《环境政策与管理》课程教学资源：The Cost of Air Pollution（World Bank 2016）

of the low-and middle-income countries.In addition to the inadequate and highly uneven coverage of ground-level measurements,measurement protocols and techniques are not stan- dardized globally,with different quality control programs and different numbers of samples to arrive at annual averages.Even for measurements made by(similar)filter-based approaches, filters are equilibrated at different relative humidity conditions prior to weighing(for example, 35 percent,40 percent,and 50 percent relative humidity in the United States,Canada,and European Union,respectively)and therefore are not completely equivalent.In addition,PM measurements and PM,/PM ratios are commonly used to infer PM,concentrations for ground-level estimates.Therefore,surface measurements,although a key component of any global assessment approach,cannot be used solely to derive global exposure estimates Satellite-based measurements can help provide estimates for areas with no ground-level mon- itoring networks.But even in North America,where monitor density is high in populated areas,studies have indicated that satellite-based estimates do provide additional useful infor- mation on spatial and temporal patterns of air pollution(Kloog et al.2011,2013;Lee et al. 2012).Furthermore,in a large population-based study in Canada the magnitudes of estimated mortality effects of PM,derived from ground measurements and satellite-based estimates were identical(Crouse et al.2012).Satellite-based estimates nonetheless must be used cau- tiously,as discussed shortly. Deriving Ambient Air Pollution Concentrations Long-term average exposure to PM,was estimated at 0.1X 0.1 resolution.Satellite-based estimates that incorporated additional information on temporal trends were applied,as well as chemical transport model simulations incorporating internally consistent emissions trends from 1990 to 2013.Available surface measurements of PM,were incorporated to calibrate the estimates based on satellite retrievals and chemical transport model simulations.Data and methodologies are summarized here and reported in more detail in Brauer et al.(2016). A series of satellite-based estimates for PM,were used,which included year-specific estimates for 1998-2012.Satellite-based PM,s estimates used aerosol optical depth (AOD)retrievals from multiple satellites to estimate near-surface PM,by applying the relationship of PM,s to AOD simulated by the GEOS-Chem chemical transport model.These updated PM,esti- mates used both“unconstrained”and“optimal-.estimation'”AOD retrievals in combination with the MODIS,MISR,and SeaWiFS satellite-borne instruments.These estimates were com- bined with information on temporal variation based on SeaWiFS and MISR to estimate global PM,at0.1°×0.1°for2000,2005,2010,and2011. TM5-FASST(FAst Scenario Screening Tool,a reduced-form version of the TM5 chemical transport model)simulations for 1990,2000,and 2010 were included,using an updated set of emissions inventories and constant meteorological inputs and emissions from dust and sea salt (see box 2.1 for the current understanding of the health impacts of dust and a discussion of why these sources were included in exposure estimates).Emissions of windblown mineral dust and sea salt were estimated in the TM5 model by incorporating information on land cover and wind speed,combined with emission factors.For simulations,a constant "typical" meteorologic year with corresponding emissions from windblown mineral dust and sea salt was used.Year-to-year variations in emissions of windblown mineral dust and sea salt were, 12 The Cost of Air Pollution:Strengthening the Economic Case for Action

12 The Cost of Air Pollution: Strengthening the Economic Case for Action of the low- and middle-income countries. In addition to the inadequate and highly uneven coverage of ground-level measurements, measurement protocols and techniques are not stan￾dardized globally, with different quality control programs and different numbers of samples to arrive at annual averages. Even for measurements made by (similar) filter-based approaches, filters are equilibrated at different relative humidity conditions prior to weighing (for example, 35 percent, 40 percent, and 50 percent relative humidity in the United States, Canada, and European Union, respectively) and therefore are not completely equivalent. In addition, PM10 measurements and PM2.5 /PM10 ratios are commonly used to infer PM2.5 concentrations for ground-level estimates. Therefore, surface measurements, although a key component of any global assessment approach, cannot be used solely to derive global exposure estimates. Satellite-based measurements can help provide estimates for areas with no ground-level mon￾itoring networks. But even in North America, where monitor density is high in populated areas, studies have indicated that satellite-based estimates do provide additional useful infor￾mation on spatial and temporal patterns of air pollution (Kloog et al. 2011, 2013; Lee et al. 2012). Furthermore, in a large population-based study in Canada the magnitudes of estimated mortality effects of PM2.5 derived from ground measurements and satellite-based estimates were identical (Crouse et al. 2012). Satellite-based estimates nonetheless must be used cau￾tiously, as discussed shortly. Deriving Ambient Air Pollution Concentrations Long-term average exposure to PM2.5 was estimated at 0.1° 3 0.1° resolution. Satellite-based estimates that incorporated additional information on temporal trends were applied, as well as chemical transport model simulations incorporating internally consistent emissions trends from 1990 to 2013. Available surface measurements of PM2.5 were incorporated to calibrate the estimates based on satellite retrievals and chemical transport model simulations. Data and methodologies are summarized here and reported in more detail in Brauer et al. (2016). A series of satellite-based estimates for PM2.5 were used, which included year-specific estimates for 1998–2012. Satellite-based PM2.5 estimates used aerosol optical depth (AOD) retrievals from multiple satellites to estimate near-surface PM2.5 by applying the relationship of PM2.5 to AOD simulated by the GEOS-Chem chemical transport model. These updated PM2.5 esti￾mates used both “unconstrained” and “optimal-estimation” AOD retrievals in combination with the MODIS, MISR, and SeaWiFS satellite-borne instruments. These estimates were com￾bined with information on temporal variation based on SeaWiFS and MISR to estimate global PM2.5 at 0.1° 3 0.1° for 2000, 2005, 2010, and 2011. TM5-FASST (FAst Scenario Screening Tool, a reduced-form version of the TM5 chemical transport model) simulations for 1990, 2000, and 2010 were included, using an updated set of emissions inventories and constant meteorological inputs and emissions from dust and sea salt (see box 2.1 for the current understanding of the health impacts of dust and a discussion of why these sources were included in exposure estimates). Emissions of windblown mineral dust and sea salt were estimated in the TM5 model by incorporating information on land cover and wind speed, combined with emission factors. For simulations, a constant “typical” meteorologic year with corresponding emissions from windblown mineral dust and sea salt was used. Year-to-year variations in emissions of windblown mineral dust and sea salt were, 1700234_Cost of Pollution.indd 12 8/29/16 1:55 PM

BOX 2.1 Dust and Dust Storm Health Effects Because there is no evidence that the dust components of PM,should be excluded when esti- mating health impacts,concentrations of windblown mineral dust and sea salt were included in estimating the health impacts of exposure to ambient PM,for the GBD 2013 study.The cur- rent positions of the U.S.Environmental Protection Agency(EPA),World Health Organization (WHO),and International Agency for Research on Cancer(IARC)are that an insufficient basis exists for using separate indicators for a specific PMscomponent or group of components asso- ciated with any source category of fine particles.Many constituents of particulate matter can be linked with differing health effects,and the evidence is not yet sufficient to allow differentiation of those constituents or sources that are more closely related to specific health outcomes(EPA 2009;IARC 2013;WHO 2014).Concentrations of windblown mineral dust and sea salt were therefore included in estimating the health impacts of exposure to ambient PM,for the GBD 2013 study. For windblown mineral dust specifically,there is substantial evidence of its association with mor- tality and morbidity during episodes of high concentrations such as Saharan dust storms that affect Europe(Perez et al.2008;Mallone et al.2011;Karanasiou et al.2012),Asian dust storms (Chen et al.2004;Bell,Levy,and Lin 2008),or regional episodes in the Middle East(Thalib and Al-Taiar 2012;Vodonos et al.,2014,2015).Most of this evidence points to the coarse fraction of particulate matter,or to PMand not to the smaller proportion of dust that is in the PM,frac- tion.In toxicology,there is no evidence that dust is more benign than other components of PMs (WHO2014). however,incorporated into the overall estimates because these sources are also captured by remote sensing observations that contribute to the satellite-based estimates.Furthermore, these sources contribute to ground-level measurements,and their influence is reflected to some degree in the calibration,as described shortly. A variety of information sources was used to collect updated ground-level PM,measurement data for 2010-13.These included national and European Union(EU)measurement databases as well as new data where available,especially from China and India.Input from an interna- tional group of GBD collaborators was sought;targeted searches for data were conducted;and measurements were compiled from a literature search and from the 2014 WHO database on ambient air pollution in cities.A final database was constructed,including measurement val- ues,year of annual average(data for 2010-13 were targeted,and other years were used only if no other data were available),site coordinates(if available,or city centroid coordinates if not available),site type(if available),International Standard Organization(ISO)3 country code, data source,and whether PM,s was measured directly or estimated from a PM,/PM ratio. The proportion of ground measurements based on direct measurement of PM,s versus esti- mated by PM,PM ratios is presented in appendix table A.1.Although the use of PM measurements is a balance between provision of spatial coverage and the uncertainty that may be introduced because of the use of a ratio to estimate PM,levels,it is important to note that in regions with either low numbers of measurements or a low percentage of direct PM,s mea- surements,ground measurements will likely be more uncertain. The Cost of Air Pollution:Strengthening the Economic Case for Action 13

The Cost of Air Pollution: Strengthening the Economic Case for Action 13 however, incorporated into the overall estimates because these sources are also captured by remote sensing observations that contribute to the satellite-based estimates. Furthermore, these sources contribute to ground-level measurements, and their influence is reflected to some degree in the calibration, as described shortly. A variety of information sources was used to collect updated ground-level PM2.5 measurement data for 2010–13. These included national and European Union (EU) measurement databases as well as new data where available, especially from China and India. Input from an interna￾tional group of GBD collaborators was sought; targeted searches for data were conducted; and measurements were compiled from a literature search and from the 2014 WHO database on ambient air pollution in cities. A final database was constructed, including measurement val￾ues, year of annual average (data for 2010–13 were targeted, and other years were used only if no other data were available), site coordinates (if available, or city centroid coordinates if not available), site type (if available), International Standard Organization (ISO) 3 country code, data source, and whether PM2.5 was measured directly or estimated from a PM2.5 /PM10 ratio. The proportion of ground measurements based on direct measurement of PM2.5 versus esti￾mated by PM2.5 /PM10 ratios is presented in appendix table A.1. Although the use of PM10 measurements is a balance between provision of spatial coverage and the uncertainty that may be introduced because of the use of a ratio to estimate PM2.5 levels, it is important to note that in regions with either low numbers of measurements or a low percentage of direct PM2.5 mea￾surements, ground measurements will likely be more uncertain. Box 2.1 Dust and Dust Storm Health Effects Because there is no evidence that the dust components of PM2.5 should be excluded when esti￾mating health impacts, concentrations of windblown mineral dust and sea salt were included in estimating the health impacts of exposure to ambient PM2.5 for the GBD 2013 study. The cur￾rent positions of the U.S. Environmental Protection Agency (EPA), World Health Organization (WHO), and International Agency for Research on Cancer (IARC) are that an insufficient basis exists for using separate indicators for a specific PM2.5 component or group of components asso￾ciated with any source category of fine particles. Many constituents of particulate matter can be linked with differing health effects, and the evidence is not yet sufficient to allow differentiation of those constituents or sources that are more closely related to specific health outcomes (EPA 2009; IARC 2013; WHO 2014). Concentrations of windblown mineral dust and sea salt were therefore included in estimating the health impacts of exposure to ambient PM2.5 for the GBD 2013 study. For windblown mineral dust specifically, there is substantial evidence of its association with mor￾tality and morbidity during episodes of high concentrations such as Saharan dust storms that affect Europe (Perez et al. 2008; Mallone et al. 2011; Karanasiou et al. 2012), Asian dust storms (Chen et al. 2004; Bell, Levy, and Lin 2008), or regional episodes in the Middle East (Thalib and Al-Taiar 2012; Vodonos et al., 2014, 2015). Most of this evidence points to the coarse fraction of particulate matter, or to PM10, and not to the smaller proportion of dust that is in the PM2.5 frac￾tion. In toxicology, there is no evidence that dust is more benign than other components of PM2.5 (WHO 2014). 1700234_Cost of Pollution.indd 13 8/29/16 1:55 PM

This combination of data from ground-level monitoring with satellite observations and chem- ical transport models provides a globally consistent estimate of PM,concentrations.The final PM,estimates used in the burden of disease estimation were calibrated against observations from ground-level monitoring from more than 75 countries.The calibration equation was estimated from 4,073 ground-level measurements of annual average concentrations,including significant interaction terms for quality and accuracy of location of ground monitors. All three sources of information incorporate strengths and limitations with different sources of uncertainty,and so were combined for the exposure estimates.The mean of the TM5-FASST and satellite-derived estimates was calculated for each grid cell,which inherently captures some of the uncertainty between these two input sources.Furthermore,the error from the calibration with ground measurements was used to propagate the uncertainty between the TM5-FASST and satellite-based estimates and the ground measurements into the burden calculations. This approach has some shortcomings,however,which should be considered when interpret- ing the modeling results.For one thing,estimates for regions of elevated windblown mineral dust have high levels of uncertainty.This uncertainty is partially driven by the TM5-FASST use of standard dust contributions that do not align with a specific year and the temporally variable levels of re-suspended mineral dust in affected regions.Even with ground-level mon- itoring,in dusty areas it is hard to get accurate measurements of the dust contribution to PM2s because much of the dust is in larger size fractions;small errors or between-monitor differ- ences in size fractionation can therefore result in large errors.That said,more surface mea- surements from such locations will nonetheless be needed to reduce uncertainties related to windblown mineral dust in the future(see appendix map A.1). Furthermore,with regard to TM5-FASST,in locations where emissions sources are highly variable and not well characterized,uncertainty is likely to be larger.This suggests,for exam- ple,greater uncertainty in rapidly developing regions with high concentration levels.But,as indicated earlier,these uncertainties are mitigated to some degree by the inclusion of both satellite-based estimates and ground-level observations,which may better capture the dynamic nature of emissions sources. Finally,underestimation of ground measurements has been reported for satellite-based esti- mates(see box 2.2),which may be more pronounced in locations that experience higher con- centrations in wintertime and nighttime,when satellite observations are limited,compared with other seasons in daytime(van Donkelaar et al.2015).Furthermore,because of the spatial resolution of TM5-FAAST,and to a lesser degree the satellite-based estimates,localized features affecting concentrations,including topography and small emissions sources,are unlikely to be well characterized.For example,underestimation of ground-level measurements in southern Poland and Ulaanbaatar may stem from higher wintertime(and in Ulaanbaatar also nighttime) emissions,when satellite retrievals are more limited because of the more frequent winter cloud cover(or unavailability at night).This underestimation has also been described for the satellite- based estimates alone by van Donkelaar et al.(2015).A similar phenomenon may also contrib- ute to underestimation in Chile,where nighttime wood burning during winter contributes to elevated PM,s concentrations.These underestimations of ground measurements in specific locations were also evident in TM5-FASST simulations,suggesting that both chemical trans- port model and satellite-based estimates may fail to accurately estimate ground-level PM,s in 14 The Cost of Air Pollution:Strengthening the Economic Case for Action

14 The Cost of Air Pollution: Strengthening the Economic Case for Action This combination of data from ground-level monitoring with satellite observations and chem￾ical transport models provides a globally consistent estimate of PM2.5 concentrations. The final PM2.5 estimates used in the burden of disease estimation were calibrated against observations from ground-level monitoring from more than 75 countries. The calibration equation was estimated from 4,073 ground-level measurements of annual average concentrations, including significant interaction terms for quality and accuracy of location of ground monitors. All three sources of information incorporate strengths and limitations with different sources of uncertainty, and so were combined for the exposure estimates. The mean of the TM5-FASST and satellite-derived estimates was calculated for each grid cell, which inherently captures some of the uncertainty between these two input sources. Furthermore, the error from the calibration with ground measurements was used to propagate the uncertainty between the TM5-FASST and satellite-based estimates and the ground measurements into the burden calculations. This approach has some shortcomings, however, which should be considered when interpret￾ing the modeling results. For one thing, estimates for regions of elevated windblown mineral dust have high levels of uncertainty. This uncertainty is partially driven by the TM5-FASST use of standard dust contributions that do not align with a specific year and the temporally variable levels of re-suspended mineral dust in affected regions. Even with ground-level mon￾itoring, in dusty areas it is hard to get accurate measurements of the dust contribution to PM2.5 because much of the dust is in larger size fractions; small errors or between-monitor differ￾ences in size fractionation can therefore result in large errors. That said, more surface mea￾surements from such locations will nonetheless be needed to reduce uncertainties related to windblown mineral dust in the future (see appendix map A.1). Furthermore, with regard to TM5-FASST, in locations where emissions sources are highly variable and not well characterized, uncertainty is likely to be larger. This suggests, for exam￾ple, greater uncertainty in rapidly developing regions with high concentration levels. But, as indicated earlier, these uncertainties are mitigated to some degree by the inclusion of both satellite-based estimates and ground-level observations, which may better capture the dynamic nature of emissions sources. Finally, underestimation of ground measurements has been reported for satellite-based esti￾mates (see box 2.2), which may be more pronounced in locations that experience higher con￾centrations in wintertime and nighttime, when satellite observations are limited, compared with other seasons in daytime (van Donkelaar et al. 2015). Furthermore, because of the spatial resolution of TM5-FAAST, and to a lesser degree the satellite-based estimates, localized features affecting concentrations, including topography and small emissions sources, are unlikely to be well characterized. For example, underestimation of ground-level measurements in southern Poland and Ulaanbaatar may stem from higher wintertime (and in Ulaanbaatar also nighttime) emissions, when satellite retrievals are more limited because of the more frequent winter cloud cover (or unavailability at night). This underestimation has also been described for the satellite￾based estimates alone by van Donkelaar et al. (2015). A similar phenomenon may also contrib￾ute to underestimation in Chile, where nighttime wood burning during winter contributes to elevated PM2.5 concentrations. These underestimations of ground measurements in specific locations were also evident in TM5-FASST simulations, suggesting that both chemical trans￾port model and satellite-based estimates may fail to accurately estimate ground-level PM2.5 in 1700234_Cost of Pollution.indd 14 8/29/16 1:55 PM

BOX 2.2 Underestimation of Ground Measurements in Locations with High Concentrations Figure B2.2.1 depicts the relationship between the fused estimates(satellite-based and from TM5-FAAST)and all available ground measurements.Comparisons of GBD super-regions indi- cated underestimation by the calibration function in North Africa and the Middle East;Cen- tral Europe,Eastern Europe,and Central Asia;and Sub-Saharan Africa.Because of the complete absence of ground-level measurements in specific regions and very limited data in others,com- parisons by regions were not feasible. FIGURE B2.2.1 Calibration Regression Simple(Pink)versus Advanced(Green)Model by Super-Region 200 Central Europe,Eastern Europe,and Central Asia High-income Latin America and Caribbean North Africa and Middle East 150 South Asia Southeast Asia,East Asia,and Oceania Sub-Saharan Africa 100 50 50 100 Fused(mean of satellite and TM5 models) Source:Brauer et al.2016. Note:The graph is based on calibration of the mean of satellite-based and TM5 grid cell estimates of annual average PM (micrograms per cubic meter,ug/m )with available ground-level monitoring data color-coded by seven super-regions. go0s8a8saderOb46stBbeteaeeab88e8aRaea802fege8e。 B=0.82,B,=0.73;residual standard error model included additional information on the ground measurements and has characteristics of B,=0.42,B.=0.87: residual standard error =0.41;multiple R-squared:0.64;adjusted R-squared:0.64.Reprinted with permission from Ambient Air Pollution Exposure Estimation for the Global Burden of Disease 2013.Brauer M,Freedman G,Frostad J,van Vos T,Forouzanfar MH,Burnett RT,Cohen A.Environ Sci Technol.2016 Jan 5;50(1):79-88.doi:10.1021/acs.est.5b03709 Copyright 2016 American Chemical Society The Cost of Air Pollution:Strengthening the Economic Case for Action 15

The Cost of Air Pollution: Strengthening the Economic Case for Action 15 Box 2.2 Underestimation of Ground Measurements in Locations with High Concentrations Figure B2.2.1 depicts the relationship between the fused estimates (satellite-based and from TM5-FAAST) and all available ground measurements. Comparisons of GBD super-regions indi￾cated underestimation by the calibration function in North Africa and the Middle East; Cen￾tral Europe, Eastern Europe, and Central Asia; and Sub-Saharan Africa. Because of the complete absence of ground-level measurements in specific regions and very limited data in others, com￾parisons by regions were not feasible. Figure B2.2.1 Calibration Regression Simple (Pink) versus Advanced (Green) Model by Super-Region 200 150 100 50 0 Ground monitors 0 50 100 Fused (mean of satellite and TM5 models) Central Europe, Eastern Europe, and Central Asia High-income Latin America and Caribbean North Africa and Middle East South Asia Southeast Asia, East Asia, and Oceania Sub-Saharan Africa Source: Brauer et al. 2016. Note: The graph is based on calibration of the mean of satellite-based and TM5 grid cell estimates of annual average PM2.5 (micrograms per cubic meter, µg/m3 ), with available ground-level monitoring data color-coded by seven super-regions. Both models are of the form Measured ln(PM2.5) 5 b0 1 b1 * ln(fused), with the “simple” model having characteristics of b0 5 0.82, b1 5 0.73; residual standard error 5 0.43; multiple R-squared: 0.60; adjusted R-squared: 0.60. The “advanced” model included additional information on the ground measurements and has characteristics of b0 5 0.42, b1 5 0.87; residual standard error 5 0.41; multiple R-squared: 0.64; adjusted R-squared: 0.64. Reprinted with permission from Ambient Air Pollution Exposure Estimation for the Global Burden of Disease 2013. Brauer M, Freedman G, Frostad J, van Donkelaar A, Martin RV, Dentener F, van Dingenen R, Estep K, Amini H, Apte JS, Balakrishnan K, Barregard L, Broday D, Feigin V, Ghosh S, Hopke PK, Knibbs LD, Kokubo Y, Liu Y, Ma S, Morawska L, Sangrador JL, Shaddick G, Anderson HR, Vos T, Forouzanfar MH, Burnett RT, Cohen A. Environ Sci Technol. 2016 Jan 5; 50(1):79-88. doi: 10.1021/acs.est.5b03709. Copyright 2016 American Chemical Society. 1700234_Cost of Pollution.indd 15 8/29/16 1:55 PM

relatively small areas having very high levels.Improvements in the spatial precision of emis- sions estimates and satellite retrievals and use of regional models will reduce these uncertainties in the future. To estimate chronic long-term exposure to ozone,the TM5-FASST chemical transport model and the same set of emissions used for PM,were applied to calculate a running three-month average of(daily one-hour maximum values)ozone concentrations for each grid cell over a full year from which the maximum of these values was selected.This metric was chosen to align with epidemiologic studies of chronic exposure,which typically employ a seasonal(sum- mer)average(Jerrett et al.2009),and to account for global variation in the timing of the ozone (summer)season.These estimates were simulated with TM5-FASST at 0.1X 0.1 for 1990, 2000,and 2010 using the same emissions and meteorological inputs as for the PM,s simula- tions.Estimates for 1995,2005,2011,and 2013 were generated with splines and extrapolations as described earlier for PM,s Population Exposure to Ambient Air Pollution Estimates of population exposure to PM,were developed in five-year intervals from 1990 to 2010 and for 2013 with0.1X0.1resolution,using estimates from satellites and chemical transport mod- els,calibrated with surface measurements.Similarly,for ozone,estimates of population exposure for the same five-year intervals and for 2013 were estimated from the TM5-FASST chemical transport model.Gridded exposure concentrations were aggregated to national-level,population-weighted means with the corresponding grid cell population value.National-level,population-weighted means and 95 percent uncertainty interval(UD)concentrations were estimated by sampling 1,000 draws of each grid cell value of the mean of the chemical transport model and satellite-based con- centration estimates,in combination with the calibration parameters and the uncertainty of the cal- ibration function.For ozone,population-weighted concentrations and 95 percent UI for each country were estimated as for PM,,but assuming a normal distribution with a UI of+6 percent of the estimated concentration. Method for Estimating Exposure to Household Air Pollution from Cooking with Solid Fuels Exposure to household air pollution is defined as the 24-hour average of exposure to PM emitted from cooking with solid fuels such as coal,wood,charcoal,dung,and agricultural res- idues.Estimates of exposure to household air pollution are not provided for high-income coun- tries.Quantifying exposure to indoor air pollution by the average PM,exposure associated with household use of solid cooking fuel makes it possible to utilize the integrated exposure- response(IER)curves needed to calculate the burden of indoor air pollution. Although solid fuel use is an indirect measure of true exposure,this information is easier to collect and more frequently reported in epidemiological studies than direct measures of household air pollution.Therefore,estimation of exposure to household air pollution starts with data on household use of solid fuels.Such data were extracted from nationally represen- tative household surveys.Fuels such as coal,wood,charcoal,dung,and agricultural residues were classified as solid fuels in this analysis.Data were extracted from 148 countries,and data 16 The Cost of Air Pollution:Strengthening the Economic Case for Action

16 The Cost of Air Pollution: Strengthening the Economic Case for Action relatively small areas having very high levels. Improvements in the spatial precision of emis￾sions estimates and satellite retrievals and use of regional models will reduce these uncertainties in the future. To estimate chronic long-term exposure to ozone, the TM5-FASST chemical transport model and the same set of emissions used for PM2.5 were applied to calculate a running three-month average of (daily one-hour maximum values) ozone concentrations for each grid cell over a full year from which the maximum of these values was selected. This metric was chosen to align with epidemiologic studies of chronic exposure, which typically employ a seasonal (sum￾mer) average (Jerrett et al. 2009), and to account for global variation in the timing of the ozone (summer) season. These estimates were simulated with TM5-FASST at 0.1° 3 0.1° for 1990, 2000, and 2010 using the same emissions and meteorological inputs as for the PM2.5 simula￾tions. Estimates for 1995, 2005, 2011, and 2013 were generated with splines and extrapolations as described earlier for PM2.5 . Population Exposure to Ambient Air Pollution Estimates of population exposure to PM2.5 were developed in five-year intervals from 1990 to 2010 and for 2013 with 0.1° 3 0.1° resolution, using estimates from satellites and chemical transport mod￾els, calibrated with surface measurements. Similarly, for ozone, estimates of population exposure for the same five-year intervals and for 2013 were estimated from the TM5-FASST chemical transport model. Gridded exposure concentrations were aggregated to national-level, population-weighted means with the corresponding grid cell population value. 1 National-level, population-weighted means and 95 percent uncertainty interval (UI) concentrations were estimated by sampling 1,000 draws of each grid cell value of the mean of the chemical transport model and satellite-based con￾centration estimates, in combination with the calibration parameters and the uncertainty of the cal￾ibration function. For ozone, population-weighted concentrations and 95  percent UI for each country were estimated as for PM2.5 , but assuming a normal distribution with a UI of ± 6 percent of the estimated concentration. Method for Estimating Exposure to Household Air Pollution from Cooking with Solid Fuels Exposure to household air pollution is defined as the 24-hour average of exposure to PM2.5 emitted from cooking with solid fuels such as coal, wood, charcoal, dung, and agricultural res￾idues. Estimates of exposure to household air pollution are not provided for high-income coun￾tries. Quantifying exposure to indoor air pollution by the average PM2.5 exposure associated with household use of solid cooking fuel makes it possible to utilize the integrated exposure￾response (IER) curves needed to calculate the burden of indoor air pollution. Although solid fuel use is an indirect measure of true exposure, this information is easier to collect and more frequently reported in epidemiological studies than direct measures of household air pollution. Therefore, estimation of exposure to household air pollution starts with data on household use of solid fuels. Such data were extracted from nationally represen￾tative household surveys. Fuels such as coal, wood, charcoal, dung, and agricultural residues were classified as solid fuels in this analysis. Data were extracted from 148 countries, and data 1700234_Cost of Pollution.indd 16 8/29/16 1:55 PM