four income regions (low, lower-middle, upper-middle, and high) to the AQG-compliant population as predicted by scenarios SSP2-4.5 and SSP1-1.9. Results are shown for selected time intervals: 2015–2019 (shown as 2020), 2025–2029 (shown as 2030), 2045–2049 (shown as 2050) and 2095–2099 (shown as 2100).

The North American population are predicted to make up the largest fraction (35%–65%) of the population exposed to PM2.5 concentrations within the AQG, under both the baseline and decarbonization scenarios (Figure 3b), despite making up only 7%–8% of the total global population (see Figure S3 in Supporting Information S1). The next largest fractions of the global AQG-compliant population are from Europe (12%–30%; whilst making up 8%–10% of the global population) and Asia (9%–20%; whilst making up 49%–60% of the global population), which increase under SSP1-1.9 toward 2100. The proportions of the populations within Europe and Asia that are compliant with the AQG increase strongly under SSP1-1.9 from 5% to <1% in 2015–2019 to 60% and 6% in 2095–2099, respectively (see Figure S10 in Supporting Information S1). The fractional contributions of Oceania (3%–11%) and South & Central America (5%–10%) to the global AQG-compliant population are relatively small and decrease toward 2100 as the contributions from Europe and Asia increase (Figure 3b). However, within South & Central America, the proportion of the AQG-compliant population increases strongly under SSP1-1.9 from 4% in 2015–2019 to 25% in 2095–2099 (Figure S10 in Supporting Information S1). Oceania has the greatest proportion of its population in compliance with the AQG of all regions (Figure S10 in Supporting Information S1; evident by the low exposures in Figure 1f), which increases under both scenarios from 58%–66% in 2015–2019 to 83%–86% in 2095–2099. The African population make up the smallest fraction of the global AQG-compliant population under both scenarios (≤5%; Figure 3b), despite a 16%–29% contribution to the global population (Figure S3 in Supporting Information S1), with little change across the century (decarbonization generally reduces future PM2.5 exposure in Africa (Figure 1d), but not to levels below the AQG). Africa has the smallest proportion of its population in compliance with the AQG of all six regions, increasing from 0.6% in 2015–2019 to just 1.5% in 2095–2099 under SSP1-1.9 (Figure S10 in Supporting Information S1).

Across the century, the high-income region accounts for the largest fraction of the global AQG-compliant population under all three scenarios, with low- and lower-middle-income regions accounting for the smallest fractions (Figure 3c). In the middle of the century, 65%–67% of the AQG-compliant population is in the high-income region, with only 8%–9% in the low-income region and 7%–8% in the lower-middle-income region. At the end of the century, the proportion of the AQG-compliant population in the high-income region is smaller (39%–40%) but remains over twice that in the low-income region (17%–18%) and the lower-middle-income region (16%). Across the century, low- and lower-middle-income populations consistently have greater proportions that remain exposed to PM2.5 concentrations above the AQG than high-income populations. Under the decarbonization scenarios, 89%–91% of the low- and lower-middle-income populations remain exposed to PM2.5 concentrations that are not in compliance with the AQG at the end of the century, compared to 75%–76% of the high-income population. Therefore, although global PM2.5 inequalities are projected to reduce in the future, they persist even under the strongest mitigation scenario.

3.4 Impacts of Decarbonization on Future Air Pollution-Associated Mortality

We estimate the global PM2.5-attributable mortality burden for 2015–2019 to be 6.61 (95% confidence interval: 5.49–7.68) million annual premature deaths (see Table S2 and Section S3 in Supporting Information S1). Under the baseline scenario (SSP2-4.5), the global PM2.5-attributable mortality burden is predicted to increase from present day toward 2075 (despite reductions in global PM2.5 exposure) following projected increases in global population and population aging, then decrease slowly toward the end of the century to 10.34 (95% CI: 8.53–12.10) million annual deaths. Projected changes in population demographics can have a strong influence on estimates of the future PM2.5-attributable mortality burden (see Section S4 and Figures S11–S13 in Supporting Information S1), as found in previous studies (Conibear, Butt, Knote, Arnold, & Spracklen, 2018; Conibear, Butt, Knote, Spracklen, & Arnold, 2018; Rafaj et al., 2021; Turnock et al., 2023; Yang et al., 2023). In general, increasing population count and age act to increase the future PM2.5-attributable mortality burden, while decreasing baseline mortality rates act to moderate this future increase, although there are interesting differences in these drivers across income regions (Figure S11 in Supporting Information S1) and continental regions (Figure S13 in Supporting Information S1).

Relative to the baseline scenario, the decarbonization scenarios consistently produce reduced global annual PM2.5-attributable mortality burdens across the 21st century (see the “all varying” line in Figure S12 in Supporting Information S1). Figure 4 shows that the annual mortality burden that could be averted by following a decarbonization scenario instead of the baseline scenario is greatest around mid-century (when decarbonization is predicted to drive the largest reductions in PM2.5 exposure; Figure 1) and then decreases toward 2100 (see also Table S2 in Supporting Information S1). Following the SSP1-2.6 scenario could avert 2.48 (95% CI: 2.09–2.84) million annual premature deaths worldwide in 2045–2049, and 0.99 (95% CI: 0.82–1.15) million annual premature deaths in 2095–2099. Following the SSP1-1.9 scenario could avert 2.95 (95% CI: 2.48–3.38) million annual premature mortalities in 2045–2049, and 1.24 (95 CI: 1.03–1.44) million annual premature deaths in 2095–2099. We note that the magnitude of future changes in total and averted PM2.5-attributable mortality burdens may be underestimated here due to the exclusion of simulated future changes in nitrate and ammonium aerosol concentrations (see Section S2 in Supporting Information S1).

Details are in the caption following the image

Global averted PM2.5-attributable premature mortality burden that could be achieved by following the decarbonization scenarios (SSP1-1.9 or SSP1-2.6) relative to the baseline scenario (SSP2-4.5) for (a) 2025–2029 shown as 2030, (b) 2045–2049 shown as 2050, and (c) 2095–2099 shown as 2100. The PM2.5-attributable premature mortality burden was calculated for adults aged 25 years and older. Error bars represent the upper and lower mortality estimates (the 95% confidence interval) due to the uncertainty in the Global Exposure Mortality Model health function.

Across the 21st century, substantial public health benefits relative to the baseline scenario could be achieved in most continental regions by following either of the decarbonization scenarios. Figure 5 shows the regional averted PM2.5-attributable premature mortality per 100,000 head of total population of all ages (or “mortality rate”). The averted mortality rate depends on the PM2.5 exposure levels predicted by the baseline and decarbonization scenarios, in addition to projected changes in baseline mortality and population aging. The mortality rate is not dependent on projected changes in future population count, allowing values to be more easily compared between continental regions.

Details are in the caption following the image

Averted annual PM2.5-attributable premature mortality burden in continental regions over 2015–2100 that could be achieved by following the decarbonization scenarios (SSP1-1.9 or SSP1-2.6) relative to the baseline scenario (SSP2-4.5). Results shown are the averted regional annual premature mortality rates (deaths per 100,000 head of total population of all ages) associated with PM2.5 exposure in 5-year intervals from 2015 to 2019 to 2095–2099 (plotted as 2020 to 2100). The PM2.5-attributable premature mortality burden was calculated for adults aged 25 years and older. The shading represents the uncertainty in the mortality estimates (the 95% confidence interval) due to the uncertainty in the Global Exposure Mortality Model health function.

Both decarbonization scenarios result in a similar temporal pattern of averted mortality rates over the 21st century in all regions except Oceania, with SSP1-1.9 generally producing greater values (Figure 5). All continental regions experience a strong increase in averted mortality rates in the early part of the 21st century, from 2015 (from 2025 onwards in Europe) up to around 2035, reflecting the strong reductions in regional PM2.5 exposure in the decarbonization scenarios relative to the baseline over the same time period (Figure 1). In the Americas and Europe, the positive trend in averted annual mortality rate flattens off beyond ∼2035 but continues to increase at a slower rate toward the latter part of the century. In Asia and Africa, the averted annual mortality rates peak around mid-century and then begin to decrease toward 2100.

The greatest averted annual mortality rates of up to 49 (95% CI: 41–55) premature mortalities per 100,000 people are predicted to occur in Asia, around the middle of the 21st century (Figure 5). The averted mortality rate in Asia decreases toward the end of the century but remains the largest of the six regions with 22 (95% CI: 18–25) averted premature mortalities per 100,000 people per year under the SSP1-1.9 scenario. In Europe, North America, and South & Central America, the greatest averted per-capita mortality burdens are achieved during the latter half of the century with up to 21 (95% CI: 18–25), 21 (95% CI: 17–25), and 16 (95% CI: 13–19) premature mortalities per 100,000 people per year, respectively. Beyond mid-century, the averted mortality rates in Europe and the Americas remain similar in magnitude year to year with small variability. In Africa, following either of the decarbonization scenarios yields health benefits up to around 2085 (averting 1–12 premature mortalities per 100,000 people per year), but leads to health penalties in the latter part of the century relative to the baseline scenario (driven by the differences in predicted PM2.5 exposures shown in Figure 1d). In Oceania, the averted mortality rates are generally small relative to the other continents (up to 6 (95% CI: 5–7)) and fluctuate between health benefits and health penalties over the century due to small variations in predicted PM2.5 exposure in this region (Figure 1f).

3.5 Impacts of Decarbonization on Future Air Pollution-Associated Mortality by Income Region

Figure 6 shows the global PM2.5-attributable mortality burden for the four income regions that could be averted by following a decarbonization pathway instead of the middle-of-the-road pathway. The greatest per-capita health benefits of reduced PM2.5 pollution through decarbonization are predicted to occur in the middle-income regions, with an average of 27 (95% CI: 22–31) averted premature mortalities per 100,000 people per year under SSP1-1.9 (Figures 6a and 6b). Meanwhile, the smallest health benefits are predicted to occur in the low-income region (beyond ∼2030), with an average of 14 (95% CI: 12–16) averted premature mortalities per 100,000 people per year under SSP1-1.9.

Details are in the caption following the image

Averted annual PM2.5-attributable premature mortality burden in four income regions (low, lower-middle, upper-middle, and high) between 2015 and 2100 that could be achieved by following the decarbonization scenarios (SSP1-2.6 or SSP1-1.9) relative to the baseline scenario (SSP2-4.5). Panels (a) and (b) show the averted annual premature mortality rates (deaths per 100,000 head of total population of all ages) associated with PM2.5 exposure. The shading represents the uncertainty in the mortality estimates (the 95% confidence interval) due to the uncertainty in the Global Exposure Mortality Model health function. Panels (c) and (d) show the percentage of the total PM2.5-attributable premature mortality burden (under SSP2-4.5) averted under SSP1-2.6 or SSP1-1.9. The PM2.5-attributable premature mortality burden was calculated for adults aged 25 years and older. All results are in 5-year intervals from 2015 to 2019 to 2095–2099 (plotted as 2020 to 2100). The income regions are calculated based on the population-weighted per-capita gross domestic product (GDP). The projected population count and GDP data, and thus the regions within each income region, vary with time in 10-year intervals.

The proportion of the total annual PM2.5-attributable mortality burden that could be averted through decarbonization is greatest around mid-century in all income regions (Figures 6c and 6d), with up to 34% of PM2.5-attributable deaths averted in the lower-middle-income region under SSP1-1.9. For low- and lower-middle income regions, the peak in the averted fraction occurs slightly earlier (during 2040–2044) than for upper-middle- and high-income regions (between 2045 and 2070), particularly under SSP1-2.6. Beyond 2025, the proportion of deaths averted through decarbonization in the low-income region (an average of 17% over 2025–2099 under SSP1-1.9) is noticeably smaller than in the other income regions (averages of 24%–25%).

The greater number of averted mortalities in the middle-income regions (particularly between ∼2030 and 2090) is largely due to the strong reductions in PM2.5 exposure predicted by the decarbonization scenarios relative to the baseline (29%–30% on average across the century compared to 20% in the low-income region and 25% in the high-income region under SSP1-1.9). However, the averted mortality rate in each income region depends not only on the changes in PM2.5 exposure between the decarbonization scenarios and the baseline, but also on the absolute PM2.5 concentrations (since the exposure-outcome association is non-linear) and on the underlying health data of the populations within each income region, all of which vary with time. Thus, in regions with higher PM2.5 exposure and older populations there may be reduced health benefits per unit exposure decrease, compared to regions with lower PM2.5 exposure and younger populations, depending on the baseline mortality rates in the different age groups. On average between 2015 and 2100, the relative reduction in the PM2.5-attributable mortality burden per 1% reduction in PM2.5 exposure between SSP2-4.5 and SSP1-1.9 is greater in high income regions (0.86%) compared to the low, lower-middle, and upper-middle income regions (0.73%–0.79%). Inequalities in the underlying health data are projected to reduce by the end of the century (Figure S14 in Supporting Information S1). Therefore, keeping population demographics fixed at 2020 values, the difference between income regions is more pronounced, with a 1% reduction in PM2.5 exposure resulting in an average 0.71%–0.77% reduction in PM2.5-attributable mortality burden in low- and middle-income regions, compared to an average 1.00% reduction in high-income regions. These results show that although PM2.5 exposure-health inequalities are predicted to remain throughout the 21st century under all three scenarios, projected reductions in PM2.5 exposure and changes in population demographics are acting to reduce these inequalities over time.

4 Discussion and Conclusions

In this study we used future projections of global PM2.5 pollution under three different pathways; a middle-of-the-road baseline scenario (SSP2-4.5) and two decarbonization scenarios with strong air pollution controls (SSP1-2.6 and SSP1-1.9), to explore the air quality and health inequalities of transitioning to a low carbon future.

We found that all three future scenarios predict reductions in global PM2.5 exposure, relative to present-day. However, immediate reduction in global PM2.5 exposure in the near term, is only achieved under a decarbonization scenario. Moving from the SSP2-4.5 scenario to a decarbonization scenario could further reduce future PM2.5 exposure by 21%–26% on average over the 21st century and will bring over a half a billion more people into compliance with the WHO AQG by 2100. Projected changes in PM2.5 exposure from decarbonization vary strongly by world region, with the largest air quality benefits predicted to occur in Asia.

Despite strong reductions in global PM2.5 exposure under the decarbonization scenarios (which include stringent air pollution controls), a large fraction of the world's population (∼85%) are projected to remain exposed to concentrations above the WHO AQG in 2100. Regional PM2.5 exposures remain particularly high in Africa and Asia, with the PM2.5 exposure remaining above the WHO AQG for over 94% of the populations in these regions in 2100. Our results are consistent with findings from previous studies that assessed the impact of removing all major anthropogenic sources on air quality in China (Conibear, Reddington, Silver, Chen, Knote, et al., 2022) and globally (Pai et al., 2022). As anthropogenic aerosol emissions are reduced, natural or semi-natural aerosol, such as mineral dust, carbonaceous aerosol from wildfires, and biogenic SOA may make increasingly important contributions to regional PM2.5 exposure in the future (Pai et al., 2022) particularly in a warming climate. Future work should seek to quantify the anthropogenic (“abatable”) and natural contributions to future PM2.5 exposure across the different income groups, particularly for lower-income populations with high PM2.5 exposures. This future work should include the response of wildfire emissions to a warming climate and the subsequent impacts on air quality and public health, which was not considered here.

We found that substantial public health benefits could be achieved by following either of the decarbonization scenarios relative to the baseline scenario. Moving from SSP2-4.5 to the more stringent decarbonization scenario, SSP1-1.9, could substantially reduce the PM2.5-attributable mortality burden, averting 2.95 (95% CI: 2.48–3.38) million annual premature deaths globally in 2050. The largest per-capita health benefits of reduced PM2.5 pollution through decarbonization are predicted to occur in Asia around mid-century.

By grouping the global population into four income groups, using projections of per-capita GDP, we found that populations in low- and lower-middle-income regions (predominantly in Africa and Asia) consistently experience the highest PM2.5 exposures across the 21st century in all three scenarios (SSP2-4.5, SSP1-2.6, and SSP1-1.9). The lowest PM2.5 exposures consistently occur in the high-income region (predominantly populations in Europe, North America, and Asia). The proportion of the low- and lower-middle-income populations that remain exposed to PM2.5 concentrations above the WHO AQG at the end of the century (89%–91%) is considerably greater than the proportion of the high-income population (75%–76%).

The number of averted PM2.5-attributable deaths from decarbonization is greatest in middle-income populations across the 21st century, with the fewest averted premature deaths in low-income populations. The magnitude of the averted PM2.5-attributable mortality burden is largely driven by reductions in PM2.5 exposure predicted by the decarbonization scenarios relative to the baseline, but it can be influenced by other underlying differences between the income regions. The relative reduction in the PM2.5-attributable mortality burden per unit reduction in PM2.5 exposure is 8%–15% less on average in low- and middle-income populations than in high-income populations due to differences in population demographics and the non-linearity of the exposure-outcome association at high exposures. Some studies show that associations between air pollution exposure and health outcomes may be stronger in groups with lower socioeconomic status (e.g., Bell et al., 2013; Fuller et al., 2017; Rodriguez-Villamizar et al., 2016), which has not been considered here and would therefore act to increase the disparity between high- and lower-income regions.

Despite the large number of deaths that could be avoided by following a decarbonization pathway, particularly in middle-income regions, the total PM2.5-attributable deaths at the end of the century are greatest in the low- and lower-middle-income regions (109 (95% CI: 89–126) annual deaths per 100,000 people under the SSP1-1.9 scenario). This means that although there are rapid and substantial health co-benefits of decarbonization through improved air quality, it is the lower-income populations that are predicted to benefit the least from climate and air pollution mitigation; and continue to be exposed to PM2.5 concentrations that are over three times that of the AQG. Overall, the PM2.5 exposure inequality is predicted to reduce by 2100, but still remain even in the strongest mitigation scenario. In order to tackle inequalities in global PM2.5 exposure and the associated health impacts, future climate change mitigation and air quality control measures should be better targeted toward lower-income regions with high PM2.5 exposures.

This study has shown that although some co-benefits arise from decarbonization, more could be done to improve non-climate outcomes, particularly in lower-income regions. To improve health outcomes, either additional air quality improvement measures could be introduced and/or health and other co-benefit metrics could be incorporated into net zero policies. The latter option will likely be more successful at minimizing trade-offs and creating a just transformation.

Acknowledgments

We gratefully acknowledge support for this work from the AIA Group Limited, the UK's Natural Environment Research Council (NERC) COP-AQ project (Grant 2021GRIP02COP-AQ), and the UK Met Office. P. Forster acknowledges support from the European Union's Horizon 2020 Research and Innovation Programme under grant agreement number 820829 (CONSTRAIN). We gratefully acknowledge B. B. Hughes, J. S. Arevalo, C. Vandenberg and colleagues at the Frederick S. Pardee Center for International Futures for providing global future baseline mortality and population age data. We gratefully acknowledge advice from S. Annenberg on estimating future multiannual PM2.5-attributable mortality.

    Data Availability Statement

    The PM2.5 concentration data used in this study was obtained from the CMIP6 data archive which is hosted at the Earth System Grid Federation and is freely available to download from https://esgf-node.llnl.gov/search/cmip6/. Future projections of global gridded population count following SSP2 are freely available to download from Jones and O'Neill (2020). Future projections of national baseline mortality rates and population age structures are freely available to download from Frederick S. Pardee Center for International Futures (2021). Future projections of global gridded GDP following SSP2 are freely available to download from Murakami et al. (2020).