The Impact of Individual Anthropogenic Emissions Sectors on the Global Burden of Human Mortality due to Ambient Air Pollution

Background: Exposure to ozone and fine particulate matter (PM2.5) can cause adverse health effects, including premature mortality due to cardiopulmonary diseases and lung cancer. Recent studies quantify global air pollution mortality but not the contribution of different emissions sectors, or they focus on a specific sector. Objectives: We estimated the global mortality burden of anthropogenic ozone and PM2.5, and the impact of five emissions sectors, using a global chemical transport model at a finer horizontal resolution (0.67° × 0.5°) than previous studies. Methods: We performed simulations for 2005 using the Model for Ozone and Related Chemical Tracers, version 4 (MOZART-4), zeroing out all anthropogenic emissions and emissions from specific sectors (All Transportation, Land Transportation, Energy, Industry, and Residential and Commercial). We estimated premature mortality using a log-linear concentration–response function for ozone and an integrated exposure–response model for PM2.5. Results: We estimated 2.23 (95% CI: 1.04, 3.33) million deaths/year related to anthropogenic PM2.5, with the highest mortality in East Asia (48%). The Residential and Commercial sector had the greatest impact globally—675 (95% CI: 428, 899) thousand deaths/year—and in most regions. Land Transportation dominated in North America (32% of total anthropogenic PM2.5 mortality), and it had nearly the same impact (24%) as Residential and Commercial (27%) in Europe. Anthropogenic ozone was associated with 493 (95% CI: 122, 989) thousand deaths/year, with the Land Transportation sector having the greatest impact globally (16%). Conclusions: The contributions of emissions sectors to ambient air pollution–related mortality differ among regions, suggesting region-specific air pollution control strategies. Global sector-specific actions targeting Land Transportation (ozone) and Residential and Commercial (PM2.5) sectors would particularly benefit human health. Citation: Silva RA, Adelman Z, Fry MM, West JJ. 2016. The impact of individual anthropogenic emissions sectors on the global burden of human mortality due to ambient air pollution. Environ Health Perspect 124:1776–1784; http://dx.doi.org/10.1289/EHP177

-Global emissions per sector for six pollutants (Tg/yr) in the RCP 8.5 Global Emissions Inventory, 2005, and the total emissions for all 12 sectors and for the sectors modeled in zero-out simulations (and percentage of the modeled sectors in total emissions in parenthesis). Table S3 -Ozone concentrations for baseline and zero-out simulations (ppb), showing the population-weighted average for each region of the average 1-hr daily maximum ozone for the consecutive 6-month period with the highest average.  Table S6 -Regional and global total population, exposed population (adults age 25 and older), and average cause-specific mortality rates (for the exposed population). Table S7 -Premature ozone-related respiratory mortality in ten world regions, and contributions from different sectors (deaths in 2005), showing the mean and 95% CI from 1000 Monte Carlo simulations. All numbers are rounded to three significant digits. Regional mean values correspond to the values shown in Figure 3 of the main paper. Table S8 -Premature ozone-related respiratory mortality in ten world regions, and contributions from different sectors (deaths in 2005), showing the deterministic mean estimates. All numbers are rounded to three significant digits. Table S9 -Premature ozone-related respiratory mortality per million people in total population in ten world regions, and contributions from different sectors (deaths per million people in 2005), showing the mean from 1000 Monte Carlo simulations. All numbers are rounded to the nearest unit.
Table S10 -Premature PM 2.5 -related mortality in ten world regions (IHD+Stroke+COPD+LC), and contributions from different sectors (deaths in 2005), showing the mean and 95% CI from 1000 Monte Carlo simulations. All numbers are rounded to three significant digits. Regional mean values correspond to the values shown in Figure 5 of the main paper.
Table S11 -Premature PM 2.5 -related mortality (IHD+Stroke+COPD+LC) in ten world regions, and contributions from different sectors (deaths in 2005), showing the deterministic mean estimates. All numbers are rounded to three significant digits. showing the deterministic mean estimates. All numbers are rounded to three significant digits.

Input emissions
Input anthropogenic and biomass burning emissions were processed from datasets at 0.5° x 0.5° horizontal resolution prepared for the IPCC's Fifth Assessment Report and obtained through the GEIA-ACCENT Emission Data Portal 1 , for the IPCC AR5 Representative Concentration Pathway 8.5 (RCP8.5) global emissions inventory (Riahi et al. 2011(Riahi et al. ) for 2005 Anthropogenic emissions were processed as described by Fry et al. (2013), including speciation of volatile organic compound (VOC) species to MOZART-4 VOC categories, adding monthly temporal variation to all emissions species from anthropogenic sources, and regridding to 0.67°x0.5° horizontal resolution. We applied 1.15 and 1.4 factors to black and organic carbon species, respectively, to adjust for the conversion of emissions from PM 1 to PM 2.5 in the Bond inventory (Bond et al. 2004;Cooke et al. 1999) used in the emissions dataset, following Liu et al.  East and South East regions (regional average bias less than 25 ppb). In Europe, bias is lower than in the US.

Simulation with zeroed-out anthropogenic emissions
For this simulation we zeroed-out emissions from all anthropogenic sectors (i.e. Energy, Industry, Land Transportation, Shipping, Aviation, Residential & Commercial, Solvents, Agriculture, Agricultural Waste Burning, Waste). Also, we set total soil NOx emissions to a preindustrial value (3.6 Tg N yr -1 , estimated by Yienger and Levy, 1995), following Horowitz (2006), and we fixed methane concentrations at 722 ppb (Myhre et al. 2013).
For biomass burning emissions, we took into account that climate was likely the main driver of global fire activity up to the early 18th century. In contrast, anthropogenic influence plays a major role afterwards; fire ignition is the greatest contributor to the increase in global fire activity up to the end of the 19th century and fire suppression is responsible for a considerable decrease in global fire activity in the 20th century (Marlon et al. 2008;Pechony and Shindell 2010;Power et al. 2012 and references therein). Considering that present-day savannah fires are mostly ignited by humans (Andrae, 1991;Schultz et al. 2008), that conditions in closed tropical forests are usually not conducive to biomass burning (Krawchuk et al. 2009) and that humanignited fires related to tropical deforestation have greatly increased in recent decades (Mieville et al. 2010), we assumed natural emissions from savannah and tropical forest fires to be 10% of present-day emissions. We took into account that fire emissions likely increased substantially between 1750 and the late 1800s due to anthropogenic fire ignitions (Pechony and Shindell 2010;Power et al. 2012 and references therein), so a 50% assumption (Moulliot et al. 2006) would not be applicable. For forest fires at extratropical latitudes (greater than 30° N and 20° S), we assumed natural emissions to be 90% of present-day emissions to account for the reduced effect of anthropogenic ignition in these regions (Andrae, 1991;Eliseev et al. 2014), although others suggest that a lower percentage may be appropriate in the boreal Eurasian forests (Mollicone et al. 2006).

Ozone and PM 2.5 surface concentrations
All averaged concentrations mentioned in this section are population-weighted averages, unless otherwise stated. All regional results refer to the regions shown in Figure S7.

Ozone
Modeled hourly ozone concentrations were processed to estimate the average 1-hr daily maximum concentration for the consecutive 6-month period with the highest average, in each grid cell.
Ozone concentrations in the baseline simulation were 56.0 ppb globally, ranging from 39.9 to 64.1 ppb across ten world regions, with the highest values in East Asia, India and US (Table S3 and Figure S8). When all anthropogenic emissions were zeroed-out, global concentrations decreased to 22.2 ppb (40% of baseline) and regional concentrations ranged from 17.0 to 24.3 ppb, reflecting a much lower spatial variability in surface ozone (Table S3).
Considering the results of the simulations with the different sectors zeroed-out (Table S3 and
When all anthropogenic emissions were zeroed-out, global concentrations decreased to 6.6 µg/m 3 (30% of baseline) and regional concentrations ranged from 1.4 to 20.6 µg/m 3 with the highest value in the Middle East (Table S4). Also, the reduction was the greatest in East Asia (less 91%) and the least in the Middle East (less 26%) and Africa (less 30%), due to the fraction of dust in these latter regions.
Considering the results of the simulations with the different sectors zeroed-out (Table S4 and Figure S12), Residential & Commercial had the highest contribution to surface PM 2.5 , globally (28%) and in several regions. Land Transportation had the greatest individual contribution in North America and Europe. Table S6 shows the global and regional total population, exposed population (adults aged 25 and older), and cause-specific baseline mortality rates obtained from the gridded values used in the health impact assessment.

Additional mortality results
Tables S7 and S10 show present-day global and regional burdens of anthropogenic ozone and PM 2.5 -related mortality.
We also obtained deterministic estimates for mortality, considering the mean values reported for each of the variables (see Tables S8 and S11). For anthropogenic ozone mortality, the global mean from the Monte Carlo simulations is 0.1% greater than the deterministic estimate (492,000 deaths/year), with less than ±1% differences for most regions, except Middle East (2%) and Africa (5%). For anthropogenic PM 2.5 mortality, there is a larger difference (-5%) between the global mean from the Monte Carlo simulation and the deterministic estimate (2.36 million deaths/year), reflecting large differences at the regional level in some regions (e.g. -15% in India, -4% in East Asia, -6% in Europe, 19% in North America).
Tables S9 and S12 show ozone-and PM 2.5 -related total mortality rates per sector, considering total population, per region and globally. India and East Asia have 113 deaths per million people per year due to exposure to anthropogenic ozone, compared to 69 deaths per million people in North America and lower rates in all other regions (

Fine vs. coarse resolution
Using the output from simulations at 2.5°x1.9° horizontal resolution, we estimate a present-day global burden of anthropogenic ozone-related mortality of 480,000 deaths/year, a slight negative bias of 2% relative to the deterministic estimate at finer resolution (Table S13); this negative bias was greater in some regions (e.g. South America, -9%; Southeast Asia, -8%; Africa, -6%; Australia, -12%). For anthropogenic PM 2.5 mortality, we estimate 2.7 million deaths/year, which corresponds to a global positive bias of 16% for the coarse resolution estimate, resulting from considerably different regional bias, including much larger positive biases in North America (52%), Europe (34%) and FSU (39%), smaller positive biases in India (4%), East Asia (11%) and Southeast Asia (11%), and a 15% negative bias in South America (Table S13).
If we regrid the output from simulations at 0.67°x0.5° horizontal resolution to 2.5°x1.9° horizontal resolution, we estimate a present-day global burden of anthropogenic ozone-related mortality of 479,000 deaths/year, a slight negative bias of 3% relative to the deterministic estimate at finer resolution, while for anthropogenic PM 2.5 mortality we estimate a burden of 2.2 million deaths /year, corresponding to a negative bias of 8% (Table S14).

Low-concentration threshold for ozone
By applying a low-concentration threshold of 33.3ppb to the ozone mortality calculation, we estimate a 27.4% decrease in the global mortality burden, since ozone concentrations for the simulation with all anthropogenic emissions zeroed-out are below the threshold, except in a few very small areas in Africa. The low-concentration threshold has a negligible effect on the contributions of each sector (less than 2%) since ozone concentrations with zeroed-out sectors are lower than the threshold only in a few populated areas, mostly in South America and Australia. Grassland burning Savanna burning, grassland fires * All Transportation includes these three sectors. ** According to Lamarque et al. (2010), includes emissions from fuelwood burning and charcoal production.   Table S6 -Regional and global total population, exposed population (adults age 25 and older), and average cause-specific mortality rates (for the exposed population).  Table S7 -Premature ozone-related respiratory mortality in ten world regions, and contributions from different sectors (deaths in 2005), showing the mean and 95% CI from 1000 Monte Carlo simulations. All numbers are rounded to three significant digits.
Regional mean values correspond to the values shown in Figure 3 of the main paper.