2 aaron van donkelaar , randall v. martin , michael brauer...

1

Global fine particulate matter concentrations from satellite for long-term exposure assessment 1

Aaron van Donkelaar1, Randall V. Martin

1,2, Michael Brauer

3, and Brian L. Boys

1 2

1Department of Physics and Atmospheric Science, Dalhousie University, 6300 Coburg Rd., Halifax, Nova 3

Scotia, Canada, B3H 4R2 4 2Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts, 5

USA. 6 3School of Environmental Health, University of British Columbia, British Columbia, Canada. 7

8

Background: More than a decade of satellite observations offers global information about the trend and 9

magnitude of human exposure to fine particulate matter (PM2.5). 10

Objective: In this study, we developed improved global exposure estimates of ambient PM2.5 mass and 11

trend using PM2.5 concentrations inferred from multiple satellites. 12

Methods: We combined three satellite-derived PM2.5 sources to produce global PM2.5 estimates at about 13

10 km × 10 km from 1998-2012 inclusive. For each source, we related total column retrievals of aerosol 14

optical depth to near-ground PM2.5 using the GEOS-Chem chemical transport model to represent local 15

aerosol optical properties and vertical profiles. We collected 210 global ground-based PM2.5 16

observations from the literature to evaluate our satellite product beyond North America and Europe. 17

Results: Over these 15 years, global population-weighted ambient PM2.5 concentrations increased 18

0.55±0.12 μg/m3/yr (2.1±0.5 %/yr). Increasing PM2.5 in some developing regions drove this global 19

change, despite decreasing PM2.5 in some developed regions. The proportion of population in East Asia 20

living above the World Health Organization (WHO) Interim Target-1 of 35 μg/m3 increased from 51% in 21

1998-2000 to 70% in 2010-2012. In contrast, the North American proportion below the WHO Air Quality 22

Guideline of 10 μg/m3 fell from 62% in 1998-2000 to 19% in 2010-2012. We found significant agreement 23

between satellite-derived and ground-based values outside North America and Europe (r=0.81; n=210), 24

but satellite-derived values are biased low (slope=0.68), implying that true concentrations could be even 25

greater. 26

Conclusions: Satellite observations provide insight into global long term changes in ambient PM2.5 27

concentrations. 28

29

30

31

of 28

2

Abbreviations 32

AOD – aerosol optical depth 33

AQG – Air Quality Guideline 34

CEISIN – Center for International Earth Science Information Network 35

GBD – Global Burden of Disease 36

IT – Interim Target 37

MISR – Multiangle Imaging Spectroradiometer 38

MODIS – Moderate Resolution Imaging Spectroradiometer 39

PM2.5 – fine particular matter with diameter less than 2.5 μm 40

SeaWiFS – Sea-viewing Wide Field-of-view Sensor 41

WHO – World Health Organisation 42

of 28

3

Long-term exposure to fine particulate matter (PM2.5) is associated with morbidity and premature 43

mortality (Dockery et al. 1993; Pope et al. 2009). The Global Burden of Disease (GBD) assessment 44

attributed 3.2 million premature deaths per year to ambient PM2.5 exposure, such that PM2.5 is one of 45

the leading risk factors for premature mortality (Lim et al. 2012). Assessments and indicators of the 46

health effects of long-term exposure to PM2.5, such as the GBD assessment, the WHO assessment 47

(http://www.who.int/gho/phe/outdoor_air_pollution/burden/en/) and the Environmental Performance 48

Index (http://epi.yale.edu), rely on an accurate representation of both magnitude and spatial 49

distribution. Long-term trends in concentration can inform whether appropriate steps are being taken 50

to mitigate health and environmental outcomes, and can motivate additional action. Global monitoring 51

can occur from a single satellite, minimizing artifacts that may result from regional differences in 52

ground-level network design and operation. Satellites also offer one of the few observationally-based 53

sources for long-term PM2.5 concentrations that can represent long-term exposure and detect significant 54

changes in many parts the world. 55

Satellite retrievals of Aerosol Optical Depth (AOD), which provide a measure of the amount of light 56

extinction through the atmospheric column due to the presence of aerosol, have a global data record 57

extending more than a decade. Differing design characteristics between satellite instruments and their 58

retrievals can benefit particular applications. For example, Collection 5 retrievals from the MODIS 59

instrument (Levy et al. 2007) provide relatively frequent (daily) global observation and accurate AOD 60

over dark surfaces, but have uncertain changes in instrument sensitivity with time which could introduce 61

artificial trends. Retrievals from the MISR instrument (Diner et al. 2005; Martonchik et al. 2009) require 62

around 6 days for global coverage, but are accurate for both AOD and trend studies (Zhang and Reid 63

2010). SeaWiFS (Hsu et al. 2013) provides sufficient radiometric stability for trends (Eplee et al. 2011), 64

but is less accurate over land for absolute AOD compared to MODIS or MISR due to the lack of a mid-65

infrared channel (Petrenko and Ichoku 2013). 66

The relationship between AOD and PM2.5 is complicated by the effects of aerosol vertical distribution, 67

humidity, and aerosol composition, which are impacted by changes in meteorology and emissions. One 68

technique of relating AOD to near-surface PM2.5 uses the ratio of PM2.5 to AOD simulated by a chemical 69

transport model. This parameter allows a ground-level PM2.5 estimate to be calculated from satellite 70

AOD retrievals. This approach was first demonstrated using the MISR instrument with the GEOS-Chem 71

chemical transport model (www.geos-chem.org) over the United States for 2001 (Liu et al. 2004), and 72

subsequently extended globally for each of the MODIS and MISR instruments for 2001-2002 at a spatial 73

resolution of about 100 km × 100 km (van Donkelaar et al. 2006). 74

The first long-term mean, global, satellite-derived PM2.5 estimates used this technique to combine 75

filtered values from both MODIS and MISR over 2001-2006 at a spatial resolution of about 10 km × 10 76

km. This dataset demonstrated promising agreement with coincident ground-based observations over 77

North America (r=0.77; slope = 1.07) and globally (r=0.83; slope = 0.86) (van Donkelaar et al. 2010). We 78

hereafter refer to this dataset as Unconstrained (UC), owing to the unrestricted freedom it gave satellite 79

AOD retrievals to represent the total aerosol column. 80

of 28

4

Improved correlation with ground-based observations for the year 2005 was achieved using optimal 81

estimation (OE) (van Donkelaar et al. 2013). OE constrained AOD retrievals from MODIS top-of-82

atmosphere reflectances based on the relative uncertainties of observational and simulated estimates 83

(van Donkelaar et al. 2013). The PM2.5 estimates produced with this dataset used vertical profile 84

information from the CALIOP satellite instrument to inform the relation of column AOD to ground-level 85

concentrations. 86

Boys et al. (submitted) created a time-series of PM2.5 anomalies by combining AOD from both SeaWiFS 87

and MISR with spatiotemporal information on the PM2.5 to AOD relationship from a GEOS-Chem 88

simulation over 1998-2012 inclusive. In this paper, we extended the OE-based PM2.5 estimates to 2004-89

2010 and combined them with the UC PM2.5 values of van Donkelaar et al. (2010) to produce a global, 90

decadal PM2.5 dataset at approximately 10 km × 10 km, with improved agreement compared to either 91

dataset alone. We then applied the temporal variation based upon SeaWiFS and MISR (Boys et al. 92

submitted) to estimate annual global PM2.5 estimates and trends over 1998-2012 at 10 km × 10 km 93

resolution. 94

95

Materials and Methods 96

97

Production of satellite-derived estimates 98

We first produced a decadal mean PM2.5 estimate over 2001-2010. Following Boys et al. (submitted), we 99

combined retrievals from SeaWiFS (supplemental) and MISR (supplemental) with time-varying GEOS-100

Chem (supplemental) simulated AOD to PM2.5 relationships to infer annual variation in PM2.5 over 1998 101

to 2012 at a spatial resolution of 0.1° x 0.1° (henceforth referred to as SeaWiFS&MISR PM2.5). We then 102

extended both OE and UC to cover the temporal range 2001-2010 by applying to each dataset the ratio 103

of a coincident SeaWiFS&MISR PM2.5 to its decadal mean. We evaluated each extended dataset using 104

ground-based PM2.5 observations over North America. The global MODIS land-cover type product 105

(MOD12; Freidl et al. 2010) was used to determine the relative weighting of each dataset over each land 106

cover type that maximized agreement with ground-level PM2.5 observations following van Donkelaar et 107

al. (2013) to produce an initial global combined decadal mean PM2.5 estimate. 108

We subsequently produced a consistent time series of PM2.5 over 1998-2012. We applied to the initial 109

decadal mean dataset the relative temporal variation of SeaWiFS&MISR PM2.5 to produce monthly 110

satellite-derived PM2.5 estimates over 1998-2012 inclusive. We calculated absolute annual trends for 111

both datasets using a regression of five month box-car filtered (i.e. median of +/- five months from the 112

center date), deseasonalized monthly mean values following Zhang and Reid (2010). We superimposed 113

these trends to create global annual PM2.5 estimates that were consistent in trend with SeaWiFS&MISR 114

and in magnitude with the initial decadal mean. We used a three-year running median to reduce noise 115

in the annual satellite-derived values. All PM2.5 concentrations are given at 35% relative humidity, 116

of 28

5

except for comparisons involving ground-level measurements outside North America, where the 50% 117

standard is adopted for consistency with the ground-level measurements. 118

Following Evans et al. (2013), we estimated dust-free and sea-salt-free PM2.5 concentrations by 119

removing the simulated relative contribution of these species from total satellite-derived PM2.5. We 120

produced satellite-derived PM2.5 surface area estimates for interpretation of the dust and seasalt-free 121

PM2.5 estimates following a similar approach as PM2.5 mass concentrations, except that the GEOS-Chem 122

model was used to relate AOD to surface area, rather than to mass (supplemental). 123

124

Collection of ground-based observations for evaluation 125

We also collected ground-based PM2.5 observations over Canada and the United States at locations 126

operational for at least 8 years between 2001-2010 (supplemental). We required European sites 127

(supplemental) to be in operation at least 3 years throughout the decade; less than North American 128

locations due to the more recent expansion of this regional network. 129

We collected global ground-based PM2.5 measurements from published values based upon a literature 130

review using the search terms “aerosol” and “PM2.5” in the Thomson Reuters Web of Knowledge, 131

yielding ca. 3500 results. We selected 541 papers for detailed evaluation from this list and in-132

publication citations, and found 342 contained relevant PM2.5 observations. We extracted mean PM2.5, 133

seasonal variation, city, country, site description and geo-coordinates as available. We approximated 134

geo-coordinates using GoogleEarth and in-reference maps at 70 locations. Geocoordinates were not 135

clear for 110 sites; we assumed measurements occurred within 0.1° of city center. When necessary, we 136

approximated seasonal variation from figures. We considered an observational period every third 137

month as sufficient for annual representation. Where possible, we inferred annual mean concentrations 138

for sites without observations every third month using the relative seasonal variation from nearby 139

published values at distances of up to 1°. We excluded industrial, traffic and military studies. We 140

combined observational PM2.5 values at locations within 0.1°, weighted by their temporal coverage, and 141

used only locations that had at least 3 months of direct observation, for a total of 210 ground-based 142

comparison sites outside of Canada, the United States and Europe. 143

We evaluated the combined fifteen year PM2.5 timeseries from MODIS, MISR, and SeaWiFS (henceforth 144

‘combined’) with annual average ground-based PM2.5 observations. We conduct the comparison versus 145

PM2.5 measurements from ground-based monitors on all days (not only days coincident with satellite 146

observations). We included in the evaluation the 110 global comparison sites from the literature 147

without clearly specified geo-coordinates; we conducted evaluations both assuming locations at city 148

center and up to 0.1° away. 149

Gridded population estimates at 2.5’ resolution from the Center for International Earth Science 150

Information Network (CEISIN 2005) at five year intervals starting from 1995, are regridded onto 0.1° x 151

0.1°. Years beyond 2005 are based upon projections. We estimated year-specific population densities 152

using linear interpolation. 153

of 28

6

154

Results 155

Figure 1 (top panel) shows decadal mean satellite-derived PM2.5 concentrations over North America. 156

Enhancements are visible in the eastern United States and in the San Joaquin Valley of California. Figure 157

1 also shows long-term mean ground-level PM2.5 measured during this period over Canada and the 158

United States and comparison with the satellite-derived estimates. Significant overall agreement is 159

found (slope=1.00, r=0.79; 1σ error=1 μg/m3+14%). Separate comparisons of OE and UC satellite-160

derived estimates with the same ground-level monitors gave similar levels of agreement compared to 161

one another (r=0.72; 1σ error=1 μg/m3+18-21%; not shown). Contributions of OE and UC to the final 162

PM2.5 estimates were approximately equal over most land cover types. 163

Figure 2 (top panel) shows decadal mean satellite-derived PM2.5 concentrations over Europe. PM2.5 is 164

generally higher in Eastern Europe than Western Europe. The Po Valley in Italy is characterized by the 165

highest regional concentrations, with average PM2.5 for some local locations exceeding 35 μg/m3 from 166

2001-2010. Figure 2 also shows available long-term mean ground-level observations which are mostly 167

for the latter part of this period. We find slightly weaker agreement with satellite-derived estimates for 168

Europe than for North America, with slope=0.78, r=0.73 and 1σ error=1 μg/m3+21%. The weaker 169

agreement likely results from the shorter temporal sampling of three years over this region, as 170

illustrated in Table S1 and Table S2 of the supplemental material. A cluster of ground-level monitors in 171

southern Poland contribute to the disagreement with annual mean concentrations above 35 μg/m3. 172

PM2.5 concentrations in Southern Poland near Katowice have pronounced wintertime enhancements 173

(Rogula-Kozlowska et al. 2013) when satellite observations are sparse. 174

Figure 3 (top panel) shows global decadal mean satellite-derived PM2.5. PM2.5 concentrations in large 175

populated regions of northern India and eastern China respectively exceed 60 μg/m3 and 80 μg/m

3. The 176

bottom panels contain the 210 locations of global mean ground-level PM2.5 concentrations outside 177

Canada, the United States and Europe. Significant agreement (r=0.81) exists, but satellite-derived values 178

tend to be lower than ground-level measurements, with an overall slope of 0.68. Some of this 179

underestimate may arise from locations such as Ulaanbataar, Mongolia that experience pronounced 180

enhancement in wintertime and nighttime PM2.5 (World Bank 2011) when satellite observations are 181

limited. Comparison in which the 110 sites with unspecified geo-coordinates are at city center yielded 182

similar, but slightly weaker agreement (r=0.78; slope=0.65). 183

Table 1 provides a summary of population-weighted satellite-derived exposure according to the regions 184

used by the Global Burden of Disease (Lim et al. 2012). Globally, population-weighted PM2.5 exposure 185

between 2001-2010 is 26.4 μg/m3 with large spatial variability (standard deviation of 21.4 μg/m

3). South 186

and East Asia have the highest population-weighted mean exposures at 34.6 and 50.3 μg/m3. 187

Figure 3 (middle) presents global estimates of satellite-derived PM2.5 with dust and sea salt 188

concentrations removed for 2001-2010. Pronounced enhancements remain over China and the Indo-189

Gangetic Plain. North African and Middle Eastern PM2.5 have large relative decreases. Some studies 190

have suggested that the toxicity of particulate matter is more directly related to particle surface area 191

of 28

7

than mass (e.g. Maynard and Maynard 2002; Oberdörster et al. 2005). Interestingly, satellite-derived 192

PM2.5 surface area (supplemental) demonstrates similar spatial patterns as dust and sea salt-free PM2.5. 193

Table 1 summarizes dust and sea salt-free PM2.5 according to GBD region. These components of PM2.5 194

are responsible for about half the population-weighted decadal mean PM2.5 concentrations in Central 195

Asia, North Africa/Middle East and East Sub-Saharan Africa and for three quarters of the concentration 196

in West Sub-Saharan Africa. Dust and sea salt account for 10% of these concentrations in East Asia and 197

20% in South Asia. Dust and sea salt have little influence over European and North American 198

concentrations. 199

Table 1 contains population-weighted PM2.5 trends over 1998-2012 for each GBD region. A 200

corresponding global trend map following Boys et al. (submitted) is in supplemental Figure S1. 201

Statistically significant increasing population-weighted trends include 1.63 ±0.54 μg/m3/yr (3.2±1.1 %/yr) 202

over East Asia and 1.02±0.25 μg/m3/yr (2.9±0.7 %/yr) over South Asia. These trends primarily follow 203

changes in anthropogenic emissions (Klimont et al. 2013) and increasing sulfate-nitrate-ammonium 204

concentrations as described in Boys et al. (submitted). Trends of 0.38±0.21 μg/m3/yr (1.5±0.8 %/yr) in 205

the Middle East are driven by mineral dust (Chin et al. 2014). Statistically significant downward 206

population-weighted trends include -0.33±0.08 μg/m3/yr (-3.3%±0.8 %/yr) over North America and -207

0.25±0.12 μg/m3/yr (-1.9±0.9 %/yr) over Western Europe. The global population-weighted trend is 208

0.55±0.12 μg/m3/yr (2.1± 0.5 %/yr). 209

Figure 4 shows time-series snapshots of PM2.5 over the four large-scale global regions that demonstrate 210

statistically significant trends. Changes are coherent over broad regions. Figure 5 shows local trends for 211

a major city within each region. Evaluation of the satellite-derived PM2.5 trends with available ground-212

level observations near Detroit yields consistent decreasing trends of 0.51-0.54 μg/m3/yr from 2001-213

2010, with a similar C.I. of ±(0.28-0.37) μg/m3/yr. The full 15 year satellite-derived PM2.5 time-series 214

decreases by 0.43±0.12 μg/m3/yr over 1998-2012. Beijing and New Delhi have significant increasing 215

trends over this time period of ca. 2 μg/m3/yr. Kuwait City has an even larger increasing trend of 3.1±0.8 216

μg/m3/yr. 217

Figure 6 gives the cumulative distribution of global annual mean PM2.5 as a function of time, and for the 218

three GBD regions of greatest positive and negative trend magnitude. Table 1 provides the percent of 219

population living in areas where concentrations are below the WHO interim targets (IT3, IT2 and IT1) 220

and guideline (AQG) for 1998-2000 and 2010-2012 for all regions. A small population-weighted global 221

improvement (1%) of those living within AQG is found over the past 15 years, predominantly driven by 222

improvements to air quality in North America which reduced its population living above this target from 223

62% to 19%. Globally, exceedance of IT1 (35 μg/m3) rose by 8% over the same time period, reaching 224

30% by 2010-2012 as driven by increasing PM2.5 concentrations in the heavily populated regions of 225

South and East Asia. The negative bias of satellite values versus ground-based monitors suggests the 226

percentage in living above WHO targets could be even higher. 227

Table 1 also shows the effect on WHO target achievement of population changes over 1998-2012. The 228

effect of population changes on WHO target achievement is less than 25% across all targets for all 229

of 28

8

regions, and less than ca. 10% in most cases. The number of people living above AQG has increased due 230

to population changes in some regions, accounting for about a quarter of the change seen in Central 231

Asia and South Sub-Saharan Africa from 1998 to 2012. About half the change in Eastern Europe is due 232

to population, although the overall change is small (2%). Population changes contributed to small 233

reductions in population-weighted mean PM2.5 concentrations for regions such as Southeast Asia and 234

North America. 235

236

Discussion 237

A broad community requires globally consistent estimates of long-term PM2.5 exposure and changes 238

over time. For example, this information is used for global burden of disease assessment (Brauer et al. 239

2012; Lim et al. 2012; World Health Organization 2014), for environmental performance indicators 240

(Environmental Performance Index 2014), and for epidemiologic studies of air pollution health effects at 241

global (Anderson et al. 2012; Fleischer et al. 2014) and regional (Chudnovsky et al. 2012; Crouse et al. 242

2012; Vinneau et al. 2013) scales. Satellite retrievals offer the most globally complete observationally-243

based data source of this information, but improvements to these estimates are needed to reduce 244

uncertainties. 245

In this work, we combined the attributes of several recent satellite-derived PM2.5 datasets to improve 246

the accuracy in estimates of long-term exposure and changes in annual concentrations from 1998 to 247

2012. We inferred decadal mean PM2.5 from Unconstrained (van Donkelaar et al. 2010) and Optimal 248

Estimation (van Donkelaar et al. 2013) based approaches utilizing the MODIS and MISR instruments. We 249

then applied the relative temporal variation from SeaWiFS and MISR observations (Boys et al. 250

submitted) to represent the annual variation over 15 years. The resultant combined dataset had 251

significant agreement with 8+ year means of ground-based observations over North America 252

(slope=1.00; r=0.79; 1σ error=1 μg/m3+14%) and 3+ year means over Europe (slope=0.78; r=0.73; 1σ 253

error =1 μg/m3+21%) in non-coincident comparisons that represent both retrieval and sampling induced 254

uncertainties. This performance was better than for any of the individual datasets. We found a 255

noteworthy difference in agreement of satellite-derived PM2.5 with ground-based monitors if they are 256

sampled coincidently (only on the days when the satellite observes) as performed in many previous 257

works or non-coincidently as used here. For example, the coincident correlation of r=0.77 over North 258

America for 2001-2006 previously given in van Donkelaar et al. (2010), drops to r=0.70 when taken non-259

coincidently. The non-coincident comparison used here offers a more rigorous test of satellite sampling 260

bias. 261

A major challenge in evaluating global satellite-derived PM2.5 is the paucity of ground-based 262

measurements. We collected a global dataset of 210 ground-based observations from the literature and 263

used them to evaluate global satellite-derived PM2.5 estimates, including many locations in India and 264

China. Significant agreement was found (r=0.81), although these new monitors revealed that satellite-265

derived PM2.5 is typically lower than ground-based observations (slope=0.68). This underestimate may 266

result from factors such as AOD bias in the MISR retrieval over South and East Asia (Kahn et al. 2009), 267

of 28

9

missing satellite observations during wintertime and/or nighttime enhancements (e.g. Katowice, Poland 268

and Ulaanbaatar, Mongolia), or coarse resolution of either the satellite-derived product or the 269

simulation used to related AOD to PM2.5 which may obscure localized features. The potential 270

underestimate in satellite-derived PM2.5 outside North America and Europe furthermore suggests true 271

PM2.5 concentrations may be even greater than we determined. 272

We found that decade-long populated-weighted ambient PM2.5 concentrations in East Asia are nearly 273

double the global mean of 26.4 μg/m3, and increase at an annual population-weighted rate of 1.63±0.54 274

μg/m3/yr (3.2±1.1 %/yr) between 1998 and 2012. Population-weighted concentrations over western 275

Europe and North America over the same period decreases by 0.25-0.33 ±0.08-0.12 μg/m3/yr (1.9-276

3.3±0.8-0.9 %/yr) in contrast with increases over South Asia (1.02±0.25 μg/m3/yr; 2.9±0.7 %/yr) and the 277

Middle East (0.38±0.21 μg/m3/yr; 1.5±0.8 %/yr). Satellite-derived estimates suggest that 30% of the 278

global population lives in regions above the WHO IT1 standard (35 μg/m3) for PM2.5 in 2010-2012, up 279

from 22% in 1998-2000. We found that most of the changes in exposure were driven by changes in 280

PM2.5 rather than changes in population location. 281

Both the satellite-derived PM2.5 estimates created in and ground-level observations collected for this 282

study are freely available as a public good on our website (http://fizz.phys.dal.ca/~atmos/martin) or by 283

contacting the authors. 284

Further developments to satellite retrievals and simulated aerosol profiles will continue to allow 285

improved representation of global exposures to PM2.5. In particular, higher resolution satellite retrievals 286

may better capture intra-urban variation (Chudnovsky et al. 2012). Recent improvements to MODIS 287

instrument calibration (Levy et al. 2013) may provide an additional data source for trends. 288

289

Acknowledgements 290

This work was supported by Health Canada, the Natural Sciences and Engineering Research Council of 291

Canada, and the National Institutes of Health. Some of the computing facilities used here were provided 292

by the Atlantic Computational Excellence Network. 293

294

295

296

297

of 28

10

References 298

299

Anderson HR, Butland BK, van Donkelaar A, Brauer M, Strachan DP, Clayton T, et al. 2012. Satellite-300

based Estimates of Ambient Air Pollution and Global Variations in Childhood Asthma Prevalence. 301

Environmental Health Perspectives 120(9): 1333-1339. 302

Boys B, Martin RV, Van Donkelaar A, MacDonell R, Hsu NC, Cooper MJ, et al. submitted. Fifteen year 303

global time series of satellite-derived fine particulate matter. Environmental Science & Technology 304

Available from http://fizz.phys.dal.ca/~atmos/publications/BOYS_2014_EST_PM_trend_submit.pdf. 305

Brauer M, Amman M, Burnett RT, Cohen A, Dentener FJ, Ezzatti M, et al. 2012. Exposure assessment for 306

the estimation of the global burden of disease attributable to outdoor air pollution Environmental 307

Science and Technology 2012(46): 652-660. 308

CEISIN. 2005. Gridded Population of the World, Version 3 (GPWv3). NASA Socioeconomic Data and 309

Applications Center (SEDAC). 310

Chin M, Diehl T, Tan Q, Prospero JM, Kahn RA, Remer LA, et al. 2014. Multi-decadal variations of 311

atmospheric aerosol from 1980-2009: sources and regional trends. Atmospheric Chemistry and Physics 312

Discussions 13: 19751-19835. 313

Chudnovsky AA, Kostinski A, Lyapustin A, Koutrakis P. 2012. Spatial scales of pollution from variable 314

resolution satellite imaging. Environmental Pollution 172: 131-138. 315

Crouse DL, Peters PA, van Donkelaar A, Goldberg MS, Villeneuve PJ, Brion O, et al. 2012. Risk of 316

Nonaccidental and Cardiovascular Mortality in Relation to Long-term Exposure to Low Concentrations of 317

Fine Particulate Matter: A Canadian National-level Cohort Study. Environmental Health Perspectives 318

120(5): 708-714. 319

Diner DD, Braswell BH, Davies R, Gobron N, Hu J, Jin Y, et al. 2005. The value of multiangle 320

measurements for retrieving structurally and radiatively consistent properties of clouds, aerosols, and 321

surfaces. Remote Sensing of Environment 97: 495-518. 322

Dockery DW, Pope CA, Xu XP, Spengler JD, Ware JH, Fay ME, et al. 1993. An assocation between air-323

pollution and mortality in 6 United-States cities. New England Journal of Medicine 329(24): 1753-1759. 324

Environmental Performance Index T. year. 2014 Environmental Performance Index Full Report and 325

Analysis. Available: http://epi.yale.edu/. 326

Eplee RE, Meister G, Patt FS, Franz BA, McClain CR. 2011. Uncertainty Assessment of the SeaWiFS On-327

Orbit Calibration. In: Earth Observing Systems Xvi, Vol. 8153, (Butler JJ, Xiong X, Gu X, eds). 328

Evans JA, van Donkelaar A, Martin RV, Burnett RT, Rainham DG, Birkett NJ, et al. 2013. Estimates of 329

global mortality attributable to particulate air pollution using satellite imagery. Environmental Research 330

120: 33-42. 331

of 28

11

Fleischer NL, Merialdi M, van Donkelaar A, Vadillo-Ortega F, Martin RV, Betran AP, et al. 2014. Outdoor 332

air pollution, preterm birth, and low birth weight: Analysis of te World Health Organization Global 333

Survey on maternal and perinatal health. Environmental Health Perspectives 122(4): 425-430. 334

Freidl MA, Sulla-Menashe D, Tan B, Schneider A, Ramankutty N, Sibley A, et al. 2010. MODIS Collection 5 335

global land cover: Algorithm refinements and characterization of new datasets. Remote Sensing of 336

Environment 114: 168-182. 337

Hsu NC, Jeong MJ, Bettenhausen C, Sayer AM, Hansell R, Seftor CS, et al. 2013. Enhanced Deep Blue 338

aerosol retrieval algorithm: The second generation. Journal of geophysical research 118: 1-20. 339

Kahn RA, Nelson D, Garay M, Levy R, Bull M, Diner DD, et al. 2009. MISR aerosol product attributes, and 340

statistical comparisons with MODIS. Transactions on Geoscience and Remote Sensing. 341

Klimont Z, Smith SJ, Cofala J. 2013. The last decade of global anthropogenic sulfur dioxide: 2000-2011 342

emissions. Environmental Research Letters 8. 343

Levy RC, Mattoo S, Munchak LA, Remer LA, Sayer AM, Hsu NC. 2013. The Collection 6 MODIS aerosol 344

products over land and ocean. Atmospheric Measurement Techniques 6: 2989-3034. 345

Levy RC, Remer LA, Mattoo S, Vermote EF, Kaufman YJ. 2007. Second-generation operational algorithm: 346

Retrieval of aerosol properties over land from inversion of Moderate Resolution Imaging 347

Spectroradiometer spectral reflectance. Journal of Geophysical Research-Atmospheres 112(D13). 348

Lim SS, Vos T, Flaxman ADF, Danaei G, Shibuya K, al. e. 2012. A comparative risk assessment of burden of 349

disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a 350

systematic analysis for the Global Burden of Disease Study 2010. The Lancet 380: 2224-2260. 351

Liu Y, Park RJ, Jacob DJ, Li QB, Kilaru V, Sarnat JA. 2004. Mapping annual mean ground-level PM2.5 352

concentrations using Multiangle Imaging Spectroradiometer aerosol optical thickness over the 353

contiguous United States. Journal of Geophysical Research-Atmospheres 109(D22). 354

Martonchik JV, Kahn RA, Diner DJ. 2009. Retrieval of Aerosol Properties over Land Using MISR 355

Observations. In: Satellite Aerosol Remote Sensing Over Land, (Kokhanovsky AA, Leeuw Gd, eds). 356

Berlin:Springer. 357

Maynard AD, Maynard RL. 2002. A derived association between ambient aerosol surface area and 358

excess mortality using historic time series data. Atmospheric Environment 36: 5561-5567. 359

Oberdörster G, Oberdörster E, Oberdörster J. 2005. Nanotoxicology: An Emerging Discipline Evolving 360

from Studies of Ultrafine Particles. Environmental Health Perspectives 113(7). 361

Petrenko M, Ichoku C. 2013. Coherent uncertainty analysis of aerosol measurements from multiple 362

satellite sensors. Atmospheric Chemistry and Physics 13: 6777-6805. 363

Pope CA, Ezzati M, Dockery DW. 2009. Fine-Particulate Air Pollution and Life Expectancy in the United 364

States. New England Journal of Medicine 360: 376-386. 365

of 28

12

Rogula-Kozlowska W, Kleijnowski K, Rolua-Kopiec P, Osrodka L, Krajny E, Blaszczak B, et al. 2013. Spatial 366

and seasonal variability of the mass concentration and chemical composition of PM2.5 in Poland. Air 367

Quality Atmosphere and Health. 368

van Donkelaar A, Martin RV, Brauer M, Kahn R, Levy R, Verduzco C, et al. 2010. Global Estimates of 369

Ambient Fine Particulate Matter Concentrations from Satellite-Based Aerosol Optical Depth: 370

Development and Application. Environmental Health Perspectives 118(6): 847-855. 371

van Donkelaar A, Martin RV, Park RJ. 2006. Estimating ground-level PM2.5 using aerosol optical depth 372

determined from satellite remote sensing. Journal of Geophysical Research-Atmospheres 111(D21). 373

van Donkelaar A, Martin RV, Spurr RJD, Drury E, Remer LA, Levy RC, et al. 2013. Optimal estimation for 374

global ground-level fine particulate matter concentrations. Journal of Geophysical Research 118: 1-16. 375

Vinneau D, de Hoogh K, Bechle MJ, Beelen R, van Donkelaar A, Martin RV, et al. 2013. Western European 376

land use regression incorporating satellite- and ground-based measurements of NO2 and PM10. 377

Environmental Science & Technology 47(22): 12903-12911. 378

World Bank T. 2011. Air Quality Anaylsis of Ulaanbaatar: Improving Air Quality to Reduce Health 379

Impacts. 380

World Health Organization T. year. Burden of disease from Ambient Air Pollution for 2012 - summary of 381

results. Available: 382

http://www.who.int/phe/health_topics/outdoorair/databases/AAP_BoD_results_March2014.pdf. 383

Zhang J, Reid JS. 2010. A decadal regional and global trend analysis of the aerosol optical depth using a 384

data-assimilation grade over-water MODIS and Level 2 MISR aerosol products. Atmos Chem Phys 10: 385

10949-10963. 386

387

388

of 28

13

Table 1: Population-weighted ambient PM2.5 and trend within GBD regionsa

Region

2001-2010

PM2.5

2001-2010

Dust and

Seasalt-free

PM2.5 1998-2012 PM2.5 Trend

Population in excess of WHO PM2.5 target [%]b

AQG IT3 IT2 IT1

19

98

-20

00

20

10

-20

12

20

10

-20

12

1

99

8-2

00

0 P

op

19

98

-20

00

20

10

-20

12

20

10

-20

12

1

99

8-2

00

0 P

op

19

98

-20

00

20

10

-20

12

20

10

-20

12

1

99

8-2

00

0 P

op

19

98

-20

00

20

10

-20

12

20

10

-20

12

1

99

8-2

00

0 P

op

Ave

Std

Dev Ave

Std

Dev

Ave C.I. Ave C.I.

[μg/m3] [μg/m

3] [μg/m

3/yr] [%]

GLOBAL 26.4 21.4 21.2 19.1 0.55 0.12 2.1 0.5 76 75 75 57 61 60 32 43 42 22 30 30

ASIA PACIFIC, HIGH INCOME 16.8 6.4 15.3 6.0 -0.06 0.14 -0.4 0.8 77 80 80 50 50 49 9 11 10 1 0 0

ASIA, CENTRAL 17.3 5.7 9.7 3.1 0.29 0.17 1.7 1.0 78 84 82 56 69 68 14 18 17 2 2 2

ASIA, EAST 50.3 24.3 45.2 22.5 1.63 0.54 3.2 1.1 95 99 99 86 95 95 67 84 84 51 70 70

ASIA, SOUTH 34.6 15.8 27.8 13.2 1.02 0.25 2.9 0.7 92 100 100 75 98 97 43 78 77 27 52 51

ASIA, SOUTHEAST 11.0 6.4 10.2 6.0 0.30 0.09 2.7 0.8 42 55 56 23 27 28 6 7 7 3 2 2

AUSTRALASIA 3.0 1.0 2.6 0.9 0.01 0.03 0.3 1.0 0 0 0 0 0 0 0 0 0 0 0 0

CARIBBEAN 7.0 2.5 4.7 1.5 -0.02 0.07 -0.3 1.0 15 27 24 2 2 2 1 0 0 0 0 0

EUROPE, CENTRAL 17.8 2.6 16.2 2.7 -0.22 0.26 -1.2 1.5 96 96 97 80 63 63 10 3 3 1 0 0

EUROPE, EASTERN 12.6 3.7 11.2 3.5 -0.04 0.21 -0.3 1.7 66 68 67 28 22 21 2 0 0 0 0 0

EUROPE, WESTERN 13.5 4.6 12.1 4.2 -0.25 0.12 -1.9 0.9 84 66 66 45 27 26 7 3 3 1 0 0

LATIN AMERICA, ANDEAN 6.6 3.7 6.6 3.7 0.09 0.14 1.4 2.1 23 26 26 10 4 4 1 0 0 0 0 0

LATIN AMERICA, CENTRAL 8.5 4.3 7.8 4.3 -0.07 0.07 -0.8 0.8 43 34 34 24 9 9 11 1 0 6 0 0

LATIN AMERICA, SOUTHERN 6.4 2.4 5.4 2.3 0.08 0.09 1.3 1.4 8 8 8 2 1 1 0 0 0 0 0 0

LATIN AMERICA, TROPICAL 5.0 2.6 4.9 2.5 0.01 0.04 0.2 0.8 15 6 6 2 0 0 0 0 0 0 0 0

NORTH AFRICA / MIDDLE EAST 25.5 10.7 11.5 3.6 0.38 0.21 1.5 0.8 93 97 97 72 80 79 35 53 51 15 28 27

NORTH AMERICA, HIGH INCOME 9.9 3.2 9.6 3.3 -0.33 0.08 -3.3 0.8 62 19 20 17 2 2 1 0 0 0 0 0

OCEANIA 2.3 1.1 2.3 1.1 0.09 0.03 3.9 1.3 0 1 0 0 0 0 0 0 0 0 0 0

SUB-SAHARAN AFRICA, CENTRAL 11.4 3.3 9.9 2.7 -0.05 0.09 -0.4 0.8 65 60 59 34 27 26 5 2 2 1 0 0

SUB-SAHARAN AFRICA, EAST 9.8 8.2 5.5 2.4 0.10 0.09 1.0 0.9 32 38 38 19 19 20 8 9 9 3 3 3

SUB-SAHARAN AFRICA, SOUTHERN 5.9 2.0 5.6 1.9 0.09 0.08 1.5 1.4 3 8 7 0 0 0 0 0 0 0 0 0

SUB-SAHARAN AFRICA, WEST 30.8 14.9 7.6 2.9 -0.04 0.29 -0.1 0.9 97 96 95 91 84 84 74 56 55 51 32 32

Abbreviations: Ave, average; Std Dev, standard devation; C.I., confidence interval; AQG, air quality guideline; IT, interim target; Pop, population.

a) Defined in Figure 3.

b) Percent above WHO PM2.5 targets for 1998-2000, 2010-2012, and 2010-2012 using 1998-2000 population distributions.

of 28

14

Figure 1: Decadal (2001-2010) mean PM2.5 concentrations over North America. White areas denote

water or missing values. The top panel displays satellite-derived values. The lower right panel contains

averages at ground-based sites in operation at least 8 years during this period. The lower left panel

provides a scatterplot of the two datasets. The 1:1 line is solid. The line of best fit is dash-dot. The

observed 1-σ error is dotted. Ground-based and satellite values are not coincidently sampled to avoid

biasing the data toward clear-sky conditions when satellite retrievals occur. A common, logarithmic

color scale is used for Figures 1-4.

Figure 2: Decadal (2001-2010) mean PM2.5 concentrations over Europe. The top panel displays satellite-

derived values. The lower right panel contains ground-based values in operation at least 3 years during

this period. The lower left panel provides a scatterplot of the two datasets, sampled on the same years

but non-coincidently on a daily basis. The 1:1 line is solid. The line of best fit is dash-dot. The observed

1-σ error is dotted. A common, logarithmic color scale is used for Figures 1-4.

Figure 3: Global decadal (2001-2010) mean PM2.5 concentrations. The top panel displays satellite-

derived PM2.5. The middle panel contains dust and sea-salt free PM2.5. Inset maps display GBD regional

population-weighted mean concentrations. The bottom right panel contains ground-based values in

operation during this period. The lower left panel provides a scatterplot of the two all-species datasets,

sampled on the same years but non-coincidently on a daily basis. The 1:1 line is solid. The line of best

fit is dash-dot. The observed 1-σ error is dotted. A common, logarithmic color scale is used for Figures

1-4.

Figure 4: Three-year running mean of satellite-derived PM2.5 over sample regions of significant trends.

Sub-regions are denoted by boxes with black circles around the city centers highlighted in Figure 5. A

common, logarithmic color scale is used for Figures 1-4.

Figure 5: PM2.5 time-series at the four locations identified in Figure 4. Dots and vertical lines denote

monthly mean and 25th

-75th

percentile. Trend and confidence interval are in the inset. Satellite-derived

(black), ground-level monitor (red), and satellite-derived coincident with ground-level monitor (blue)

PM2.5 are given, as available.

Figure 6: Cumulative distribution of regional, annual mean PM2.5 for 1998-2012. AQG, IT3, IT2, and IT1

refer to the WHO air quality guidelines of 10, 15, 25 and 35 μg/m3.

of 28

Figure 1: Decadal (2001-2010) mean PM2.5 concentrations over North America. White areas denote water or missing values. The top panel displays satellite-derived values. The lower right panel contains averages

at ground-based sites in operation at least 8 years during this period. The lower left panel provides a scatterplot of the two datasets. The 1:1 line is solid. The line of best fit is dash-dot. The observed 1-σ

error is dotted. Ground-based and satellite values are not coincidently sampled to avoid biasing the data toward clear-sky conditions when satellite retrievals occur. A common, logarithmic color scale is used for

Figures 1-4.

of 28

Figure 2: Decadal (2001-2010) mean PM2.5 concentrations over Europe. The top panel displays satellite-derived values. The lower right panel contains ground-based values in operation at least 3 years during this period. The lower left panel provides a scatterplot of the two datasets, sampled on the same years but non-

coincidently on a daily basis. The 1:1 line is solid. The line of best fit is dash-dot. The observed 1-σ error is dotted. A common, logarithmic color scale is used for Figures 1-4.

of 28

Figure 3: Global decadal (2001-2010) mean PM2.5 concentrations. The top panel displays satellite-derived PM2.5. The middle panel contains dust and sea-salt free PM2.5. Inset maps display GBD regional

population-weighted mean concentrations. The bottom right panel contains ground-based values in

operation during this period. The lower left panel provides a scatterplot of the two all-species datasets, sampled on the same years but non-coincidently on a daily basis. The 1:1 line is solid. The line of best fit is

dash-dot. The observed 1-σ error is dotted. A common, logarithmic color scale is used for Figures 1-4. 165x186mm (300 x 300 DPI)

of 28

Figure 4: Three-year running mean of satellite-derived PM2.5 over sample regions of significant trends. Sub-regions are denoted by boxes with black circles around the city centers highlighted in Figure

5. A common, logarithmic color scale is used for Figures 1-4.

of 28

Figure 5: PM2.5 time-series at the four locations identified in Figure 4. Dots and vertical lines denote monthly mean and 25th-75th percentile. Trend and confidence interval are in the inset. Satellite-derived

(black), ground-level monitor (red), and satellite-derived coincident with ground-level monitor (blue) PM2.5 are given, as available.

of 28

Figure 6: Cumulative distribution of regional, annual mean PM2.5 for 1998-2012. AQG, IT3, IT2, and IT1 refer to the WHO air quality guidelines of 10, 15, 25 and 35 µg/m3.

of 28

1

Supplemental Material: Unified global fine particulate matter concentrations from satellite for long-term 1

exposure assessment 2

Aaron van Donkelaar, Randall Martin, Brian Boys, and Mike Brauer 3

4

Description of ground-level monitor sources from established networks 5

Established PM2.5 networks provide a robust source of evaluation for satellite-derived PM2.5 6

concentrations due to their long-term observation period and consistent measurement practices. 7

Ground-level Canadian PM2.5 observations were obtained from the National Air Pollution Surveillance 8

network (NAPS; http://www.etc.cte.ec.gc.ca/NAPS/index_e.html), excluding industrial sites. American 9

observations were taken from sites of the Interagency Monitoring of Protected Visual Environments 10

network (IMPROVE; http://vista.cira.colostate.edu/improve/Data/data.htm) and from the 11

Environmental Protection Agency Air Quality System that employ the Federal Reference Method (FRM; 12

http://www.epa.gov/air/data/index.html). PM2.5 measurements at background sites from the European 13

air quality database (Airbase; http://acm.eionet.europa.eu/databases/airbase/) and European 14

Monitoring and Evaluation Programme (EMEP; Torseth et al. 2012) were used over Europe. 15

16

Description of satellite instrumentation 17

As described in van Donkelaar et al. (2010; 2013), the Unconstrained (UC) and Optimal Estimation (OE) 18

PM2.5 datasets use data from the MODIS (MODerate resolution Imaging Spectroradiometer) 19

instruments. UC uses MODIS onboard the Terra satellite, while OE uses MODIS onboard both Terra and 20

Aqua. Both MODIS instruments provide near-daily global AOD coverage in the absence of clouds from a 21

polar orbiting, sun-synchronous orbit. Quality assured collection (version) 5 MODIS AOD at 10 km × 10 22

km over land (Levy et al. 2007) has been validated such that at least two-thirds of its retrievals are 23

within ±(0.05 + 15%) using Aerosol Robotic Network (Holben et al. 2001) measurements of AOD (Remer 24

et al. 2008). Concerns have been raised about drift in MODIS collection 5 over land (Zhang and Reid 25

2010. We use this dataset only for long-term averages (not trends). 26

The MISR (Multi-angle Imaging SpectroRadiometer) instrument onboard the Terra satellite is used for 27

the UC dataset (van Donkelaar et al. 2010) and trends (Boys et al. submitted). MISR observes radiation 28

leaving the top of the atmosphere in four spectral bands (0.446, 0.558, 0.672 and 0.866 μm), each at 29

nine viewing angles (±70.5º, ±60.0º, ±45.6º, ±25.1º and nadir). MISR typically takes 6 to 9 days for 30

complete global in the absence of clouds. The MISR AOD retrieval algorithm (Diner et al. 2005; 31

Martonchik et al. 2002; Martonchik et al. 2009) has been validated such that two-thirds of retrievals fall 32

within the maximum of ±(0.05 or 20%) of ground truth observations (Kahn et al. 2005), and has reliable 33

trend information over land Zhang and Reid 2010. 34

The SeaWiFS (Sea-viewing Wide Field-of-view Sensor) instrument provides near-daily global coverage at 35

8 wavelengths from a sun-synchronous orbit. The Deep Blue algorithm has recently been applied to 36

of 28

2

SeaWiFS AOD retrieval at a resolution of 13.5 km (Hsu et al. 2013), providing a well-calibrated retrieval 37

of global AOD from 1998-2010 suitable for trend studies (Hsu et al. 2012). High quality SeaWiFS AOD 38

has been validated such that at least two-thirds of retrievals are within ±(0.05 + 20%) (Sayer et al. 2012). 39

40

Description of the GEOS-Chem chemical transport model 41

The GEOS-Chem chemical transport model (http://geos-chem.org) solves for the spatial and temporal 42

evolution of atmospheric aerosol and gaseous compounds using meteorological data sets, emission 43

inventories, and equations that represent the physics and chemistry of the atmosphere. We used GEOS-44

Chem to relate AOD to PM2.5 mass and surface area, and to provide prior estimates with which to 45

constrain the OE satellite retrievals of AOD. 46

Detailed simulation descriptions are contained within the corresponding publications for UC (van 47

Donkelaar et al. 2010), OE (van Donkelaar et al. 2013) and SeaWiFS&MISR (Boys et al. submitted). A 48

major distinction between these simulations are the assimilated meteorological fields used for UC 49

(GEOS-4), OE (GEOS-5) and SeaWiFS&MISR (MERRA). All fields were provided by the Goddard Earth 50

Observing System and represented current versions of available meteorology at the original time of 51

each publication. All simulations were performed globally at 2° × 2.5°. OE additionally used three 52

nested 1/2° × 2/3° regions overs North America, Europe and eastern Asia. 53

All simulations share a similar treatment of aerosol that include the sulphate-ammonium-nitrate-water 54

system (Park et al. 2004), primary carbonaceous aerosols (Park et al. 2003), secondary organic aerosols 55

(Henze et al. 2008), sea salt (Alexander et al. 2005), and mineral dust (Fairlie et al. 2007). 56

57

Descripton of satellite-derived PM2.5 surface area 58

AOD is more directly related to PM2.5 surface area than PM2.5 mass since light extinction is proportional 59

to particle surface area (not volume) and surface area does not require assumptions about particle 60

densities. Satellite-derived estimates of surface area can, therefore, be readily created following the 61

approaches established for PM2.5 mass. We produced such estimates of surface area by applying to 62

satellite (MODIS, MISR and SeaWiFS) GEOS-Chem simulations of coincident AOD to ground-level surface 63

area of particles with aerodynamic diameter smaller than 2.5 μm which we refer to as PM2.5 surface 64

area. OE, UC, and SeaWiFS&MISR-based surface area was produced using the simulations and methods 65

described for PM2.5 in van Donkelaar et al. (2013) and van Donkelaar et al. (2010), respectively, and 66

combined following the approach outlined in the main manuscript. Figure S2 shows the resultant 67

decadal mean PM2.5 surface area for comparison with Figure 4. 68

69

70

of 28

3

71

72

Sub-annual agreement of three-month mean satellite-derived PM2.5 73

Table S1 summarizes the variation in seasonal agreement between the satellite-derived and ground-74

based PM2.5 at approximately 1000 locations in North America. Seasonal agreement varies with 75

expected patterns of AOD retrieval accuracy, with improved agreement during summer months when 76

surface reflectance is better characterized and when seasonal PM2.5 enhancements increase the aerosol 77

signal in satellite observations. We also provide the agreement found when applying GEOS-Chem 78

seasonality to the satellite-derived annual means. Simulated seasonal variation improves monthly 79

satellite-derived PM2.5, particularly in the winter season when satellite retrievals can be inhibited by 80

snow-cover. Seasonal cycles will vary globally, but these results suggest that the impact of snow, cloud 81

and reduced sampling may increase the uncertainty of seasonal decadal mean PM2.5 estimates by up to 82

a factor of two relative to annual mean values. 83

Table S2 evaluates the impact of temporal range on accuracy, comparing mean satellite-derived and 84

ground-based PM2.5 over a varying number of years at ca. 1000 locations in North America. On average, 85

annual performance is degraded significantly from decadal mean values (r=0.68 vs. r=0.79; slope=1.07 86

vs. 1.00; 1σ error = 20% vs. 14%). Errors in long-term exposure assessment increase with decreasing 87

number of measurements from satellite. Sub-annual agreement of three-month running means further 88

increases error by up to a factor of two (supplemental). Significant improvement, however, is found 89

when using as few as three years (r = 0.73; slope = 1.05; 1σ error = 17%), although still well below 90

decadal agreement. As a result, the spatial correlations obtained over European and global regions 91

(Figures 2 and 3) may indicate comparable significance to North America, only reduced by the limited 92

sampling period of available ground-level observations for comparison. 93

94

95

96

of 28

4

Figure S1: PM2.5 annual trend over 1998-2012. The intensity of the colorscale provides a measure of statistical significance. Inset gives

population-weighted mean values within GBD-defined regions. Grey areas denote water or missing data.

of 28

5

Figure S2: Global decadal (2001-2010) mean PM2.5 surface areas. The inset map displays GBD regional population-weighted mean surface area.

The logarithmic color scale follows that used for Figures 1-4.

of 28

6

Table S1: Effect of season on satellite-derived and ground-level PM2.5 agreement over North America,

2001-2010. Mean and standard deviation of year-specific monthly mean agreement is given. The

agreement of simulated seasonality applied to annual mean satellite-derived PM2.5 is also given.

Monthly values represent the center month of a three-month temporal range. Approximately 1000

locations are used.

Satellite Seasonality Simulated Seasonality

Time

Period

1σ error

[% + 1

μg/m3]

Slope Offset Pearson

Coefficient

1σ error

[% + 1

μg/m3]


Coefficient

Annual 20 ± 2 1.07 ± 0.10 -1.4 ± 0.7 0.68 ± 0.07 20 ± 2 1.07 ± 0.10 -1.4 ± 0.7 0.68 ± 0.07

January 36 ± 3 1.47 ± 0.27 -4.9 ± 2.4 0.37 ± 0.06 26 ± 3 0.87 ± 0.11 0.6 ± 0.7 0.48 ± 0.04

February 33 ± 4 1.54 ± 0.23 -5.4 ± 2.2 0.45 ± 0.08 24 ± 2 0.91 ± 0.09 0.1 ± 0.7 0.57 ± 0.06

March 28 ± 3 1.49 ± 0.21 -4.7 ± 2.0 0.51 ± 0.08 20 ± 2 0.99 ± 0.10 -0.6 ±0.8 0.65 ± 0.07

April 24 ± 3 1.30 ± 0.18 -2.6 ± 1.4 0.59 ± 0.06 19 ± 2 1.03 ± 0.13 -0.7 ± 0.9 0.66 ± 0.08

May 22 ± 3 1.24 ± 0.18 -2.3 ± 1.5 0.62 ± 0.09 21 ± 3 1.00 ± 0.13 -0.5 ± 0.9 0.62 ± 0.12

June 23 ± 3 1.23 ± 0.18 -2.7 ± 1.5 0.66 ± 0.12 23 ± 4 0.98 ± 0.12 -0.6 ±1.0 0.64 ± 0.15

July 23 ± 3 1.24 ± 0.16 -3.2 ± 1.6 0.68 ± 0.11 24 ± 4 0.94 ± 0.08 -0.4 ±0.9 0.68 ± 0.14

August 24 ± 4 1.14 ± 0.21 -2.3 ± 1.6 0.68 ± 0.11 22 ± 4 0.98 ± 0.10 -0.7 ± 0.8 0.71 ± 0.11

September 25 ± 5 1.06 ± 0.23 -1.5 ± 1.5 0.64 ± 0.10 21 ± 2 1.06 ± 0.16 -0.9 ± 1.0 0.69 ± 0.08

October 30 ± 5 0.94 ± 0.20 -0.7 ± 1.1 0.53 ± 0.10 24 ± 3 1.10 ± 0.17 -1.2 ± 1.1 0.60 ± 0.06

November 34 ± 5 0.95 ± 0.15 -0.7 ± 0.8 0.42 ± 0.08 27 ± 4 1.02 ± 0.11 -0.5 ± 0.7 0.49 ± 0.06

December 37 ± 4 1.05 ± 0.15 -1.3 ± 1.7 0.37 ± 0.07 27 ± 3 0.87 ± 0.10 0.6 ± 0.7 0.45 ± 0.05

Table S2: Effect of temporal range on satellite-derived and ground-level PM2.5 agreement over North

America. Mean and standard deviation of individual temporal comparisons are given (e.g. mean and

standard deviation of annual agreement when temporal range is 1 year). Sites must be active for at

least 80% of the temporal range, resulting in ca. 1000 locations used.

Temporal

Range (yrs)

1σ error

[% + 1 μg/m3]


Coefficient

1 20 ± 2 1.07 ± 0.10 -1.4 ± 0.7 0.68 ± 0.07

2 17 ± 2 1.05 ± 0.08 -1.2 ± 0.7 0.72 ± 0.06

3 17 ± 2 1.04 ± 0.08 -1.2 ± 0.7 0.73 ± 0.05

4 15 ± 2 1.02 ± 0.06 -1.0 ± 0.5 0.74 ± 0.04

5 16 ± 1 1.02 ± 0.04 -1.0 ± 0.3 0.75 ± 0.03

6 15 ± 2 1.01 ± 0.04 -0.9 ± 0.3 0.77 ± 0.03

7 15 ± 2 1.00 ± 0.04 -0.8 ± 0.3 0.77 ± 0.03

8 14 ± 1 1.00 ± 0.03 -0.8 ± 0.2 0.78 ± 0.02

9 14 ± 1 1.00 ± 0.02 -0.8 ± 0.2 0.78 ± 0.02

10 14 ± 0 1.00 ± 0.00 -0.7 ± 0.0 0.79 ± N/A

of 28

7

References

Alexander, B, RJ Park, DJ Jacob, QB Li, RM Yantosca, J Savarino, et al. 2005. Sulfate formation in sea-salt

aerosols: Constraints from oxygen isotopes. Journal of Geophysical Research-Atmospheres 110(D10).

Boys, B, RV Martin, A Van Donkelaar, R MacDonell, NC Hsu, MJ Cooper, et al. submitted. Fifteen year

global time series of satellite-derived fine particulate matter. Environmental Science & Technology.

Available from http://fizz.phys.dal.ca/~atmos/publications/BOYS_2014_EST_PM_trend_submit.pdf.

Diner, DD, BH Braswell, R Davies, N Gobron, J Hu, Y Jin, et al. 2005. The value of multiangle

measurements for retrieving structurally and radiatively consistent properties of clouds, aerosols, and

surfaces. Remote Sensing of Environment 97: 495-518.

Fairlie, TD, DJ Jacob and RJ Park. 2007. The impact of transpacific transport of mineral dust in the United

States. Atmospheric Environment 41(6): 1251-1266.

Henze, DK, JH Seinfeld, NL Ng, JH Kroll, TM Fu, DJ Jacob, et al. 2008. Global modeling of secondary

organic aerosol formation from aromatic hydrocarbons: high- vs. low-yield pathways. Atmospheric

Chemistry and Physics 8: 2405-2421.

Holben, BN, D Tanre, A Smirnov, TF Eck, I Slutsker, N Abuhassan, et al. 2001. An emerging ground-based

aerosol climatology: Aerosol optical depth from AERONET. Journal of Geophysical Research-

Atmospheres 106(D11): 12067-12097.

Hsu, NC, R Gautam, AM Sayer, C Bettenhausen, C Li, MJ Jeong, et al. 2012. Global and regional trends of

aerosol optical depth over land and ocean using SeaWiFS measurements from 1997-2010. Atmos. Chem.

Phys. 12: 8037-8053.

Hsu, NC, MJ Jeong, C Bettenhausen, AM Sayer, R Hansell, CS Seftor, et al. 2013. Enhanced Deep Blue

aerosol retrieval algorithm: The second generation. Journal of geophysical research 118: 1-20.

Kahn, RA, BJ Gaitley, JV Martonchik, DJ Diner, KA Crean and B Holben. 2005. Multiangle Imaging

Spectroradiometer (MISR) global aerosol optical depth validation based on 2 years of coincident Aerosol

Robotic Network (AERONET) observations. Journal of Geophysical Research-Atmospheres 110(D10).

Levy, RC, LA Remer and O Dubovik. 2007. Global aerosol optical properties and application to Moderate

Resolution Imaging Spectroradiometer aerosol retrieval over land. Journal of Geophysical Research

112(D13210).

Martonchik, JV, DJ Diner, KA Crean and MA Bull. 2002. Regional aerosol retrieval results from MISR. Ieee

Transactions on Geoscience and Remote Sensing 40(7): 1520-1531.

Martonchik, JV, RA Kahn and DJ Diner. 2009. Retrieval of Aerosol Properties over Land Using MISR

Observations. In: Satellite Aerosol Remote Sensing Over Land, (AA Kokhanovsky and Gd Leeuw, eds).

Berlin:Springer.

Park, RJ, DJ Jacob, M Chin and RV Martin. 2003. Sources of carbonaceous aerosols over the United

States and implications for natural visibility. Journal of Geophysical Research-Atmospheres 108(D12).

of 28

8

Park, RJ, DJ Jacob, BD Field, RM Yantosca and M Chin. 2004. Natural and transboundary pollution

influences on sulfate-nitrate-ammonium aerosols in the United States: Implications for policy. Journal of

Geophysical Research-Atmospheres 109(D15).

Remer, LA, RG Kleidman, RC Levy, YJ Kaufman, D Tanre, S Mattoo, et al. 2008. Global aerosol climatology

from the MODIS satellite sensors. Journal of Geophysical Research-Atmospheres 113(D14).

Sayer, AM, NC Hsu, C Bettenhausen, M-J Jeong and J Zhang. 2012. Global and regional evaluation of

over-land spectral aerosol optical depth retrievals from SeaWiFS. Atmospheric Measurement

Techniques 5: 1761-1778.

Torseth, K, W Aas, K Breivik, AM Fjaeraa, M Fiebig, AG Hjellbrekke, et al. 2012. Introduction to the

European Monitoring and Evaluation Programme (EMEP) and observated atmosperhic composition

change during 1972-2009. Atmos. Chem. Phys. 12: 5447-5481.

van Donkelaar, A, RV Martin, M Brauer, R Kahn, R Levy, C Verduzco, et al. 2010. Global Estimates of

Ambient Fine Particulate Matter Concentrations from Satellite-Based Aerosol Optical Depth:

Development and Application. Environmental Health Perspectives 118(6): 847-855.

van Donkelaar, A, RV Martin, RJD Spurr, E Drury, LA Remer, RC Levy, et al. 2013. Optimal estimation for

global ground-level fine particulate matter concentrations. Journal of Geophysical Research 118: 1-16.

Zhang, J and JS Reid. 2010. A decadal regional and global trend analysis of the aerosol optical depth

using a data-assimilation grade over-water MODIS and Level 2 MISR aerosol products. Atmos. Chem.

Phys. 10: 10949-10963.

of 28

2 aaron van donkelaar , randall v. martin , michael brauer...

Documents