{"id":564210,"date":"2026-07-01T23:39:20","date_gmt":"2026-07-01T23:39:20","guid":{"rendered":"https:\/\/www.europesays.com\/ie\/564210\/"},"modified":"2026-07-01T23:39:20","modified_gmt":"2026-07-01T23:39:20","slug":"residual-emissions-may-perpetuate-community-scale-inequalities-in-us-air-pollution","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ie\/564210\/","title":{"rendered":"Residual emissions may perpetuate community-scale inequalities in US air pollution"},"content":{"rendered":"

We quantify the community-specific PM2.5 air pollution and health-related impacts of one reference scenario and two net-zero scenarios in the USA in five steps. First, we run an integrated assessment model (GCAM-USA) to estimate future emissions from the energy sector in three scenarios. Second, we downscale state level emissions to a resolution consistent with the Emissions & Generation Resource Integrated Database (eGRID) and the NEI. Third, we run WRF-CMAQ to estimate ambient PM2.5 concentrations in each scenario to 2050. Fourth, we run BenMAP to quantify premature mortality, and the costs associated with death from our scenarios. Fifth, we identify the demographic and economic characteristics of people living in each pixel with census data at the census block group level. Extended Data Fig. 1<\/a> illustrates the steps of our research; the sections below detail each step, with further explanations in the Supplementary Information<\/a>. Code and data related to this research are publicly available via Zenodo at https:\/\/doi.org\/10.5281\/zenodo.20076464<\/a> (ref. 68<\/a>) and at https:\/\/doi.org\/10.5281\/zenodo.20045877<\/a> (ref. 69<\/a>).<\/p>\n

Step 1: estimate future energy and emissions (GCAM)<\/p>\n

We use the release version of GCAM-USA v6.070<\/a>, a model that has been used in the Intergovernmental Panel on Climate Change scenarios71<\/a> and in the White House 2021 Long-term Strategy72<\/a>. For a detailed model explanation, refer to the \u2018Methodological notes\u2019 section in the Supplementary Information<\/a>.<\/p>\n

We model one reference BAU scenario and two net-zero GHG scenarios varying the amount of CDRs allowed at a national level (high and low). By limiting carbon removals, the model finds different market structures to solve for the net-zero emissions constraint, including higher decarbonization of other end-use sectors (buildings, transportation and industry), earlier\/later fossil power plant retirements and small demand reductions when alternative technologies are not available. We focus on emissions from the energy system from all relevant pollution sectors: electricity generation, industrial activity, transportation, resource production and buildings73<\/a>. Emissions from these sources are tracked at a state level in 5-year intervals from 2015 to 2050.<\/p>\n

For the reference scenario, we model a default GCAM v6.0 scenario, without modifying any assumptions. The reference scenario follows historical trends where the gross domestic product and population continue growing, thus increasing the demand for services. The historical electrification of end-use sectors continues, which leads to an increase in total electricity generation. The reference scenario also has state-specific assumptions about coal and nuclear retirements, and state-specific assumptions on hydro generation based on recent trends. Natural gas electricity generation grows into the future, following historical trends, displacing coal. This scenario includes the Clean Air Act section 111 (b), which limits CO2 emissions from new steam-generating electricity and base-load natural gas plants. This reference scenario also assumes no new development of coal-fired power plants without CCS. There is electricity trade between states. Refining follows historical trends and can therefore only occur in states with an existing history of refining. Biomass refining can only be developed in states where suitable feedstocks are available. For further information about reference scenario assumptions, refer to ref. 70<\/a>.<\/p>\n

The two net-zero scenarios follow the US nationally determined contribution for 2030 of 50% reduction below 2005 GHG emissions, followed by a net-zero GHG emissions target by 2050. The model solves for this GHG constraint by applying a shadow carbon price in the economy that forces net-zero emissions by mid-century (Supplementary Fig. 9<\/a>). To avoid leakages in this global model, we include a net-zero CO2 emissions constraint for the rest of the world by 2060, consistent with a 2\u2009\u00b0C world, assuming some regions achieve this sooner and others later. We have additionally reduced the emission factors for electricity generation technologies with carbon capture and storage (CCS) for SO2, NOx, PM2.5 and particulate matter with a diameter of 10\u2009\u00b5m or smaller (PM10) following Technical Report No. 14 from the European Environment Agency74<\/a> (Supplementary Table 17<\/a>), which were previously assumed in GCAM to be the same as the non-CCS technology counterparts.<\/p>\n

We include the CDR portfolio available in the release version of GCAM v6.0 used here, which models anthropogenic CDR from BECCS in refining, electricity generation and hydrogen production, as well as DAC, and CDR from AR. The two net-zero scenarios differ in their amount of CDR: there is an unrestricted scenario that leads to higher amounts of CDR, and there is a restricted scenario that leads to lower amounts of CDR. To restrict CDR, we have artificially inflated the cost of carbon storage in GCAM, which also constrains fossil-CCS (Supplementary Table 6<\/a>). The three scenarios are listed in Supplementary Table 2<\/a>, and additional scenario outputs are presented in Supplementary Tables 3<\/a>\u20135<\/a>. We run these scenarios from 2015 (last historical year in GCAM) to 2050 (net-zero target year). We downscale future emissions for 2050 and for all the energy sectors.<\/p>\n

Step 2: downscale future emissions (eGRID and NEI)<\/p>\n

Given GCAM modelling details and data availability, we use two approaches for downscaling emissions: one for electricity generation and another for other energy sectors. We also test different assumptions to electricity generation siting and retiring, with details in the \u2018Sensitivities\u2019 section in the Supplementary Information<\/a>.<\/p>\n

Electricity generation<\/p>\n

Based on plant-by-plant details from GCAM-USA for the US electricity sector and eGRID database, we downscale electricity at a point-source level (that is, power plant) following retirement and addition quotas from GCAM. We downscale GCAM outputs based on 2020 eGRID data at the generator level. GCAM models electricity generation at the state level in each model period (that is, every 5\u2009years) by fuel and technology. eGRID provides current electricity generation in the USA for each power plant by fuel and technology. From the 30,193 generation sources in eGRID 2020 dataset75<\/a>, we filter for 10,707 that represent combustion fuels (coal, gas, oil and biomass), that are operating or planned to operate and that have positive generation (Supplementary Fig. 11<\/a>). These generators serve as our baseline.<\/p>\n

From GCAM, we calculate the gross electricity generation additions (that is, addition quota) and retirements (that is, retirement quota) in each state by scenario, fuel and technology by 2050 (net-zero target year) compared with 2020 (eGRID data). Given that GCAM is not calibrated to eGRID, we scale GCAM quotas to eGRID generation by state, fuel and technology. This gives us a total amount of generation to be added and to be retired in each state by scenario, fuel and technology. Once we have the final scaled retirement and addition quotas, we establish a retiring and addition schedule as follows:<\/p>\n

    \n
  1. \n (1)<\/p>\n

    We first retire generators following the Energy Information Administration\u2019s retirement schedule58<\/a>. Note that we do not allow over-retirements; we only retire generators until the GCAM retirement quota is met.<\/p>\n<\/li>\n

  2. \n (2)<\/p>\n

    If this is not sufficient\u2014that is, if the GCAM retirement quota is not met after following the Energy Information Administration retirement schedule\u2014we begin retiring generators based on age, prioritizing the retirement of older units first. This is consistent with the findings of Mills et al.59<\/a>.<\/p>\n<\/li>\n

  3. \n (3)<\/p>\n

    If the GCAM retirement quota is larger than existing generation, we subtract this from the addition quota, so that the net change in generation is not affected.<\/p>\n<\/li>\n<\/ol>\n

    Once the retirement quota is met, we proceed to work with GCAM addition quotas.<\/p>\n

      \n
    1. \n (1)<\/p>\n

      We first max out existing generators in that state of the same fuel and technology, assuming a potential maximum capacity factor of 85%.<\/p>\n<\/li>\n

    2. \n (2)<\/p>\n

      If this was not enough to meet the addition quota, then we build new generators in existing plants that have the same technology based on a weighted distribution, so that larger power plants receive larger generators and smaller power plants receive smaller generators. This is consistent with previous studies56<\/a>,57<\/a>.<\/p>\n<\/li>\n<\/ol>\n

      The previous steps give us total electricity generation at a generator and plant level by scenario, fuel and technology by 2050. We then apply emission factors from GCAM by state, fuel, technology, pollutant and period to estimate total emissions for nine different pollutants: black carbon (BC), carbon monoxide (CO), ammonia (NH3), nitrogen oxides (NOx), non-methane volatile organic compounds (NMVOC), organic carbon (OC), PM2.5, PM10, and sulfur dioxide (SO2) (see Supplementary Fig. 12<\/a> for a state example for PM2.5). These 2050 point-source electricity emissions are adjusted to a 2019 baseline following the EPA air pollutant emissions trends76<\/a>. We use 2019 as the base-year case, instead of 2020, because the global pandemic disrupted emission trends. These are the electricity emissions used to drive the WRF-CMAQ air quality simulations. Note that we provide a sensitivity test to these assumptions in the \u2018Sensitivities\u2019 section in the Supplementary Information<\/a>.<\/p>\n

      Other energy sectors<\/p>\n

      Other energy sectors in GCAM include industry, transportation, refining, resource production, buildings (commercial and residential) and urban areas. Given that GCAM does not have detailed technology information in these energy sectors, we simulate the emission change into the future (that is, 2050) for a given sector compared with 2020 and apply this factor to the EPA\u2019s NEI 2020 dataset40<\/a> to estimate total emissions in 2050.<\/p>\n

      We match 51 non-point sectors and 84 facility types in NEI to sectors in GCAM and exclude emissions from power plants. It is noted that we keep emissions from six non-point sectors (that is, agriculture and livestock dust, road dust, biogenic, prescribed fires and wildfires) and three facility types (that is, crematory animal, crematory human and military base) constant with the base-year level, as they are not included in GCAM and are outside the scope for this study. In addition, the changes in emissions from transportation are applied to NEI in 2017, as opposed to 2020, because of the impact of the 2020 global pandemic on transportation emissions. Finally, these 2050 emissions are also scaled to a 2019 baseline following values based on the EPA air pollutant emissions trends76<\/a> and used to drive the WRF-CMAQ air quality simulations.<\/p>\n

      Anthropogenic emissions processing for CMAQ input<\/p>\n

      Anthropogenic emissions for the USA are based on the NEI 2020 and scaled to a 2019 baseline using EPA annual emissions trends. For future-year scenarios, sector- and state-specific scaling factors were applied to the corresponding NEI sectors.<\/p>\n

      To generate gridded, hourly anthropogenic emissions for CMAQ, we processed all NEI-based inventories with the Sparse Matrix Operator Kernel Emissions (SMOKE) system using the US EPA 2020 Emissions Modeling Platform ancillary inputs (including spatial surrogates, temporal profiles and chemical speciation profiles)77<\/a>,78<\/a>. Detailed emission input files include (1) plant-level power-sector point-source inventory (see \u2018Electricity generation\u2019 section for details) and (2) NEI 2020 inventories (excluding power plants) scaled to a 2019 baseline and projected to 2050 using GCAM-derived gas-, state- and sector-specific scaling factors (see \u2018Other energy sectors\u2019 section for details). Domain-specific gridding was performed for both 9-km CONUS domain and 1-km nested domains by using grid-specific spatial surrogates and applying SMOKE spatial allocation to produce gridded emissions for each modelling grid. Temporal allocation converts annual totals to hourly emissions using sector-specific temporal profiles and from the EPA platform. Specifically, hourly timing for power sector was based on base-year CEMS- and heat-input-derived patterns, and the same profiles were applied in future scenarios to distribute projected power-sector emissions to the hourly scale. Accordingly, future power-sector emissions in our analysis vary in magnitude and spatial distribution, but not in assumed sub-annual dispatch timing.<\/p>\n

      Chemical speciation is then applied to map inventory pollutants (for example, NMVOC and primary PM) to CMAQ mechanism species consistent with CB05 gas-phase chemistry and the AE6 aerosol scheme. We use SPECIATE-based speciation profiles and SMOKE speciation inputs generated with the Speciation Tool and coupled with platform-maintained speciation cross-references79<\/a>. Then, the NMVOC emission changes from GCAM are applied to baseline NMVOC totals and are apportioned within SMOKE into CB05 VOC model species (for example, FORM and ISOP) using the same source-category profiles as the base year. Finally, CMAQ-ready emissions for each domain and scenario-year are produced as NetCDF files containing gridded, hourly, speciated emissions ready for WRF-CMAQ simulations.<\/p>\n

      Step 3: model PM2.5 concentrations (WRF-CMAQ)<\/p>\n

      In the third step, we employ the Weather Research and Forecasting (WRF, version 4.0.1) and the Community Multiscale Air Quality (CMAQ, version 5.2.1) to estimate changes in long-term PM2.5 air quality under different scenarios. We focus on PM2.5 because of its large health impacts and, therefore, its potential to deliver the greatest health benefits when reduced46<\/a>,80<\/a>.<\/p>\n

      We design double-nested simulations, with the first domain covering the contiguous USA at a 9-km scale, and 12 nested domains covering the 15 most populous metropolitan statistical areas in the USA at a 1-km scale (Supplementary Fig. 13<\/a> and Supplementary Table 18<\/a>). The vertical resolution is 23 sigma levels from surface to tropopause (about 100\u2009mb) for WRF simulations, and 14 sigma levels for the CMAQ model. We conduct a total of four groups of experiments, including one base-year case (that is, 2019) and three future 2050 emission scenarios (that is, reference, net-zero high-CDR and net-zero low-CDR). All the simulations are conducted throughout the whole year and with a 1-month spin-up. In addition, we run 36 simulations at a 1-km resolution for the 15 cities of interest (12 base-year case 2019, 12 net-zero high-CDR scenario in 2050 and 12 net-zero low-CDR scenario in 2050).<\/p>\n

      We use NCEP final analysis data to drive WRF simulations and provide meteorological inputs fixed in 2019 for all simulations81<\/a>. That is, all WRF-CMAQ simulations are driven by the same meteorological year (2019; NCEP FNL), consistent with an emissions-only experimental design commonly used to isolate scenario-driven changes in air quality38<\/a>,82<\/a>,83<\/a>,84<\/a>,85<\/a>; consequently, our interscenario differences reflect emissions-driven responses under present-day meteorology rather than climate-driven meteorological change.<\/p>\n

      Historical anthropogenic emissions for the USA are obtained from NEI 2020 and are scaled to the 2019 level according to the EPA air pollutant annual emission trends. As noted earlier, we used 2019 as a base-year case as opposed to 2020 because the global pandemic disrupted emission trends. Future emission trends (that is, 2050) are provided by GCAM downscaling process (see \u2018Step 2: downscale future emissions\u2019 section for details). Historical and future anthropogenic emissions for other bordering countries in the first 9-km domain are derived from the Community Emissions Data System (CEDS)86<\/a> and the Coupled Model Intercomparison Project (CMIP)87<\/a> databases, respectively, and scaled following GCAM regional values for future years. Natural source emissions, including dust, open biomass burning and biogenic emissions are also incorporated and fixed at 2019 levels. Besides, the chemical initial and boundary conditions for the first domain are interpolated from the dynamic outputs of GEOS-Chem model, which are driven by future gridded CMIP6 emissions. Detailed model configurations, including meteorological and chemical schemes, anthropogenic and natural emission sources, chemical initial and boundary conditions are listed in Supplementary Table 19<\/a>.<\/p>\n

      We evaluate our base-year PM2.5 simulations with in situ observations, which were collected from Air Quality System (AQS) monitoring network, maintained by the US EPA. The evaluations of annual PM2.5 simulations for the contiguous USA (Supplementary Fig. 14<\/a>) and 15 most populated cities (Supplementary Fig. 15<\/a>) suggest a reliable performance of our air quality modelling system.<\/p>\n

      To reduce the uncertainty of systematic CMAQ bias in absolute PM2.5 levels, we apply a high-resolution and observation-informed ground-level PM2.5 dataset (GlobalHighPM2.5)88<\/a>, which was developed with a combination of ground measurements, satellite retrievals, atmospheric reanalysis and simulations, to systematically calibrate WRF-CMAQ simulations. Equation (1<\/a>) denotes the calibration process, where i and j represent the specific simulation case and year, respectively; C and CSIM refer to the calibrated and original simulated PM2.5 concentrations, respectively. This formulation anchors the baseline spatial pattern and magnitude to an observation-constrained field while applying the CMAQ-modelled fractional change (\\({{C}_{\\mathrm{SIM}}}_{i,j}\/{{C}_{\\mathrm{SIM}}}_{\\mathrm{base},\\,2019}\\)) to project scenario-year concentrations, thereby reducing CMAQ-induced uncertainties while preserving scenario-driven contrasts.<\/p>\n

      $${C}_{i,\\,j}={{C}_{\\mathrm{GlobalHighPM}2.5}}_{2019}\\times \\frac{{{C}_{\\mathrm{SIM}}}_{i,\\,j}}{{{C}_{\\mathrm{SIM}}}_{\\mathrm{base},2019}}.$$<\/p>\n

      \n (1)\n <\/p>\n

      Step 4: model PM2.5-related mortality (BenMAP)<\/p>\n

      In the fourth step, we use the Benefits Mapping and Analysis Program\u2013Community Edition (BenMAP-CE, version 1.5) to estimate the burden to human health of total air pollution from each scenario in the contiguous USA at a 9-km resolution and in the 15 most populated metropolitan statistical areas at a 1-km resolution. BenMAP-CE is an epidemiological model developed by the EPA to estimate the number and economic value of air pollution-related deaths and illness. The model uses health impact functions derived from the published epidemiology literature, considering air quality changes, population, baseline incidence rates and an effect estimate. BenMAP calculates the health impact based on the following function:<\/p>\n

      $$\\Delta Y=\\left(1-{{\\rm{e}}}^{-\\beta \\times \\Delta \\mathrm{AQ}}\\right)\\times {Y}_{{\\rm{o}}}\\times \\mathrm{Pop},$$<\/p>\n

      \n (2)\n <\/p>\n

      where \u0394Y is the estimated health impact attributed to air pollution, \u03b2 is the beta coefficient from an epidemiologic study, \u0394AQ is a defined change in air quality, Yo is the baseline rate for the health effect of interest and Pop is the population exposed to air pollution. The reductions in premature mortality are expressed in total deaths and in monetary terms based on the VSL. VSL is calculated based on the aggregate dollar amount that people would be willing to pay for a small reduction in their individual risk of dying in a given year89<\/a>,90<\/a>.<\/p>\n

      The PM2.5 pollution data are provided by WRF-CMAQ, as explained in step 3, while the population data are from the American Community Survey (ACS) from the US Census Bureau from 2019 (specifically table ACSDT5Y2019.B01003). The population data were rescaled to the desired resolutions (9-km grids for the contiguous USA and 1-km grids for each city analysis). In addition, we run a sensitivity analysis for three cities using future population projections from Wang et al.91<\/a>, who estimate future population for each SSP at a 1-km resolution. Details are provided in the \u2018Sensitivities\u2019 section in the Supplementary Information<\/a>.<\/p>\n

      For health baseline incidence, we use BenMAP 2015 and 2020 values and linearly extrapolate for 2019 and then rescale these 2019 values from the county level to our 9-km and 1-km grids based on a weighted average per pixel based on overlap area. For our health analysis, we use the health impact function from Pope et al.92<\/a> included in BenMAP. The authors in this study examined the relationship between long-term PM2.5 exposure and mortality in the contiguous USA for 1,599,329 adults aged 18\u201384\u2009years who were interviewed by the National Health Interview Surveys between 1986 and 2014. For the economic valuation, we use the EPA\u2019s Standard Valuation Functions on a current undiscounted VSL of US$8.7 million (2015 US dollars), which represents the mean of a distribution fitted to 26 VSL estimates and is used by the EPA in Regulatory Impact Analyses93<\/a>.<\/p>\n

      We run three main BenMAP scenarios for the 9-km resolution WRF-CMAQ results (reference scenario in 2050; net-zero high-CDR scenario in 2050; and net-zero low-CDR scenario in 2050), and include four additional runs in Extended Data Figs. 2<\/a> and 3<\/a> (base-year case 2019; reference scenario in 2050 versus net-zero high-CDR scenario in 2050; reference scenario in 2050 versus net-zero low-CDR scenario in 2050; and net-zero high-CDR scenario in 2050 versus net-zero low-CDR scenario in 2050). For the 1-km resolution results, we run 15 main BenMAP scenarios, one per city, to see the difference between net-zero high-CDR and net-zero low-CDR scenarios in 2050, and run BenMAP 45 more times to represent base-year 2019, net-zero high-CDR and net-zero low-CDR mortality for each city to help interpret results and estimate population-weighted mortality. In addition, we test three sensitivities to changes in the baseline incidence mortality rates following different assumptions on income and race\u2013ethnicity for the city of Philadelphia, with further details in the \u2018Sensitivities\u2019 section in the Supplementary Information<\/a>.<\/p>\n

      We note that, while GCAM-USA does not represent sub-annual electricity system dynamics such as unit-level dispatch, transmission constraints or intra-annual variability in plant utilization, this limitation is less consequential for the long-term health impact assessment conducted here. BenMAP estimates mortality based on changes in annual average PM2.5 exposure and concentration\u2013response relationships derived from long-term epidemiological studies. As a result, the analysis is primarily sensitive to differences in total annual emissions and their spatial distribution rather than the precise timing of emissions within a year. Although sub-annual variability in power plant dispatch may influence short-term pollution episodes and acute health outcomes, these effects are not the focus of this study and are expected to have a limited influence on the relative changes in long-term PM2.5 exposure and associated mortality estimated across scenarios63<\/a>,94<\/a>.<\/p>\n

      The CIs in mortality and economic valuation come from BenMAP, which performs a full Monte-Carlo analysis by sampling an uncertainty distribution around the incidence coefficients or willingness to pay estimates.<\/p>\n

      Step 5: assess distributional pollution and health outcomes<\/p>\n

      In the fifth step, we identify the race\u2013ethnicity and median household income for people living in each pixel at a 1-km scale for each of the 15 cities analysed. To identify demographic information, we use data from the Census Bureau, specifically from the ACS. The ACS is a nationwide survey that collects and produces information on social, economic, housing, and demographic characteristics in the USA every year. The survey is conducted by the US Census Bureau to gather information at a community level that helps determine how US$675 billion in federal and state funds are distributed every year. About 3.5 million US households (1 in 38) per year receives an invitation to participate in the survey, and the participant is required to fill the questionnaire. The household is selected to statistically represent other households in the surrounding community. The USA is divided into regions, divisions, states, counties, census tracts and block groups. The smallest unit available in the ACS is the block group, and it generally contains between 600 and 3,000 people95<\/a>. Other studies have used the ACS to analyse air pollution exposure disparities in the US population and income groups18<\/a>, race, age and poverty19<\/a>,22<\/a>.<\/p>\n

      In this project, we use the 2015\u20132019 ACS 5-Year Data Products, as this dataset includes smaller population groups and provides higher spatial resolution, which is important for assessing the distribution of air pollution impacts. The 5-year products are not just an average of 5\u2009years, but rather data pooled over 60\u2009months, weighted to produce estimates controlling for age, race and Hispanic ethnicity. The 5-year estimates are thus more robust than 1- or 3-year estimates for analysing data for small population groups. We focus on demographic and socioeconomic variables, including race\u2013ethnicity (table ACSDT5Y2019.B03002) and median household income (table ACSDT5Y2019.B19013).<\/p>\n

      We use data at the block group level, and when data are missing for income because of privacy issues we use data at higher census levels. We grouped race\u2013ethnicity into the following: \u2018Hispanic or Latino\u2019, \u2018non-Hispanic white\u2019, \u2018non-Hispanic Black or African American\u2019 and \u2018Other\u2019, which includes non-Hispanic Native Hawaiian and Other Pacific Islander alone, non-Hispanic some other race alone, and non-Hispanic two or more races. To simplify our analysis, we then group our data into three bins for race\u2013ethnicity, and three bins for median household income. The race\u2013ethnicity bins relate to the percent of non-Hispanic whites in each group: 0\u201330% non-Hispanic white, 31\u201360% non-Hispanic white and 61\u2013100% non-Hispanic white. The income bins are created based on income percentiles in each city: 0\u201333rd percentile, 33rd percentile plus one US dollar to the 66th percentile, and 66th percentile plus one US dollar and higher. Supplementary Fig. 16<\/a> shows the racial distribution of each city, and Supplementary Table 20<\/a> contains the median household income cut-off point.<\/p>\n

      In our analysis, we rescale census block group data to calculate the amount of people by race\u2013ethnicity in each pixel and the median household income, using R. For population, we assume an equal distribution across the pixel, and for income, we calculated the weighted mean for the pixel based on the area of overlap. We then calculate the population-weighted PM2.5 and population-weighted mortality, similar to refs. 18<\/a>,24<\/a>,26<\/a>, as follows:<\/p>\n

      $$\\mathrm{weighted}\\,\\mathrm{average}=\\frac{{\\sum }_{i=1}^{n}\\left({X}_{i}\\times {W}_{i}\\right)}{{\\sum }_{i=n}^{n}{W}_{i}},$$<\/p>\n

      \n (3)\n <\/p>\n

      where Xi is the PM2.5 concentrations or mortality in each pixel and Wi is the population in that pixel. The population-weighted value is thus the sum of the product of either PM2.5 concentration or mortality in each pixel and the population in that pixel, divided by the total population in the city (for the 1-km analysis). We calculated this for the population overall, and then for each race\u2013ethnicity group and income group. For further details, refer to methodological notes in the Supplementary Information<\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"We quantify the community-specific PM2.5 air pollution and health-related impacts of one reference scenario and two net-zero scenarios…\n","protected":false},"author":2,"featured_media":564211,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","_share_on_mastodon":"0"},"categories":[269],"tags":[442,96230,6134,74466,18,440,6129,29959,910,19,17,133],"class_list":["post-564210","post","type-post","status-publish","format-standard","has-post-thumbnail","category-environment","tag-climate-change","tag-climate-change-policy","tag-climate-change-climate-change-impacts","tag-climate-change-mitigation","tag-eire","tag-environment","tag-environmental-impact","tag-environmental-law-policy-ecojustice","tag-general","tag-ie","tag-ireland","tag-science"],"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@ie\/116847363213693390","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/564210","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/comments?post=564210"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/564210\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media\/564211"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media?parent=564210"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/categories?post=564210"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/tags?post=564210"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}