local and remote impacts of aerosol climate forcing on

Local and Remote Impacts of Aerosol Climate Forcing on Tropical Precipitation*

CHIA CHOU

Research Center for Environmental Changes, Academia Sinica, Taipei, Taiwan

J. DAVID NEELIN

Department of Atmospheric Sciences, and Institute of Geophysics and Planetary Physics, University of California, Los Angeles,Los Angeles, California

ULRIKE LOHMANN

Institute of Atmospheric and Climate Science, ETH Zürich, Zürich, Switzerland

JOHANN FEICHTER

Max Planck Institute for Meteorology, Hamburg, Germany

(Manuscript received 18 October 2004, in final form 10 May 2005)

ABSTRACT

Mechanisms that determine the direct and indirect effects of aerosols on the tropical climate involvemoist dynamical processes and have local and remote impacts on regional tropical precipitation. Thesemechanisms are examined in a climate model of intermediate complexity [quasi-equilibrium tropical cir-culation model (QTCM)] forced by prescribed aerosol forcing, which is obtained from a general circulationmodel (ECHAM4). The aerosol reflection is the dominant aerosol forcing, while the aerosol absorption hascomplex but much weaker influences on the regional tropical precipitation based on the ECHAM4 aerosolforcing. The local effect associated with aerosols contributes negative precipitation anomalies over convec-tive regions by affecting the net energy flux into the atmospheric column. This net energy flux is controlledby the radiative forcing at the top of the atmosphere on time scales where surface heat flux is nearequilibrium, balancing anomalous solar radiation by evaporation, longwave radiation, and sensible heat.Considering the aerosol absorption effect alone, the associated precipitation anomalies are slightly negativebut small when surface heat fluxes are near equilibrium. Two effects found in global warming, the upped-ante mechanism and the anomalous gross moist stability mechanism, occur with opposite sign in the aerosolcase. Both act as remote effects via the widespread cold tropospheric temperature anomalies induced by theaerosol forcing. In the upped-ante mechanism in global warming, a warm troposphere increases the low-level moisture “ante” required for convection, creating spatially varying moisture anomalies that disfavorprecipitation on those margins of convective zones where the mean flow imports air from nonconvectiveregions. In the aerosol case here, a cool troposphere preferentially decreases moisture in convective regions,creating positive precipitation anomalies at inflow margins. In the anomalous gross moist stability mecha-nism for the aerosol case, the decrease in moisture in convective regions acts to enhance the gross moiststability, so convection and the associated precipitation are reduced. The partitioning between the aerosollocal and remote effects on regional tropical precipitation differs spatially. Over convective regions thathave high aerosol concentration, such as the South American region, the aerosol local effect contributesmore negative precipitation anomalies than the anomalous gross moist stability mechanism in the QTCMsimulations. On the other hand, the remote effect is more important over convective regions with smallaerosol concentrations, such as the western Pacific Maritime Continent. Remote effects of midlatitudeaerosol forcing have a substantial contribution to tropical anomalies.

* Institute of Geophysics and Planetary Physics, University of California, Los Angeles Contribution Number 6213.

Corresponding author address: Dr. Chia Chou, Research Center for Environmental Changes, Academia Sinica, P.O. Box 1-48, Taipei11529, Taiwan.E-mail: [email protected]

15 NOVEMBER 2005 C H O U E T A L . 4621

© 2005 American Meteorological Society

JCLI3554

1. Introduction

Aerosols create great uncertainties in studying cli-mate change under global warming. Aerosols directlyaffect the climate system by scattering and absorbingsolar radiation. These “direct” effects can cool both theatmosphere and surface by reflecting solar radiationback into space. At the same time, the direct effects canalso warm the atmosphere by absorbing solar radiation,but this absorption cools the surface. These effectschange the temperature structure and then influencethe strength of convection (Kaufman et al. 2002). Theabsorption effect mainly induced by black carbonwarms the atmosphere (Jacobson 2001) but is hypoth-esized to reduce the convection contribution to warm-ing the atmosphere (Ramanathan et al. 2001; V. Ra-manathan 2004, personal communication). Aerosolscan also increase cloud drop number concentration andreduce cloud drop effective radius, so clouds existlonger and precipitation efficiency is decreased. These“indirect” effects can lead to a weaker hydrologicalcycle (Feichter et al. 2004; Liepert et al. 2004; Ra-manathan et al. 2001). The feedback of smaller clouddrop effective radius and longer lifetimes of cloudstends to reflect more solar radiation and further coolthe climate system (Lohmann and Lesins 2002; Penneret al. 2004). Overall, the aerosol climate forcing tends topartly offset the global warming impacts associatedwith greenhouse gases (GHGs). However, the aerosollifetime is much shorter than that of most GHGs andthe aerosol concentration varies strongly with space, sothe aerosol climate impact cannot be treated as a mirrorimage of global warming (Houghton et al. 2001; Rot-stayn and Lohmann 2002). Finally, Feichter et al.(2004) noted that the nonlinear interaction betweenaerosol particles and GHG weakens the warming ten-dency dominated by GHG.

Considering the GHG warming alone, the radiativeeffects have a very broad spatial pattern. The associ-ated tropospheric temperature anomalies are relativelyuniform, while the precipitation anomalies are muchmore complex and have strong spatial variations. Stud-ies of global warming impacts on precipitation showpoor agreement among climate model simulations on aregional basis (Allen and Ingram 2002; Neelin et al.2003, hereafter NCS03). To understand the mecha-nisms of global warming impacts on regional tropicalprecipitation, Chou and Neelin (2004, hereafter CN04)have proposed two dominant mechanisms: the anoma-lous gross moist stability (M�) mechanism and theupped-ante mechanism. As the troposphere warms, thelow-level moisture tends to increase to maintain con-vective quasi equilibrium (QE) in convective regions.

This effect linking the tropospheric temperature andthe low-level moisture is termed “QE mediation” [Nee-lin and Su 2005; CN04 focused on convective availablepotential energy (CAPE), but QE mediation is a moregeneral term for this effect]. Over nonconvective re-gions, the low-level moisture is controlled by differentprocesses that cause smaller increases in the low-levelmoisture, such as the balance between divergence andevaporation. Thus, a spatial gradient of anomalous low-level moisture is induced. This coexistence of thesmooth anomalous temperature distribution and thestrong horizontal gradient of moisture anomalies iscritical to inducing regional tropical precipitationchanges. In the M� mechanism, gross moist stability isreduced by the increases in low-level moisture, so con-vection is enhanced and precipitation increases overconvective regions. In the upped-ante mechanism, in-flow of air with less positive moisture anomalies fromnonconvective regions into margins of convective re-gions opposes the low-level moisture increases. As aresult, the low-level moisture does not increase enoughto meet the higher “convective ante” needed to main-tain the convection in the face of warmer tropospherictemperature aloft. The precipitation is thus reducedover margins of convective regions wherever there isstrong inflow. This upped-ante mechanism is also foundin the reduction of precipitation over equatorial SouthAmerica and the Atlantic intertropical convergencezone (ITCZ) during El Niño events (NCS03).

Regional tropical precipitation changes induced bythe aerosol effects are also complex (Feichter et al.2004) and have aspects similar to the global warmingcase. Aerosol distribution is rather regional and thedistribution of greenhouse gases is global. To what ex-tent are mechanisms of tropical precipitation impactssimilar or different between the global warming andaerosol cases? To answer this question, examination ofthe aerosol effects without interaction with globalwarming should be conducted. The separation of aero-sol and GHG effects may imply caveats because of non-linear interaction between aerosol effects and GHG ef-fects, such as in Feichter et al. (2004). However, thiskind of experiment can be a prelude to examining morecomplex cases of simultaneously competing greenhouseand aerosol effects. In this study, an atmospheric modelof intermediate complexity (Neelin and Zeng 2000;Zeng et al. 2000) is used. This model is coupled with amixed layer ocean model (Chou et al. 2001) and isforced by prescribed aerosol forcing. Throughout thispaper, “aerosol effects” will be used to mean combineddirect and indirect effects as estimated from anthropo-genic sulfate and carbonaceous aerosols in Feichter etal. (2004). The quasi-equilibrium tropical circulation

4622 J O U R N A L O F C L I M A T E VOLUME 18

model (QTCM) is a special purpose primitive equationmodel designed to aid understanding of interaction oftropical deep convection with large-scale dynamics. Anumber of cases of such interaction have already beenworked out, including ENSO teleconnections (Su et al.2001; Su and Neelin 2002) and greenhouse gas response(Chou and Neelin 2004; Neelin et al. 2003). To theextent that the ECHAM4 tropical response to aerosolscan be reproduced in the QTCM when given forcingfrom ECHAM4, theory developed in prior work canaid in unraveling the causal pathways. Furthermore, theforcing can be decomposed in a number of ways sincerunning a series of long experiments with different sub-sets of the forcing fields is straightforward in QTCM.

Considering the aerosol case without any influence ofgreenhouse gases, the aerosol forcing cools the climatesystem in general. Figure 1 shows the aerosol climateimpacts in the boreal winter, including direct and indi-rect effects, from the fourth-generation Max Planck In-stitute for Meteorology model, ECHAM4, coupled to aslab ocean and thermodynamic sea ice model (Feichteret al. 2004). The troposphere does become colder glo-bally. The temperature anomalies over the NorthernHemisphere are stronger than those in the SouthernHemisphere, and the temperature anomalies show little

longitudinal gradient. This broad cooling pattern hintsat a potential connection with the global warming casein terms of mechanisms affecting regional tropical pre-cipitation since widespread tropospheric temperatureanomalies occur in both cases. The precipitation anoma-lies in Fig. 1a show strong spatial variations and occurmostly in the Tropics, especially over the tropical oceans.This spatial pattern of the precipitation anomalies doesexhibit qualitative similarities in some regions to theglobal warming case (e.g., Fig. 2a of CN04), but withopposite sign. Negative precipitation anomalies occurover convective regions and positive anomalies are foundover margins of convective regions for the aerosol case,and vice versa for the global warming case. To whatextent mechanisms have similar pathways between theglobal warming and the aerosol cases presents an inter-esting question and will be examined in this study.

Figure 2 shows the aerosol solar radiative forcing inthe boreal winter. The associated longwave radiativeforcing (not shown) is much smaller. Thus, only thesolar component of the aerosol radiative effect is pre-

FIG. 1. (a) DJF precipitation and (b) tropospheric temperature(850–200 hPa) differences between the ECHAM4 present day(PD) and preindustrial (PI) simulations. The thick dashed line in(a) is 150 W m�2 contour from the DJF precipitation climatology.Contour interval is 10 W m�2 for precipitation. Dark shading isabove 30 W m�2 and light shading is below �30 W m�2. Contourinterval is 0.2°C for temperature. Dark shading is above 1.2°C andlight shading is between 0.8° and 1.2°C. Precipitation is given in Wm�2 for comparison to energy budgets. Divide by 28 to obtain mmday�1.

FIG. 2. ECHAM4 aerosol forcing for DJF. (a) Solar absorptionby the atmosphere (S↓

t � S↑t � S↓

s � S↑s ), (b) solar reflection at the

TOA (S↑t � S↓

t ), and (c) net solar absorption by the surface (S↓s �

S↑s ). Contour interval is 4 W m�2. Dark shading is above 8 W m�2

and light shading is below �8 W m�2.


scribed as forcing in this study. The aerosol forcing,which includes reflection and absorption, has a muchmore complex spatial distribution than the forcing as-sociated with greenhouse gases, which is fairly uniformin space. Figure 2a shows solar absorption, which isdominated by absorbing aerosols such as carbonaceousparticles. Two maximum regions are found over South-east Asia and West Africa with values around 8 W m�2.Figure 2b shows solar reflection at the top of the atmo-sphere (TOA), which is controlled by nonabsorbingaerosols such as sulfate particles, and the associatedindirect feedback of this interaction with cloud micro-physics. The distribution of the aerosol reflection dif-fers from that of the absorption, and there are fivemaximum areas: Southeast Asia, the southeastern partof North America, South America, South Africa, andthe western Indian Ocean. The maximum amplitude ofthe aerosol reflection is similar to that of absorption,but the global average of the aerosol reflection is stron-ger since its distribution is much broader than that ofthe aerosol absorption. Both aerosol absorption andreflection reduce downward solar insolation, so the netsolar absorption by the surface is reduced more thanthe reflection at the TOA.

With the aerosol forcing in Fig. 2, we examine theaerosol effects on regional tropical precipitation. Themodels and the forcing strategy used in this study aredescribed in section 2. Simulated impacts of aerosolabsorption and reflection effects are examined togetherand separately in section 3. In section 4, a moist staticenergy budget analysis approach is outlined with pro-posed mechanisms and diagnostics of these. Experi-ments to separate aerosol local and remote effects arepresented in section 5, with discussion in section 6.

2. Models and forcing strategy

a. Models

The aerosol model used here is coupled with the at-mospheric general circulation model (GCM) ECHAM4;a complete description can be found in Feichter et al.(2004). Below is a brief description of the aerosolmodel. The aerosol concentration is predicted from theaerosol budget, which is controlled by processes ofemissions, transport, chemical transformation, dry andwet deposition, and sedimentation. The aerosol com-ponents include sulfate, black carbon, organic carbon,mineral dust, and sea salt. The physical and opticalaerosol properties that are used to calculate the directaerosol effect are specified from observations for eachcomponent. The aerosol number concentration canmodify cloud microphysics and induces the indirectaerosol effect; for example, it enhances cloud albedo

through reduction of the droplet effective radius andincrease of cloud droplet number concentration as de-scribed by Lohmann et al. (1999, 2000).

The ECHAM4 atmospheric GCM with the aerosolmodel is coupled to a slab ocean model with a depth of50 m (Roeckner et al. 1995) and a thermodynamic seaice model. Using the ECHAM4 model, a pair of experi-ments is used to examine aerosol impacts on climate.The first experiment is for preindustrial (PI) conditionswith PI aerosol and aerosol precursor emissions of sul-fate and organic carbon. The second experiment is forpresent-day (PD; mid-1980s) conditions with PD aero-sol and aerosol precursor emissions of sulfate and blackand organic carbon. The greenhouse gases are constantat the PD level in both experiments. Thus, differencesbetween the PD and PI experiments are purely causedby aerosol effects, including both direct and indirecteffects. The results shown in this study are for 50-yraverages after 20 yr of integration to reach approximateclimate equilibrium.

To examine the aerosol effects on regional precipi-tation, a coupled ocean–atmosphere–land model of in-termediate complexity (Neelin and Zeng 2000; Zeng etal. 2000, hereafter ZNC) with prescribed divergence ofocean heat transport (Q flux) is used (Chou et al. 2001).Based on the analytical solutions derived from theBetts–Miller moist convective adjustment scheme(Betts and Miller 1993), typical vertical structures oftemperature, moisture, and winds for deep convectionare used as leading basis functions for a Galerkin ex-pansion (Neelin and Yu 1994; Yu and Neelin 1994).The resulting primitive equation model makes use ofconstraints on the flow by quasi-equilibrium thermody-namic closures and is referred to as QTCM1 (quasi-equilibrium tropical circulation model with a single ver-tical structure of temperature and moisture for deepconvection). Because the basis functions are based onvertical structures associated with convective regions,these regions are expected to be well represented andsimilar to a GCM with the Betts–Miller moist convec-tive adjustment scheme. Far from convective regions,QTCM1 is a highly truncated Galerkin representationequivalent to a two-layer model. A cloud–radiationscheme (Chou and Neelin 1996; ZNC), simplified fromthe full GCM radiation schemes (Harshvardhan et al.1987; Fu and Liou 1993), is included. Deep convectionand cirrocumulus/cirrostratus cloud fraction is esti-mated by an empirical parameterization (Chou andNeelin 1999). A simple formula is used to obtain atmo-spheric boundary layer winds under assumptions of asteady state and a vertically homogeneous mixed layerwith fixed height (Stevens et al. 2002). An intermediateland surface model (ZNC) is used to simulate interac-


tion between the atmosphere and land surface. Thismodel simulates processes such as evapotranspirationand surface hydrology in a single land surface layer forcalculating energy and water budgets. Soil moisture isbalanced by precipitation, evaporation, surface runoff,and ground runoff. QTCM version 2.3 is used here withthe solar radiation scheme slightly modified using thenew Fu and Liou radiation scheme.

b. Forcings and experiments

Motivated by the strong spatial pattern of the tropi-cal precipitation change induced by aerosol effects, weattempt to understand the mechanisms of these aerosolimpacts using the QTCM forced by prescribedECHAM4 aerosol radiative forcing (Feichter et al.2004). Because indirect aerosol effects are included, di-rect implementation of aerosol particles in QTCM isnot appropriate. Thus, the aerosol forcing needs to in-clude changes in solar radiation associated with changesin cloud properties, such as cloud droplet effective ra-dius. A portion of the cloud change is due to dynamicalfeedback, but it is hard to partition this effect. There-fore, we use total solar radiation changes as forcing and“freeze” the QTCM solar radiation deep-convectioncloud feedback by using fixed cloud cover to avoidcounting this effect twice. Moreover, since the role ofdeep-convection cloud feedback is reasonably under-stood in QTCM and it is secondary in global warmingsimulations (CN04), this assumption of freezing solarradiation cloud feedback is a reasonable case in whichto examine aerosol effects. All longwave radiative feed-backs are treated with the standard QTCM version 2.3parameterizations. Since mechanisms for inducing theprecipitation anomalies do not depend significantly onseason when considering the first-order effects of aero-sol, we present the December–February (DJF) seasononly in this study. Moreover, DJF can avoid some com-plications involved with the Asian summer monsoons,which are not simulated well by most climate models.

We first aim to reproduce general features of theECHAM4 runs in QTCM. Two components of theaerosol forcing associated with solar radiation, net solarradiative fluxes at the top of the atmosphere (TOA)and at the surface, are prescribed in QTCM. Since thereis only one thermodynamic equation in QTCM1, this issufficient to get projection onto the temperature basisfunction response. To examine changes associated withabsorption versus reflection of solar radiation, the aero-sol forcing is decomposed into two cases: One case as-sumes pure absorption by aerosols and no change dueto the aerosol forcing at TOA. The converse case hasno absorption in the atmosphere by aerosols withchanges in net surface solar radiation identical to TOA

reflection. In the case of aerosol absorption, the netsolar absorption in the atmosphere is matched to theECHAM4 runs and surface solar radiation is reducedcorrespondingly. For the aerosol reflection case, we usethe TOA solar radiation changes from the ECHAM4simulations, and surface solar radiation anomalies arereduced by the same amount so that atmospheric ab-sorption is zero. Another useful aspect of this decom-position is that solar radiative forcing anomalies of thetwo cases add up to the total ECHAM4 forcing. Thus,if the responses are approximately linear, they can beused to compare to the full forcing response.

One caveat on this decomposition of the forcing isthat it is not identical to running a full radiative transfercalculation with absorption and reflection set to zero inturn. Setting absorption to zero in such a calculationwould change TOA forcing slightly and, in the conversecase, pure absorption would have a small TOA compo-nent due to surface albedo. However, the decomposi-tion used here appears useful for understanding theimpacts of each process.

Other cases are used to evaluate local versus remoteresponse. Over a region of interest, such as SouthAmerica, the solar radiative forcing is set to zero, soremote response over this region to the outside forcingcan be examined. To examine local response, a con-verse experiment with the aerosol forcing only over thisregion is performed. The target region considered hereis tropical South America, which has a large deep con-vection zone and large aerosol forcing. Another experi-ment is to examine impacts of the midlatitude aerosolforcing on tropical climate. In this experiment, theaerosol forcing over the band region between 20° and50°N where most aerosol particles are emitted is pre-scribed, but the aerosol forcing outside the band regionis set to zero. To identify how the aerosol forcing re-motely affects the tropical precipitation, an experimentis also conducted with modified tropospheric tempera-ture obtained from the full aerosol experiment speci-fied in the convective scheme.

To have a signal with statistical significance, espe-cially for those experiments with small changes, a long-term average is needed. Moreover, the long-term aver-age can help us to identify regional climate changesmore precisely. Thus, results of all experiments withQTCM are 400-yr averages.

3. Response to aerosol solar forcing andabsorption versus reflection components

a. Total response

The first experiment examines how well QTCM canreproduce the response to the aerosol effects found in


the ECHAM4 simulation. Forced by the prescribed netsolar radiation anomalies at the TOA and the surface,the results of the QTCM simulation are shown in Fig. 3.Comparing to the ECHAM4 runs in Fig. 1, the exactpositions of the precipitation anomalies are not thesame, but the positions relative to the preindustrialmean precipitation in each model are very similar. Thenegative precipitation anomalies are located over deepconvective regions, while the positive anomalies arefound over the margins of deep convective regions.Two areas with anomalies of opposite sign to the globalwarming case are well simulated: South America andthe central Pacific. The anomalies over South Americawill be discussed as an example in the following sec-tions. The amplitudes of the precipitation anomalies inthe QTCM and ECHAM4 runs are comparable. Thetropospheric temperature anomalies in the QTCMsimulations are similar to those in the ECHAM4 simu-lations with a smoother spatial pattern and slightlyweaker amplitude. Overall, QTCM reproduces theanomalies induced by the aerosol effects found in theECHAM4 simulations, so QTCM can be used to fur-ther examine the aerosol effects.

b. Absorption versus reflection components

To examine the aerosol absorption and reflection ef-fects separately, two experiments described in section2b are performed, and the results are shown in Figs. 4and 5. Forced only by the reflection effect, the results inFig. 4 show similar patterns of precipitation and tropo-

spheric temperature anomalies to those in Fig. 3. How-ever, the results, in Fig. 5, that are forced only by theabsorption effect are much weaker than those in Fig. 4.This indicates that the aerosol solar reflection is theleading effect on changing tropical precipitation and ismuch more effective than the aerosol absorption. Thetropospheric temperature anomalies induced by theaerosol absorption (not shown) are also weaker thanthose induced by the solar reflection. Their magnitudeis less than 0.05°C in the Tropics, so little remote effectassociated with the tropospheric temperature changesshould be expected. These results are surprising be-cause both aerosol absorption and reflection effectscool the surface and their amplitudes are relativelycomparable (Fig. 2). The reduced surface solar radia-tion is balanced by evaporation and surface longwaveradiation anomalies, so the net surface heat fluxanomalies are close to zero at a time scale when land

FIG. 5. As in Fig. 3 except for precipitation of the experimentwith the aerosol absorption only. Contour interval is 2 W m�2

(different from Fig. 3).

FIG. 3. As in Fig. 1 except for the QTCM simulations of PD �PI differences: (a) precipitation and (b) temperature. Dark shad-ing is above 1.0°C and light shading is between 0.6° and 1.0°C(different from Fig. 1).

FIG. 4. As in Fig. 3 except for the experiment with the aerosolreflection only.


and ocean surfaces reach equilibrium (Liepert et al.2004). Thus, the net radiation anomaly at the TOAbecomes crucial in determining the aerosol effects. Forthe reflection experiment (Fig. 4), the aerosol TOAforcing is similar to the control experiment in Fig. 3, sothe aerosol climate impacts in both experiments aresimilar. For the absorption experiment (Fig. 5), theaerosol TOA forcing is zero by definition, so changesinduced by the aerosol absorption are small. Themechanism associated with the net radiative fluxanomalies at the TOA will be discussed in detail insection 4b. The combination of the precipitationanomalies induced by the aerosol absorption and re-flection, respectively (Figs. 4 and 5), tends to be close tothat in Fig. 3. This implies that the aerosol absorptionand reflection effects are relatively linear in their im-pact on atmospheric dynamics.

4. Diagnosis of mechanisms

a. Moist static energy budget

In general, the aerosol effects cool the climate systemand reduce global precipitation. On a regional basis,both positive and negative precipitation anomalies arefound. This indicates that more complex processes areinvolved in determining regional precipitation anoma-lies. Thus, we analyze moisture and moist static energybudgets to understand mechanisms of the regionaltropical precipitation anomalies caused by the aerosoleffects. Under the quasi-equilibrium convective clo-sure, the anomaly moisture budget integrated throughthe troposphere for a steady state can be written as

P� � Mq� · v�1 � M�q� · v1 � �v · �q�� E�, �1�

where an overbar denotes averaged variables withoutaerosol impacts, primes represent the anomalies causedby aerosol effects, P is precipitation, E is evaporation, vis horizontal velocity, and g is gravity. ngle bracketsdenote a vertical integral over the troposphere with pT

as the depth of the troposphere:

�X� 1g �ps

ps�pT

X dp, �2�

where ps is surface pressure. The specific humidity q isin energy units (absorbing the latent heat per unit mass,L). The horizontal velocity v1 is the wind componentassociated with baroclinic structure under convectivequasi-equilibrium constraints, and Mq is the gross mois-ture stratification (Neelin and Yu 1994; Yu et al. 1998)and is given by

Mq ��pq�, �3�

where �(p) is the vertical structure of vertical velocityfrom baroclinic wind and

��x,y,p,t� ��p�� · v1�x,y,t�, �4�

where � is pressure velocity. Thus, (1) uses

��pq� � Mq� · v1, �5�

with Mq giving the amount of moisture convergenceproduced per unit of low-level convergence/upper-leveldivergence in the baroclinic flow.

The precipitation anomaly in (1) is dominated by(Mq� · v1)�, which is mainly determined by � · v�1 in theTropics. Note that � · v1 0 indicates low-level con-vergence and upper-level divergence. To estimate � · v�1the vertically integrated moist static energy budget isused. Under similar assumptions to (1), the verticallyintegrated moist static energy equation can be writtenas

M� · v�1 Fnet� � M�� · v1 � �v · ��q � T��, �6�

where T is atmospheric temperature (J kg�1, absorb-ing the heat capacity at constant pressure, Cp). Thenet energy input into the atmospheric column F net isgiven by

F net Ft � Fs, �7�

where the net heat flux at the TOA can be obtained by

Ft St↓ � St

↑ � Rt↑, �8�

and the net heat flux at the surface is from

Fs Ss↓ � Ss

↑ � Rs↓ � Rs

↑ � E � H. �9�

Subscripts s and t on the solar (S↓ and S↑) and longwave(R↑ and R↓) radiative terms denote surface and modeltop, and R↓

t � 0 has been used; H is sensible heat flux.Positive values of Ft and Fs indicate downward heatfluxes, and M is the gross moist stability (Neelin and Yu1994; Yu et al. 1998), defined by

M ��ph��, �10�

where h s � q is moist static energy and s T � �is dry static energy, with � the geopotential. Thus, (6)uses

��ph� � M� · v1. �11�

More detailed derivations of the moisture and moiststatic energy budget equations can be found in CN04and Su and Neelin (2002). Following Neelin and Su(2005), eliminating � · v�1 between (6) and (1) yields


P� Mq

M��v · �T�� F�t � F�s� � E� � �Mq

M� 1�

� �v · �q�� Mq

MM� � M�q�� · v1. �12�

The enhancement of terms in the moist energy budget(by a factor of Mq /M) in forcing precipitation is termedthe gross moist stability multiplier effect. In deep con-vective regions, Mq /M � 4 in this model, consistentwith estimate in Yu et al. (1998), where values in therange 3–7 were obtained.

b. Implications for the solar absorption forcing case

The aerosol absorption effect is highly uncertain (Ra-manathan et al. 2001) and is dependent on the verticaldistribution of aerosol (Hansen et al. 1997). The surfaceboundary condition is also crucial. For instance, Menonet al. (2002) use fixed SST and find that aerosol absorp-tion heats the atmosphere and enhances regional con-vection and the associated precipitation. This is becauseany solar radiation reaching the surface is assumed lostinto an infinite heat capacity ocean. In the fixed SSTcase, aerosol absorption produces heating in the atmo-sphere that is not compensated by changes in surfacefluxes and can thus affect precipitation. When the sur-face layer is allowed to equilibrate, as in the QTCMsimulations with a mixed layer ocean here, very differ-ent balances apply. An increase in solar absorption inthe atmosphere implies an equal reduction in down-ward insolation at the surface. When the surface layerequilibrates, this is compensated by reductions in theupward flux into the atmosphere by the sum of othersurface fluxes. The net effect on the column-integratedenergy input is zero. The leading-order impact in thefixed SST case is thus removed, as elaborated below,and any impacts on precipitation must come from moresubtle effects, such as the partition among evaporationand other surface fluxes. In the QTCM experiments,even though the solar absorption forcing (Fig. 2a) isalmost as strong as the TOA forcing (Fig. 2b), the pre-cipitation anomalies are much weaker (Fig. 5 comparedto Fig. 4a). Based on (6), F net� plays an important rolein determining low-level divergence associated withconvection. Since F�s 0 is often found over land and issmall under oceanic equilibrium, F net� � F�t . The TOAforcing associated with F�t becomes the leading role inmoist dynamical processes. The solar radiative fluxanomalies at the TOA are zero by definition in thesolar absorption forcing case, so the precipitationanomalies are small. The precipitation anomalies donot fit well with respect to the effect of aerosol solarabsorption in tropical convective zones, which would be

expected from one-dimensional radiative–convectivebalance (e.g., Cess et al. 1985; Ramanathan et al. 2001;V. Ramanathan 2004, personal communication). In thisbalance, convection does not have to transport as muchheat upward, so convection becomes weaker to com-pensate increased solar absorption in the atmosphere.The sign of the precipitation anomalies in the QTCMsimulations is consistent with this argument, but theanomalies are strongly offset in space from the regionsof strongest solar absorption (Fig. 2a).

Table 1 contrasts the three-dimensional (3D) caserelevant here with the one-dimensional (1D) case oftenconsidered for aerosol impacts for the case of land or anequilibrated mixed layer ocean where F�s 0. In the 1Dcase, F�t 0 in equilibrium due to longwave balancingaerosol shortwave effects. In the 3D case, F�t can havesubstantial local anomalies since wave dynamicsspreads temperature anomalies, so longwave balance oflocalized shortwave anomalies occurs over broad re-gions. Thus TOA aerosol impacts can be more impor-tant in 3D than in 1D, especially since they induce amultiplier effect via the (M� · v1)� and (Mq� · v1)�transport terms according to (12). However, in theaerosol absorption forcing case, such a multiplier effectdisappears when the solar radiation anomalies at TOAare zero by definition. Under this circumstance, re-gional tropical precipitation anomalies can be createdby evaporation, although with no multiplier effect. Fur-thermore, the evaporation is constrained by the surfaceheat flux balance and, thus, is just a fraction of theaerosol solar absorption with opposite sign. Therefore,the aerosol absorption impact on precipitation is lesseffective than that of aerosol reflection in convectiveregions when surface heat fluxes are in approximateequilibrium. In the 1D case, the changes in E� due tothe surface energy balance are the only way to changeprecipitation, so attention has focused on these, eventhough they are small compared to changes due totransport terms. In the solar absorption case in Fig. 5,

TABLE 1. Summary of moist static energy and moisture budgetbalances that make tropical regional precipitation anomalies be-have differently from a 1D radiative convective model, especiallywith respect to absorption. The case for land or equilibratedmixed layer with F�s 0 is used.

3D case 1D case

P� (Mq� · v1)� � �v · �q�� E� P� E�F�t (M� · v1)� � �v · �q�� F�t 0Tropical regional precipitation

anomalies mostly due to energyand moisture transports; littleaffected by aborption;E� secondary

Surface energy partitionaffects E�


the evaporation changes are less than 1 W m�2, so donot account for the precipitation anomalies.

c. Diagnosis

Before using QTCM to diagnose the aerosol effectsin the ECHAM4 simulations, other components of theQTCM runs, shown in Fig. 3, are compared to theECHAM4 simulations. In general, the variations of theECHAM4 runs (Fig. 6) are mimicked by the QTCMsimulations (Fig. 7), but the amplitudes of surface tem-perature, low-level moisture, and evaporation in theQTCM runs are slightly weaker than those in theECHAM4 runs. Note that the values of the shading inFigs. 7a and 7b are different from those in Figs. 6a and6b. The near-surface temperature is decreased globally,more so in the Northern Hemisphere than in the South-ern Hemisphere since most aerosols are emitted in theNorthern Hemisphere (Figs. 6a and 7a). The surfacetemperature anomalies are relatively smooth longitudi-nally. Comparing to the tropospheric temperatureanomalies (Figs. 1b and 3b), the surface temperature

anomalies are colder at higher latitudes where aerosolconcentration is high, but warmer (less negative) atlower latitudes.

Unlike the temperature anomalies, the moistureanomalies in the lower troposphere (Fig. 6b forECHAM4 and Fig. 7b for QTCM) are correlated withregions of mean convection (the thick dashed lines inFigs. 1 and 3), having the larger negative anomalies inconvective regions and much smaller negative anoma-lies over nonconvective regions. These spatial patternsof the moisture and temperature anomalies stronglysuggest that similar mechanisms to the global warmingcase might influence the regional tropical precipitationthrough interaction with moist dynamical processes, aswill be discussed later. Another interesting point is thatthe strongest response in moisture is often found overregions where aerosol concentration is not high. Thisfurther indicates that remote effects through moist dy-namical processes are active in the response to aerosolforcing.

Over regions where the precipitation anomalies are

FIG. 6. As in Fig. 1 except for other ECHAM4 variables: (a) near-surface temperature (1000 hPa), (b) low tropospheric moisture(1000–700 hPa), (c) Fs�, (d) F net�, (e) OLR, and (f) evaporation. The moisture is in energy units, i.e., with latent heat per unit mass,L, absorbed. Contour interval for (a) and (b) is 0.3°C with dark shading above 1.2°C and light shading between 0.9° and 1.2°C. Contourinterval is 4 W m�2 for (c), (d), and (f) and 2 W m�2 for (e), with dark shading above 8 W m�2 and light shading below �8 W m�2.


large, net surface flux F�s is smaller than other terms inthe moist static energy budget equation (Figs. 6c and7c) because evaporation, sensible heat, and net long-wave radiation balance the net solar radiation anoma-lies. This balance holds particularly well over land (overoceans, some Fs anomalies occur due to seasonal varia-tions). Essentially, the surface temperature anomaliesare just a by-product of the surface heat flux balanceand should not be viewed as a forcing. This is not likea case with strong ocean heat transports, which wouldyield large F net� associated with the surface tempera-ture anomalies. This surface heat flux balance is animportant step in the pathway of inducing the regionaltropical precipitation anomalies, but its net impact onclimate heating in the moist static energy budget issmall. Since F�s is small, F net� (Figs. 6d and 7d) is pri-marily associated with the net heat flux at the TOA, F�t ,in regional imbalances. Here F�t is dominated by theaerosol solar reflection, which is partially canceled byoutgoing longwave radiation (OLR) anomalies on a re-gional basis. On broader scales, such as the global scale,the aerosol solar reflection and OLR anomalies willcancel each other (Figs. 6e and 7e). To balance the

reduced downward solar radiation, evaporation is gen-erally suppressed (Figs. 6f and 7f) and its horizontaldistribution coincides well with the downward solar ra-diation anomalies at the surface (Fig. 2c; Liepert et al.2004). Evaporation anomalies have a direct effect as-sociated with moisture budget on the precipitationanomalies but, when F�s � 0, not on the low-level con-vergence � · v�1.

To further examine impacts of the aerosol forcing onregional precipitation, the moisture and moist static en-ergy budgets in QTCM are analyzed. The precipita-tion anomalies are determined by four terms on theright of (1). From the moisture budget analysis shownin Figs. 7 and 8, the combination of the first two terms(Mq� · v1)�: that is, the moisture divergence, dominatesthe regional tropical precipitation changes and its spa-tial pattern is very similar to the distribution of the pre-cipitation anomalies (Figs. 3a and 8a). This (Mq� · v1)�term is central to the dynamic feedback that can bediagnosed using the moist static energy budget to esti-mate � · v�1. In the Tropics, the direct effects of evapo-ration anomalies and anomalous horizontal transport ofmoisture �(v · �q)� (�6 W m�2) are not negligible in

FIG. 7. As in Fig. 6 except for the QTCM simulations and surface temperature in (a). Dark shading is above 0.9°C and lightshading is between 0.6° and 0.9°C.


the moisture budget (Figs. 8b and Fig. 7f), but they aresmaller than those effects via � · v�1 (Fig. 8a) in themoist static energy budget (�35 W m�2). Note that�(v · �q)� has a multiplier effect (Mq/M � 4) throughthe moist static energy budget (Yu et al. 1998) accord-ing to (12).

The value of � · v�1 can be calculated from M� · v�1 inthe moist static energy budget, which is similar to thedistribution of (Mq� · v1)� in the Tropics. Thus, thoseeffects associated with � · v�1 [the rhs of (6)] will beenhanced by a factor of Mq/M. For instance, �(v · �q)�� 6 W m�2 over Northeast Brazil, but it contributes 24W m�2 (6 � 4) of the regional precipitation anomaliesvia � · v�1. The analysis of the QTCM moist static en-ergy budget indicates that � M�� · v1, � [v · �(T � q)]�and F net� are all contributing to M� · v�1 (Fig. 8c), butthe importance of each term is different in differentlocations (see Figs. 8d, 8e, and 7d). In the Tropics, par-ticularly over the summer hemisphere, the term�(v · �T)� tends to be small (not shown), so the valuesof �[v · �(T � q)]� (Fig. 8e) are mainly determined by�(v · �q)� (Fig. 8b). Thus, �(v · �q)� is used in the dis-cussion instead of �[v · �(T � q)]�. The term�M�� · v1 is associated with the M� mechanism dis-cussed in CN04, while �(v · �q)� is associated with the

upped-ante mechanism. Negative values of �M�� · v1

(Fig. 8d) are found over convective regions, so the M�mechanism reduces precipitation over convective re-gions. Positive values of �(v · �q)� are found over mar-gins of convective regions, so the upped-ante mecha-nism enhances precipitation over those regions. The M�and upped-ante mechanisms discussed here are similarto those in the global warming case, but with oppositeeffects on regional tropical precipitation. This is be-cause the aerosol forcing cools the troposphere throughradiative feedbacks and wave dynamics, while green-house gases warm the troposphere. The colder tropo-spheric temperature (Figs. 1b and 3b) induces negativelow-level moisture anomalies in convective regions(Figs. 6b and 7b) through interaction with convection.The gross moist stability is increased due to the nega-tive low-level moisture anomalies producing � M�� · v1

(Fig. 8d), so the precipitation in convective regions isreduced by the M� mechanism. Controlled by differentmechanisms, such as evaporation, the low-level mois-ture over nonconvective regions is decreased less thanover convective regions (Figs. 6b and 7b). The gradientof the anomalous low-level moisture induces �(v · �q)�(Fig. 8b) where transport occurs from nonconvectiveregions to convective regions and thus positive precipi-

FIG. 8. As in Fig. 7 except for moisture and moist static energybudget terms (a) (Mq� · v1)�, (b) ��v · �q��, (c) M� · v1�, (d)�M�� · v1, and (e) ��v · � (q � T )��. Contour interval is 10 Wm�2 with dark shading above 30 W m�2 and light shading below�30 W m�2 for (a). Contour interval is 4 W m�2 with dark shad-ing above 8 W m�2 and light shading below �8 W m�2 for (b)–(e).


tation over these margins of convective regions by theupped-ante mechanism.

Here F net� plays two roles in processes changing re-gional tropical precipitation. The first role is associatedwith a local effect that induces local reductions in low-level convergence where there is increased solar reflec-tion by aerosols. The second effect is remote, commu-nicated by the anomalous temperature response. In asteady state, the net solar radiation should balanceOLR at the TOA on a global scale. Aerosols reflectsolar radiation back to space, so OLR must be reducedto balance the increased outgoing solar radiationanomalies on a global average. This remote effectyields colder temperature anomalies globally. The tro-pospheric temperature anomalies give �M�� · v1 and(v · �q)� terms locally via the M� and upped-antemechanisms.

5. Local versus remote impact experiments

The analysis above provides evidence that aerosolclimate forcing has both local and remote impacts onthe regional precipitation in the Tropics. The local im-pact is associated with local aerosol absorption and re-flection, so the impact should be found over regionswith high aerosol concentration. The remote impact isassociated with global distributions of temperature andmoisture anomalies, which induce the M� and upped-ante mechanisms found in the QTCM. To provide fur-ther tests of the hypothesized local and remote aerosoleffects respectively, several experiments were con-ducted. The first pair uses a target area over SouthAmerica. During the boreal winter (DJF), the SouthAmerican region has strong aerosol forcing (Fig. 2) andis also dominated by deep convection, so the SouthAmerican region is a good place to compare local andremote aerosol effects on regional precipitation. More-over, both climate models simulate the South Americanconvective zones relatively well, so South America is agood target region to examine both aerosol effects. Inthe first of this pair, solar reflection and absorption arespecified only within the South American region. Sincethe tropospheric temperature anomalies (not shown) inthis experiment are small (less than 0.2°C), the remoteeffects associated with the M� and upped-ante mecha-nisms are weak. Thus, the precipitation anomalies overSouth America are contributed mainly by the aerosollocal effects. The pattern of the precipitation anomaliesover South America is similar to M� · v�1, which is as-sociated with low-level convergence anomalies, that is,� · v�1. The low-level convergence anomalies over SouthAmerica are dominated by Fnet� (Fig. 9). ComparingFig. 9a to the control experiment in Fig. 3a, the negative

precipitation anomalies over the South American con-vective region are weaker than those in Fig. 3a, and thepositive precipitation anomalies near Northeast Brazildisappear. The local forcing creates negative precipita-tion anomalies but does not induce positive precipita-tion anomalies. Over land, F�s is close to zero, so F net� isdetermined by F�t , which is determined by the negativesolar radiation anomalies at TOA (Fig. 2b). The nega-tive F net� disfavors convection, so the associated pre-cipitation is decreased.

In the second pair of the South American experi-ment, the aerosol forcing associated with the solar re-flection and absorption is specified only outside the boxover South America (Fig. 10). This is used to examinethe aerosol remote effect on South America. Over thetarget region, positive (negative) precipitation anoma-lies are found over the east (west) (Fig. 10a). The hori-zontal distribution of temperature and moistureanomalies is similar to that in Figs. 3 and 7, but theamplitudes are slightly weaker. Thus, the M� andupped-ante mechanisms induced by the tropospherictemperature and moisture anomaly patterns are re-

FIG. 9. As in Fig. 3 except for the experiment with the aerosolforcing only within the box around South America. (a) Precipita-tion with contour interval 10 W m�2, (b) Fnet� with contour inter-val 4 W m�2, and (c) M� · v�1, with contour interval 4 W m�2.


sponsible for the anomalous precipitation over theSouth American region in this experiment. Positiveprecipitation anomalies over Northeast Brazil arefound in Fig. 10a but not in Fig. 9a. This indicates thatthe upped-ante mechanism is the dominant mechanisminducing the positive precipitation anomalies. The mag-nitude of the negative precipitation anomalies is lessthan half of those in Fig. 9a, so the local effect is largerthan the remote effect in producing the negative pre-cipitation anomalies over the South American convec-tive region. The sum of the South American precipita-tion anomalies in Figs. 9 and 10 is roughly equal to thatof the precipitation anomalies in Fig. 3. Overall, thepositive precipitation anomalies near Northeast Brazilare contributed mostly by the remote effect associatedwith the upped-ante mechanism, while the negativeprecipitation anomalies in the west are contributed byboth the aerosol local effect and the aerosol remoteeffect associated with the M� mechanism.

Another experiment examining the aerosol remoteeffect is performed with the prescribed aerosol forcingonly within a zonal band between 20° and 50°N, where

most aerosol emissions are taking place. Widespreadcold temperature anomalies occur over the Tropics(Fig. 11b), almost half the magnitude of those in the fullforcing case (Fig. 3b). Tropospheric temperatureanomalies due to the aerosol cooling effects at midlati-tudes are transmitted to the Tropics presumably bytransient heat fluxes limiting temperature gradients.The temperature anomalies then propagate throughoutthe Tropics by wave dynamics. The colder tropospherictemperature anomalies in the Tropics cause horizontalgradients of low-level moisture anomalies through theQE-mediated effect. The M� and upped-ante mecha-nisms then affect regional tropical precipitation. Thelargest precipitation anomalies in this experiment (Fig.11a) occur over the Tropics despite the aerosol forcingbeing set to zero. Comparing to the full experimentshown in Fig. 4, the midlatitude aerosol effect contrib-utes half of the precipitation anomalies over most tropi-cal regions. Some caveats on this experiment are thatgradients at the edge of the applied aerosol region ap-pears to affect transient moisture transport by winterstorms, contributing negative precipitation anomaliesin the Northern Hemisphere subtropics, such as in thetropical North Atlantic and western Pacific.

To further examine the remote effects associatedwith tropospheric temperature, an experiment is con-ducted in which the tropospheric temperature anoma-lies obtained from the full aerosol experiment shown inFig. 3b are specified in the convective scheme. The pat-tern of the precipitation anomalies in Fig. 12a is simi-lar to those in the full aerosol experiment (Fig. 3b),

FIG. 10. As in Fig. 9 except for the experiment with the aerosolforcing only outside the box around South America: (a) precipi-tation, (b) tropospheric temperature (850–200 hPa), and (c) low-level moisture (1000–700 hPa).

FIG. 11. As in Fig. 3 except for the experiment with the aerosolforcing only within 20°–50°N: (a) precipitation and (b) tropo-spheric temperature (850–200 hPa).


and the magnitude of the anomalies is also compa-rable over most regions except for those with high aero-sol concentration, such as South America and Africa.The associated low-level moisture anomalies are alsosimilar to those in Fig. 7b. This demonstrates that theQE-mediated impacts that depend fundamentally ondeep convection initiate low-level moisture changesand the change of the low-level moisture is not simplya Clausius–Clapeyron response to surface warming. Wealso note that surface temperature anomalies (notshown) in this experiment are less than half those in thecontrol, consistent with our claim that surface tempera-ture is essentially a by-product.

6. Discussion

The intermediate atmospheric model QTCM quali-tatively reproduces aspects of the ECHAM4 simula-tions of present-day minus preindustrial climateanomalies due to aerosol direct and indirect effectswhen forced by the aerosol-induced solar radiative fluxanomalies from ECHAM4. The widespread coolingdue to aerosol reflection of solar radiation is compa-rable in surface temperature anomalies and tropo-spheric temperature. Since there is no sea ice in QTCM,amplitudes of the temperature anomalies at high lati-tudes are weaker in the QTCM simulations than in theECHAM4 simulations. The spatial pattern of tropo-spheric temperature anomalies—which is key to someof the effects below—is well reproduced, with dynami-

cal effects smoothing and spreading the cooling relativeto the aerosol forcing. The associated precipitationanomalies present a much more complicated spatial dis-tribution. Negative precipitation anomalies are foundover convective regions, while positive precipitationanomalies occur over margins of convective regions.Thus, the aerosol effects on precipitation are complexon a regional scale.

We note some caveats on this approach. Cloudinessin the QTCM simulations is fixed so as not to doublecount cloud–radiative feedbacks, which are included inthe solar forcing from ECHAM4. In QTCM, the verti-cal profile of the temperature variations is prescribed,so the change of the vertical temperature structure in-duced by the aerosol absorption effects cannot be simu-lated. However, the large-scale temperature anomaliesin ECHAM4 tend to have a deep structure and this iscaptured.

A pair of experiments was conducted to approxi-mately separate impacts of aerosol absorption and re-flection, respectively, on the climate system. Both aero-sol absorption and reflection effects cool the surfacewith similar amplitudes, but the reflection effect ismuch larger than the absorption effect in cooling theclimate system and producing anomalous precipitation.This occurs because at the time scales considered here,both land and ocean are equilibrated to have a smallnet surface heat flux anomaly. Absorption thus mainlyaffects the partitioning of surface heat flux to the at-mosphere but does not affect the net flux heating of theatmospheric column since other surface fluxes compen-sate. It also does not change the TOA anomaly (in thedefinition used here), so it does not change the overallheat absorbed in the climate system.

We distinguish between local and remote effects ofaerosol forcing. Aerosol effects that involve the dy-namically direct impact of the change in radiative fluxeson the column impacting the energy budget (and hencevertical motion in the Tropics) tend to be highly local.On the other hand, tropospheric temperature anoma-lies tend to spread to broad spatial scales and this, inturn, has remote impacts on tropical precipitation in amanner akin to ENSO teleconnections (NCS03) orglobal warming (CN04). We note that the terms localand remote are most appropriate when the aerosolforcing is localized. If the aerosol were evenly distrib-uted, the effects would overlap spatially.

The local effect can be understood from the aerosolTOA forcing (Fig. 2b). Aerosols absorb and reflect so-lar radiation, so the solar radiation reaching the surfaceis reduced. This anomalous solar radiation is balancedby evaporation, sensible heat, and longwave radiationemitted by the surface associated with colder surface

FIG. 12. (a) Precipitation and (b) low-level moisture differencesbetween the QTCM PD and PI period for the experiment withmodified tropospheric temperature anomalies specified in theconvection scheme.


temperature, so the associated surface heat flux anoma-lies are close to zero. Thus, the net radiative fluxes atthe TOA, F�t , determine F net�, which results in regionalimbalances in the energy budget and induces the pre-cipitation anomalies through interaction with convec-tion. In convective zones, the partition between anoma-lous evaporation, sensible heat, and surface longwaveradiation is secondary since convection links moistureand energy budgets.

The remote effects are associated with two mecha-nisms: the anomalous gross moist stability (M�) mecha-nism and the upped-ante mechanism. The widespreadcold tropospheric temperature anomalies create a hori-zontal moisture gradient between convective regionsand nonconvective regions. In convective zones, thecolder temperature yields reduced low-level moisturein order to maintain convective quasi equilibrium, aprocess referred to as QE mediation (CN04; Neelin andSu 2005). Negative low-level moisture anomalies tendto increase the gross moist stability and stabilize theatmosphere. Therefore, convection is suppressed andnegative precipitation anomalies are induced over con-vective regions. This effect is referred to as the M�mechanism. Over nonconvective regions where mois-ture is controlled by different mechanisms, such as thebalance between divergence and evaporation, the low-level moisture does not decrease as much as in convec-tive regions. The mean circulation thus transports lessnegative low-level moisture from nonconvective re-gions into convective regions where negative low-levelmoisture anomalies are larger. The horizontal advec-tion of moisture enhances the low-level moisture at themargins of convective regions, strengthening convec-tion. Thus, precipitation is increased by this upped-antemechanism over margins of convective regions wherethere is strong inflow. As noted in CN04, uncertaintiesin M� are a potential source of climate model disagree-ment, so effects associated with the M� mechanismshould be regarded as model dependent.

Over regions with strong aerosol forcing coincidingwith convection, the local effect of aerosol contributesa large portion of the negative precipitation anomalies,while the M� mechanism contributes only a small por-tion of the negative precipitation anomalies. This isseen, for instance, in local versus remote experimentsover a South American target region. The upped-antemechanism is responsible for the positive precipitationanomalies over margins of convective regions, such asNortheast Brazil. The importance of the aerosol localand remote effects differs from region to region. Forinstance, over the Maritime Continent where the aero-sol concentration is low, most precipitation changes arecontributed by remote effects. An experiment that

specifies tropospheric temperature anomalies from thecontrol run within the convection scheme—and thusincludes only the remote effects—produces many re-gions of precipitation anomalies resembling those ofthe control. A surprising portion of the tropical tem-perature and precipitation response is reproduced byspecifying midlatitude aerosol forcing alone, indicatingthe importance of midlatitude to Tropics remote ef-fects.

Acknowledgments. This work was supported underNational Science Council Grant 92-2111-M-001-001,the National Science Foundation Grant ATM-0082529,and National Oceanic and Atmospheric AdministrationGrant NA05OAR4310007. Comments from Dr. ChrisBretherton and two anonymous reviewers were helpfulfor improving the quality of this paper.

REFERENCES

Allen, M. R., and W. J. Ingram, 2002: Constraints on futurechanges in climate and the hydrologic cycle. Nature, 419, 224–232.

Betts, A. K., and M. J. Miller, 1993: The Betts–Miller scheme. TheRepresentation of Cumulus Convection in Numerical Modelsof the Atmosphere, Meteor. Monogr., No. 46, Amer. Meteor.Soc., 107–121.

Cess, R. D., G. L. Potter, S. J. Ghan, and W. L. Gates, 1985: Theclimatic effects of large injections of atmospheric smoke anddust: A study of climate feedback mechanisms with one- andthree-dimensional climate models. J. Geophys. Res., 90,12 937–12 950.

Chou, C., and J. D. Neelin, 1996: Linearization of a longwaveradiation scheme for intermediate tropical atmospheric mod-els. J. Geophys. Res., 101, 15 129–15 145.

——, and ——, 1999: Cirrus detrainment–temperature feedback.Geophys. Res. Lett., 26, 1295–1298.

——, and ——, 2004: Mechanisms of global warming impacts onregional tropical precipitation. J. Climate, 17, 2688–2701.

——, ——, and H. Su, 2001: Ocean–atmosphere–land feedbacksin an idealized monsoon. Quart. J. Roy. Meteor. Soc., 127,1869–1891.

Feichter, J., E. Roeckner, U. Lohmann, and B. Liepert, 2004:Nonlinear aspects of the climate response to greenhouse gasand aerosol forcing. J. Climate, 17, 2384–2398.

Fu, Q., and K. N. Liou, 1993: Parameterization of the radiativeproperties of cirrus clouds. J. Atmos. Sci., 50, 2008–2025.

Hansen, J., M. Sato, and R. Ruedy, 1997: Radiative forcing andclimate response. J. Geophys. Res., 102, 6831–6864.

Harshvardhan, R. Davies, D. A. Randall, and T. G. Corsetti, 1987:A fast radiation parameterization for general circulationmodels. J. Geophys. Res., 92, 1009–1016.

Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van derLinden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 1995:Climate Change 2001: The Scientific Basis. Cambridge Uni-versity Press, 881 pp.

Jacobson, M. Z., 2001: Strong radiative heating due to the mixingstate of black carbon in atmospheric aerosols. Nature, 409,695–697.


Kaufman, Y. J., D. Tanré, and O. Boucher, 2002: A satellite viewof aerosols in the climate system. Nature, 419, 215–223.

Liepert, B. G., J. Feichter, U. Lohmann, and E. Roeckner, 2004:Can aerosols spin down the water cycle in a warmer andmoister world? Geophys. Res. Lett., 31, L06207, doi:10.1029/2003GL019060.

Lohmann, U., and G. Lesins, 2002: Stronger constraints on theanthropogenic indirect aerosol effect. Science, 298, 1012–1016.

——, J. Feichter, C. C. Chuang, and J. E. Penner, 1999: Predictingthe number of cloud droplets in the ECHAM4 GCM. J. Geo-phys. Res., 104, 9169–9198.

——, ——, J. E. Penner, and W. R. Leaitch, 2000: Indirect effectof sulfate and carbonaceous aerosols: A mechanistic treat-ment. J. Geophys. Res., 105, 12 193–12 206.

Menon, S., J. Hansen, L. Nazarenko, and Y. Luo, 2002: Climateeffects of black carbon aerosols in China and India. Science,297, 2250–2253.

Neelin, J. D., and J.-Y. Yu, 1994: Modes of tropical variabilityunder convective adjustment and the Madden–Julian oscilla-tion. Part I: Analytical theory. J. Atmos. Sci., 51, 1876–1894.

——, and N. Zeng, 2000: A quasi-equilibrium tropical circulationmodel—Formulation. J. Atmos. Sci., 57, 1741–1766.

——, and H. Su, 2005: Moist teleconnection mechanisms for thetropical South American and Atlantic sector. J. Climate, 18,3928–3950.

——, C. Chou, and H. Su, 2003: Tropical drought regions in globalwarming and El Niño teleconnections. Geophys. Res. Lett.,30, 2275, doi:10.1029/2003GL018625.

Penner, J. E., X. Dong, and Y. Chen, 2004: Observational evi-

dence of a change in radiative forcing due to the indirectaerosol effect. Nature, 427, 231–234.

Ramanathan, V., P. J. Crutzen, J. T. Kiehl, and D. Rosenfeld,2001: Aerosols, climate, and the hydrological cycle. Science,294, 2119–2124.

Roeckner, E., T. Siebert, and J. Feichter, 1995: Climatic responseto anthropogenic sulfate forcing simulated with a general cir-culation model. Aerosol Forcing of Climate, R. J. Charlsonand J. Heintzenberg, Eds., John Wiley and Sons, 349–362.

Rotstayn, L. D., and U. Lohmann, 2002: Tropical rainfall trendsand the indirect aerosol effect. J. Climate, 15, 2103–2116.

Stevens, B., J. Duan, J. C. McWilliams, M. Münnich, and J. D.Neelin, 2002: Entrainment, Rayleigh friction, and boundarylayer winds over the tropical Pacific. J. Climate, 15, 30–44.

Su, H., and J. D. Neelin, 2002: Teleconnection mechanisms fortropical Pacific descent anomalies during El Niño. J. Atmos.Sci., 59, 2694–2712.

——, ——, and C. Chou, 2001: Tropical teleconnection and localresponse to SST anomalies during the 1997–1998 El Niño. J.Geophys. Res., 106 (D17), 20 025–20 043.

Yu, J.-Y., and J. D. Neelin, 1994: Modes of tropical variabilityunder convective adjustment and the Madden–Julian oscilla-tion. Part II: Numerical results. J. Atmos. Sci., 51, 1895–1914.

——, C. Chou, and J. D. Neelin, 1998: Estimating the gross moiststability of the tropical atmosphere. J. Atmos. Sci., 55, 1354–1372.

Zeng, N., J. D. Neelin, and C. Chou, 2000: A quasi-equilibriumtropical circulation model—Implementation and simulation.J. Atmos. Sci., 57, 1767–1796.


local and remote impacts of aerosol climate forcing on

Documents