Environmental Research 195 (2021) 110754

Contents lists available at ScienceDirect

Environmental Research
journal homepage: www.elsevier.com/locate/envres

Global mortality from outdoor fine particle pollution generated by fossil

fuel combustion: Results from GEOS-Chem
Karn Vohra a, *, Alina Vodonos b, Joel Schwartz b, Eloise A. Marais c, 1, Melissa P. Sulprizio d,
Loretta J. Mickley d
a
School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham, UK
b
Harvard T.H. Chan School of Public Health, Department of Environmental Health, Harvard University, Boston, MA, USA
c
Department of Physics and Astronomy, University of Leicester, Leicester, UK
d
John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA

A R T I C L E I N F O A B S T R A C T

Keywords: The burning of fossil fuels – especially coal, petrol, and diesel – is a major source of airborne fine particulate
Particulate matter matter (PM2.5), and a key contributor to the global burden of mortality and disease. Previous risk assessments
Fossil fuel have examined the health response to total PM2.5, not just PM2.5 from fossil fuel combustion, and have used a
Mortality
concentration-response function with limited support from the literature and data at both high and low con­
Health impact assessment
centrations. This assessment examines mortality associated with PM2.5 from only fossil fuel combustion, making
use of a recent meta-analysis of newer studies with a wider range of exposure. We also estimated mortality due to
lower respiratory infections (LRI) among children under the age of five in the Americas and Europe, regions for
which we have reliable data on the relative risk of this health outcome from PM2.5 exposure. We used the
chemical transport model GEOS-Chem to estimate global exposure levels to fossil-fuel related PM2.5 in 2012.
Relative risks of mortality were modeled using functions that link long-term exposure to PM2.5 and mortality,
incorporating nonlinearity in the concentration response. We estimate a global total of 10.2 (95% CI: − 47.1 to
17.0) million premature deaths annually attributable to the fossil-fuel component of PM2.5. The greatest mor­
tality impact is estimated over regions with substantial fossil fuel related PM2.5, notably China (3.9 million),
India (2.5 million) and parts of eastern US, Europe and Southeast Asia. The estimate for China predates sub­
stantial decline in fossil fuel emissions and decreases to 2.4 million premature deaths due to 43.7% reduction in
fossil fuel PM2.5 from 2012 to 2018 bringing the global total to 8.7 (95% CI: − 1.8 to 14.0) million premature
deaths. We also estimated excess annual deaths due to LRI in children (0–4 years old) of 876 in North America,
747 in South America, and 605 in Europe. This study demonstrates that the fossil fuel component of PM2.5
contributes a large mortality burden. The steeper concentration-response function slope at lower concentrations
leads to larger estimates than previously found in Europe and North America, and the slower drop-off in slope at
higher concentrations results in larger estimates in Asia. Fossil fuel combustion can be more readily controlled
than other sources and precursors of PM2.5 such as dust or wildfire smoke, so this is a clear message to poli­
cymakers and stakeholders to further incentivize a shift to clean sources of energy.

1. Introduction association between exposure to air pollution and adverse health out­
comes (Brook et al., 2010), even at low exposure levels (<10 μg m− 3, the
The burning of fossil fuels – especially coal, petrol, and diesel – is a current World Health Organization, WHO, guideline) (Di et al., 2017).
major source of airborne particulate matter (PM) and ground-level The Global Burden of Diseases, Injuries, and Risk Factors Study 2015
ozone, which have both been implicated as key contributors to the (GBD, 2015) identified ambient air pollution as a leading cause of the
global burden of mortality and disease (Apte et al., 2015; Dedoussi and global disease burden, especially in low-income and middle-income
Barrett, 2014; Lim et al., 2012). A series of studies have reported an countries (Forouzanfar et al., 2016). Recent estimates of the global

* Corresponding author.
E-mail address: [email protected] (K. Vohra).
1
Now at: Department of Geography, University College London, London, UK.

https://doi.org/10.1016/j.envres.2021.110754
Received 29 July 2019; Received in revised form 12 January 2021; Accepted 14 January 2021
Available online 9 February 2021
0013-9351/© 2021 Elsevier Inc. All rights reserved.
K. Vohra et al. Environmental Research 195 (2021) 110754

burden of disease suggest that exposure to PM2.5 (particulate matter Lin et al., 2014), Europe (Protonotariou et al., 2013; Veefkind et al.,
with an aerodynamic diameter < 2.5 μm) causes 4.2 million deaths and 2011), and Africa (Lacey et al., 2017; Marais et al., 2014a, 2014b, 2019;
103.1 million disability-adjusted life-years (DALYs) in 2015, repre­ Marais and Wiedinmyer, 2016). The model has also been applied to
senting 7.6% of total global deaths and 4.2% of global DALYs, with 59% previous studies quantifying the global burden of disease from particu­
of these in east and south Asia Cohen et al. (2017). late matter from all sources (Brauer et al., 2016; Cohen et al., 2017).
A series of newer studies conducted at lower concentrations and at In this analysis we used GEOS-Chem with fossil fuel emissions from
higher concentrations have reported higher slopes than incorporated multiple sectors (power generation, industry, ships, aircraft, ground
into the GBD using the integrated exposure–response (IER) curve (Bur­ transportation, backup generators, kerosene, oil/gas extraction),
nett et al., 2014). These studies examined mortality due to exposure to detailed oxidant-aerosol chemistry, and reanalysis meteorology from
PM2.5 at concentrations below 10 μg m− 3 in North America (Di et al., the NASA Global Modeling and Assimilation Office. Fossil fuel emissions
2017; Pinault et al., 2016) and above 40 μg m− 3 in Asia (Katanoda et al., are from regional inventories where these are available for the US,
2011; Tseng et al., 2015; Ueda et al., 2012; Wong et al., 2015, 2016; Yin Europe, Asia, and Africa, and from global inventories everywhere else
et al., 2017). Here we have used a concentration-response curve from a (such as Mexico, Australia, South America and Canada). More details of
recently published meta-analysis of long-term PM2.5 mortality associa­ the specific fossil fuel inventories used in GEOS-Chem are in Table S1.
tion among adult populations which incorporates those new findings at Global-scale simulations in GEOS-Chem were carried out on a coarse
high and low PM2.5 concentrations (Vodonos et al., 2018). We also focus spatial grid (2 ◦ × 2.5 ◦ , about 200 km × 250 km). Four regional sim­
our study on the health impacts of fossil-fuel derived PM2.5. In contrast, ulations were also performed at fine spatial scale (0.5 ◦ × 0.67 ◦ , about
GBD reports only the health impacts of total PM2.5 and does not 50 km × 60 km) for North America, Europe, Asia, and Africa using
distinguish mortality from fossil-fuel derived PM2.5 and that from other boundary conditions from the global model. The regional simulations
kinds of PM2.5, including dust, wildfire smoke, and biogenically-sourced allow for a better match with the spatial distribution of population, thus
particles. We focus only on PM2.5 since recent studies have provided enhancing the accuracy of the estimates of health impacts. All simula­
mixed results on the link between ozone and mortality (Atkinson et al., tions were set up to replicate 2012 pollution conditions. As described in
2016) and there does not exist a global coherent concentration-response the Supplemental Material, we find that globally, GEOS-Chem captures
function (CRF) for ozone. observed annual mean PM2.5 concentrations with a spatial correlation of
The developing fetus and children younger than 5 years of age are 0.70 and mean absolute error of 3.4 μg m− 3, values which compare well
more biologically and neurologically susceptible to the many adverse with those from other models (Shindell et al., 2018; Xing et al., 2015).
effects of air pollutants from fossil-fuel combustion than adults. This We performed two sets of simulations: one set with fossil fuel emissions
differential susceptibility to air pollution is due to their rapid growth, turned on and the other with such emissions turned off. We then
developing brain, and immature respiratory, detoxification, immune, assumed that the difference between the two sets of simulations repre­
and thermoregulatory systems (Bateson and Schwartz, 2008; Perera, sents the contribution of fossil fuel combustion to surface PM2.5. More
2018). Children also breathe more air per kilogram of body weight than information on our choice of GEOS-Chem, the model setup, details of
adults, and are therefore more exposed to pollutants in air (WHO, 2006; relevant anthropogenic emissions, and model validation is described in
Xu et al., 2012). The WHO estimated that in 2012, 169,000 global deaths the Supplemental material.
among children under the age of 5 were attributable to ambient air
pollution (WHO, 2016). Further estimation of the burden of mortality 2.2. Population and health data
due to PM2.5 (particularly from anthropogenic sources) among the
young population would highlight the need for intervention aimed at We used population data from the Center for International Earth
reducing children’s exposure. Science Information Network (CIESIN) (CIESIN, 2018). The Gridded
Using the chemical transport model GEOS-Chem, we quantified the Population of the World, Version 4 Revision 11 (GPWv4.11) is gridded
number of premature deaths attributable to ambient air pollution from with an output resolution of 30 arc-seconds (approximately 1 km at the
fossil fuel combustion. Improved knowledge of this very immediate and equator). Since the population data are provided only at five-year in­
direct consequence of fossil fuel use provides evidence of the benefits to tervals, we applied 2015 population statistics to the results of our 2012
current efforts to cut greenhouse gas emissions and invest in alternative GEOS-Chem simulation. CIESIN population data was then aggregated to
sources of energy. It also helps quantify the magnitude of the health the spatial scale of the model for the exposure estimates. Country/region
impacts of a category of PM2.5 that can be more readily controlled than level data on baseline mortality rates were from GBD data for 2015
other kinds of PM2.5 such as dust or wildfire smoke. (based on the 2017 iteration) (IHME, 2017). USA state-specific mortality
rates were obtained from the CDC Wide-ranging Online Data for
2. Materials and methods Epidemiologic Research (WONDER) compressed mortality files (CDC,
2016). Canada death estimates by province were obtained from Statis­
2.1. Calculation of surface PM2.5 concentrations tics Canada, CANSIM (Canada, 2018).

Previous studies examining the global burden of disease from out­ 2.3. PM2.5 mortality concentration –response model
door air pollution have combined satellite and surface observations with
models to obtain improved estimates of global annual mean concen­ The risk of air pollution to health in a population is usually estimated
trations of PM2.5 (Shaddick et al., 2018). However, the goal of such by applying a concentration–response function (CRF), which is typically
studies was to quantify the health response to PM2.5 from all sources, based on Relative Risk (RR) estimates derived from epidemiological
both natural and anthropogenic (Brauer et al., 2016; Cohen et al., 2017). studies. CRFs are necessary elements for the quantification of health
Here the focus of our study is on surface ambient PM2.5 generated by impacts due to air pollution and require regular evaluation and update
fossil fuel combustion, and for that we rely solely on the chemical to incorporate new developments in the literature.
transport model GEOS-Chem since current satellite and surface mea­ Global assessments of air pollution risk often use the Integrated
surements cannot readily distinguish between the sources of PM2.5. Exposure-Response model (IER) (Burnett et al., 2014), which combined
Results from GEOS-Chem have been extensively validated against sur­ information on PM2.5–mortality associations from non-outdoor PM2.5
face, aircraft, and space-based observations around the world, including sources, including secondhand smoke, household air pollution from use
simulation of surface pollution over the United States (Drury et al., of solid fuels, and active smoking. The IER used data from active
2010; Ford and Heald, 2013; Heald et al., 2012; Leibensperger et al., smoking and passive smoking to address the limited number of outdoor
2012; Marais et al., 2016; Zhang et al., 2012), Asia (Koplitz et al., 2016; PM2.5 epidemiologic studies at PM2.5 > 40 μg m− 3 available at the time.

2
K. Vohra et al. Environmental Research 195 (2021) 110754

The IER formed the basis of the estimates of disease burden attributable in each grid cell over all grid cells in that country. To estimate the
to PM2.5 (e.g., 4 million deaths in 2015 in GBD, 2015). This function was change in deaths between the two scenarios (with and without fossil fuel
then updated in 2018 using the Global Exposure Mortality Model combustion), we computed the change in deaths in each grid cell, based
(GEMM). In GEMM, data from 41 epidemiological cohort studies were on its population, baseline rate, and exposure under the two scenarios
applied (Burnett et al., 2018). Independently conducted analyses were (Equation (1)). The attributable fraction (AF), or proportion of deaths
conducted on 15 of these cohorts to characterize the shapes of estimated as due to long-term exposure to PM2.5 fossil fuel air pollution,
PM2.5–mortality associations in each cohort, using a specified functional was calculated using the concentration-response estimate, following the
form of the CRF. For the remaining 26 cohorts, the form shown in Equation (2) (Figure S5 in Supplemental material).
concentration-response was examined with a linear concentration haz­ Because these estimates of mortality concentration response (β) are a
ard ratio model. A recent meta-analysis of the association between nonlinear function of concentration, we used the penalized spline model
long-term PM2.5 and mortality (Vodonos et al., 2018) applied techniques predictions from this meta-analysis to integrate the concentration-
involving flexible penalized spline CRF in a multivariate random effects specific β in each grid cell from the low PM2.5 scenario (without fossil
and meta-regression model. This approach allows the data to specify the fuel emissions) to the high PM2.5 scenario (with all emissions, including
shape of the CRF. The meta-regression pooled 135 estimates from 53 fossil fuel). In this way, we could calculate a mean value of β for each
studies examining long-term PM2.5 and mortality of cohorts aged 15 grid cell. There exist insufficient epidemiological data to calculate a
years and older. The estimate of the confidence intervals about the CRF robust health response function specific to fossil-fuel PM2.5. GEOS-Chem
includes a random variance component. This meta-analysis provided is a deterministic model. Therefore, our 95% confidence intervals (CI)
evidence of a nonlinear association between PM2.5 exposure and mor­ for our estimates reflect only the 95% CI for the concentration response
tality in which the exposure-mortality slopes decreases at higher con­ function.
centrations (Figure S5 in Supplemental Material). We have chosen to use
the dose-response function from the meta-analysis rather than the 2.5. Secondary analysis among children <5 years old
GEMM function as the meta-regression approach is more flexible and
does not constrain the CRF to a specific functional form, it incorporates a Lower respiratory infections (LRI), including pneumonia and bron­
random variance component in estimating the uncertainty around that chiolitis of bacterial and viral origin, are the largest single cause of
curve, it is derived with more studies than previous approaches, and its mortality among young children worldwide and thus account for a
estimates at high and low exposures are closer to the estimates in cohorts significant global burden of disease worldwide (Nair et al., 2010). As
restricted to only very high and very low exposures. To ensure consis­ mentioned previously, young children are more susceptible to the
tency with the concentration-response curve, premature mortality rates adverse effects of particulate air pollution than adults. Mehta et al.
for the portion of the population >14 years of age were determined (2013) estimated the overall impact of PM2.5 concentration with Rela­
using the population and baseline mortality rates for different age tive Risk (RR) of 1.12 for LRI mortality per 10 μg m− 3 increase in annual
groups from GBD data for 2015. average PM2.5 concentration, as compared to RR of 1.04 for respiratory
mortality among adults (Vodonos et al., 2018). We estimated the num­
2.4. Health impact calculations ber of premature deaths attributable to PM2.5 among children under the
age of 5 years due to a range of LRI classifications (ICD-10, International
We estimated the number of premature deaths attributable to fossil Classification of Diseases codes: A48.1, A70, J09-J15.8, J16-J16.9,
fuel PM2.5 using: (1) GEOS-Chem PM2.5 estimated with all emission J20-J21.9, P23.0-P23.4). Baseline numbers of deaths due to LRI were
sources and GEOS-Chem PM2.5 estimated without fossil fuel emissions, obtained from the GBD for 2015 (IHME, 2017). We used the Relative
as a comparison against the first simulation, (2) total population above Risk (RR) of 1.12 (1.03–1.30) for LRI occurrence per 10 μg m− 3 increase
the age of 14 gridded to the GEOS-Chem grid resolution, (3) baseline all- in annual average PM2.5 concentration (Mehta et al., 2013). Studies of
cause mortality rates for population above the age of 14 (per country or longer-term exposure of PM2.5 and LRI in that meta-analysis were con­
per state in the US and province in Canada), and (4) the meta-analysis ducted in only a few developed countries with relatively low levels of
CRF (Vodonos et al., 2018). All health impacts were calculated on a annual mean PM2.5 (<25 μg m− 3), specifically the Netherlands, Czech
per-grid basis at the spatial resolution of the model. We applied the Republic, Germany, Canada and USA. We therefore calculated the
following health impact function to estimate premature mortality number of premature LRI deaths attributable to PM2.5 only in North
related to exposure to fossil fuel PM2.5 in each GEOS-Chem grid cell: America, South America, and Europe.
∑
Δy = yo ∗p ∗ AF (1) 3. Results

exp(β∗Δx) − 1 3.1. Impact of fossil fuel use on PM2.5

AF ​ = (2)
exp(β∗Δx)
Fig. 1 shows the difference between global GEOS-Chem PM2.5 with
PM2.5 all∫​ emissions
and without fossil fuel emissions, plotted as the annual mean for 2012.
β(PM2.5 ) = β(PM2.5 ) (3) Results show large contributions of 50–100 μg m− 3 in PM2.5 over China
PM2.5 no ​ fossil ​ fuel and India, with smaller increments of 10–50 μg m− 3 over large swaths of
the United States and Europe, industrialized countries in Africa (South
where Δy is the change in the number of premature deaths due to Africa and Nigeria), and along the North African coastline due to Eu­
exposure to fossil fuel PM2.5, yo is the country/state/province specific ropean pollution.
baseline (all-cause) mortality rate, p is the total population above the
age of 14, AF is the attributable fraction of deaths (the fraction of total 3.2. Global assessment of mortality attributable to PM2.5
deaths attributable to PM2.5 exposure), β is the mean estimate for long-
term PM2.5 mortality concentration-response over a range of concen­ Based on the annual PM2.5 simulation with and without global fossil
trations from the penalized spline model in the recent meta-analysis, and fuel emissions, we estimated the excess deaths and attributable fraction
Δx is the change in PM2.5 concentration, calculated as the difference (AF %) for the population above 14 years old. Fig. 2 shows the simulated
between GEOS-Chem PM2.5 with all emissions and GEOS-Chem PM2.5 annual global premature mortality due to exposure to ambient PM2.5
without fossil fuel emissions. from fossil fuel emissions. Greatest mortality is simulated over regions
For each country, we summed the change in premature deaths (Δy) with substantial influence of fossil-fuel related PM2.5, notably parts of

3
K. Vohra et al. Environmental Research 195 (2021) 110754

Fig. 1. Contribution of fossil fuel combustion to surface PM2.5, as calculated by the chemical transport model GEOS-Chem. The plot shows the difference in surface
PM2.5 concentrations from GEOS-Chem with and without fossil fuel emissions.

Fig. 2. Estimated annual excess deaths due to exposure to ambient PM2.5 generated by fossil fuel combustion.

Eastern North America, western Europe, and South-East Asia. higher than the average PM2.5 concentrations for each country, since
We estimated a total global annual burden premature mortality due fossil-fuel PM2.5 is mainly emitted in populous areas. The two countries
to fossil fuel combustion in 2012 of 10.2 million (95% CI: − 47.1 to 17.0 with the highest premature mortality are China with 3.91 million and
million). Table 1 reports the baseline number of deaths for people >14 India with 2.46 million. Supplemental Table S2 provides extended data
years old, the annual PM2.5 simulation with and without global fossil of the health impact calculations for each country. For comparison,
fuel emissions, the estimated excess deaths, and the attributable fraction Table 1 also reports the number of premature deaths attributable to
for the populated continents. As shown in Table 1, we calculated fossil fuel PM2.5 when the GEMM function is applied to the GEOS-Chem
483,000 premature deaths in North America (95% CI: output. For most regions, the number of premature deaths calculated
284,000–670,000), 187,000 deaths in South America (95% CI: with GEMM is significantly lower than that calculated with the new
107,000–263,000), 1,447,000 deaths in Europe (95% CI: function from Vodonos et al. (2018). Globally, the GEMM function
896,000–1,952,000), 7,916,000 deaths in Asia (95% CI: − 48,106,000 to yields 6.7 million deaths in 2012 due to fossil fuel combustion.
13,622,000), and 194,000 deaths in Africa (95% CI: − 237,000 to
457,000). The wide confidence intervals in Asia and Africa are due to the 3.3. Assessment of children (under the age of 5) LRI mortality
lack of data for areas where the exposure remains outside the range of attributable to PM2.5
the concentration response curve (PM2.5 > 50 μg m− 3; Figure S5). The
population-weighted pollution concentrations presented in Table 1 are We estimated the number of premature deaths attributable to PM2.5

4
K. Vohra et al. Environmental Research 195 (2021) 110754

Table 1
Number of deaths attributable to exposure to fine particulate matter (PM2.5) generated by fossil fuel combustion for the population >14 years old.
GEOS-Chem Regionb Total Population-weighted annual mean PM2.5 Mean Deaths attributable to GEMM function
spatial grid deaths concentration, μg m− 3 attributable fossil-fuel related PM2.5, deaths attributable
resolutiona >14 years fraction of in thousands (95% CI)c to fossil-fuel related
PM2.5 from PM2.5 Estimated
old, in deaths, % (95% PM2.5, in thousands
all emission without PM2.5 from
thousands CI)d (95% CI)e
sources fossil fuel fossil fuel, %

Fine North Central 1148 10.06 3.03 7.03 (69.9) 8.2 (4.5–11.6) 94 (52–133) 80 (62–98)
America America &
the
Caribbean
USA 2705 11.81 2.15 9.66 (81.8) 13.1 (7.8–18.1) 355 (212–490) 305 (233–375)
Canada 250 12.01 1.76 10.25 (85.4) 13.6 (8.0–18.7) 34 (20–47) 28 (22–35)
Coarse South America 2389 8.66 3.02 5.65 (65.2) 7.8 (4.5–11.0) 187 (107–263) 159 (121–195)
Fine Europe 8626 19.22 4.68 14.54 (75.7) 16.8 1447 (896–1952) 1033 (798–1254)
(10.4–22.6)
Fine Asia Eastern Asia 25,468 51.72 8.68 43.05 (83.2) 30.7 (− 189.1- 7821 (− 48,150–13,478) 4945 (3943–5826)
52.9)
Coarse Western 1456 26.95 20.73 6.22 (23.1) 6.5 (3.0–9.9) 95 (44–144) 54 (43–65)
Asia & the
Middle East
Fine Africa 5274 32.98 28.98 4.00 (12.1) 3.7 (− 4.5-8.7) 194 (-237-457) 102 (81–121)
Coarse Australia & Oceania 189 4.17 2.19 1.98 (47.4) 3.2 (1.6–4.8) 6.0 (2.9–9.0) 6.4 (4.8–7.9)
Global 47,506 38.01 11.14 26.87 (70.7) 21.5 (− 99.0- 10,235 6713 (5308–7976)
35.7) (− 47,054–16,972)
a
Fine spatial scale is 0.5 ◦ × 0.67 ◦ , or about 50 km × 60 km. Coarse spatial scale is 2 ◦ × 2.5 ◦ , or about 200 km × 250 km.
b
List of countries for each region and subregion is provided in supplemental Table S2.
c
Annual number of deaths attributable to long-term exposure to PM2.5 derived from fossil fuel combustion. CI is the confidence interval.
d
Mean proportion of all deaths which can be attributed to long-term exposure to PM2.5 generated by fossil fuel combustion, averaged over the country or region. CI;
confidence interval.
e
Attributable deaths calculated with the Global Exposure Mortality Model (GEMM) concentration-response function.44.

among children under the age of 5 due to LRI only for those countries or 4. Discussion
regions with levels of annual PM2.5 concentrations below 25 μg m− 3.
These include North America, South America, and Europe. Based on the We used the chemical transport model GEOS-Chem to quantify the
annual PM2.5 simulation with and without fossil fuel emissions, we global mortality attributed to PM2.5 air pollution from fossil fuel com­
calculated 876 excess deaths due to LRI in North and Central America, bustion. Using the updated concentration response relationship between
747 in South America, and 605 in Europe (Table 2). Using the GBD es­ relative mortality and airborne PM2.5, we estimated global premature
timate of total deaths due to LRI (Institute for Health Metrics and mortality in 2012 of 10.2 million per year from fossil fuel combustion
Evaluation), we estimate that PM2.5 from fossil fuel combustion alone. China has the highest burden of 3.91 million per year, followed by
accounted on average for 7.2% of LRI mortality among children under India with 2.46 million per year. These estimates carry large uncertainty
the age of 5 in these regions, with the largest proportion of 13.6% in (e.g., 95% CI of − 47.1 to 17.0 million for the global estimate) from the
Europe (95% CI -0.4 to 25.3%). concentration-response curve, as it is an improved function that pro­
vides a more realistic picture of the health consequences of PM2.5
compared to previous studies.
Our estimate is for the year when fossil fuel emissions in China
peaked and so predates large and dramatic reductions in fossil fuel
emissions due to strict mitigation measures. These reductions led to a
Table 2 30–50% decline in annual mean PM2.5 across the country from 2013 to
Number of deaths due to lower respiratory infection (LRI) attributable to 2018 (Zhai et al., 2019). If we apply a 43.7% reduction in GEOS-Chem
exposure to fine particulate matter (PM2.5) from fossil fuel combustion for the PM2.5 concentrations from the simulation with all emission sources,
population <5 years old.
premature mortality in China decreases from 3.91 million to 2.36
Region Total deaths for LRI deaths Mean attributable million. India has recently imposed controls on pollution sources, but
children <5 attributable to fraction of deaths, there is not yet evidence of air quality improvements in densely popu­
years old due to fossil-fuel PM2.5 % (95% CI)b
LRI (95% CI)a
lated cities like Delhi (Vohra et al., 2020). Consideration of the
2012–2018 decrease in PM2.5 exposure in China reduces the total global
North 13,230 876 (-26-1657) 6.6 (¡0.2-12.5)
premature mortality due to fossil fuel PM2.5 from 10.2 million premature
America
Central 12,507 802 (-23-1516) 6.4 (− 0.2-12.1) deaths each year to 8.7 (95% CI: − 1.8 to 14.0) million.
America & In 2012, the population-weighted PM2.5 is 72.8 μg m− 3 for China and
the 52.0 μg m− 3 for India from all sources and 9.9 μg m− 3 for China and 9.0
Caribbean μg m− 3 for India without fossil fuel emissions. The low value of non-
USA 672 69 (-2-131) 10.2 (− 0.3-19.5)
Canada 50 5 (0–10) 10.8 (− 0.3-20.5)
fossil fuel PM2.5 is reasonable for southern India (Dey et al., 2012) but
South 13,231 747 (-21–1443) 5.7 (-0.2–10.9) may be an underestimate in the Indo-Gangetic Plain where crop residue
America burning contributes to high levels of PM2.5 (100–200 μg m− 3) during the
Europe 4446 605 (-18–1126) 13.6 (-0.4–25.3) post-monsoon season (Ojha et al., 2020). An increase in the concentra­
a
Annual number of deaths attributed to long-term exposure to PM2.5 derived tion of non-fossil-fuel PM2.5 would decrease our estimate of the number
from fossil fuel combustion. of premature deaths due to fossil fuel PM2.5 in India and China, as this
b
Mean proportion of deaths due to long-term exposure to PM2.5 generated by would decrease the risk of premature mortality with a unit change in
fossil fuel combustion. CI is the confidence interval.

5
K. Vohra et al. Environmental Research 195 (2021) 110754

PM2.5 (Figure S5). for mortality from five specific causes (ischemic heart disease, stroke,
chronic obstructive pulmonary disease, lung cancer and acute respira­
4.1. Comparison with previous estimates of global mortality attributable tory infections), in the current analysis we estimated changes in deaths
to outdoor PM2.5 from all causes. Fourth, some of the difference in the mortality estimates
may come from differences in the age range. Our approach considers a
Previous estimates of the GBD for 2015 suggest that exposure to total wider population age range of over 14 years old (Vodonos et al., 2018)
PM2.5 causes 4.2 million deaths (Cohen et al., 2017), whereas here we compared to the other studies, which considered a population age range
estimate more than double (10.2 million) the number of premature of over 25 years (Burnett et al., 2018; Cohen et al., 2017; Lelieveld et al.,
deaths from fossil fuel combustion alone in 2012. Differences between 2019). Our approach has wider age range since the age range for the
the current study and the 2015 GBD lower estimates are related mainly studies in the meta-analysis (Vodonos et al., 2018) included people
to the choice of the shape of the concentration-response function and the younger than 25 years old (Hart et al., 2011; Pinault et al., 2016).
relative risk estimate. First, to provide information about exposure Finally, the finer spatial resolution that GEOS-Chem utilizes over much
response at higher concentrations, the 2015 GBD study used the inte­ of the globe improves co-location of PM hotspots and population cen­
grated exposure–response (IER) model in which active and second-hand ters, yielding higher estimates of excess mortality compared to Lelieveld
smoking exposures were converted to estimated annual PM2.5 exposure et al. (2019).
equivalents using inhaled doses of particle mass (Burnett et al., 2014).
Recent cohort studies from Asia indicate that this substantially un­ 4.2. Limitations
derestimates the CRF at high concentrations. In contrast, in the current
study we applied a CRF that was directly estimated from PM2.5 studies There are a number of limitations that must be acknowledged. First,
alone, as described in a recent meta-analysis that included estimates vulnerability to PM2.5 exposure may vary by population characteristics
from studies in countries like China with higher PM2.5 concentrations such as ethnicity, socio-economic status (SES), risk behaviors such as
than are included in previous derivations of CRFs (Vodonos et al., 2018). smoking and underlying comorbidities (Krewski et al., 2000; Pope et al.,
The CRF from this recent meta-analysis flattens out at higher concen­ 2004; Wang et al., 2017) and by different exposure characteristics. We
trations, as does the IER curve. However, this flattening is not as great as were limited in our ability to undertake a comprehensive analysis of
in the IER, as Asian cohort studies at high PM2.5 concentrations report factors influencing the association between PM2.5 and mortality since
larger effects than would be expected from the IER. Hence estimates of the global mortality data were not available by detailed age, ethnicity,
the global attributable fraction of deaths due to air pollution using the SES, lifestyle, and underlying disease strata. In addition, the 95% CI of
function from the recent meta-analysis are higher than the estimates our estimates reflect the lower and upper bound of the CRF, which
using the IER function. In addition, at much lower concentrations (<10 flattens out at higher concentrations. Regions with very high concen­
μg m− 3), we applied higher slopes than assumed in the IER function. trations (>50 μg m− 3) are beyond the data range in the meta-analysis;
Recent studies at very low concentrations similarly show that the IER thus, the lower limit of the CI for those regions (China, West and
underestimated effects in this range (Pinault et al., 2016). Since North Africa; Table 1) are much less than zero. Second, for LRI in chil­
GEOS-Chem estimated quite low concentrations in developed countries dren, we have restricted our analysis to developed countries with annual
in Europe and North America, the number of premature deaths from PM2.5 < 25 μg m− 3, in accordance with the geographical locations of the
PM2.5 in these countries is greater than previous estimates. studies included in the meta-analysis by Mehta et al. (2013). Developing
Following an approach similar to the recent meta-analysis (Vodonos countries have much higher LRI mortality rates, and this restriction
et al., 2018), Burnett et al. (2018) modeled the shape of the association doubtless results in an underestimate. Finally, GEOS-Chem estimates of
between PM2.5 and non-accidental mortality using data from 41 cohorts PM2.5 concentrations almost certainly contains errors in estimates of
from 16 countries with GEMM. In that study, the uncertainty in a subset emissions of pollution precursors, meteorological effects on air quality,
(15 cohorts) was characterized in the shape of the and representation of the complex physical and chemical formation
concentration-response parameter by calculating the Shape-Constrained pathways. In the absence of systematic bias, such model error may not
Health Impact Function, a prespecified functional form. These estimated produce large aggregate errors in the mortality burden of PM2.5, but bias
shapes varied across the cohorts included in the function. GEMM pre­ may be present as well. In any event, it is challenging to estimate the true
dicted 8.9 million (95% CI: 7.5–10.3) deaths in 2015 attributable to size of this error.
long-term exposure to PM2.5 from all sources; 120% higher excess deaths
than previous estimates, but still lower than our estimate of mortality 5. Conclusions
from exposure to fossil-fuel derived PM2.5 for 2012. Lelieveld et al.
(2019) estimated the global and regional mortality burden of fossil fuel The effects of CO2-driven climate change on human health and
attributable PM2.5 by applying the GEMM CRF to a global welfare are complex, ranging from greater incidence of extreme weather
chemistry-climate model that is overall coarser (~1.9◦ latitude and events, more frequent storm-surge flooding, and increased risk of crop
longitude) than the model used in this work. The authors reported 3.61 failure (Duffy et al., 2019). One consequence of increasing reliance on
million deaths per year attributable to pollution from fossil fuel com­ fossil fuel as an energy source that has thus far received comparatively
bustion and 5.55 million deaths per year due to pollution from all little attention is the potential health impact of the pollutants co-emitted
anthropogenic sources. The estimated deaths from fossil fuel combus­ with the greenhouse gas CO2. Such pollutants include PM2.5 and the
tion are much lower than those in the current study for several reasons. gas-phase precursors of PM2.5. This study demonstrates that the fossil
First, the meta-analysis function used in our work includes 135 co­ fuel component of PM2.5 contributes a large global mortality burden. By
efficients of all-cause mortality for adults aged 14–64 years old, together quantifying this sometimes overlooked health consequence of fossil fuel
with cause-specific mortality and all-cause mortality among adults aged combustion, a clear message is sent to policymakers and stakeholders of
65 and older, thus incorporating many more studies in a meta-regression the co-benefits of a transition to alternative energy sources.
framework than the 41 cohorts and coefficients in the GEMM function.
Second, the approach used to estimate the CRF in Vodonos et al. (2018) Author contribution
allows for additional flexibility in the shape of the function because of its
use of penalized splines. In contrast, the GEMM pooled CRF integrates a K. Vohra and A. Vodonos carried out the health impact calculations
set of 26 log-linear functions and 15 functions characterized by three guided by J. Schwartz. E. A. Marais and M. P. Sulprizio performed GEOS-
parameters governing the shape of the function. Third, while Cohen Chem simulations. L. J. Mickley oversaw the project. All authors
et al. (2017), Lelieveld et al. (2019) and Burnett et al. (2018) accounted contributed to writing the manuscript.

6
K. Vohra et al. Environmental Research 195 (2021) 110754

Data availability Duffy, P.B., Field, C.B., Diffenbaugh, N.S., et al., 2019. Strengthened scientific support
for the Endangerment Finding for atmospheric greenhouse gases. Science 363, 597–
+. https://doi.org/10.1126/science.aat5982.
GEOS-Chem code and output are available at the GEOS-Chem web­ Ford, B., Heald, C.L., 2013. Aerosol loading in the Southeastern United States:
site (http://acmg.seas.harvard.edu/geos_chem.html) and upon request. reconciling surface and satellite observations. Atmos. Chem. Phys. 13, 9269–9283.
https://doi.org/10.5194/acp-13-9269-2013.
Forouzanfar, M.H., Afshin, A., Alexander, L.T., et al., 2016. Global, regional, and
national comparative risk assessment of 79 behavioural, environmental and
Declaration of competing interest occupational, and metabolic risks or clusters of risks, 1990-2015: a systematic
analysis for the Global Burden of Disease Study 2015. Lancet 388, 1659–1724.
https://doi.org/10.1016/S0140-6736(16)31679-8.
The authors declare that they have no known competing financial
Hart, J.E., Garshick, E., Dockery, D.W., et al., 2011. Long-term ambient multipollutant
interests or personal relationships that could have appeared to influence exposures and mortality. Am J Resp Crit Care 183, 73–78. https://doi.org/10.1164/
the work reported in this paper. rccm.200912-1903OC.
Heald, C.L., Collett, J.L., Lee, T., et al., 2012. Atmospheric ammonia and particulate
inorganic nitrogen over the United States. Atmos. Chem. Phys. 12, 10295–10312.
Acknowledgments https://doi.org/10.5194/acp-12-10295-2012.
IHME, 2017. Institute for Health Metrics and Evaluation. http://ghdx.healthdata.
org/gbd-results-tool.
This study was funded by the Wallace Global Fund, the Environment Katanoda, K., Sobue, T., Satoh, H., et al., 2011. An association between long-term
and Health Fund (EHF) Israel, and a University of Birmingham Global exposure to ambient air pollution and mortality from lung cancer and respiratory
Challenges Fund PhD studentship awarded to KV. This publication was diseases in Japan. J. Epidemiol. 21, 132–143. https://doi.org/10.2188/jea.
JE20100098.
made possible by USEPA grant RD-835872. Its contents are solely the Koplitz, S.N., Mickley, L.J., Marlier, M.E., et al., 2016. Public health impacts of the severe
responsibility of the grantee and do not necessarily represent the official haze in Equatorial Asia in September-October 2015: demonstration of a new
views of the USEPA. Further, USEPA does not endorse the purchase of framework for informing fire management strategies to reduce downwind smoke
exposure. Environ. Res. Lett. 11 https://doi.org/10.1088/1748-9326/11/9/094023.
any commercial products or services mentioned in the publication.
Krewski, D., Burnett, R.T., Goldberg, M.S., et al., 2000. Special Report Reanalysis of the
Harvard Six Cities Study and the American Cancer Society Study of Particulate Air
Appendix A. Supplementary data Pollution and Mortality Part II: Sensitivity Analyses Appendix C. Flexible Modeling
of the Effects of Fine Particles and Sulphate on Mortality. Health Effects Institute.
https://www.healtheffects.org/system/files/SR-PartIIAppC.pdf.
Supplementary data to this article can be found online at https://doi. Lacey, F.G., Marais, E.A., Henze, D.K., et al., 2017. Improving present day and future
org/10.1016/j.envres.2021.110754. estimates of anthropogenic sectoral emissions and the resulting air quality impacts in
Africa. Faraday Discuss 200, 397–412. https://doi.org/10.1039/c7fd00011a.
Leibensperger, E.M., Mickley, L.J., Jacob, D.J., et al., 2012. Climatic effects of 1950-2050
References changes in US anthropogenic aerosols - Part 1: aerosol trends and radiative forcing.
Atmos. Chem. Phys. 12, 3333–3348. https://doi.org/10.5194/acp-12-3333-2012.
Lelieveld, J., Klingmuller, K., Pozzer, A., et al., 2019. Effects of fossil fuel and total
Apte, J.S., Marshall, J.D., Cohen, A.J., et al., 2015. Addressing global mortality from
anthropogenic emission removal on public health and climate. P Natl Acad Sci USA
ambient PM2.5. Environ. Sci. Technol. 49, 8057–8066. https://doi.org/10.1021/acs.
116, 7192–7197. https://doi.org/10.1073/pnas.1819989116.
est.5b01236.
Lim, S.S., Vos, T., Flaxman, A.D., et al., 2012. A comparative risk assessment of burden of
Atkinson, R.W., Butland, B.K., Dimitroulopoulou, C., et al., 2016. Long-term exposure to
disease and injury attributable to 67 risk factors and risk factor clusters in 21
ambient ozone and mortality: a quantitative systematic review and meta-analysis of
regions, 1990-2010: a systematic analysis for the Global Burden of Disease Study
evidence from cohort studies. Bmj Open 6. https://doi.org/10.1136/bmjopen-2015-
2010. Lancet 380, 2224–2260, 628-628, ://WOS:000315189300032.
009493.
Lin, J.T., van Donkelaar, A., Xin, J.Y., et al., 2014. Clear-sky aerosol optical depth over
Bateson, T.F., Schwartz, J., 2008. Children’s response to air pollutants. J. Toxicol.
East China estimated from visibility measurements and chemical transport modeling.
Environ. Health A 71, 238–243. https://doi.org/10.1080/15287390701598234.
Atmos. Environ. 95, 258–267. https://doi.org/10.1016/j.atmosenv.2014.06.044.
Brauer, M., Freedman, G., Frostad, J., et al., 2016. Ambient air pollution exposure
Marais, E.A., Jacob, D.J., Guenther, A., et al., 2014a. Improved model of isoprene
estimation for the global burden of disease 2013. Environ. Sci. Technol. 50, 79–88.
emissions in Africa using Ozone Monitoring Instrument (OMI) satellite observations
https://doi.org/10.1021/acs.est.5b03709.
of formaldehyde: implications for oxidants and particulate matter. Atmos. Chem.
Brook, R.D., Rajagopalan, S., Pope, C.A., et al., 2010. Particulate matter air pollution and
Phys. 14, 7693–7703. https://doi.org/10.5194/acp-14-7693-2014.
cardiovascular disease an update to the scientific statement from the American heart
Marais, E.A., Jacob, D.J., Jimenez, J.L., et al., 2016. Aqueous-phase mechanism for
association. Circulation 121, 2331–2378. https://doi.org/10.1161/
secondary organic aerosol formation from isoprene: application to the southeast
CIR.0b013e3181dbece1.
United States and co-benefit of SO2 emission controls. Atmos. Chem. Phys. 16,
Burnett, R., Chen, H., Szyszkowicz, M., et al., 2018. Global estimates of mortality
1603–1618. https://doi.org/10.5194/acp-16-1603-2016.
associated with long-term exposure to outdoor fine particulate matter. P Natl Acad
Marais, E.A., Jacob, D.J., Wecht, K., et al., 2014b. Anthropogenic emissions in Nigeria
Sci USA 115, 9592–9597. https://doi.org/10.1073/pnas.1803222115.
and implications for atmospheric ozone pollution: a view from space. Atmos.
Burnett, R., Pope, C.A., Ezzati, M., et al., 2014. An integrated risk function for estimating
Environ. 99, 32–40. https://doi.org/10.1016/j.atmosenv.2014.09.055.
the global burden of disease attributable to ambient fine particulate matter exposure.
Marais, E.A., Silvern, R.F., Vodonos, A., et al., 2019. Air quality and health impact of
Environ. Health Perspect. 122, 397–403. https://doi.org/10.1289/ehp.1307049.
future fossil fuel use for electricity generation and transport in Africa. Environ. Sci.
Canada, S., 2018. Government of Canada. https://www150.statcan.gc.ca/n1/en/type/
Technol. 53, 13524–13534. https://doi.org/10.1021/acs.est.9b04958.
data.
Marais, E.A., Wiedinmyer, C., 2016. Air quality impact of diffuse and inefficient
CIESIN, Center for International Earth Science Information Network - Columbia
combustion emissions in Africa (DICE-Africa). Environ. Sci. Technol. 50,
University, 2018. Gridded Population of the World, Version 4 (GPWv4): Population
10739–10745. https://doi.org/10.1021/acs.est.6b02602.
Count Adjusted to Match 2015 Revision of UN WPP Country Totals, Revision 11.
Mehta, S., Shin, H., Burnett, R., et al., 2013. Ambient particulate air pollution and acute
NASA Socioeconomic Data and Applications Center (SEDAC), Palisades, NY. https://
lower respiratory infections: a systematic review and implications for estimating the
doi.org/10.7927/H4PN93PB.
global burden of disease. Air Qual Atmos Hlth 6, 69–83. https://doi.org/10.1007/
Cohen, A.J., Brauer, M., Burnett, R., et al., 2017. Estimates and 25-year trends of the
s11869-011-0146-3.
global burden of disease attributable to ambient air pollution: an analysis of data
Nair, H., Nokes, D.J., Gessner, B.D., et al., 2010. Global burden of acute lower respiratory
from the Global Burden of diseases study 2015. Lancet 389, 1907–1918. https://doi.
infections due to respiratory syncytial virus in young children: a systematic review
org/10.1016/S0140-6736(17)30505-6.
and meta-analysis. Lancet 375, 1545–1555. https://doi.org/10.1016/S0140-6736
Dedoussi, I.C., Barrett, S.R.H., 2014. Air pollution and early deaths in the United States.
(10)60206-1.
Part II: attribution of PM2.5 exposure to emissions species, time, location and sector.
Ojha, N., Sharma, A., Kumar, M., et al., 2020. On the widespread enhancement in fine
Atmos. Environ. 99, 610–617. https://doi.org/10.1016/j.atmosenv.2014.10.033.
particulate matter across the Indo-Gangetic Plain towards winter. Sci Rep-Uk 10.
Dey, S., Di Girolamo, L., van Donkelaar, A., et al., 2012. Variability of outdoor fine
https://doi.org/10.1038/s41598-020-62710-8.
particulate (PM2.5) concentration in the Indian Subcontinent: a remote sensing
Perera, F., 2018. Pollution from fossil-fuel combustion is the leading environmental
approach. Remote Sens. Environ. 127, 153–161. https://doi.org/10.1016/j.
threat to global pediatric health and equity: solutions exist. Int. J. Environ. Res. Publ.
rse.2012.08.021.
Health 15. https://doi.org/10.3390/ijerph15010016.
Di, Q., Wang, Y., Zanobetti, A., et al., 2017. Air pollution and mortality in the medicare
Pinault, L., Tjepkema, M., Crouse, D.L., et al., 2016. Risk estimates of mortality
population. N. Engl. J. Med. 376, 2513–2522. https://doi.org/10.1056/
attributed to low concentrations of ambient fine particulate matter in the Canadian
NEJMoa1702747.
community health survey cohort. Environ Health-Glob 15. https://doi.org/10.1186/
Drury, E., Jacob, D.J., Spurr, R.J.D., et al., 2010. Synthesis of satellite (MODIS), aircraft
s12940-016-0111-6.
(ICARTT), and surface (IMPROVE, EPA-AQS, AERONET) aerosol observations over
Pope, C.A., Burnett, R.T., Thurston, G.D., et al., 2004. Cardiovascular mortality and long-
eastern North America to improve MODIS aerosol retrievals and constrain surface
term exposure to particulate air pollution - epidemiological evidence of general
aerosol concentrations and sources. J. Geophys. Res. Atmos. 115 https://doi.org/
10.1029/2009jd012629.

7
K. Vohra et al. Environmental Research 195 (2021) 110754

pathophysiological pathways of disease. Circulation 109, 71–77. https://doi.org/ WHO, World Health Organization, 2016. Ambient Air Pollution: A Global Assessment of
10.1161/01.Cir.0000108927.80044.7f. Exposure and Burden of Disease. https://www.who.int/phe/publications/air-poll
Protonotariou, A.P., Bossioli, E., Tombrou, M., et al., 2013. Air Pollution in Eastern ution-global-assessment/en/.
Mediterranean: Nested-Grid GEOS-CHEM Model Results and Airborne Observations. Wong, C.M., Lai, H.K., Tsang, H., et al., 2015. Satellite-based estimates of long-term
Advances in Meteorology, Climatology and Atmospheric Physics. Springer exposure to fine particles and association with mortality in elderly Hong Kong
Atmospheric Sciences, Springer, Berlin, Heidelberg, pp. 1203–1209. residents. Environ. Health Perspect. 123, 1167–1172. https://doi.org/10.1289/
Shaddick, G., Thomas, M.L., Green, A., et al., 2018. Data integration model for air ehp.1408264.
quality: a hierarchical approach to the global estimation of exposures to ambient air Wong, C.M., Tsang, H., Lai, H.K., et al., 2016. Cancer mortality risks from long-term
pollution. J R Stat Soc C-Appl 67, 231–253. https://doi.org/10.1111/rssc.12227. exposure to ambient fine particle. Cancer Epidem Biomar 25, 839–845. https://doi.
Shindell, D., Faluvegi, G., Seltzer, K., et al., 2018. Quantified, localized health benefits of org/10.1158/1055-9965.Epi-15-0626.
accelerated carbon dioxide emissions reductions. Nat. Clim. Change 8. https://doi. Xing, J., Mathur, R., Pleim, J., et al., 2015. Can a coupled meteorology-chemistry model
org/10.1038/s41558-018-0108-y. reproduce the historical trend in aerosol direct radiative effects over the Northern
Tseng, E., Ho, W.C., Lin, M.H., et al., 2015. Chronic exposure to particulate matter and Hemisphere? Atmos. Chem. Phys. 15, 9997–10018. https://doi.org/10.5194/acp-
risk of cardiovascular mortality: cohort study from Taiwan. BMC Publ. Health 15. 15-9997-2015.
https://doi.org/10.1186/s12889-015-2272-6. Xu, Z.W., Sheffield, P.E., Hu, W.B., et al., 2012. Climate change and children’s health-A
Ueda, K., Nagasawa, S., Nitta, H., et al., 2012. Exposure to particulate matter and long- call for Research on what works to protect children. Int. J. Environ. Res. Publ. Health
term risk of cardiovascular mortality in Japan: NIPPON DATA80. J. Atherosclerosis 9, 3298–3316. https://doi.org/10.3390/ijerph9093298.
Thromb. 19, 246–254. https://doi.org/10.5551/jat.9506. Yin, P., Brauer, M., Cohen, A., et al., 2017. Long-term fine particulate matter exposure
Veefkind, J.P., Boersma, K.F., Wang, J., et al., 2011. Global satellite analysis of the and nonaccidental and cause-specific mortality in a large national cohort of Chinese
relation between aerosols and short-lived trace gases. Atmos. Chem. Phys. 11, men. Environ. Health Perspect. 125 https://doi.org/10.1289/Ehp1673.
1255–1267. https://doi.org/10.5194/acp-11-1255-2011. Zhai, S.X., Jacob, D.J., Wang, X., et al., 2019. Fine particulate matter (PM2.5) trends in
Vodonos, A., Abu Awad, Y., Schwartz, J., 2018. The concentration-response between China, 2013-2018: separating contributions from anthropogenic emissions and
long-term PM2.5 exposure and mortality; A meta-regression approach. Environ. Res. meteorology. Atmos. Chem. Phys. 19, 11031–11041. https://doi.org/10.5194/acp-
166, 677–689. https://doi.org/10.1016/j.envres.2018.06.021. 19-11031-2019.
Vohra, K., Marais, E.A., Suckra, S., et al., 2020. Long-term trends in air quality in major Zhang, L., Jacob, D.J., Knipping, E.M., et al., 2012. Nitrogen deposition to the United
cities in the UK and India: a view from space. Atmos. Chem. Phys. Discuss. https:// States: distribution, sources, and processes. Atmos. Chem. Phys. 12, 4539–4554.
doi.org/10.5194/acp-2020-342. https://doi.org/10.5194/acp-12-4539-2012.
Wang, Y., Shi, L.H., Lee, M., et al., 2017. Long-term exposure to PM2.5 and mortality CDC, 2016. Centers for Disease Control and Prevention Wide-ranging ONline Data for
among older adults in the southeastern US. Epidemiology 28, 207–214. https://doi. Epidemiologic Research https://www.cdc.gov/nchs/data/nvsr/nvsr64/nvsr64_02.
org/10.1097/Ede.0000000000000614. pdf.
WHO, World Health Organization, 2006. Principles for Evaluating Health Risks in
Children Associated with Exposure to Chemicals. https://apps.who.int/iris
/handle/10665/43604.