Global indirect aerosol effects: a review

Atmos. Chem. Phys., 5, 715–737, 2005
Atmospheric
www.atmos-chem-phys.org/acp/5/715/
SRef-ID: 1680-7324/acp/2005-5-715
Chemistry
European Geosciences Union
and Physics
Global indirect aerosol effects: a review
U. Lohmann1 and J. Feichter2
1ETH Institute of Atmospheric and Climate Science, Schafmattstr. 30, CH-8093 Zurich, Switzerland
2Max Planck Institute for Meteorology, Bundesstr. 53, D-20146 Hamburg, Germany
Received: 7 October 2004 – Published in Atmos. Chem. Phys. Discuss.: 17 November 2004
Revised: 28 January 2005 – Accepted: 18 February 2005 – Published: 3 March 2005
Abstract. Aerosols affect the climate system by changing
Clouds themselves are an important regulator of the
cloud characteristics in many ways. They act as cloud con-
Earth’s radiation budget. About 60% of the Earth’s surface is
densation and ice nuclei, they may inhibit freezing and they
covered with clouds. Clouds cool the Earth-atmosphere sys-
could have an in?uence on the hydrological cycle. While the
tem on a global average basis at the top-of-the-atmosphere.
cloud albedo enhancement (Twomey effect) of warm clouds
Losses of 48 W m?2 in the solar spectrum are only par-
received most attention so far and traditionally is the only
tially compensated (30 W m?2) by trapped infrared radiation.
indirect aerosol forcing considered in transient climate sim-
Measurements of the Earth Radiation Budget Experiment
ulations, here we discuss the multitude of effects. Different
(ERBE) (Collins et al., 1994) indicate that small changes
approaches how the climatic implications of these aerosol ef-
to macrophysical (coverage, structure, altitude) and micro-
fects can be estimated globally as well as improvements that
physical properties (droplet size, phase) have signi?cant ef-
are needed in global climate models in order to better repre-
fects on climate. For instance a 5% increase of the short-
sent indirect aerosol effects are discussed in this paper.
wave cloud forcing would compensate the increase in green-
house gases between the years 1750–2000 (Ramaswamy
et al., 2001). Consequently the growing interest in the im-
pact of aerosols on climate stimulated the development of
1
Introduction
better physically based parameterizations in climate models.
Nevertheless, the lack of understanding feedbacks of external
forcings on clouds remains one of the largest uncertainties in
Anthropogenic aerosol particles such as sulfate and carbona-
climate modeling and climate change prediction (Cess et al.,
ceous aerosols have substantially increased the global mean
1990; Houghton et al., 1996).
burden of aerosol particles from preindustrial times to the
present-day. Aerosol particles affect the climate system via
A summary of the different anthropogenic aerosol effects
the following physical mechanisms: First, they scatter and
on clouds is given in Table 1 while the effects are discussed
can absorb solar radiation. Second, they can scatter, ab-
in detail in the subsequent chapters. Most transient climate
sorb and emit thermal radiation.
Third, aerosol particles
model simulations allow for a cooling by aerosols in order
act as cloud condensation nuclei (CCN) and ice nuclei (IN).
to achieve good agreement with the observed temperature
The ?rst two mechanisms are referred to as direct effects
record. However, these studies usually ignore aerosol indi-
and are not subject of this paper but are discussed in detail
rect effects beyond the Twomey effect (Roeckner et al., 1999;
in e.g., Haywood and Boucher (2000). The last one is re-
Boer et al., 2000). Here we illustrate that radiative forcings of
ferred to as indirect effect. It will be the subject of this re-
other indirect aerosol effects exist and need to be considered
view together with other atmospheric properties in?uenced
in future transient simulations. A positive forcing is associ-
by aerosols (e.g. semi-direct effect, suppression of convec-
ated with a warming or energy gain of the Earth-atmosphere
tion). Even though the semi-direct effect is a consequence
system while a negative forcing represents a cooling or en-
of the direct effect of absorbing aerosols, it changes cloud
ergy loss. When available from the literature, we focus on the
properties in response to these aerosols and therefore is part
global aspect of these various anthropogenic indirect aerosol
of this review on aerosol-cloud-interactions.
effects because a review of all regional studies on indirect
aerosol effects would be beyond the scope of this study. We
Correspondence to: U. Lohmann
concentrate on studies that have been published since the
(ulrike.lohmann@env.ethz.ch)
2001 International Panel on Climate Change (IPCC) report.
© 2005 Author(s). This work is licensed under a Creative Commons License.

---

716
U. Lohmann and J. Feichter: Indirect aerosol effects: a review
Table 1. Overview of the different aerosol indirect effects and range of the radiative budget perturbation at the top-of-the atmosphere (FT OA)
[W m?2], at the surface (FSF C ) and the likely sign of the change in global mean surface precipitation (P) as estimated from Fig. 2 and from
the literature cited in the text.
Effect
Cloud type
Description
FT OA
FSF C
P
Indirect aerosol effect for
All clouds
The more numerous smaller
?0.5
similar
n/a
clouds with ?xed water amounts
cloud particles re?ect
to
to
(cloud albedo or Twomey effect)
more solar radiation
?1.9
FT OA
Indirect aerosol effect with
All clouds
Smaller cloud particles
?0.3
similar
decrease
varying water amounts
decrease the precipitation
to
to
(cloud lifetime effect)
ef?ciency thereby prolonging
?1.4
FT OA
cloud lifetime
Semi-direct effect
All clouds
Absorption of solar radiation
+0.1
larger
decrease
by soot may cause evaporation
to
than
of cloud particles
?0.5
FT OA
Thermodynamic effect
Mixed-phase
Smaller cloud droplets delay
?
?
increase or
clouds
the onset of freezing
decrease
Glaciation indirect effect
Mixed-phase
More ice nuclei increase the
?
?
increase
clouds
precipitation ef?ciency
Riming indirect effect
Mixed-phase
Smaller cloud droplets decrease
?
?
decrease
clouds
the riming ef?ciency
Surface energy
All clouds
Increased aerosol and cloud
n/a
?1.8
decrease
budget effect
optical thickness decrease the
to
net surface solar radiation
?4
2
Aerosol effects on water clouds
black carbon with respect to the cloud as discussed in chap-
ter 6. Both the cloud lifetime effect and the semi-direct effect
The IPCC Third Assessment Report concluded that the
involve feedbacks because the cloud lifetime and cloud liq-
Twomey effect of anthropogenic aerosol particles amounts
uid water content change. Therefore they were not included
to 0 to ?2 W m?2 in the global mean (Ramaswamy et al.,
in the radiative forcing bar chart of the IPCC (2001) assess-
2001). The Twomey effect refers to the enhanced re?ec-
ment.
tion of solar radiation due to the more but smaller cloud
droplets in a cloud whose liquid water content remains con-
2.1
Evidence of aerosol effects on warm clouds from ob-
stant (Twomey, 1959). Based on studies since the 2001 IPCC
servational data
report as shown in Fig. 1, the upper negative bound is slightly
reduced to ?1.9 W m?2. On the other hand, there is no cli-
The indirect aerosol effect of changing cloud albedo and
mate model that suggests that the Twomey effect is close to
cloud lifetime due to anthropogenic emissions of aerosols
zero, but the smallest cooling is ?0.5 W m?2 (Table 1).
and their precursors has been evaluated from observational
In addition, the more but smaller cloud droplets reduce the
studies, starting with observations of ship tracks perturb-
precipitation ef?ciency and therefore enhance the cloud life-
ing marine stratus cloud decks off the coast of California,
time and hence the cloud re?ectivity, which is referred to as
e.g. Ferek et al. (1998) and lately also over continental ar-
the cloud lifetime or second indirect effect (Albrecht, 1989).
eas (Feingold et al., 2003; Penner et al., 2004). Investiga-
This effect is estimated to be roughly as large as the Twomey
tions by Brenguier et al. (2000) and Schwartz et al. (2002)
effect as will be discussed below. Absorption of solar ra-
over the Atlantic Ocean showed that the cloud droplets were
diation by aerosols leads to a heating of the air, which can
smaller in the polluted clouds than in the clean clouds. This
result in an evaporation of cloud droplets. It is referred to as
contrast between polluted and clean clouds is partially off-
semi-direct effect (Graßl, 1979; Hansen et al., 1997). This
set because both papers found that the polluted clouds were
warming can partially offset the cooling due to the indirect
thinner as they originated over the continents, which causes
aerosol effect. Conversely, as shown by Penner et al. (2003),
them to be drier than their counterpart marine clean clouds
Johnson et al. (2004) and indicated in Table 1 the semi-direct
(Lohmann and Lesins, 2003). Since the cloud albedo de-
effect can result in a cooling depending on the location of the
pends on both the cloud droplet size and the cloud thickness
Atmos. Chem. Phys., 5, 715–737, 2005
www.atmos-chem-phys.org/acp/5/715/

---

U. Lohmann and J. Feichter: Indirect aerosol effects: a review
717
these competing effects partially cancel each other making it
more dif?cult to detect an indirect aerosol effect.
These systematic differences in cloud thickness between
clean and polluted clouds also affect the correlation between
optical thickness and effective radius as investigated by Bren-
guier et al. (2003). This correlation is negative, as anticipated
by Twomey (1977) if only cases of comparable values of ge-
ometrical thickness are considered. On the other hand, if the
most polluted cases are also accounted for, the trend suggests
a positive correlation, because the most polluted cloud sys-
tems sampled during ACE-2 were slightly drier, hence thin-
ner, than the marine and intermediate cases. Likewise, Peng
et al. (2002) showed that the slope between optical thickness
and effective radius is positive for polluted clouds due to the
increase in liquid water content and absence of drizzle size
drops and vice versa for clean clouds.
Feingold et al. (2003) studied the indirect aerosol effect
from ground-based remote sensing at the Atmospheric Radi-
ation Measurement (ARM) site in Oklahoma using observa-
tions of subcloud Raman lidar aerosol extinction ? at 355 nm
and cloud droplet effective radius to de?ne the aerosol in-
direct effect (IE) as the partial derivative of the logarithm
Fig. 1. Global mean Twomey effect and its contribution on the
Northern and Southern Hemisphere (NH, SH) and the ratio SH/NH
of cloud droplet radius with respect to the logarithm of the
of anthropogenic sulfate aerosols (red bars) from Rotstayn and Pen-
aerosol extinction:
ner (2001), Rotstayn and Liu (2003) and Jones et al. (2001), of
? ln re
anthropogenic sulfate and organic carbon (blue bars) from Menon
I E = ?
(1)
? ln ?
et al. (2002a); Quaas et al. (2004), of anthropogenic sulfate and
black, and organic carbon (turquoise bars) from Chuang et al.
Feingold et al. (2003) obtained IE values between 0.07 and
(2002) and the mean plus standard deviation from all simulations
0.11 over the ARM site for liquid water paths between 100
(olive bars). The results from Menon et al. (2002a) are averaged
and 130 g m?2. They showed that for a homogeneous cloud
over both simulations of the Twomey effect.
with a constant liquid water content for which cloud opti-
1/3
cal depth is proportional to N
, one can bracket IE to be
d
between 0 and 0.33. This derivative of the indirect aerosol
purely from observations permits estimates of the anthro-
effect can be used for model validation. For example, model
pogenic indirect aerosol effect globally.
simulations by Lohmann and Lesins (2003) obtained larger
slopes than observed, suggesting an overestimate of the in-
2.2
Global estimates of indirect aerosol effects on warm
direct aerosol effect. There is, however, some uncertainty in
clouds
this estimate related to different observing platforms. The
estimate of the indirect effect by Feingold et al. (2003) at the
Aerosol indirect effects are estimated from general circula-
ARM site is larger than estimated from POLDER satellite
tion models (GCMs) by conducting a present-day simulation
data by Br´eon et al. (2002). Rosenfeld and Feingold (2003)
and a pre-industrial simulation in which the anthropogenic
pointed out that limitations of the POLDER satellite retrieval
emissions are set to zero. The difference in the top-of-the-
could explain this discrepancy. Penner et al. (2004) com-
atmosphere radiation budget of these multi-year simulations
bined ARM data together with a Lagrangian parcel model at
is then taken to be the anthropogenic indirect aerosol effect.
the ARM sites in Oklahoma as a surrogate for a polluted site
The aerosol mass or number is then either empirically re-
and Alaska as a surrogate for a clean site to provide obser-
lated to the cloud droplet number concentration (Boucher and
vational evidence of a change in radiative forcing due to the
Lohmann, 1995; Menon et al., 2002a) or is obtained by us-
anthropogenic indirect aerosol effect.
ing a physically-based parameterization (Abdul-Razzak and
Long-term observations from satellites over Europe and
Ghan, 2002; Nenes and Seinfeld, 2003). Warm clouds form
China show evidence for the semi-direct effect, i.e. a reduc-
precipitation-size particles by the collision/coalescence pro-
tion in planetary albedo that can be attributed to absorbing
cess. In GCMs this is divided into the autoconversion (col-
aerosols in winter (Kr¨uger and Graßl, 2002, 2004; Kr¨uger
lisions and coalescence among cloud droplets) and the ac-
et al., 2004). In summer, on the other hand, when more sul-
cretion of rain drops with cloud droplets. The former is ei-
fate is produced, the Twomey effect is larger. So far, none
ther solely a function of the liquid water content (Sundqvist,
of the techniques used to derive the indirect aerosol effect
1978) and the cloud droplet size or concentration (Khairout-
www.atmos-chem-phys.org/acp/5/715/
Atmos. Chem. Phys., 5, 715–737, 2005

---

718
U. Lohmann and J. Feichter: Indirect aerosol effects: a review
the-atmosphere. The difference between these de?nitions is
small because the contribution of the longwave radiation to
the Twomey effect is below 0.1 W m?2 (Menon et al., 2002a;
Rotstayn and Penner, 2001). Also the clear-sky radiation will
remain the same in the absence of changes in temperature
and differences in ice and snow cover between pre-industrial
and present-day conditions. This latter constraint does not
apply any longer when feedback processes are included. In
general, the ratio of cooling of the Northern Hemisphere to
the cooling of the Southern Hemisphere is larger when only
sulfate aerosols are considered because biomass burning is
only a minor source for sulfate but a large source for car-
bonaceous aerosols (Fig. 1). The lowest ratio is simulated
by Quaas et al. (2004) who used the empirical Boucher and
Lohmann (1995) relationship but using the maximum of the
three hydrophilic species (sulfate, sea salt, organic carbon)
instead of just sulfate aerosols as a surrogate for all species.
Here sea salt has a more prominent role causing the anthro-
pogenic emissions on the Southern Hemisphere to only play
Fig. 2. Global mean Twomey effect, lifetime effect, both effect
a minor role.
and the ratio lifetime effect/Twomey effect of anthropogenic sulfate
aerosols (red bars) from Williams et al. (2001), Rotstayn and Penner
2.4
Twomey versus cloud lifetime effect
(2001), Ghan et al. (2001) and Jones et al. (2001), of anthropogenic
sulfate and black carbon (green bars) from Kristj´ansson (2002), of
Climate model estimates of the cloud lifetime effect and the
anthropogenic sulfate and organic carbon (blue bars) from Menon
semi-direct aerosol effect are at least as uncertain as of the
et al. (2002a); Quaas et al. (2004), of anthropogenic sulfate and
Twomey effect. As shown in Fig. 2, Kristj´ansson (2002) and
black, and organic carbon (turquoise bars) from Lohmann et al.
(2000); Takemura et al. (2005) and the mean plus standard devia-
Williams et al. (2001) concluded that the Twomey effect at
tion from all simulations (olive bars). The results from Menon et al.
the top-of-the atmosphere is four times as important as the
(2002a) and Ghan et al. (2001) are taken to be the averages of the
cloud lifetime effect whereas Lohmann et al. (2000), Ghan
simulations for only the Twomey effect and for both effects.
et al. (2001) and Quaas et al. (2004) simulated a cloud life-
time effect that is larger than the Twomey effect. This dis-
crepancy is independent of the chemical nature of the anthro-
dinov and Kogan, 2000; Liu et al., 2004). Once the autocon-
pogenic aerosol species that are used in these different sim-
version rate depends on the size or number of cloud droplets,
ulations. Likewise, the estimates of both indirect aerosol ef-
the Twomey and cloud lifetime effect cannot be calculated
fects are smallest for the climate models that use the most an-
separately any longer without changing the reference state.
thropogenic species (Fig. 2). One reason for the large aerosol
Estimates of the separate effects are then conducted by either
indirect effects obtained by Menon et al. (2002a) could be
prescribing a constant cloud droplet number concentration
due to their empirical treatment between the aerosol mass
(Lohmann et al., 2000) or by calculating the cloud water con-
and the cloud droplet number because sensitivity simulations
tent three times, once for advancing the model, and twice for
by Lohmann et al. (2000) yielded a higher total indirect effect
diagnostic purposes. The difference in the latter two stems
when an empirical relationship instead of a mechanistic rela-
from the different precipitation ef?ciencies of the clouds in
tionship was used. However, other models that use an empiri-
response to pre-industrial and present-day aerosol concentra-
cal relation such as Williams et al. (2001) obtain a smaller in-
tions (Kristj´ansson, 2002).
direct aerosol effect. Another reason for the discrepancy be-
tween models could be the dependence of the indirect aerosol
2.3
Twomey effect
effect on the background aerosol concentration. Sensitivity
studies by Lohmann et al. (2000) showed that reducing the
The estimates of the global mean Twomey effect and its divi-
minimum number of cloud droplets (which can be regarded
sion into the Northern and Southern Hemisphere are shown
as a surrogate for the background aerosol number concentra-
in Fig. 1. Note that the de?nition of the Twomey effect is not
tion) from 40 cm?3 to 10 cm?3 increased the indirect aerosol
unique. While Chuang et al. (2002), Rotstayn and Liu (2003)
effect from ?1.1 W m?2 to ?1.9 W m?2. Likewise differ-
and Quaas et al. (2004) de?ne the Twomey effect as the net
ences in the cloud microphysics scheme, especially in the au-
change in the shortwave ?ux at the top-of-the-atmosphere,
toconversion rate, cause uncertainties in the indirect aerosol
Menon et al. (2002a) de?ned the Twomey effect in terms of
effect estimates (Lohmann and Feichter, 1997; Jones et al.,
the change in the net cloud radiative forcing at the top-of-
2001; Menon et al., 2002a, 2003).
Atmos. Chem. Phys., 5, 715–737, 2005
www.atmos-chem-phys.org/acp/5/715/

---

U. Lohmann and J. Feichter: Indirect aerosol effects: a review
719
2.5
Land versus ocean
Most models suggest that the total indirect effect is at least
as large over land as over the oceans (Fig. 3). The only ex-
ception are the simulations by Rotstayn and Penner (2001)
in which different empirical formulas were used for relating
sulfate mass as a surrogate for all aerosols to cloud droplet
number (relationship A from Boucher and Lohmann (1995)).
This parameterization causes clouds over oceans to be more
susceptible to increases in CCN. The ?atter form of this pa-
rameterization over land than over ocean is broadly consis-
tent with the idea that continental clouds are less susceptible
to the effects of anthropogenic increases in CCN, because
there are more natural CCN over land than over ocean. This
conclusion is also consistent with the estimate of the total
indirect aerosol effect as derived from combining POLDER
satellite data and ECHAM4 GCM results (Lohmann and
Lesins, 2002) that suggest a larger indirect aerosol effect over
the oceans than over land. These satellite data, nevertheless,
have to be viewed with caution, because the retrieval from the
Fig. 3. Global mean total indirect aerosol effects and their contribu-
POLDER satellite is limited to clouds with a rather narrow
tion over the oceans, over land and the ratio ocean/land of anthro-
cloud droplet size distribution that produce a glory (Rosen-
pogenic sulfate (red bars) from Jones et al. (2001), from anthro-
feld and Feingold, 2003).
pogenic sulfate and black carbon (green bars) from Kristj´ansson
(2002), of anthropogenic sulfate and organic carbon (blue bars)
2.6
Constraints on the indirect aerosol effect
from Menon et al. (2002a); Quaas et al. (2004), of anthro-
pogenic sulfate and black, and organic carbon (turquoise bars) from
The cooling from both indirect effects on water clouds of
Lohmann and Lesins (2002); Takemura et al. (2005), from a com-
bination of ECHAM4 GCM and POLDER satellite results (black
sulfate and carbonaceous aerosols has been estimated from
bars) by Lohmann and Lesins (2002) and the mean plus standard
climate models since the last IPCC report to be ?1 to
deviation from all simulations (olive bars). The results from Menon
?4.4 W m?2 in the global mean (Ghan et al., 2001; Jones
et al. (2002a) are averaged over the three simulations for both ef-
et al., 2001; Lohmann and Feichter, 2001; Williams et al.,
fects.
2001; Menon et al., 2002a). This is larger than estimated
from inverse calculations which start from historical climate
record data of oceanic and atmospheric warming. They typ-
rect effect. Assuming that the column aerosol number con-
ically use ensembles of simulations with climate models of
centration increased by 30%, the total global mean indirect
reduced complexity and estimate a smaller anthropogenic in-
effect on warm clouds is estimated to be between ?0.6 and
direct aerosol effect within the range of 0 to ?2 W m?2 (For-
?1.2 W m?2. A smaller indirect aerosol effect is also ob-
est et al., 2002; Knutti et al., 2002; Anderson et al., 2003).
tained when constraining the total indirect aerosol effect by
If internal variability is thought to be averaged out over the
taking the difference in the slope of the cloud droplet effec-
anthropocene, then the total aerosol effect can solely be de-
tive radius-aerosol index relationship between the POLDER
duced from the greenhouse gas forcing and the increase in
satellite data (Br´eon et al., 2002) and the ECHAM GCM
land surface temperature and ocean heat content (Crutzen
results into account (Lohmann and Lesins, 2002). This re-
and Ramanathan, 2003). They obtain a cooling effect of
duces the total global mean aerosol effect from ?1.4 W m?2
aerosols between ?0.7 and ?1.7 W m?2 similar to the values
to ?0.85 W m?2. Indirect evidence for the existence of a
obtained by inverse models. The constraints from these in-
cloud lifetime effect on a global scale was reached by Suzuki
verse models or thermodynamic considerations are, however,
et al. (2004) when comparing simulations with and without a
not restricted to the indirect aerosol effect on water clouds
cloud lifetime effect with AVHRR satellite data of liquid wa-
only even though it is traditionally understood in this con-
ter path as a function of column aerosol number (Nakajima
text. Instead, the range encompasses all indirect aerosol ef-
et al., 2001).
fects and other effects currently not included in climate mod-
els. We will revisit this issue in the conclusions and outlook
2.7
Dispersion effect
section.
Sekiguchi et al. (2002) used different correlations between
Liu and Daum (2002) estimated that the magnitude of the
aerosol and cloud parameters derived from satellite remote
Twomey effect can be reduced by 10–80% by including the
sensing to estimate the radiative forcing of the aerosol indi-
in?uence that an increasing number of cloud droplets has
www.atmos-chem-phys.org/acp/5/715/
Atmos. Chem. Phys., 5, 715–737, 2005

---

720
U. Lohmann and J. Feichter: Indirect aerosol effects: a review
3
Aerosol effects on mixed-phase clouds
ÐÓÙ
Ð
Ó
¢
Æ
¢
3.1
Aerosol effects on large-scale mixed-phase clouds
ß
·
¢
¢
¢
Since most precipitation originates via the ice phase (Lau
ÐÓÙ
Ö
Ø
ÓÒ
Ò
Ð
Ø
Ñ
¢
and Wu, 2003), aerosol effects on ice clouds might have
¢
·
¢
larger consequences for the hydrological cycle than aerosol
ß
·
¢
¢
effects on water clouds. Precipitation originating from su-
¢
ÈÖ
Ô
Ø
Ø
ÓÒ
percooled liquid water clouds where the temperatures are too
 
¢
 
¢
 
ß
warm for homogeneous freezing of supercooled aerosols or
¢
 
·
cloud droplets to occur (T>?35?C) requires an aerosol sur-
¢
 
·
·
face to provide a substrate for ice initiation. This in?uence
Å
Ü
Ô
×
ÐÓÙ
Ý
ÖÓÑ
Ø
ÓÖ×
Æ
ÁÈ
of aerosol particles on changing the properties of ice forming
nuclei (IN) is poorly understood because of the variety of het-
·
·
erogeneous ice crystal nucleation modes. Aerosols can act as
·
·
Æ
ÖÓ×ÓÐ
Ñ
××
ÁÆ
IN by coming into contact with supercooled cloud droplets
(contact freezing), or by initiating freezing from within a
·
cloud droplet by immersion or condensation freezing, or by
acting as deposition nuclei. Ice nuclei that initiate freezing
Ä
Ò
×ÙÖ
are also referred to as freezing nuclei. Contact nucleation
is usually the most ef?cient process at slight supercoolings,
Fig. 4.
Schematic diagram of the warm indirect aerosol effect
while at lower temperatures immersion freezing can be more
(solid arrows) and glaciation indirect aerosol effect (dotted arrows)
prevalent. Deposition nuclei are generally least ef?cient be-
(adapted from Lohmann, 2002a). CDNC denotes the cloud droplet
cause the energy barrier that needs to be overcome for the
number concentration and IP the number concentration of ice parti-
cles.
phase change of water vapor to ice is larger than that re-
quired for the freezing nuclei modes. Unlike CCN, ice nu-
clei are generally insoluble particles, such as certain mineral
on the shape of the cloud droplet spectrum (dispersion ef-
dusts, soot, as well as some biological materials, e.g., Levin
fect). Taking this dispersion effect in global climate models
and Yankofsky (1983); Diehl et al. (2001); Gorbunov et al.
into account, this reduction is rather moderate and amounts
(2001). Ice nuclei may lose their nucleability, if foreign gases
only to 15–35% (Peng and Lohmann, 2003; Rotstayn and
such as sulfur dioxide (SO2) or ammonia (NH3) occupy their
Liu, 2003). Rotstayn and Liu (2005) obtained a similar re-
active sites (Pruppacher and Klett, 1997).
duction also for the cloud lifetime effect when including the
If some cloud droplets freeze in a supercooled water cloud,
dispersion effect.
then ice crystals will grow at the expense of cloud droplets
½
because of the lower saturation vapor pressure over ice than
2.8
Semi-direct aerosol effect
over water (the so-called Bergeron-Findeisen process). This
leads to a rapid glaciation of the supercooled water cloud.
Lohmann and Feichter (2001); Kristj´ansson (2002), and Pen-
Because the precipitation formation via the ice phase is more
ner et al. (2003) concluded that the semi-direct effect is only
ef?cient than in warm clouds, these glaciated clouds have a
marginally important at the top of the atmosphere in the
shorter lifetime than supercooled water clouds (Rogers and
global mean whereas Jacobson (2002) pointed out that the
Yau, 1989).
climatic effect of black carbon is strongly positive. The in?u-
Lohmann (2002a) showed that if, in addition to mineral
ence of black carbon is dominated via its absorption of solar
dust, a fraction of the hydrophilic soot aerosol particles is
radiation within the atmosphere, which also leads to a large
assumed to act as contact ice nuclei at temperatures between
negative forcing at the surface. The net reduction in short-
0?C and ?35?C, then increases in aerosol concentration from
wave radiation at the surface from all aerosol direct and indi-
pre-industrial times to present-day pose a new indirect effect,
rect effects is estimated to be between ?1.8 and ?4 W m?2
a “glaciation indirect effect”, on clouds as shown in Fig. 4.
(Ramanathan et al., 2001a; Lohmann and Feichter, 2001;
Here increases in contact ice nuclei in the present-day cli-
Liepert et al., 2004).
mate result in more frequent glaciation of supercooled clouds
and increase the amount of precipitation via the ice phase.
This reduces the cloud cover and the cloud optical depth of
mid-level clouds in mid- and high latitudes of the Northern
Hemisphere and results in more absorption of solar radiation
within the Earth-atmosphere system. Therefore, this effect
can at least partly offset the cloud lifetime effect. Laboratory
Atmos. Chem. Phys., 5, 715–737, 2005
www.atmos-chem-phys.org/acp/5/715/

---

U. Lohmann and J. Feichter: Indirect aerosol effects: a review
721
measurements by Gorbunov et al. (2001) yield evidence for
hydrophilic soot as ice nuclei. In addition, evidence of ef-
fective ice nuclei was recently measured with the continu-
ous ?ow diffusion chamber when sampling Asian dust parti-
cles (DeMott et al., 2003). In case of Saharan African dust,
mildly supercooled clouds at temperatures between ?5 to
?9?C were already glaciated (Sassen et al., 2003).
Observations by Borys et al. (2003) in midlatitude oro-
graphic clouds show that for a given supercooled liquid water
content, both the riming and the snowfall rates are smaller if
the supercooled cloud has more cloud droplets as, for exam-
ple, caused by anthropogenic pollution. Examination of this
effect in global climate model simulations with pre-industrial
and present-day aerosol concentrations showed that while the
riming rate in stratiform clouds has indeed decreased due
to the smaller cloud droplets in polluted clouds, the snow-
fall rate has actually increased (Fig. 5). This is caused by
the pollution induced increase in aerosol and cloud optical
thickness, which reduces the solar radiation at the surface
and causes a cooling that favors precipitation formation via
the ice phase (Lohmann, 2004).
Fig. 5.
Schematic diagram of the effect of pollution on snow
3.2
Aerosol effects on deep convective clouds (thermody-
showing the microphysical and climatic implications (adapted from
namic effect)
Lohmann, 2004).
Andronache et al. (1999) showed that an increase in sul-
fate loading during the TOGA-COARE experiment causes a
Khain et al. (2004)1 postulate that smaller cloud droplets,
signi?cant decrease of the effective radius of cloud droplets
such as originating from anthropogenic activity, would re-
(changes up to 2 µm on average) and an increase in the num-
duce the production of drizzle drops. When these droplets
ber concentration of cloud droplets of 5–20 cm?3 over a lim-
freeze, the associated latent heat release results in more vig-
ited domain of 500 km. The change in the average net short-
orous convection. In a clean cloud, on the other hand, driz-
wave radiation ?ux above the clouds was estimated to be on
zle would have left the cloud so that less latent heat is re-
average ?1.5 W m?2, with signi?cant spatial and temporal
leased when the cloud glaciates resulting in less vigorous
variations. The changes in the average net longwave radi-
convection. Therefore, no squall line is formed with mar-
ation ?ux above the clouds were negligible, but signi?cant
itime aerosol concentrations, but the squall line arises under
variations between ?10 W m?2 and 10 W m?2 near the sur-
continental aerosol concentrations and results in more pre-
face associated with changes in cloud water path of about
cipitation after 2 h of simulations with a detailed cloud mi-
10–20% were simulated.
crophysics model. More precipitation from polluted clouds
Rosenfeld (1999) and Rosenfeld and Woodley (2000) an-
was also simulated for different three-week periods over the
alyzed aircraft data together with satellite data suggesting
Atmospheric Radiation Measurement Program (ARM) site
that pollution aerosols suppress precipitation by decreasing
in Oklahoma (Zhang et al., 2005) as well as for multicell
cloud droplet size. This hypothesis was con?rmed by a mod-
cloud systems by (Seifert and Beheng, 2005).
eling study with a cloud resolving model by Khain et al.
On the contrary, cloud resolving model simulations of
(2001) who showed that aircraft observations of highly su-
mixed-phase shallow cumuli revealed decreasing precipita-
percooled water in deep convective clouds can only be repro-
tion ef?ciency with increasing atmospheric concentrations of
duced if large concentrations of small droplets exist but not
CCN because of the dominance of warm-rain processes in
if the cloud is rather clean. Taking these results to the global
these simulations (Phillips et al., 2002). Likewise precipita-
scale, Nober et al. (2003) evaluated the sensitivity of the gen-
tion from single mixed-phase clouds is reduced under conti-
eral circulation to the suppression of precipitation by anthro-
nental and maritime conditions when aerosol concentrations
pogenic aerosols by implementing a simple warm cloud mi-
are increased (Khain et al., 20041).
crophysics scheme into convective clouds. They found large
instantaneous local aerosol forcings reducing the warm phase
precipitation, but the precipitation change at the surface was
1Khain, A., Rosenfeld, D., and Pokrovsky, A.: Aerosol impact
guided by feedbacks within the system. Hence, no estimate
on the dynamics and microphysics of convective clouds, Q. J. R.
of the aerosol forcing on convective clouds can be given.
Meteorol. Soc., submitted, 2004.
www.atmos-chem-phys.org/acp/5/715/
Atmos. Chem. Phys., 5, 715–737, 2005

---

722
U. Lohmann and J. Feichter: Indirect aerosol effects: a review
Tropical biomass burning aerosols could have led to a re-
?ed from analyzing ISCCP data over Europe (Stordal et al.,
duction of ice crystal size in tropical deep convective clouds
2004).
(Sherwood, 2002). These smaller and more numerous ice
Ponater et al. (2002) and Marquart et al. (2003) studied
crystals would then lead to more scattering of solar radiation,
the climate effect of contrails using a global climate model,
i.e. exert a Twomey effect as discussed above. However, no
but so far related the contrail formation only to relative hu-
global mean radiative forcing was deduced by Sherwood for
midity but did not link it to aerosol properties. The study
this effect. He used this hypothesis to explain the increase
by Lohmann and K¨archer (2002) that parameterized homo-
in stratospheric water vapor, which by being a greenhouse
geneous freezing of supercooled aerosols suggests that the
gas, provides a positive radiative forcing that would partially
impact of aircraft sulfur emissions on cirrus properties via
offset the Twomey effect associated with the smaller ice crys-
homogeneous freezing of sulfate aerosols is small. Hence
tal size in these deep convective clouds. This demonstrates
the question has been raised whether aircraft-generated black
the complex interactions between the different forcing agents
carbon particles serving as heterogeneous ice nuclei (Str¨om
that need to be understood and the dif?culties to disentan-
and Ohlsson, 1998) may have a signi?cant impact on cirrus
gle forcings and feedbacks. Both issues will be revisited in
cloudiness and cirrus microphysical properties.
Sect. 7.
Hendricks et al. (2004) performed climate model simu-
lations that revealed that the large-scale impact of aviation
black carbon (BC) emissions on the upper troposphere/lower
stratosphere (UTLS) BC mass concentration is small. Nev-
4
Aerosol effects on ice clouds
ertheless, the simulations suggest a signi?cant aviation im-
pact on the number concentrations of UTLS BC particles and
Condensation (con) trails left behind jet aircrafts form when
potential heterogeneous IN (BC and mineral dust particles).
hot humid air from jet exhaust mixes with environmental air
Large-scale increases of the potential heterogeneous IN num-
of low vapor pressure and low temperature. The mixing is a
ber concentration of up to 50% were simulated. Provided that
result of turbulence generated by the engine exhaust. Con-
BC particles from aviation serve as ef?cient heterogeneous
trails cannot be distinguished any longer from cirrus clouds
IN, maximum increases or decreases in ice crystal number
once they lose their line-shape. While there are only a few
concentrations of more than 40% were simulated assuming
general studies on aerosol effects on cirrus, many investi-
that the “background” (no aviation impact) cirrus cloud for-
gations analyzed the effect of aircraft emissions on climate.
mation is dominated by heterogeneous or homogeneous nu-
Therefore we will discuss these two effects separately below.
cleation, respectively (Hendricks et al., 2003).
4.1
Aerosol effects on contrails
4.2
Aerosol effects on cirrus clouds
The IPCC aviation report (Penner et al., 1999) identi?ed the
An increase in the n