Consistency Between Satellite-Derived and Modeled Estimates of the Direct Aerosol Effect

Abstract

In the Intergovernmental Panel on Climate Change Fourth Assessment Report, the direct aerosol effect is reported to have a
radiative forcing estimate of –0.5 Watt per square meter (W m−2), offsetting the warming from CO2 by almost one-third. The uncertainty, however, ranges from –0.9 to –0.1 W m−2, which is largely due to differences between estimates from global aerosol models and observation-based estimates, with the
latter tending to have stronger (more negative) radiative forcing. This study demonstrates consistency between a global aerosol
model and adjustment to an observation-based method, producing a global and annual mean radiative forcing that is weaker than
–0.5 W m−2, with a best estimate of –0.3 W m−2. The physical explanation for the earlier discrepancy is that the relative increase in anthropogenic black carbon (absorbing
aerosols) is much larger than the overall increase in the anthropogenic abundance of aerosols.

Full text

DOI: 10.1126/science.1174461
, 187 (2009); 325Science et al.Gunnar Myhre,
Estimates of the Direct Aerosol Effect
Consistency Between Satellite-Derived and Modeled
www.sciencemag.org (this information is current as of July 21, 2009 ):
The following resources related to this article are available online at
http://www.sciencemag.org/cgi/content/full/325/5937/187
version of this article at: including high-resolution figures, can be found in the onlineUpdated information and services,
http://www.sciencemag.org/cgi/content/full/1174461/DC1
can be found at: Supporting Online Material
found at: can berelated to this articleA list of selected additional articles on the Science Web sites
http://www.sciencemag.org/cgi/content/full/325/5937/187#related-content
http://www.sciencemag.org/cgi/content/full/325/5937/187#otherarticles
, 3 of which can be accessed for free: cites 23 articlesThis article
http://www.sciencemag.org/cgi/content/full/325/5937/187#otherarticles
1 articles hosted by HighWire Press; see: cited byThis article has been
http://www.sciencemag.org/cgi/collection/atmos
Atmospheric Science : subject collectionsThis article appears in the following
http://www.sciencemag.org/about/permissions.dtl
in whole or in part can be found at: this article permission to reproduce of this article or about obtaining reprintsInformation about obtaining
registered trademark of AAAS. is aScience2009 by the American Association for the Advancement of Science; all rights reserved. The title CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
on July 21, 2009 www.sciencemag.orgDownloaded from

Desulfobulbus (36)ore-Proteobacteria, could be
the underlying process observed, indirectly link-
ing AOM to metal reduction. Although the shift
in the bacterial community from known sulfate-
reducing bacteria to putative metal-reducing micro-
organisms in the birnessite incubations supports
the idea that the AOM is directly linked to metal
reduction, the observed shift in microbial com-
munity could also be a result of the stimulation of
heterotrophic metal reduction. If metal reduction
is indirectly linked to AOM in marine sediments,
then the realized energy gain for the microorga-
nisms directly catalyzing AOM would be much
lower than that suggested in Table 1. Regardless
of mechanism, the stimulation of AOM with Mn
and Fe has important implications for capacity of
CH
4
oxidation.
It is estimated that AOM consumes most
methane released in marine settings, equaling 5
to 20% of today’s total global methane flux (37),
making this process an important part of the
global carbon cycle today. However, before Earth
became oxygenated, growth of methanotrophs
was limited by their ability to find electron ac-
ceptors. Based on the column-integrated photo-
oxidation rates of 5 mg/cm
2
/year of manganese
and 200 mg/cm
2
/year of iron (38), on the order of
10,000 Tg/year of methane could be oxidized
during this time period by manganese- and iron-
dependent AOM, irrespective of whether the pro-
cesses directly link metal reduction to methane
oxidation. Estimates of the methane flux to the
atmosphere during the Proterozoic are on the
order of 1,000 to 10,000 Tg/year (39), meaning
that manganese- and iron-dependent AOM had
the oxidative potential to oxidize the entire early
Earth methane flux. Thus, manganese- and iron-
dependent AOM could have been extremely im-
portant methane sinks, as well as energy sources,
for the early biosphere.
References and Notes
1. K. F. Ettwig et al., Environ. Microbiol. 10, 3164 (2008).
2. A. Raghoebarsing et al., Nature 440, 918 (2006).
3. A. J. B. Zehnder, T. D. Brock, Appl. Environ. Microbiol.
39, 194 (1980).
4. K. Nauhaus, T. Treude, A. A. Boetius, M. Krüger,
Environ. Microbiol. 7, 98 (2005).
5. S. D'Hondt et al., Science 306, 2216 (2004).
6. S. B. Joye et al., Chem. Geol. 205, 219 (2004).
7. J. R. Hein, R. A. Koski, Geology 15, 722 (1987).
8. J. Reitner, J. Peckmann, A. Reimer, G. Schumann,
V. Thiel, Facies 51, 66 (2005).
9. S. W. Poulton, R. Raiswell, Mar. Chem. 72, 17 (2000).
10. J.-M. Martin, M. Meybeck, Mar. Chem. 7, 173 (1979).
11. There are minor contributions of manganese and iron
from hydrothermal systems and aeolian input. See SOM
text for further discussion.
12. R. A. Wheatcroft, C. K. Sommerfield, D. E. Drake,
J. C. Borgeld, C. A. Nittrouer, Geology 25, 163 (1997).
13. D. E. Canfield, B. Thamdrup, J. W. Hansen, Geochim.
Cosmochim. Acta 57, 3867 (1993).
14. Materials and methods are available as supporting
material on Science Online.
15. J. J. Moran, C. H. House, K. H. Freeman, J. G. Ferry,
Archaea 1, 303 (2005).
16. J. J. Moran, C. H. House, B. Thomas, K. H. Freeman,
J. Geophys. Res. 112, G02011 (2007).
17. See relevant SOM text for discussion regarding dissolved
manganese and iron concentrations.
18. B. Thamdrup, R. Rosselló-Mora, R. Amann, Appl. Environ.
Microbiol. 66, 2888 (2000).
19. V.J.Orphan,C.H.House,K.U.Hinrichs,K.D.McKeegen,
E. F. DeLong, Science 293, 484 (2001).
20. K. Knittel, T. Loesekann, A. Boetius, R. Kort, R. Amann,
Appl. Environ. Microbiol. 71, 467 (2005).
21. V. J. Orphan et al., Appl. Environ. Microbiol. 67, 1922 (2001).
22. V.J.Orphan,C.H.House,K.-U.Hinrichs,K.D.McKeegan,
E. F. DeLong, Proc. Natl. Acad. Sci. U.S.A. 99, 7663 (2002).
23. H. Niemann et al., Nature 443, 854 (2006).
24. A. Pernthaler et al., Proc. Natl. Acad. Sci. U.S.A. 105,
7052 (2008).
25. K. H. Nealson, D. Saffarini, Annu. Rev. Microbiol. 48,
311 (1994).
26. B. M. Tebo, A. Y. Obraztsova, FEMS Microbiol. Lett. 162,
193 (1998).
27. M. L. Coleman, D. B. Hedrick, D. R. Lovley, D. C. White,
K. Pye, Nature 361, 436 (1993).
28. D. E. Holmes et al., Microb. Ecol. 48, 178 (2004).
29. J. Reitner et al., Palaeogeogr. Palaeoclimatol.
Palaeoecol. 227, 18 (2005).
30. D. C. Gillan, P. Pernet, Biofouling 23, 1 (2007).
31. F. Gremion, A. Chatzinotas, H. Harms, Environ. Microbiol.
5, 896 (2003).
32. R. J. Ellis, P. Morgan, A. J. Wheightman, J. C. Fry,
Appl. Environ. Microbiol. 69, 3223 (2003).
33. J. He, Geomicrobiol. J. 25, 14 (2008).
34. R. C. Aller, P. D. Rude, Geochim. Cosmochim. Acta 52,
751 (1988).
35. D. J. Burdige, K. H. Nealson, Geomicrobiol. J. 4, 361 (1986).
36. D. R. Lovley, E. J. P. Phillips, Appl. Environ. Microbiol. 60,
2394 (1994).
37. D. L. Valentine, W. S. Reeburgh, Environ. Microbiol. 2,
477 (2000).
38. A. D. Anbar, H. D. Holland, Geochim. Cosmochim. Acta
56, 2595 (1992).
39. A. A. Pavlov, M. T. Hurtgen, J. F. Kasting, M. A. Arthur,
Geology 31, 87 (2003).
40. We would like to thank M. Arthur for the use of his
mass spectrometer, Z. Zhang and S. Goffredi for
laboratory assistance, D. Walizer for technical assistance,
and D. Jones for help with phylogenetics. We also thank
the shipboard scientists, crew, and pilots of R/V Atlantis
and R/V Western Flyer. Funding for this project has come
from the National Science Foundation (MCB-0348492),
National Aeronautics and Space Administration (NASA)
Astrobiology Institute under NASA-Ames Cooperative
Agreement NNA04CC06A, and the Penn State
Biogeochemical Research Initiative for Education (BRIE)
funded by NSF (IGERT) grant DGE-9972759. Sequences
were submitted to GenBank and have accession numbers
FJ264513 to FJ264602 and FJ264604 to FJ264884.
Supporting Online Material
www.sciencemag.org/cgi/content/full/325/5937/184/DC1
Materials and Methods
SOM Text
Figs. S1 to S3
References
18 December 2008; accepted 2 June 2009
10.1126/science.1169984
Consistency Between Satellite-Derived
and Modeled Estimates of the
Direct Aerosol Effect
Gunnar Myhre
In the Intergovernmental Panel on Climate Change Fourth Assessment Report, the direct aerosol
effect is reported to have a radiative forcing estimate of –0.5 watt per square meter (W m
−2
),
offsetting the warming from CO
2
by almost one-third. The uncertainty, however, ranges from
–0.9 to –0.1 W m
−2
, which is largely due to differences between estimates from global aerosol
models and observation-based estimates, with the latter tending to have stronger (more negative)
radiative forcing. This study demonstrates consistency between a global aerosol model and
adjustment to an observation-based method, producing a global and annual mean radiative forcing
that is weaker than –0.5 W m
−2
, with a best estimate of –0.3 W m
−2
. The physical explanation
for the earlier discrepancy is that the relative increase in anthropogenic black carbon (absorbing
aerosols) is much larger than the overall increase in the anthropogenic abundance of aerosols.
The complex influence of atmospheric aero-
sols on the climate system and the influ-
ence of humans on aerosols are among
the key uncertainties in the understanding of cur-
rent climate change (1–3). The direct aerosol ef-
fectmayhavecontributedtoacoolinginthemid
20th century (4)andmayhavemaskedacon-
siderable degree of current global warming (5),
potentially leading to more rapid warming in the
future because of stricter controls on aerosol emis-
sions (5). In addition, the direct aerosol effect
is probably responsible for a substantial part of
the observed dimming and the later reversal to
brightening (6) at many locations. The direct aero-
sol effect includes both scattering and absorp-
tion of solar light, and evaluating its magnitude
is complicated by the fact that some atmospheric
aerosols are predominately scattering, whereas
others are mainly absorbing. The main anthropo-
genic scattering components are sulfate, nitrate,
and organic carbon (OC), whereas black carbon
(BC) absorbs solar radiation.
Previously, two approaches have been em-
ployed in the calculation of the total direct aero-
sol effect. The first approach uses global aerosol
models to simulate the change in the aerosol abun-
dance that is attributable to human activities and
radiative transfer models to calculate the radia-
tive forcing on the basis of simulations for pre-
Center for International Climate and Environmental Research–
Oslo (CICERO), Post Office Box 1129 Blindern, N-0318 Oslo,
Norway. E-mail: gunnar.myhre@cicero.uio.no
www.sciencemag.org SCIENCE VOL 325 10 JULY 2009 187
REPORTS
on July 21, 2009 www.sciencemag.orgDownloaded from

industrial and present time (7). It has been known
for some time that aerosol optical properties
(especially the single scattering albedo) and the
location of aerosols in relation to clouds strongly
affect the radiative forcing of the direct aerosol
effect (8–11). In the second approach, observa-
tional data have been used with some minor in-
puts from global aerosol models (12–15). The
observation-based method has used satellite
data of aerosol optical depth (AOD) in combi-
nation with either aerosol optical properties from
ground-based sunphotometers from AERONET
(Aerosol Robotic Network) (12,16) or observation-
derived aerosol radiative efficiency (14,15). In
the observation-based method, satellite retrieval
constraints over land necessitate the use of in-
formation from global aerosol models in order to
derive the change in AOD that is due to an-
thropogenic activity. The aerosol optical proper-
ties (in this respect, single scattering albedo and
asymmetry factor) have been held constant be-
tween preindustrial and present time, and ra-
diative forcing calculations have been performed
with radiative transfer schemes (12,16)orthe
aerosol radiative efficiency has been assumed to
be constant over the industrial era (14,15).
Recent analysis performed with the observation-
based method by use of updated AODs from
new satellite-retrieved MODIS (Moderate Reso-
lution Imaging Spectroradiometer) data (collec-
tion 5 rather than collection 4) show a weakened
radiative forcing, from –0.8 to –0.65 W m
−2
.In
addition, differences in observational coverage
and the assumption that no anthropogenic aerosol
existed before industrial times [defined as before
1750 (1)] could also explain some of the differ-
ence between a global aerosol model and the
observational method (16). In the observation-
based method of Bellouin et al.(12,16), it is
also assumed that there is no radiative effect of
the aerosols within cloudy-sky areas, whereas
models diagnose differing signs and magnitudes
of the cloudy-sky radiative forcing (7). Although
the study of Bellouin et al.(16) contributes to an
improved understanding of differences between
model- and observation-based estimates of the
direct aerosol effect, large differences remain and
obstruct a better quantification of this important
effect counteracting global warming.
In this study, I used both the estimates from
a global aerosol model and from an observation-
AB
Fig. 1. Annual mean anthropogenic AOD from (A) the observation-based
method and (B) the global aerosol model Oslo CTM2. The anthropogenic
AOD in the observation-based method is derived from the total AOD from
MODIS (shown for MODIS onboard Aqua) combined with the anthropogenic
fraction of all types of aerosols from Oslo CTM2. Regions with missing MODIS
data because of bright surfaces are shown in white. The aerosol species
included in the derivation of modeled anthropogenic AOD are sulfate, nitrate,
BC, and OC (primary and secondary).
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
MODIS (Screen)
MODI
S
MODIS (Model)
MODEL EXT
MODEL INT
MODEL (SA T)
MODEL (SAT & OPT OBS)
all sky cloud mask
(Oa)
(Ob)
(Oc)
(Mc
ext
)
(Mc
int
)
(Mb’)
(Mb)
Radiative forcing (Wm-2)
Fig. 2. Global and annual mean radiative forcing of the direct aerosol effect for observation-based
method (O) and the global aerosol model Oslo CTM2 (M). The results for cloud mask do not take
into account radiative effects of aerosols in cloudy regions, whereas all sky takes into account the
direct aerosol effect both in cloudy and clear-sky conditions. The results denoted MODIS are from
the observation-based method as described in (21). All the results for MODIS represent the mean of
two calculations with satellite data from MODIS onboard Aqua and Terra. MODIS (Screen) (Oa) do
not take into account the direct aerosol effect in regions of missing MODIS data, whereas MODIS
(Ob) uses model information in regions with missing MODIS data. MODIS (Model) (Oc) uses
modeled aerosol optical properties and its anthropogenic change. The model versions MODEL EXT
(Mc
ext
) and MODEL INT (Mc
int
) represents two versions of the model, one with BC externally mixed
and the other with hydrophilic BC internally mixed, respectively (21). MODEL (SAT) (Mb′)usethe
observation-based approach with constant aerosol optical properties; otherwise, it is the same as
MODEL Mc
int
. MODEL (SAT & OBS OPT) (Mb) is the same as in the previous case, except that observed
aerosol optical properties are used.
10 JULY 2009 VOL 325 SCIENCE www.sciencemag.org
188
REPORTS
on July 21, 2009 www.sciencemag.orgDownloaded from

based method in order to achieve consistency
between these two methods, an important step
toward quantification of the direct aerosol effect
with the use of two independent approaches. The
global aerosol model Oslo CTM2 is a state-of-the-
art model that includes all the main atmospheric
aerosol components (17,18). The observation-
based method is based on satellite-retrieved AOD
from MODIS (19) and aerosol optical properties
from AERONET (20,21). Figure 1 shows that
the anthropogenic AOD is quite similar in the
observation-based method and the Oslo CTM2
model, with some regional differences. The stan-
dard MODIS aerosol retrieval is unable to retrieve
data over surfaces with high reflectance. In Fig. 1,
this can be seen over desert areas and snow- and
ice-covered regions. The global and annual mean
anthropogenic AOD is quite similar in the two
approaches (21).
Because of the importance of clouds in the
calculations of the direct aerosol effect, the ra-
diative forcing is calculated for two cases: “cloud
mask,”in which the aerosol direct effect is as-
sumed to be zero in the cloudy regions, such as
in recent observation-based estimates (12,16),
and “all sky,”in which the radiative forcing in
clear and cloudy regions is taken into account.
The latter case is used as the radiative forcing in
(1) so as to maintain consistency with the radia-
tive forcing calculated for other forcing mech-
anisms. The radiative forcing is calculated as the
change in the solar irradiance at the top of the
atmosphere. Figure 2 and Table 1 show radiative
forcing from the observation-based method (O
indicates the observational estimates) and the
model (M indicates the model estimates) for the
cloud-mask and all-sky cases for the simulations
(a) to (c), as described below. Using modeled
aerosol fields in regions where MODIS is not
able to retrieve aerosol information reduces the
observational estimate of the cloud-mask forc-
ing from ~–0.6 W m
−2
(Oa) to –0.5 W m
−2
(Ob).
Furthermore, changing the preindustrial aerosol
optical properties in the observation-based meth-
od so that they match the less absorbing pre-
industrial aerosols from the model (21)reduces
the cloud-mask forcing to slightly weaker than
–0.4 W m
−2
(Oc). Thus, the observation-based
method provides cloud-masked results (Oc) that
are now statistically indistinguishable from the
model results (Mc
ext
and Mc
int
).
Given the modeling framework, it is possible
to also relax the modeling assumptions to those
used in the observational methodology. Fixing the
model aerosol optical properties for preindustrial
Table 1. Radiative forcing (W m
−2
) for the experiments shown in Fig. 2 and explained in the Fig. 2
caption for cloud mask and all sky.
Cloud mask (W m
–2
) All sky (W m
–2
)
MODEL (SAT & OPT OBS) –0.48 –0.63
MODEL (SAT) –0.46 –0.60
MODEL INT –0.35 –0.28
MODEL EXT –0.38 –0.35
MODIS (Model) –0.37 –0.28
MODIS –0.51 –0.62
MODIS (Screen) –0.61 –0.71
0.06 0.4
0.3
0.2
0.1
0.0
1.0
2.0
1.5
1.0
0.5
0.0
0.8
0.6
0.4
0.2
0.0
BC OC
0.05
0.04
0.03
0.02
0.01
0.00
-90 -60 -30 0
Latitude
Present
Pre-industrial
AOD
ABS AOD
SO2
30 60 90
-90 -60 -30 0
Latitude
30 60 90
-90 -60 -30 0
Latitude
30 60 90
-90 -60 -30 0
Latitude
Anthropogenic fraction
30 60 90
Emission (g m-2 yr-1) Emission (g m-2 yr-1)
Emission (g m-2 yr-1)
AB
DC
Fig. 3. Present time and preindustrial emissions of (A)BC,(B) primary OC, (C)SO
2
,and(D) anthropogenic fraction of AOD (550 nm) and absorption AOD
(ABS AOD) (550 nm). The anthropogenic fraction is shown as the ratio of the difference in AOD between 1750 and 2004 due to anthropogenic activity, to the
total AOD (anthropogenic and natural).
www.sciencemag.org SCIENCE VOL 325 10 JULY 2009 189
REPORTS
on July 21, 2009 www.sciencemag.orgDownloaded from

simulation to those modeled for the present time
(Mb′) provides results for cloud-masked forcing
that are very similar to the observation-based re-
sults (Ob). The final model simulation (Mb) uses
aerosol optical properties for both present and
preindustrial times that are fixed to those of the
observations, and the results (Mb) are statistically
indistinguishable from those from observations
(Ob). Thus, it is here demonstrated that the
differences in the cloud-masked forcing between
the model and the observation-based method can
be reconciled by including aerosols in regions
without MODIS retrievals and including change
in the aerosol optical properties from pre-
industrial to present time.
When considering all-sky simulations, the
observation-based (Oa) and model-based (Mc
ext
and Mc
int
) data initially differed by approximately
a factor of two. However, once the direct radia-
tive forcing that was missing from satellite re-
trievals and the modeled change in the aerosol
optical properties between preindustrial and present-
day are accounted for (Oc), the results agree well
with those from the model-based estimates (Mc
ext
and Mc
int
). Whereas the radiative forcing strengthens
between the cloud-mask and all-sky cases for the
observation-based results (Oa and Ob), the forcing
weakens when modeled changes in the aerosol
optical properties are taken into account (Oc). This
is evident also in the model results, in which the
model with internal mixture (Mc
int
) has greater
weakening from cloud-mask to all-sky than does
the external mixing case (Mc
ext
), in which the
latter has a weaker aerosol absorption (21). This
difference in the absorption may explain the dif-
ference in sign of the forcing in cloudy regions that
varies among global aerosol models (7).
That aerosol optical properties, in particular
the absorption, has changed so substantially be-
tween preindustrial and present times deserves
some discussion. The main contributor to atmo-
spheric aerosol absorption are BC aerosols and
BC emissions, which have increased by more than
a factor of six (22,23), whereas emissions of
anthropogenic-influenced scattering aerosol com-
ponents and their precursors have had a smaller
increase with a factor of three to four (23). Figure 3
shows the preindustrial and present levels (rep-
resentative for year 2000) of emissions of main
anthropogenic aerosols and aerosol precursors.
A substantial increase in the BC emission can
be seen in Fig. 3, and fig. S3 shows that the
increase in AOD for BC is also much stronger
than for sulfate and OC. A consequence of the
stronger increase in BC than other aerosol com-
ponents is that the anthropogenic fraction of ab-
sorption AOD is much stronger than the increase
in the anthropogenic fraction of total AOD, as
shown in Fig. 3. The much stronger increase in
the anthropogenic fraction of absorption AOD
than total AOD will to some extent be model-
dependent but will be robust because of the
much stronger growth in the BC emissions than
the growth in the rest of the aerosols and their
precursors. This suggests that the global annual
average single scattering albedo of the aerosols
has been reduced because of human activity (18).
The global mean annual average single scatter-
ing albedo computed in the model for all aerosols
at 0.55 mm is 0.986 at preindustrial conditions
and 0.970 at present-day conditions. Thus,
aerosol in present times is approximately twice
as absorbing as that in preindustrial conditions.
I find that change in the aerosol optical prop-
erties due to anthropogenic activity is a main
reason for the difference between observation-
based and model estimates of the radiative forc-
ing. The agreement between these two methods,
when similar aerosol optical properties are used
in the calculations of the direct aerosol effect,
allows to a much better extent a quantification
of this radiative forcing mechanism. Thus, the
bottom-up approach whereby anthropogenic
aerosols and their precursors are modeled via
emission, transportation, and deposition so as
to estimate an aerosol optical depth and a ra-
diative forcing is now in accordance with the
top-down approach, in which satellite observa-
tions of aerosol optical depths are apportioned
into anthropogenic fractions in order to derive a
radiative forcing. The observation-based method
of Chung et al.(13) relies much more on model
information than the other observation-based
studies (12,14–16) and has an anthropogenic
change in the aerosol optical properties. Their
resulting radiative forcing is weaker than other
observation-based results but is in close agree-
ment with the best estimate in this study. Thus,
this study not only explains differences between
model-based studies and observation-based stud-
ies but also within the group of observation-based
studies. Although the observational and model
methods for determining the direct radiative
forcing indicate very close agreement in this study,
it is acknowledged that the observational method
relies on the model for determining the anthro-
pogenic fraction, whereas other studies (12,16)
attempt to use independent methods over ocean
areas on the basis of the fine-mode fraction
reported in the MODIS retrievals.
The previous disagreement between the two
methods caused a disconcerting uncertainty that
undermined our confidence in estimates of the
direct aerosol effect. Understanding the differ-
ences between the two independent methods
substantially enhances our confidence in the es-
timates of the direct radiative forcing of aero-
sols. The results in this study also show that the
anthropogenically induced changes in the aero-
sol optical properties causes the earlier discrep-
ancies, rather than the assumed aerosol optical
properties in either the modeling or observation-
al studies. This conclusion can be made because
the results show very similar radiative forcing
for modeled and observed aerosol optical prop-
erties. With this work, the uncertainty in the ra-
diative forcing of the direct aerosol effect is
substantially reduced, and the remaining uncer-
tainty is probably related to the vertical profile
of the aerosols and their location in relation to
clouds. The differences among the global aero-
sol models in terms of the vertical profile of the
aerosols are considerable (24,25), but results in
this study show that the radiative forcing of the
direct effect from the observation-based method
has previously been too strong, and the forcing
is weaker than –0.5 W m
−2
. The best estimate
from this study (–0.3 T0.2 W m
−2
) suggests that
the direct aerosol effect offsets only a modest
10% of the radiative forcing that is due to in-
creases in well-mixed greenhouse-gas concen-
trations at their current concentrations.
Further, radiative forcing estimates of the di-
rect aerosol effect must, at least partly, rely on
global aerosol models, and pure estimates from
observational data are currently not possible.
References and Notes
1. P. Forster et al., in Climate Change 2007: The Physical
Science Basis—Contribution of Working Group I to the
Fourth Assessment Report of the Intergovernmental Panel
on Climate Change, S. Solomon et al., Eds. (Cambridge
Univ. Press, New York, 2007), pp. 129–134.
2. Y. J. Kaufman, D. Tanre, O. Boucher, Nature 419, 215 (2002).
3. V. Ramanathan, P. J. Crutzen, J. T. Kiehl, D. Rosenfeld,
Science 294, 2119 (2001).
4. G. Myhre, A. Myhre, F. Stordal, Atmos. Environ. 35, 2361
(2001).
5. V. Ramanathan, Y. Feng, Proc. Natl. Acad. Sci. U.S.A.
105, 14245 (2008).
6. M. Wild et al., Science 308, 847 (2005).
7. M. Schulz et al., Atmos. Chem. Phys. 6, 5225 (2006).
8. J. Hansen, M. Sato, R. Ruedy, J. Geophys. Res. 102, 6831
(1997).
9. J.M.Haywood,K.P.Shine,Geophys. Res. Lett. 22, 603 (1995).
10. J. M. Haywood, D. L. Roberts, A. Slingo, J. M. Edwards,
K. P. Shine, J. Clim. 10, 1562 (1997).
11. G. Myhre, F. Stordal, K. Restad, I. S. A. Isaksen, Tellus 50,
463 (1998).
12. N. Bellouin, O. Boucher, J. Haywood, M. S. Reddy, Nature
438, 1138 (2005).
13. C. E. Chung, V. Ramanathan, D. Kim, I. A. Podgorny,
J. Geophys. Res. 110, D24207 (2005).
14. H. Yu et al., Atmos. Chem. Phys. 6, 613 (2006).
15. J. Quaas, O. Boucher, N. Bellouin, S. Kinne, J. Geophys.
Res. 113, D05204 (2008).
16. N. Bellouin, A. Jones, J. Haywood, S. A. Christopher,
J. Geophys. Res. 113, D10205 (2008).
17. G. Myhre et al., Tellus 59, 115 (2007).
18. G. Myhre et al., Atmos. Chem. Phys. 9, 1365 (2009).
19. L. A. Remer et al., J. Geophys. Res. 113, D14S07 (2008).
20. B. N. Holben et al., Remote Sens. Environ. 66, 1 (1998).
21. Materials and methods are available as supporting
material on Science Online.
22. T. C. Bond et al., J. Geophys. Res. 109, D14203 (2004).
23. F. Dentener et al., Atmos. Chem. Phys. 6, 4321 (2006).
24. J. P. Schwarz et al., J. Geophys. Res. 111, D16207 (2006).
25. C. Textor et al., Atmos. Chem. Phys. 6, 1777 (2006).
26. I highly appreciate the comments and valuable suggestions
from J. Haywood in improving the manuscript. Comments
and discussions with N. Bellouin, O. Boucher, and L. Nygaard
are also appreciated. M. Johnsrud provided the observational
data. I thank the principal investigators and their staffs for
establishing and maintaining all of the AERONET sites used in
thisstudy.IalsothanktheOsloCTMgroup.Theresearchhas
been funded by the Norwegian Research Council.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1174461/DC1
Materials and Methods
Table S1
Figs. S1 to S3
References
3 April 2009; accepted 22 May 2009
Published online 18 June 2009;
10.1126/science.1174461
Include this information when citing this paper.
10 JULY 2009 VOL 325 SCIENCE www.sciencemag.org190
REPORTS
on July 21, 2009 www.sciencemag.orgDownloaded from