Introduction

Large parts of South and East Asia suffer from poor air quality, mainly due to high concentrations of PM2.5 (fine particulate matter with a diameter <2.5 µm), leading to millions of excess deaths every year1. Both through short-term and long-term exposure, PM2.5 can cause a number of detrimental effects on human health, such as respiratory and cardiovascular diseases, and other pollutants, such as ozone (O3), add to the problem2.

Local anthropogenic emissions are the most important factor influencing air quality in the highly polluted Eastern China and Indo-Gangetic plains of India, but meteorological conditions also play a key role3,4,5. Absorption and scattering of solar radiation by particles, i.e., aerosol-radiation interactions (ARI), have been found to enhance near-surface pollution by modifying meteorology in the planetary boundary layer (PBL)6,7,8,9, leading to an increased number of PM2.5 related deaths in China10 and India11.

The enhancement of surface air pollution through ARI is a consequence of reduced turbulence and lowering of the PBL height, leading to the reduced sink of surface air pollution from turbulent transport to higher altitudes12. This is particularly important for black carbon (BC) aerosols12,13,14, due to their ability to absorb solar radiation and thereby heat the atmosphere on a short time scale15. BC in the PBL also leads to higher relative humidity, enhancing cloud development and potentially affecting climate12. Elevated relative humidity combined with high temperatures can also have negative health impacts16,17. The impact on PBL suppression is, however, dependent on the vertical distribution of BC concentrations18,19,20. Aerosols that scatter solar radiation, cool the surface and, therefore, also modify the vertical temperature profile and stability, although the aerosol-PBL feedback has been found to be small for sulfate (SO4) aerosols9. In addition to ARI, aerosols can alter the Earth’s radiation budget through aerosol-cloud interactions where aerosols act as cloud condensation nuclei21.

ARI have been found to influence tropospheric ozone through two mechanisms: meteorological changes by aerosol-radiation feedback, and changes in the photolysis rates by aerosol-photolysis interactions22. The net effect is a reduction of near-surface ozone, mainly due to aerosol-photolysis interactions23, but the daily maximum ozone response is dependent on the season24. In addition to detrimental health effects, elevated near-surface ozone concentrations have negative impacts on agricultural crop production25.

Aerosol abundances over China have been reduced in recent years due to air quality regulations while India has seen an increase26,27,28,29. However, regional trends in anthropogenic emissions vary between individual aerosols and aerosol precursors30, and there are uncertainties in how natural emissions, such as dust aerosols, respond to climate change31. Thus, it is important to gain knowledge on aerosol-PBL feedback due to individual aerosol pollutants. While BC has, in previous studies, been emphasized as particularly important for triggering aerosol-PBL feedback and exacerbating air pollution9,12,13,32, there is a lack of knowledge on the quantification of its role compared to other aerosols. Additionally, most previous studies have focused on relatively short haze episodes12,13,14,32, but in terms of health effects it is important to also investigate long-term exposure to air pollution2.

In this study, we investigate ARI (including adjustments) due to total aerosols and separately ARI due to BC based on multi-year simulations, spanning the first two decades of this century, using the Weather Research and Forecasting model with chemistry (WRF-Chem33; see Methods). The simulations cover a relatively large region in Asia, but we focus particularly on regions with large anthropogenic aerosol emissions; Northern India and Eastern China. Additional single-year simulations are performed to investigate ARI due to individual aerosol compounds. We focus on aerosol-PBL feedback and resulting health impacts. Modeled near-surface concentrations of PM2.5 and gas pollutants are evaluated against >1000 air quality monitoring stations, and aerosol optical depth (AOD) is evaluated against satellite observations.

Results and discussion

Model evaluation

The WRF-Chem model is able to reproduce satellite-retrieved annual and monthly mean AOD values over South and East Asia relatively well (Fig. 1a–c). While the modeled values are generally lower than Moderate Resolution Imaging Spectroradiometer (MODIS-Aqua) in Northern India and, especially, in Eastern China, the model overestimates AOD over these regions when compared against the Multi-Angle Imaging Spectroradiometer (MISR) satellite retrieval (Supplementary Fig. 1). This confirms that there are substantial uncertainties in AOD from individual satellite products34. It should be noted, however, that the modeled AOD is based on diurnal means for all-sky conditions, while the satellite-retrieved AOD is for a given overpass time and clear-sky conditions.

Fig. 1: Evaluation of AOD and PM2.5.
figure 1

Annual mean AOD from WRF-Chem (a) and MODIS-Aqua (b), monthly mean AOD density scatter plot between MODIS-Aqua and WRF-Chem (c), annual mean PM2.5 from WRF-Chem and air quality monitoring stations (d, e), and probability density function of hourly PM2.5 for all stations shown in d (f). The WHO dataset in e includes stations that are outside of China but within the model domain (see Supplementary Fig. 2).

Full size image

Comparisons of modeled annual mean near-surface PM2.5 against >1000 air quality stations in China show generally good correspondence (Fig. 1d, e). It is worth noting that despite extensive policies to improve air quality in China since 201335, near-surface PM2.5 in Eastern China was still high in 2019 (Fig. 1d) (to put it in perspective, an air quality guideline of 5 μg m−3 has been set by the World Health Organization36). Probability density functions of hourly PM2.5 measurements are also well reproduced by WRF-Chem, especially for the year 2019, and a shift towards lower concentrations between 2015 and 2019 is seen in both the measurements and WRF-Chem (Fig. 1f). Annual mean PM2.5 concentrations are better reproduced by the model in countries that are outside of China but within the model domain (red dots in Fig. 1e; Supplementary Fig. 2).

Comparisons between measured and modeled near-surface gas concentrations (SO2, CO, NO2, and O3) over China show more variable results (Supplementary Figs. 2 and 3), but this can be expected given that the model grid spacing is 45 km while the measurements often take place in urban areas and are influenced by local emission sources37. It should also be noted that the model has been run only with spectral nudging of horizontal winds outside of the boundary layer (see Methods), and therefore it cannot be expected that the model perfectly reproduces day-by-day historical conditions for such a large domain. Nevertheless, the model is able to reasonably well reproduce the seasonal and diurnal cycle of the gas-phase compounds, but with an underestimation for CO and NO2, and too high night-time concentrations of O3 (Supplementary Fig. 3). Modeled annual and monthly mean near-surface temperature and diurnal temperature range compares well against CRU observations (Supplementary Fig. 4).

Regional distribution of ARI

The modeled pattern of changes in near-surface turbulent kinetic energy due to total ARI largely reflects the distribution of near-surface PM2.5, which has its highest values over India, Eastern China, and the Gobi and Taklamakan deserts in Central Asia (Fig. 2a, b). ARI reduce turbulence in nearly the whole region, with substantial decreases of around 30% in Northern India, and exceeding 10% reductions in the rest of India and most of Eastern China. Changes in PBL height show a very similar pattern as changes in turbulence but with a somewhat smaller magnitude, e.g., with ~20% reduction in PBL height over N. India (Fig. 2c). The reduced turbulence and PBL height due to ARI imply that near-surface air pollution will be subject to less mixing and dilution. In fact, the model shows that near-surface PM2.5 concentrations are enhanced due to ARI by up to approx. 15% in regions that are already heavily polluted (Fig. 2d). The desert areas are an exception, showing almost no change in PM2.5, indicating that dust aerosols may be less efficient in enhancing air pollution through ARI. Also, ARI due to total aerosols change the mix of different PM2.5 aerosol pollutants, e.g., with dust aerosols mostly decreasing and carbonaceous aerosols (BC/OA), which are suspected of being more toxic than other aerosols1,38, showing a relative increase that is larger than for total PM2.5 (Supplementary Fig. 5). Nitrate aerosols increase due to ARI by 50% or more in large parts of India. The change due to ARI in the mix of different PM2.5 species is likely linked to their height distribution, for instance, with nitrate particles mainly formed from agricultural ammonia (NH3) emissions that are emitted at the surface level.

Fig. 2: Regional distributions of PM2.5 and turbulence.
figure 2

5-year averages of near-surface total PM2.5 and BC-only PM2.5 in BASE (a, e), and relative changes in near-surface turbulent kinetic energy (b, f), PBL height (c, g), and near-surface total PM2.5 (d, h) due to ARI from all aerosols (BASE-noARI) and due to ARI from BC-only (BASE-noBCARI). Stippling indicates where at least 4 of the 5 years agree on the sign of change.

Full size image

ARI due to BC also lead to widespread reductions in turbulence and PBL height, although with a smaller magnitude than due to ARI from all aerosols (Fig. 2e–g). Interestingly, BC has been considered particularly efficient in enhancing air pollution through ARI9,13, but here, it induces only modest changes in PM2.5 (Fig. 2h). There are likely several factors explaining these different conclusions. For instance, Ding et al. 13 also found strong influence of ARI due to BC on weakened turbulence and lowered PBL height using WRF-Chem, but the modeled influence on PM2.5 concentrations was only shown explicitly for ARI due to total aerosols, and not BC specifically; a direct comparison with our results is therefore difficult. In addition, their model study was limited to one winter month (December 2013), while our results are averaged over five years. In Stjern et al. 9, different methods are likely to explain the different conclusions. They used a global climate model to simulate a tenfold increase in BC emissions, and were therefore unable to directly quantify the influence of ARI due to BC on PM2.5 concentrations. When averaged over the whole land region shown in Fig. 2, the BC share of total ARI is 40% of the reduction in turbulent kinetic energy, 41% of the reduction in PBL height, but only 13% of the increase in PM2.5. The two regions with strongest enhancement of PM2.5 due to total ARI, Northern India and Eastern China, are densely populated and with high near-surface BC concentrations (Fig. 2d, e), and will be given particular emphasis in the further analysis to better understand the processes involved.

Vertical distribution of ARI

Figure 3 shows that impacts of ARI due to BC are substantial in both Northern India and Eastern China, despite BC mass mixing ratios being small compared to other aerosol compounds. BC leads to strong atmospheric heating due to solar absorption (Fig. 3b, j), up to around 0.5 K at 800 hPa over N. India and around half of that over E. China (Fig. 3c, k). Aerosols other than BC mainly scatter solar radiation and, therefore, cool the surface, and both BC and non-BC ARI act to decrease the lapse rate and thereby strengthen lower atmospheric stability. This stabilization effect weakens turbulence (Fig. 3f, n), and increases relative humidity near the surface (Fig. 3d, l). A possible explanation for higher relative humidity near the surface in polluted conditions is that the decrease in turbulence leads to weakened transport of moist and polluted air out of the surface mixed layer and weakened transport of dry and clean air into the mixed layer from above12. The modeled increase in cloud fraction throughout the atmospheric column for N. India, and the small change in cloud fraction for E. China (Fig. 3e, m), agree with a former multi-model study of BC impacts (ref. 39, their Fig. 7).

Fig. 3: Regionally averaged vertical profiles.
figure 3

Mass mixing ratios of PM2.5 aerosol compounds in BASE (a, i) and changes in meteorological variables (b–g and j–o) and total PM2.5 (h, p) due to ARI from black carbon only (BCARI: BASE-noBCARI), organic aerosols only (OAARI: BASE-noOAARI), sulfate only (SO4ARI: BASE-noSO4ARI), nitrate only (NO3ARI: BASE-noNO3ARI), ammonium only (NH4ARI: BASE-noNH4ARI), sea salt only (SSARI: BASE-noSSARI), dust only (DUSTARI: BASE-noDUSTARI), and all aerosols (ARI: BASE-noARI) for year 2015 over the two regions indicated by the black rectangles in Fig. 2a. Variable “W” is vertical velocity. Shadings for BCARI and ARI represent the range of all 5 simulated years.

Full size image

Over both regions, ARI lead to increases in PM2.5 in the PBL, with >5% near the surface, and decreases in the middle troposphere, but BC ARI alone show the opposite effect: decrease in PM2.5 in the PBL (except near the surface) and increase above (Fig. 3h, p). The self-lofting ability of BC has been shown to increase ascent over southern Asia40, and this is indeed the case here as well (Fig. 3g, o), largely explaining how BC ARI influence the PM2.5 vertical profile. In addition, the model shows that BC ARI increase precipitation in N. India, and partly also E. China, likely leading to stronger removal of aerosols through wet scavenging due to BC compared to the effect of ARI due to total aerosols (Supplementary Fig. 6). Several studies have tried to isolate the role of BC to precipitation over India and China and results are somewhat contrasting. While some find that Indian BC emissions cause local drying41,42,43, others find insignificant precipitation change44,45. However, monsoon season precipitation over India is found to increase in response to BC aerosols over China, or when perturbing BC globally or at least over both China and India simultaneously41,42,43,46,47—consistent with our findings. In the same studies, precipitation over China was found to decrease locally following an increase in BC, although comparisons to this study are more difficult due to larger differences in the East China domain.

Figure 3 also shows that ARI, due to all individual aerosol types, lead to increased PM2.5 concentrations in the PBL in the two regions, with organic aerosols being the largest contributor. Dust aerosols have a different vertical profile than other aerosols, with relatively large concentrations in the middle troposphere in the two regions, likely because dust is transported into the region while other aerosols have more local sources. This leads to a strong stabilization effect and reduction of turbulence due to ARI from dust only, and subsequent effects on the vertical profile of relative humidity and clouds.

Seasonal changes in ARI

Figure 4 shows that the effects of ARI are unevenly distributed throughout the year for N. India, but less so for E. China. ARI from all aerosols lead to an increase in relative humidity and a decrease in turbulence throughout the year for both regions, and a large part is due to BC ARI (Fig. 4e, f, i, j). Over N. India, the decreased turbulence and increased PM2.5 due to ARI from all aerosols are particularly strong in the post-monsoon (SON) and winter (DJF) seasons, when the concentrations of anthropogenic aerosols are high (Fig. 4a, i, k). The PM2.5 enhancement reached 25% in November and December 2015 with 7% contribution from BC ARI, and approximately the same contribution from ARI due to organic aerosols (Fig. 4k). In the monsoon season (JJA), both ARI due to all aerosols and ARI due to BC lead to enhanced precipitation, and PM2.5 concentrations are reduced due to more wet scavenging (Fig. 4g, k). Over E. China, the PM2.5 enhancement due to ARI from all aerosols is within 12% for 2015, and that due to ARI from BC is near-zero, throughout the year (Fig. 4l). In both regions, most effects of ARI are fairly constant between the simulated years (ranging from 2003 to 2019), except that PM2.5 enhancements over N. India has increased since the beginning of the time period (Supplementary Fig. 7).

Fig. 4: Regionally averaged seasonal cycle.
figure 4

PM2.5 aerosol compounds in BASE (a, b) and changes in meteorological variables (c–j) and total PM2.5 (k, l) due to ARI from black carbon only (BCARI: BASE-noBCARI), organic aerosols only (OAARI: BASE-noOAARI), sulfate only (SO4ARI: BASE-noSO4ARI), nitrate only (NO3ARI: BASE-noNO3ARI), ammonium only (NH4ARI: BASE-noNH4ARI), sea salt only (SSARI: BASE-noSSARI), dust only (DUSTARI: BASE-noDUSTARI), and all aerosols (ARI: BASE-noARI) for year 2015 over the two regions indicated by the black rectangles in Fig. 2a. Bars show annual means. Shadings for BCARI and ARI represent the range of all 5 simulated years.

Full size image

Impacts of ARI on health-related metrics

Figure 5a shows that the change in near-surface concentrations of PM2.5 due to ARI from all aerosols leads to excess mortality in several regions, with total changes in annual excess deaths due to PM2.5 of +27,200, +24,900, and +76,900 in N. India, E. China, and the whole domain, respectively. Corresponding numbers for ARI due to BC are +1930, +426, and −8190, respectively, and the distribution is shown in Fig. 5e. For comparison, the estimated number of annual excess deaths due to PM2.5 in the BASE simulation is 0.613 [0.478–0.792; 95% CI], 0.665 [0.531–0.834], and 2.74 [2.11–3.59] million in N. India, E. China, and the whole domain, respectively (Supplementary Fig. 8). This means that ARI constitute 4% of the annual excess deaths due to PM2.5 in each of the two regions and 3% when aggregated over the whole domain. While ARI, due to all aerosols, affect excess deaths of elderly people the most in the domain as a whole, a relatively large share (31%) of excess deaths in N. India is among middle adulthood and younger populations (age <60 years) (Supplementary Fig. 9), partly due to a younger population. In this region, excess mortality due to ARI from BC has increased substantially in the later years (2015 and 2019). In both N. India and E. China, ARI due to all aerosols, and especially due to BC, lead to a particularly large increase in deaths due to chronic obstructive pulmonary disease. It is worth noting that while excess deaths of adults in the domain as a whole are reduced due to ARI from BC, the BC still leads to an increase in excess deaths caused by lower respiratory infection among children.

Fig. 5: Regional distributions of health-related metrics.
figure 5

5-year average of the change in excess deaths due to long-term exposure to PM2.5 (a, e), change in the number of haze days (PM2.5 > 75 μg m−3) (b, f), the relative change in annual mean ozone (c, g), and change in days with a maximum hourly heat index exceeding 41 °C (d, h) due to ARI from all aerosols (BASE-noARI; top) and ARI from BC-only (BASE-noBCARI; bottom). Stippling indicates where at least 4 of the 5 years agree on the sign of change.

Full size image

The above health impact calculations are based on annual concentrations of PM2.5, but short-term exposure to elevated air pollution is also associated with a number of detrimental health effects2. Results show that the number of haze days, defined here as PM2.5 daily mean concentration higher than the WHO interim target 1 of 75 µg m−3 36, increases by >30 days in large parts of N. India due to ARI from all aerosols, with smaller but non-negligible contribution from BC ARI (Fig. 5b, f).

The effect of ARI on near-surface ozone is to reduce the concentrations throughout the domain, and BC is the main cause (Fig. 5c, g). The amount of solar radiation reaching the surface is reduced by >5% in large parts of India and E. China due to ARI from BC (Supplementary Fig. 10), and this would reduce ozone formation due to photolysis attenuation. However, dynamical changes could also influence ozone levels24, including the impact of relative humidity changes on ozone dry deposition48.

Heat index, or apparent temperature, is a measure of the temperature felt by the human body when accounting for relative humidity49. A heat index >41 °C is categorized as dangerous and involves the risk of sunstroke and heatstroke50. Results show that ARI from BC substantially increased the number of days with maximum heat index exceeding 41 °C throughout India and part of E. China, with more than 10 days increase in parts of N. India (Fig. 5h). This occurs because BC enhances both surface temperature and relative humidity (Fig. 4c, e; Supplementary Fig. 11). In contrast, ARI from total aerosols cool the surface and show only small influences on the number of days with maximum heat index exceeding 41 °C (Fig. 5d), implying that scattering aerosols reduce this number of days considerably. For heat indices with lower thresholds, the number of days with exceedance is reduced due to ARI from total aerosols but increased considerably when analyzing ARI due to BC (Supplementary Fig. 12).

Diurnal temperature range is the difference between the daily maximum and minimum temperature, and its variation could influence crop yields51 and human health, mostly with greater diurnal temperature range being associated with increased mortality52. Results show that ARI from all aerosols reduce the diurnal temperature range by 0.5–1 °C in nearly all of India, largely due to a decrease in daily maximum temperature caused by scattering aerosols (Supplementary Fig. 13). ARI due to BC lead to an increase in daily maximum temperature in most of India, and an even stronger increase in daily minimum temperature. Some studies indicate that enhanced night-time temperatures are of particular concern regarding mortality and morbidity risk, including humid-hot nights53,54.

In conclusion, our results highlight the multitude of impacts from individual aerosol species on meteorology and air pollution. Most notably, BC is particularly efficient in reducing near-surface turbulence through ARI but, in contrast to ARI from non-BC aerosols, reduces the number of excess deaths caused by long-term PM2.5 exposure in southern and eastern Asia as a whole. At the same time, ARI from BC lead to increases in both near-surface temperature and relative humidity over India, in contrast to non-BC aerosols that cool the surface. Consequently, if potential future air pollution mitigation over India would involve reductions of scattering aerosols without simultaneous reductions in black carbon, as has partly been the case in China (ref. 30, their Figure 8), this could cause a substantial increase in the number of days exceeding heat index levels that are dangerous to human health.

Methods

Air pollution monitoring data

China’s air pollution monitoring data was from the national monitoring sites (https://quotsoft.net/air/). Air pollution measurements of annual mean PM2.5 and NO2 in countries other than China have been taken from the WHO database v6.155.

AOD satellite observations

Satellite observations of AOD at 550 nm (level 3) have been taken from the MODIS instrument on board the Aqua (dataset MYD08_M3) and Terra (dataset MOD08_M3) platforms56, and the MISR instrument on board the Terra platform (dataset MIL3MAEN_004)57.

Surface temperature observations

Monthly mean observations of surface temperature and diurnal temperature range have been taken from the Climatic Research Unit gridded Time Series version 4.0858.

WRF-Chem model simulations

The WRF-Chem model33 version 3.9.1.1 has been used at a horizontal resolution of 45 km × 45 km, covering a large part of Asia. The extent of the model domain is shown in Fig. 1a, except that an outer boundary of 10 grid boxes has been removed in the analysis to avoid spurious boundary effects. In the vertical, 50 layers were used, extending from the surface and up to 50 hPa. Meteorological initial and 6-hourly boundary conditions, including sea-surface temperatures, were taken from the ERA5 reanalysis data59. Spectral nudging of horizontal winds was applied outside of the PBL. Physics schemes applied include the RRTMG radiation scheme60, Morrison 2-moment microphysics scheme61, MYNN PBL and surface layer scheme62, Unified Noah land surface model63, and Grell 3D ensemble cumulus parameterization scheme64. Although the WRF-Chem model version is 3.9.1.1, relevant bug fixes have been applied up until version 4.4. The cloud fraction diagnosis was modified to follow Xu and Randall65.

Gas-phase chemistry is represented by the MOZART mechanism66 and aerosols by the MOSAIC 4-bin sectional aerosol module67 including aqueous chemistry and aerosol-cloud interactions. Chemical initial and boundary conditions are from CESM2.1/CAM-Chem68,69. Anthropogenic emissions are from the Community Emissions Data System (CEDS) version of April 2021, which goes until 2019 and builds upon the CEDS system described in ref. 30. Sector-dependent factors have been applied to crudely account for vertical distribution70 and temporal (weekly and diurnal variations) profiles71. Biomass burning emissions are from the CMIP6 inventory, which goes until 201572, and 2015 emissions are assumed also for 2019. The biomass burning emissions have been distributed evenly between the surface and 1.2 km height in the vertical, and a diurnal cycle has been assumed as in ref. 73. (their Table 9). The calculation of biogenic emissions is done online using the Model of Emissions of Gases and Aerosols from Nature (MEGAN) version 2.0474. Natural dust and sea salt emissions are also calculated online, while dimethyl sulfide emissions use the sea-air exchange fluxes of ref. 75. and ocean climatology of ref. 76.

The three core simulations, BASE, noARI, and noBCARI, are performed for five years spanning the first two decades of the 21st century: 2003, 2007, 2011, 2015, and 2019, each with two months of spin-up. BASE is the reference simulation, while in noARI, ARI are switched off by setting aer_ra_feedback=0 in the namelist. In noBCARI, ARI for BC are switched off by setting the mass of BC to zero in the file module_optical_averaging.F. Similarly, in simulations for the year 2015 (+ 2 months of spin-up) only, ARI have been switched off for organic aerosols (noOAARI), sulfate (noSO4ARI), nitrate (noNO3ARI), ammonium (noNH4ARI), sea salt (noSSARI), and dust (noDUSTARI). When showing map plots of differences between simulations (in Figs. 2 and 5), stippling indicates where at least 4 of the 5 simulation years agree on the sign of change. However, a stricter requirement of all 5 years agreeing on the sign of change does not substantially change the area of stippling in the plots (not shown).

Health impact calculations

In combination with the modeled data on ambient PM2.5 exposure and sensitivity ARI experiments, we apply the MR-BRT (meta-regression-Bayesian, regularized, trimmed) exposure-response function77,78, which was also used in our previous studies1,79 and the most recent iteration of the Global Burden of Disease study80. Using the cause-specific exposure-response function, we estimate excess deaths from ischemic heart disease (IHD), stroke (both ischemic and hemorrhagic), chronic obstructive pulmonary disease (COPD), lung cancer (LC), and Type II diabetes (T2DM) among adults (aged 25 and above), as well as acute lower respiratory tract infections among children (under 5 years old). The excess death burden was calculated at a 5 × 5 km spatial resolution by interpolating the modeled exposure data, and the estimates were stratified by age and disease category, following the approach of our previous studies:

$${M}_{c,a,d}=B{M}_{a,d}\times {P}_{a}\times \frac{R{R}_{c,a,d}-1}{R{R}_{c,a,d}}$$
(1)

Excess deaths were estimated separately for adults and children at each 5-year interval. RR (c, a, d) were derived using MR-BRT functions for all diseases by age, where c, a, d denotes the concentration of PM2.5, population age, and disease, respectively. Age-specific RRs (Relative Risks) for IHD and stroke, are obtained using MR-BRT. For LC, T2DM, and COPD, uniform RR (c, d) were used across all age groups among adults. BM (a, d) is the baseline mortality rate per 100,000 population, obtained from the GBD (http://ghdx.healthdata.org/gbd-results-tool) for India and China for respective years. We considered BM to remain uniform within a country at 5 × 5 km resolution by age and disease. P(a) is the exposed population in a grid by age; the age distributions at 5-year intervals (adults > 25 years), and <5 years for children were obtained from the GBD, which are then merged with the gridded population data at about 5 × 5 km horizontal resolution from Global Human-Settlement Layer (https://human-settlement.emergency.copernicus.eu/) to obtain the age-specific population (P(a)) at each 5 × 5 km grid.