This con en has been downloaded om IOPscience. Please sc oll down o see he ull ex .
Download de ails:
IP Add ess: 158.36.47.18
This con en was downloaded on 24/08/2017 a 08:33
Please no e ha e ms and condi ions apply.
Global p ema u e mo ali y due o an h opogenic ou doo ai pollu ion and he con ibu ion o
pas clima e change
View he able o con en s o his issue, o go o he jou nal homepage o mo e
2013 En i on. Res. Le . 8 034005
(h p://iopscience.iop.o g/1748-9326/8/3/034005)
Home Sea ch Collec ions Jou nals Abou Con ac us My IOPscience
You may also be in e es ed in:
Public heal h impac s o excess NOX emissions om Volkswagen diesel passenge ehicles in Ge many
Guillaume P Chossiè e, Robe Malina, Akshay Ashok e al.
Global heal h and economic impac s o u u e ozone pollu ion
N E Selin, S Wu, K M Nam e al.
Global bu den o mo ali ies due o ch onic exposu e o ambien PM2.5 om open combus ion o
domes ic was e
John K Kod os, Ch is ine Wiedinmye , Bonne Fo d e al.
Es ima ing sou ce-a ibu able heal h impac s o ambien ine pa icula e ma e exposu e: global
p ema u e mo ali y om su ace anspo a ion emissions in 2005
S E Chambliss, R Sil a, J J Wes e al.
Is clima e change a ec ing human heal h?
K is ie L Ebi
The impac o Eu opean legisla i e and echnology measu es o educe ai pollu an s on ai quali y,
human heal h and clima e
S T Tu nock, E W Bu , T B Richa dson e al.
In e annual a iabili y o summe ime ae osol op ical dep h o e Eas Asia du ing 2000–2011: a
po en ial in luence om El Niño Sou he n Oscilla ion
Yikun Liu, Jun eng Liu and Shu Tao
Municipal solid was e and dung cake bu ning: discolo ing he Taj Mahal and human heal h impac s in
Ag a
Raj M Lal, Ajay S Nagpu e, Lina Luo e al.
IOP PUBLISHING ENVIRONMENTAL RESEARCH LETTERS
En i on. Res. Le . 8(2013) 034005 (11pp) doi:10.1088/1748-9326/8/3/034005
Global p ema u e mo ali y due o
an h opogenic ou doo ai pollu ion and
he con ibu ion o pas clima e change
Raquel A Sil a1, J Jason Wes 1,23, Yuqiang Zhang1, Susan C Anenbe g2,
Jean-F anc¸ois Lama que3, D ew T Shindell4, William J Collins5,
S ig Dalso en6, G eg Falu egi4, Ge d Folbe h7, La y W Ho owi z8,
Ta suya Nagashima9, Vaishali Naik10, S e en Rumbold7, Ragnhild Skeie6,
Kengo Sudo11, Toshihiko Takemu a12, Daniel Be gmann13,
Philip Came on-Smi h13, I ene Cionni14, Ru h M Dohe y15,
Ve onika Ey ing16, Bea ice Josse17, I A MacKenzie15, Da id Plumme 18,
Ma ia Righi16, Da id S S e enson15, Sa ah S ode19,20, Sophie Szopa21
and Guang Zeng22
1En i onmen al Sciences and Enginee ing, Uni e si y o No h Ca olina, Chapel Hill, NC 27599, USA
2US En i onmen al P o ec ion Agency, Washing on, DC 20004, USA
3NCAR Ea h Sys em Labo a o y, Na ional Cen e o A mosphe ic Resea ch, Boulde , CO 80301, USA
4NASA Godda d Ins i u e o Space S udies and Columbia Ea h Ins i u e, New Yo k, NY, USA
5Depa men o Me eo ology, Uni e si y o Reading, Reading, UK
6CICERO, Cen e o In e na ional Clima e and En i onmen al Resea ch-Oslo, Oslo, No way
7Hadley Cen e o Clima e P edic ion, Me O ice, Exe e , UK
8NOAA Geophysical Fluid Dynamics Labo a o y, P ince on, NJ 08540, USA
9Na ional Ins i u e o En i onmen al S udies, Tsukuba, Japan
10 UCAR/NOAA Geophysical Fluid Dynamics Labo a o y, P ince on, NJ 08540, USA
11 Ea h and En i onmen al Science, G adua e School o En i onmen al S udies, Nagoya Uni e si y,
Nagoya, Japan
12 Resea ch Ins i u e o Applied Mechanics, Kyushu Uni e si y, Fukuoka, Japan
13 Law ence Li e mo e Na ional Labo a o y, Li e mo e, CA, USA
14 Agenzia Nazionale pe le Nuo e Tecnologie, l’Ene gia e lo S iluppo Economico Sos enibile (ENEA),
Bologna, I aly
15 School o GeoSciences, Uni e si y o Edinbu gh, Edinbu gh, UK
16 Deu sches Zen um ¨
u Lu - und Raum ah (DLR) Ins i u ¨
u Physik de A mosph¨
a e,
Obe p a enho en, Ge many
17 GAME/CNRM, Me eo-F ance, CNRS—Cen e Na ional de Reche ches Me eo ologiques, Toulouse,
F ance
18 Canadian Cen e o Clima e Modeling and Analysis, En i onmen Canada, Vic o ia, BC, Canada
19 NASA Godda d Space Fligh Cen e , G eenbel , MD, USA
20 Uni e si ies Space Resea ch Associa ion, Columbia, MD, USA
21 Labo a oi e des Sciences du Clima e de l’En i onnemen , LSCE-CEA-CNRS-UVSQ,
Gi -su -Y e e, F ance
22 Na ional Ins i u e o Wa e and A mosphe ic Resea ch, Laude , New Zealand
E-mail: [email p o ec ed]
Recei ed 25 Ma ch 2013
Accep ed o publica ion 23 May 2013
Published 11 July 2013
Online a s acks.iop.o g/ERL/8/034005
23 Au ho o whom any co espondence should be add essed.
Con en om his wo k may be used unde he e ms o he C ea i e Commons A ibu ion 3.0 licence. Any u he dis ibu ion o his wo k
mus main ain a ibu ion o he au ho (s) and he i le o he wo k, jou nal ci a ion and DOI.
11748-9326/13/034005+11$33.00 c
2013 IOP Publishing L d P in ed in he UK
En i on. Res. Le . 8(2013) 034005 R A Sil a e al
Abs ac
Inc eased concen a ions o ozone and ine pa icula e ma e (PM2.5) since p eindus ial imes
e lec inc eased emissions, bu also con ibu ions o pas clima e change. He e we use
modeled concen a ions om an ensemble o chemis y–clima e models o es ima e he global
bu den o an h opogenic ou doo ai pollu ion on p esen -day p ema u e human mo ali y, and
he componen o ha bu den a ibu able o pas clima e change. Using simula ed
concen a ions o 2000 and 1850 and concen a ion– esponse unc ions (CRFs), we es ima e
ha , a p esen , 470 000 (95% con idence in e al, 140 000 o 900 000) p ema u e espi a o y
dea hs a e associa ed globally and annually wi h an h opogenic ozone, and 2.1 (1.3 o 3.0)
million dea hs wi h an h opogenic PM2.5- ela ed ca diopulmona y diseases (93%) and lung
cance (7%). These es ima es a e smalle han ones om p e ious s udies because we use
modeled 1850 ai pollu ion a he han a coun e ac ual low concen a ion, and because o
di e en emissions. Unce ain y in CRFs con ibu es mo e o o e all unce ain y han he
sp ead o model esul s. Mo ali y a ibu ed o he e ec s o pas clima e change on ai quali y
is conside ably smalle han he global bu den: 1500 (−20 000 o 27 000) dea hs y −1due o
ozone and 2200 (−350 000 o 140 000) due o PM2.5. The small mul i-model means a e
coinciden al, as he e a e la ge anges o esul s o indi idual models, e lec ed in he la ge
unce ain ies, wi h some models sugges ing ha pas clima e change has educed ai pollu ion
mo ali y.
Keywo ds: clima e change, ai pollu ion, ozone, pa icula e ma e , human heal h, p ema u e
mo ali y
SOnline supplemen a y da a a ailable om s acks.iop.o g/ERL/8/034005/mmedia
1. In oduc ion
Since he indus ial e olu ion, human ac i i ies ha e
signi ican ly inc eased he concen a ions o ozone and ine
pa icula e ma e (wi h ae odynamic diame e less han
2.5µm, PM2.5) in bo h u ban and u al egions (Schulz e al
2006, Pa ish e al 2012). These changes ha e been d i en
by di ec changes in ai pollu an emissions, and, because
clima e change also in luences ai quali y, a componen
o he changes in an h opogenic ai pollu ion may esul
om pas clima e change. Clima e change in luences ai
quali y h ough se e al mechanisms, including changes in
pho ochemical eac ion a es, biogenic emissions, deposi ion,
and a mosphe ic ci cula ion (Jacob and Winne 2009, Wea e
e al 2009, Fio e e al 2012).
Epidemiological s udies ha e shown ha ozone and
PM2.5ha e signi ican in luences on human heal h, including
p ema u e mo ali y. E idence o mo ali y in luences comes
om a la ge numbe o daily ime se ies s udies (e.g., Bell e al
2004, HEI 2004). The e is also e idence o ch onic e ec s on
mo ali y h ough se e al la ge coho s udies o PM2.5(Hoek
e al 2002, K ewski e al 2009, Lepeule e al 2012), while
e idence o ch onic e ec s o ozone de i es mainly om one
s udy (Je e e al 2009).
Pas esea ch o es ima e he global bu den o disease
due o ou doo ai pollu ion has used a a ie y o me hods.
Cohen e al (2004) es ima ed 800 000 p ema u e dea hs
annually a ibu ed o u ban PM2.5globally, based on
su ace measu emen s. Accoun ing o bo h u ban and u al
egions globally, Anenbe g e al (2010) used ou pu om
a global a mosphe ic model o es ima e 3.7 ±1.0 million
dea hs annually due o an h opogenic (p esen -day ela i e o
p eindus ial) changes in PM2.5and 0.7 ±0.3 million due
o ozone. B aue e al (2012) used high- esolu ion sa elli e
obse a ions o PM2.5 oge he wi h a global a mosphe ic
model and an ex ensi e compila ion o su ace measu emen s
o be e ep esen global ai pollu ion exposu e. These
exposu e es ima es we e hen used o es ima e 3.2 ±0.4
million p ema u e dea hs due o PM2.5and 150 000 (50 000
o 270 000) due o ozone (Lim e al 2012).
Few s udies ha e assessed he e ec s o clima e change
on human heal h ia changes in ai quali y. O hose, he ocus
has been on he in luence o u u e clima e change, including
assessmen s on a me opoli an scale (Knowl on e al 2004,
She ield e al 2011), in he US (Bell e al 2007, Taga is
e al 2009, Pos e al 2012), and globally (Wes e al 2007,
Selin e al 2009). O hese s udies, only Pos e al (2012)
use a mul i-model ensemble, showing signi ican a iabili y
in es ima es o ozone- ela ed mo ali y a ibu ed o clima e
change depending on he a mosphe ic model esul s used.
Fo he e ec s o pas clima e change on ai quali y and
human heal h, O u e al (2013) e alua ed egional e ec s
o ozone in Eu ope o bo h he ecen pas and he u u e.
Fang e al (2013) conduc ed a global analysis o pas clima e
change based on simula ions om a single a mosphe ic model
(GFDL-AM3); hose model simula ions a e included in he
mul i-model ensemble used he e.
He e we assess he bu den o global an h opogenic
ai pollu ion on p ema u e human mo ali y, and he
con ibu ion o pas changes in clima e o he o al bu den,
using simula ions om an ensemble o global coupled
chemis y–clima e models (Lama que e al 2013). Ou
app oach o es ima e he global bu den o ai pollu ion on
mo ali y expands on ha o Anenbe g e al (2010) by
using an ensemble o model es ima es o bo h p esen -day
and p eindus ial ai pollu ion. We hen use simula ions ha
2
En i on. Res. Le . 8(2013) 034005 R A Sil a e al
combine p esen -day emissions and p eindus ial clima e o
sepa a e he in luences o pas clima e change on ai quali y
and human heal h.
2. Me hods
2.1. Modeled ozone and PM2.5su ace concen a ions
The ensemble o global model simula ions was conduc ed
unde he A mosphe ic Chemis y and Clima e Model
In e compa ison P ojec (ACCMIP) (Lama que e al 2013,
Fio e e al 2012, S e enson e al 2013), including 14
models, 10 o which ully couple me eo ological and
chemical p ocesses. He e we only analyze his o ical ACCMIP
simula ions, and no u u e simula ions unde di e en
emissions scena ios. All models in ACCMIP used nea ly
iden ical an h opogenic emissions o bo h he p esen -day
(2000) and p eindus ial (1850), bu di e in na u al emissions
(Lama que e al 2010,2013, Young e al 2013). Compa ison
wi h obse a ions sugges s ha he models ep oduce ae osol
op ical dep h well, hough wi h a endency o unde es ima e
pa icula ly in Eas Asia (Shindell e al 2013). Fo ozone,
he models also ag ee well wi h sa elli e and ozonesonde
obse a ions, bu wi h a endency o o e es ima e in he
No he n Hemisphe e and unde es ima e in he Sou he n
Hemisphe e (Young e al 2013). Di e ences in na u al
emissions (biogenic VOCs), model chemical mechanisms,
and ozone anspo om he s a osphe e con ibu e o he
sp ead o ozone concen a ions ac oss models (Young e al
2013).
Fo ozone, we use ou pu om 14 models ha epo
esul s om bo h 1850 and 2000 simula ions; o hese,
9 models also epo esul s om an expe imen whe e
2000 emissions a e used oge he wi h 1850 clima e
(‘Em2000Cl1850’), o sepa a e he in luence o pas clima e
change on ai quali y. Fo PM2.5, 6 models epo esul s
om 1850 and 2000, and 5 o hese also epo esul s o
Em2000Cl1850.
We e e o he absolu e di e ence in concen a ions
be ween 1850 and 2000 as ‘an h opogenic’ ai pollu ion,
al hough 1850 includes some an h opogenic emissions, such
as om biomass bu ning (Lama que e al 2010), and he
simula ed pas clima e change includes some na u al o cings
as well as an h opogenic o cings. In a ibu ing ai pollu ion
changes o pas clima e change, his app oach accoun s
o e ec s o clima e change on a mosphe ic p ocesses
and na u al emissions, bu igno es e ec s on an h opogenic
emissions. To educe he e ec s o in e -annual a iabili y,
models ypically epo se e al yea s o ou pu o each
simula ion; we use he a e age o all yea s epo ed by mos
models ( a ying be ween 1 and 10 yea s), and use 10 yea s o
models ha epo ed mo e han 10 yea s. In all cases, modeled
concen a ions om he lowes e ical coo dina e a e aken o
ep esen su ace concen a ions.
Modeled concen a ions a e p ocessed by calcula ing
me ics consis en wi h he unde lying epidemiological
s udies we use o es ima e p ema u e mo ali y. Fo PM2.5,
his is he simple annual a e age concen a ion (K ewski e al
2009). Fo ozone, his is he 6-mon h ozone season a e age
o he 1-h daily maximum ozone concen a ion (Je e e al
2009); we es ima e he ozone season in each g id cell as he
consecu i e 6-mon h pe iod wi h highes a e age 1-h daily
maximum ozone. Model esul s o hese wo me ics a e hen
eg idded om each model’s na i e g id esolu ion ( a ying
om 1.9◦×1.2◦ o 5◦×5◦) o a common 0.5◦×0.5◦
esolu ion used o es ima e mo ali y. Fo ozone, 5 o he 14
models epo only mon hly a e age ozone concen a ions;
we calcula e he a e age a io o he 6-mon h 1-h maximum
ozone o he annual a e age o he emaining 9 models and
apply his a io o hese 5 models. Fo PM2.5, 6 models epo
esul s o PM2.5species, and 4 o hese models also epo
a PM2.5me ic, es ima ed by each model as a sum o species
using a o mula unique o ha model. Fo all 6 models, we
es ima e o al PM2.5as a sum o species using a common
o mula (see suppo ing in o ma ion a ailable a s acks.iop.
o g/ERL/8/034005/mmedia), and as a sensi i i y analysis, we
es ima e mo ali y using he PM2.5 epo ed by 4 models.
Fo he bu den o disease esul s, mo ali y is es ima ed
o each model based on he change in concen a ion be ween
he 2000 and 1850 simula ions. This app oach models
an h opogenic ai pollu ion as a esul o bo h an h opogenic
ai pollu an emissions and pas clima e change, in con as
o Anenbe g e al (2010) who did no include pas clima e
change. Fo mo ali y due o pas clima e change, we
use he change in concen a ion be ween he 2000 and
Em2000Cl1850 simula ions.
2.2. Heal h impac assessmen
Mo ali y due o long- e m exposu e o ai pollu ion is
es ima ed ollowing he me hods o Anenbe g e al (2010),
wi h upda ed inpu da a. Like Anenbe g e al (2010),
we es ima e an h opogenic ai pollu ion as a modeled
change in concen a ion be ween he p esen -day and
p eindus ial, a he han e alua ing mo ali y ela i e o a
coun e ac ual low concen a ion (no mally a single alue
ep esen ing unpollu ed condi ions o below which changes in
concen a ion a e assumed no o a ec mo ali y, e.g., Cohen
e al 2004). We use epidemiological concen a ion– esponse
unc ions (CRFs, see suppo ing in o ma ion a ailable a
s acks.iop.o g/ERL/8/034005/mmedia) o ch onic mo ali y
om he Ame ican Cance Socie y (ACS) s udy o PM2.5
ca diopulmona y disease (CPD) and lung cance (LC)
mo ali y (K ewski e al 2009), and o ozone espi a o y
mo ali y (Je e e al 2009). We selec CRFs om he ACS
s udy because his coho includes he la ges popula ion o he
a ailable long- e m PM2.5s udies (Hoek e al 2002, Lepeule
e al 2012), and i is he only s udy ha epo s long- e m
ozone mo ali y (Je e e al 2009). By analyzing PM2.5and
ozone mo ali y based on he same s udy, we achie e g ea e
consis ency and educe he po en ial o double-coun ing o
mo ali y om bo h pollu an s. Rela i e isks om he ACS
s udy di e om o he coho s udies because o di e ences
in popula ion cha ac e is ics, pollu an concen a ions, and
epidemiological me hods. CRFs om he US a e applied
globally, as a ailable s udies o he e ec s o ozone and
3
En i on. Res. Le . 8(2013) 034005 R A Sil a e al
Figu e 1. Es ima es o he cu en global bu den o an h opogenic
ozone (2000–1850) on espi a o y mo ali y om 14 models and he
mul i-model a e age, wi hou and wi h a low-concen a ion
h eshold (33.3 ppb). Unce ain y o indi idual models e lec s only
he 95% con idence in e als on he CRF (Je e e al 2009).
Unce ain y o he mul i-model a e age is a 95% CI including
unce ain y in he CRF and ac oss models. See suppo ing
in o ma ion ( able S1 a ailable a s acks.iop.o g/ERL/8/034005/
mmedia) o summa y in o ma ion on each model.
PM2.5on mo ali y ou side o he US a e b oadly consis en
(Hoek e al 2002, HEI 2004,2010), and CRFs a e no
s ongly dependen on sex, age, o ace (K ewski e al 2009,
Je e e al 2009). None heless, di e ences in popula ion
exposu e (including pollu an concen a ions, he composi ion
o PM2.5and ai pollu an mix u es, and ac i i y pa e ns) and
suscep ibili y (including unde lying heal h s a us) may cause
di e ences in esponses o ai pollu ion in e na ionally.
No low-concen a ion h esholds a e assumed, as he e is
no clea e idence o he p esence o h esholds. We analyze
he sensi i i y o he esul s o low-concen a ion h esholds o
33.3 ppb o ozone and 5.8µg m−3 o PM2.5, below which
changes in concen a ion a e assumed o ha e no e ec , as
hese a e he lowes measu ed le els in ACS.
Consis en wi h ACS, we limi ou analysis o adul s
aged 30 and olde (see suppo ing in o ma ion, able S10 and
igu e S6 a ailable a s acks.iop.o g/ERL/8/034005/mmedia).
Popula ion da a comes om LandScan (Dobson e al 2000)
o he yea 2008 a app oxima ely 1 km2 esolu ion, which is
hen eg idded o 0.5◦×0.5◦. The ac ion o he popula ion
aged 30 and olde in he yea 2008 is aken om UN s a is ics
o indi idual coun ies. No e ha p esen -day popula ion is
used in all cases, so ha we e alua e he e ec o 2000 ai
pollu ion ela i e o 1850 o ela i e o he Em2000Cl1850
simula ions, o p esen -day mo ali y. Baseline mo ali y
a es (also o 30 and olde ) a e om he WHO o indi idual
coun ies, using he mos ecen da a a ailable o each
coun y be ween 2000 and 2008, and when una ailable,
epo ed egional a es a e used (see suppo ing in o ma ion).
Figu e 2. Es ima es o he cu en global bu den o an h opogenic
PM2.5(2000–1850) on CPD and LC mo ali y wi h no
low-concen a ion h eshold, o PM2.5calcula ed as a sum o
species o 6 models (da k blue), and o PM2.5as epo ed by 4
models (da k g een). The co esponding es ima es wi h a
low-concen a ion h eshold (5.8µg m−3) a e shown o PM2.5
calcula ed as a sum o species (ligh blue), and o epo ed PM2.5
(ligh g een). Unce ain y o indi idual models e lec s only he
95% con idence in e als on he CRF (K ewski e al 2009).
Unce ain y o he mul i-model a e age is a 95% CI including
unce ain y in he CRF and ac oss models.
Baseline mo ali y a es o indi idual coun ies we e g idded
o he 0.5◦g id using A cGIS10 geop ocessing ools.
3. Global mo ali y bu den o an h opogenic ai
pollu ion
Figu es 1and 2show es ima es o p ema u e mo ali y due o
an h opogenic ozone and PM2.5 o each model, and changes
in concen a ion unde lying hese es ima es a e p esen ed
in he suppo ing in o ma ion (a ailable a s acks.iop.o g/
ERL/8/034005/mmedia). The a e age es ima e ac oss he 14
models sugges s ha 470 000 p ema u e espi a o y dea hs
occu globally and annually due o an h opogenic inc eases in
ozone, wi h no low-concen a ion h eshold. Accoun ing o
bo h he 95% con idence in e al (CI) on he CRF, epo ed
by Je e e al (2009), and he dis ibu ion o esul s om
he 14 models, using Mon e Ca lo sampling, yields a 95% CI
o 140 000 o 900 000 (unce ain y anges epo ed he ea e
ollow he same me hods). Global ozone mo ali y is abou
20% lowe when a low-concen a ion h eshold is used. In
igu e 3and able 1, ozone- ela ed mo ali y is widesp ead
globally, as ozone has inc eased essen ially e e ywhe e om
human ac i i ies, bu is g ea es in highly popula ed and highly
pollu ed a eas o India and Eas Asia, which accoun o 68%
o he global o al.
Fo PM2.5es ima ed as a sum o species, he 6-model
a e age indica es ha 2.1 (1.3 o 3.0) million p ema u e
CPD and LC dea hs occu globally and annually due o
an h opogenic inc eases, wi h no low-concen a ion h eshold.
4
En i on. Res. Le . 8(2013) 034005 R A Sil a e al
Figu e 3. Cu en p ema u e mo ali y due o an h opogenic ai pollu ion (2000–1850), in dea hs y −1(1000 km2)−1, o he mul i-model
mean in each g id cell, o ( op) ozone ( espi a o y mo ali y) o 14 models and (bo om) PM2.5(CPD +LC) o he sum o species o 6
models.
O hese dea hs, 93% a e ela ed o CPD and 7% o LC.
Rela i e o ozone, he e is less sca e among he models, wi h
a coe icien o a ia ion (σ/µ) among models o 0.10 o
PM2.5, compa ed o 0.26 o ozone. Fo bo h PM2.5and ozone,
he unce ain y in he CRF is g ea e han he unce ain y o e
he ange o models. Global PM2.5mo ali y is 11% lowe
o he mul i-model a e age when using a low-concen a ion
h eshold o 5.8µg m−3, and is 19% lowe when using
PM2.5as epo ed by 4 models. While he o mulas o
es ima ing PM2.5di e be ween models, he la ge change
in concen a ions when adding species is mainly due o he
omission o ni a e in he PM2.5 epo ed by he models.
La ge di e ences may also esul om di e ences in how
dus and sea sal a e added o PM2.5, as models ha calcula e
PM2.5use size- esol ed in o ma ion and so a e likely mo e
accu a e han he common o mula used he e. PM2.5- ela ed
mo ali y is widesp ead in popula ed egions, p incipally in
Eas Asia and India, bu also in Sou heas Asia, Eu ope,
and he Fo me So ie Union. Howe e , some loca ions a e
modeled as ha ing a dec ease in PM2.5 ela i e o 1850,
including he sou heas US and pa s o La in Ame ica, and
small egions elsewhe e. In he sou heas US, his dec ease is
caused by educ ions in biomass bu ning ela i e o 1850, as
changes in p ima y o ganic ca bon a e p ima ily esponsible
o he dec ease, which also is appa en in he adia i e o cing
due o biomass bu ning ae osols (Shindell e al 2013). Local
dec eases in India and A ica likely ela e o pas clima e
change (see suppo ing in o ma ion a ailable a s acks.iop.
o g/ERL/8/034005/mmedia).
5
En i on. Res. Le . 8(2013) 034005 R A Sil a e al
Table 1. Regional p ema u e annual dea hs om an h opogenic ou doo ai pollu ion (2000–1850), o ozone ( espi a o y) and PM2.5
calcula ed as a sum o species (CPD +LC), shown o he mean and ull ange ac oss 14 models o ozone and 6 models o PM2.5(3
signi ican igu es shown). Also shown a e dea hs pe million people (p esen -day exposed popula ion aged 30 and o e ) in each egion, in
pa en hesis.
Region
Ozone PM2.5
Mean Low High Mean Low High
No h Ame ica 34 400 12 300 52 200 43 000 12 200 77 000
(121) (44) (184) (152) (43) (272)
Eu ope 32 800 13 700 46 200 154 000 105 000 193 000
(96) (40) (135) (448) (306) (562)
Fo me So ie Union 10 600 5180 14 600 128 000 91 000 168 000
(66) (32) (91) (793) (568) (1044)
Middle Eas 16 200 10 300 22 100 88 700 80 900 95 100
(68) (43) (93) (371) (339) (398)
India 118 000 76 800 208 000 397 000 205 000 549 000
(212) (138) (376) (715) (370) (989)
Eas Asia 203 000 62 900 311 000 1049 000 908 000 1240 000
(230) (71) (353) (1191) (1031) (1406)
Sou heas Asia 33 300 20 900 49 300 158 000 118 000 187 000
(119) (75) (176) (564) (422) (669)
Sou h Ame ica 6970 5180 8950 16 800 11 900 24 900
(38) (28) (49) (92) (65) (137)
A ica 17 300 14 400 19 900 77 500 65 400 91 100
(73) (61) (84) (327) (276) (385)
Aus alia 469 273 698 1250 911 2350
(29) (17) (44) (78) (57) (147)
Global 472 000 229 000 720 000 2110 000 1880 000 2380 000
(149) (72) (227) (665) (590) (748)
These es ima es o he global bu den a e smalle han
hose epo ed by Anenbe g e al (2010). Since Anenbe g
e al (2010) used he same CRFs and only small di e ences
in global popula ion and baseline mo ali y a es, he lesse
es ima ed mo ali y is mainly due o di e ences in modeled
concen a ions. While he model used in ha s udy di e s
om he ensemble used he e, he g ea e di e ence is likely
o be he di e en emissions used o bo h he p esen -day
and p eindus ial simula ions (Lama que e al 2010, Fang e al
2013).
These global bu den es ima es a e also g ea e o ozone
bu less o PM2.5 han we e es ima ed in he mos ecen
Global Bu den o Disease s udy (Lim e al 2012). Fo ozone,
hese di e ences a e likely explained by he ac ha modeled
1850 ozone ( able S5 a ailable a s acks.iop.o g/ERL/8/
034005/mmedia) is lowe han he assumed coun e ac ual
low concen a ion o Lim e al (2012) o 37.6 ppb. Fo
PM2.5, he modeled 1850 concen a ions a e close o he
coun e ac ual concen a ions used by Lim e al (2012) o
7.3µg m−3; he smalle es ima e he e may be due o
di e ences in CRFs.
Ou es ima es using esul s om he GFDL-AM3
simula ions o Fang e al (2013) a e 45% highe o ozone
mo ali y and 24% highe o PM2.5mo ali y han hose
epo ed by Fang e al (2013); his di e ence is accoun ed
o mainly by he smalle global popula ion aged 30 yea s
and olde o he yea 2000 used in ha s udy and, o a lesse
ex en , di e ences in baseline mo ali y a es.
Figu e 4. Es ima es o he cu en global ozone espi a o y
mo ali y a ibu ed o pas clima e change (2000–Em2000Cl1850),
o 9 models and he mul i-model a e age, wi h and wi hou a
low-concen a ion h eshold (33.3 ppb). Unce ain y o indi idual
models e lec s only he 95% con idence in e als on he CRF
(Je e e al 2009). Unce ain y o he mul i-model a e age is a
95% CI including unce ain y in he CRF and ac oss models.
4. Ai pollu ion mo ali y a ibu ed o pas clima e
change
The 9-model a e age es ima e o he e ec o pas clima e
change on ozone espi a o y mo ali y is 1500 (−20 000 o
27 000) dea hs annually wi h no h eshold ( igu e 4). The e
is la ge a iabili y among models, wi h six o nine models
6
En i on. Res. Le . 8(2013) 034005 R A Sil a e al
Figu e 5. P ema u e mo ali y a ibu able o pas clima e change (2000–Em2000Cl1850), in dea hs y −1(1000 km2)−1, o he
mul i-model mean in each g id cell, o ( op) ozone ( espi a o y mo ali y) o 9 models and (bo om) PM2.5(CPD +LC mo ali y) o he
sum o species o 5 models.
sugges ing ha pas clima e change caused ozone mo ali y
o inc ease. In igu e 5and able 2, dea hs a e g ea es in Eas
Asia o he mul i-model a e age, bu also posi i e in No h
Ame ica and pa s o India. In igu e 6, mos models p edic
ozone dec eases due o clima e change in opical egions
and o e oceans. This likely esul s om inc eases in wa e
apo , which causes g ea e p oduc ion o HOx adicals and
g ea e des uc ion o ozone. O e pollu ed egions, howe e ,
ozone inc eases om as e eac ion a es and me eo ological
changes (Jacob and Winne 2009). Because mos models
epo ed se e al yea s o simula ions, he a iabili y be ween
models is no likely a esul o in e -annual me eo ological
a iabili y.
Fo PM2.5, he 5-model a e age mo ali y (CPD +LC)
a ibu ed o pas clima e change is 2200 (−350 000 o
140 000) dea hs annually, wi h no h eshold and es ima ing
PM2.5as a sum o species ( igu e 7). Fou o he i e
models es ima e an inc ease in dea hs, bu he a e age is
dec eased by one model (HadGEM2) ha es ima es −283 000
dea hs om PM2.5dec eases due o clima e change. The
mul i-model median mo ali y is 61 000 dea hs annually, and,
i HadGEM2 is excluded, he mul i-model a e age is 74 000
(30 000 o 140 000) dea hs y −1. A e age mo ali y is highe
when using PM2.5 om he ou models ha epo ed i , and
wi hou he la ge nega i e unce ain y, as HadGEM2 did no
epo PM2.5. In igu e 5and able 2, he 5-model a e age
shows ha pas clima e change caused he la ges inc eases in
7
En i on. Res. Le . 8(2013) 034005 R A Sil a e al
Table 2. Regional p ema u e annual dea hs a ibu able o pas clima e change (2000–Em2000Cl1850), o ozone ( espi a o y) and PM2.5
calcula ed as a sum o species (CPD +LC), shown o he mean and ull ange ac oss 9 models o ozone and 5 models o PM2.5(3
signi ican igu es shown). Also shown a e dea hs pe million people (p esen -day exposed popula ion aged 30 and o e ) in each egion, in
pa en hesis.
Region
Ozone PM2.5
Mean Low High Mean Low High
No h Ame ica 621 −1110 2360 3700 −6560 18 800
(2) (−4) (8) (13) (−23) (67)
Eu ope −541 −1520 774 583 −27 100 10 700
(−2) (−4) (2) (2) (−79) (31)
Fo me So ie Union −74 −674 489 2090 −16 500 9570
(0) (−4) (3) (13) (−102) (59)
Middle Eas −90 −851 377 136 −6410 12 300
(0) (−4) (2) (1) (−27) (51)
India 871 −10 700 11 000 −27 700 −248 000 59 400
(2) (−19) (20) (−50) (−447) (107)
Eas Asia 1490 −5720 11 500 23 700 −32 500 112 000
(2) (−6) (13) (27) (−37) (128)
Sou heas Asia 290 −852 1730 3300 −7620 8330
(1) (−3) (6) (12) (−27) (30)
Sou h Ame ica −215 −694 260 1000 495 2390
(−1) (−4) (1) (6) (3) (13)
A ica −794 −2930 301 −4790 −43 000 16 200
(−3) (−12) (1) (−20) (−181) (68)
Aus alia −15 −78 25 193 39 520
(−1) (−5) (2) (12) (2) (32)
Global 1540 −15 700 21 400 2200 −283 000 111 000
(0) (−5) (7) (1) (89) (35)
PM2.5p ema u e mo ali y in Eas Asia, and no able inc eases
elsewhe e including No h Ame ica, bu dec eases in India
and pa s o A ica, he Fo me So ie Union and Eu ope.
The s ong nega i e mo ali y es ima e om HadGEM2 is
he esul o PM2.5dec eases o e India, d i en by changes
in dus , as India has a la ge exposed popula ion. Figu e 6
shows ha mos models p edic an inc ease in PM2.5o e
India due o clima e change, wi h only HadGEM2 being a
s ong excep ion. Mos models p edic inc eases in PM2.5o e
land, bu i is di icul o explain he di e en egional pa e ns
o concen a ion changes due o pas clima e change ac oss
he di e en models. These in e -model di e ences o bo h
ozone and PM2.5a e likely d i en by he p ocesses included
in he di e en models, such as whe he and how emissions
om dus , ege a ion, and ligh ning a e modi ied as a esul o
clima e change, and di e ences in how he models ep esen
pas clima e change and i s in luences on pho ochemis y and
pollu an anspo and emo al.
5. Conclusions
We es ima e ha in he p esen -day, an h opogenic changes
o ai pollu an concen a ions since he p eindus ial a e
associa ed annually wi h 470 000 (95% CI, 140 000 o
900 000) p ema u e espi a o y dea hs ela ed o ozone, and
2.1 (1.3 o 3.0) million CPD and LC dea hs ela ed o PM2.5.
Ou es ima es di e om hose o Lim e al (2012) in ha
we es ima e mo ali y o changes in ai pollu ion ela i e
o he modeled p eindus ial condi ions, a he han using a
coun e ac ual low concen a ion. Rela i e o Anenbe g e al
(2010), ou esul s also di e mainly because o he di e en
emissions used in he models o p eindus ial and p esen -day
condi ions, and by using modeled concen a ions om an
ensemble o models a he han a single model.
The e is signi ican a iabili y in mo ali y es ima es
d i en by di e en a mosphe ic models, e en hough
hese models used e y simila an h opogenic emissions,
highligh ing he unce ain y in basing esul s on a single
model. Va iabili y among models is highe o ozone han o
PM2.5, bu o bo h pollu an s, i con ibu es less o o e all
unce ain y han he unce ain y in CRFs. The unce ain y in
CRFs is unde s a ed because i does no accoun o he ull
ange o e he li e a u e—e.g., use o he CRF o PM2.5 om
Lepeule e al (2012) would lead o highe mo ali y es ima es.
The ela i e magni ude o esul s using di e en CRFs and
wi h low-concen a ion h esholds, analyzed by Anenbe g
e al (2010), would also apply he e. As o p e ious s udies
ha es ima e he mo ali y bu den o ou doo ai pollu ion,
ou me hods likely unde es ima e he ue bu den because
we ha e limi ed he e alua ion o adul s aged 30 and olde ,
and base he analysis on coa se- esolu ion models ha likely
unde es ima e exposu e, pa icula ly o PM2.5in u ban a eas
(Punge and Wes 2013). On he o he hand, ecen s udies
sugges ha he ela ionship be ween PM2.5and mo ali y may
la en a high concen a ions (Pope e al 2011), sugges ing
ha we may o e es ima e PM2.5mo ali y in egions wi h
8
