Impact of aerosol indirect effect on surfacetemperature over East AsiaYan Huang*†, Robert E. Dickinson*, and William L. Chameides‡

*School of Earth and Atmospheric Sciences, Georgia Institute of Technology, Atlanta, GA 30332; and ‡Environmental Defense, New York Office,New York, NY 10010

Edited by James E. Hansen, Goddard Institute for Space Studies, New York, NY, and approved January 24, 2006 (received for review May 27, 2005)

A regional coupled climate–chemistry–aerosol model is developedto examine the impacts of anthropogenic aerosols on surfacetemperature and precipitation over East Asia. Besides their directand indirect reduction of short-wave solar radiation, the increasedcloudiness and cloud liquid water generate a substantial down-ward positive long-wave surface forcing; consequently, nighttimetemperature in winter increases by �0.7°C, and the diurnal tem-perature range decreases by �0.7°C averaged over the industrial-ized parts of China. Confidence in the simulated results is limitedby uncertainties in model cloud physics. However, they are broadlyconsistent with the observed diurnal temperature range decreaseas reported in China, suggesting that changes in downward long-wave radiation at the surface are important in understandingtemperature changes from aerosols.

anthropogenic aerosols � diurnal temperature range � long-wave radiativeforcing � regional climate change � second indirect effect

A tmospheric aerosols influence the climate directly by scat-tering and absorbing incoming solar radiation and indirectly

by acting as cloud condensation nuclei and�or ice nuclei, there-fore modifying the microphysics, radiative properties, and life-time of clouds. Consequently, they alter the net radiation bothat the top and bottom of the atmosphere (1–5). Since preindus-trial times, anthropogenic aerosols, consisting mainly of sulfateand carbonaceous aerosols [black carbon (BC) and organiccarbon], have substantially increased, especially over urban�industrial regions (6–8). This perturbation in aerosol concen-trations is believed to have had significant climatic impacts,especially at the regional scale (7, 9–11).

Recently, Zhou et al. (12), following the technique advancedby Kalnay and Cai (13), found a larger decrease in the diurnaltemperature range (DTR) over the industrialized parts of Chinausing the land-surface air temperature data recorded at 194meteorological stations of China from 1979 to 1998 than thatusing the National Centers for Environmental Prediction�Department of Energy Atmospheric Model IntercomparisonProject (AMIP)-II Reanalysis data (R-2) (14). The authorsinterpreted their results as an indication of the climatic effectfrom urbanization and�or land use changes through the modi-fications of boundary conditions. However, the aerosol indirecteffect includes changes in cloud properties, which possibly leadto a long-wave surface warming at night, in addition to daytimecooling from aerosol–solar radiation interaction. The decreaseof DTR has been hypothesized to result from increasing cloud-iness and, hence, the reduction of the daytime solar heating atthe surface (15–17). However, cloud cover has only increasedslightly in southern China (18). Could some other cloud changescontribute to the observed decrease of DTR?

To address the question, we present an attempt to use acoupled regional climate–chemistry–aerosol model to assess theeffects of anthropogenic aerosols on cloud properties and henceon regional surface temperature and examine how these effectsmight contribute to the nighttime temperature and DTR changesover East Asia. This region is one of the most populous andrapidly developing regions of the globe (19) and has a large

atmospheric loading of anthropogenic aerosols because of itsrapid industrialization, urbanization, and domestic heating (20,21). Although there is now extensive literature estimating theindirect radiative effect of observed aerosols (10, 22), relativelyfew of those studies have been coupled with the models of thechemical production and transport of the aerosols (e.g., refs. 23and 24).

Model ResultsTo assess the model-simulated aerosol impacts on climate, inparticular on the surface temperature over East Asia, we con-ducted two pairs of yearlong simulations (Table 1) spanning aperiod from June 1, 1994, to August 31, 1995, with the first 2months serving as spin-up for the subsequent 13-month period[see Fig. 5, which is published as supporting information on thePNAS web site, for the evolution of domain-averaged precipi-tation, cloudiness, and aerosol optical depth (AOD) showingthat the simulations have spun-up reasonably]. Two types ofautoconversion schemes are used in our experiments: the Kesslertype (25) (hereafter referred as KS69) and the Beheng type (26)(hereafter referred as BH94). In the two control runs (CONTand BHCONT), aerosol radiative and cloud effects are inactiveusing the KS69 and BH94 autoconversion schemes, respectively.In the experiment INDIR1, the direct and first indirect effectswere included using the KS69 scheme, whereas in the experimentBHIND2, only the second indirect effect was included using theBH94 parameterization, with the cloud radiation calculatedusing cloud liquid water content wL, and cloud effective radiusre kept at the same value as in BHCONT.

The simulated climatic conditions over East Asia and theircomparison with observations were discussed in detail by Giorgiet al. (27). They show a summer monsoon season with apredominant low-level southerly and southwesterly circulation, amaximum in precipitation and cloudiness, and a winter dryseason with prevailing westerly winds in the mid and highlatitudes and an easterly wind in the subtropical regions. Thesimulated monthly average temperature generally agrees withthe observations, with a cold bias of �1°C in summer andapproximately 2°C in winter. The model also captures theseasonal variations of the observed precipitation (which is low inwinter and high in summer) as well as the onset of the monsoonalrains in the spring, but it tends to overpredict the spring andsummer precipitation by 20�40%. The spatial and seasonalpatterns of the observed overall cloudiness as estimated fromNational Centers for Environmental Prediction�Department ofEnergy Atmospheric Model Intercomparison Project (AMIP)-IIReanalysis data (14) are generally simulated by the model, in

Conflict of interest statement: No conflicts declared.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: AOD, aerosol optical depth; BC, black carbon; DTR, diurnal temperaturerange; BH94, Beheng type experiment; KS69, Kessler type experiment.

†To whom correspondence should be addressed at: School of Earth and AtmosphericSciences, Georgia Institute of Technology, 311 Ferst Drive, Atlanta, GA 30332-0340. E-mail:yhuang@eas.gatech.edu.

© 2006 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0504428103 PNAS � March 21, 2006 � vol. 103 � no. 12 � 4371–4376

ENV

IRO

NM

ENTA

LSC

IEN

CES

GEO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

June

29,

202

0

particular over the continental region (see Figs. 6 and 7, whichare published as supporting information on the PNAS web site).However, the model underestimates cloudiness by 10% over theSichuan basin and a broad region covering the southern andeastern coast area of China.

Can the model reasonably simulate the aerosol distributionand abundance over East Asia? Fig. 1 illustrates the model-simulated AOD derived from online-simulated concentrationsof all aerosol species (sulfate, BC, and organic carbon) and theirradiative properties for January, April, July, and October fromthe CONT run. It shows that AOD is highest in January with abroad region of �0.4 extending from the Sichuan Basin to theeast coast of China and lowest in July with a region of �0.2extending from Sichuan Basin toward the northeastern area ofChina. This seasonal variation in AOD is largely due to theseasonality in the rate of wet removal, which in turn is caused bythe monsoon-driven variations in precipitation over the region(warm, wet summers and cold, dry winters) and, to a lesserextent, by the higher emission rates in winter as a result ofdomestic heating. To evaluate the simulated AOD, Fig. 2

presents a scatterplot between the model-simulated (fromCONT run) and observed annually averaged AODs over east ofChina (102E � 130E) where our model predicts the largestaerosol abundance. The observed AODs are taken from Zhouet al. (28) and were derived from surface solar irradiancemeasurements made at 31 meteorological stations from 1979 to1990. Fig. 2 reveals that the spatial pattern of AOD is reasonablyreproduced (R2 � 0.43), suggesting that the model is capable ofsimulating the spatial variation of aerosols over this region. Theslope of 0.55 indicates that the model underestimates the totalAOD by �45%. Because the calculated aerosol size hygroscopicgrowth neglects the unresolved variability of relative humidity,the AODs will be underestimated (e.g., ref. 29). The smallnegative intercept (�0.07) in the regression illustrated in Fig. 2

Table 1. List of experiments

Experiments Description

CONT Standard Model (KS69) without radiative andcloud effects of aerosols

BHCONT Control run (BH94) without radiative andcloud effects of aerosols

INDIR1 Direct and first indirect effects (KS69)BHIND2 Second indirect effect only (BH94; re remains

the same as BHCONT)

Fig. 1. Model-simulated AOD derived from on-line simulated aerosol column burden and the corresponding aerosol radiative properties, including sulfate,BC, and organic carbon for January 1995 (a), April 1995 (b), July 1995 (c), and October 1994 (d) (CONT experiment). Also shown in a is a subregion (22N � 35Nand 104E � east coast) covering the Sichuan Basin, southern China, and eastern coastal area of China where the aerosol loadings are largest.

Fig. 2. Annual mean AOD at 550 nm averaged over 31 stations of easternChina. Model-simulated AOD is derived from the aerosol column burden fromSeptember 1994 to August 1995 using the CONT run at the grids closest to thestation locations and observed AOD from surface solar irradiance measure-ment during 1979–1990.

4372 � www.pnas.org�cgi�doi�10.1073�pnas.0504428103 Huang et al.

Dow

nloa

ded

by g

uest

on

June

29,

202

0

likely arises from the inclusion of only anthropogenic aerosolsemitted from within the model domain in our simulations,whereas the observed AOD is derived from natural and anthro-pogenic aerosols emitted from both inside and outside the modeldomain. The model-simulated AODs over the western half ofChina (data not shown) are quite a bit smaller than observed,presumably because of the occurrence of other sources ofaerosols (i.e., dust) that are not included here (30).

Table 2 summarizes the model-calculated changes in cloudi-ness, net radiation at the top of the atmosphere, surface airtemperature at 2 m, and precipitation annually averaged over theentire model land region except for 12 lateral buffer grids forthe indirect experiments relative to their corresponding con-trol experiments (i.e., INDIR1 vs. CONT and BHIND2 vs.BHCONT). The direct and first indirect effects (INDIR1)produce a negative net radiative forcing at the top of theatmosphere of �4.6 W�m�2 and decrease the surface tempera-ture by �0.35°C. The spatial and seasonal distributions of thissurface cooling (mainly during daytime), with a peak over theSichuan Basin and winter, are seen in observational data asnoted (31, 32). Along with the atmospheric heating from BCabsorption (also included in INDIR1), the surface coolingincreases the atmospheric thermal stability and apparently in-hibits cloud development (there is less cloudiness at approxi-mately �0.008 and less cloud liquid water at approximately�2.3%) and decreases precipitation (�9.3%). The second indi-rect effect by itself (BHIND2) generates a negative net forcingof approximately �3.2 W�m�2 at the top of the atmosphere, butthe surface temperature actually increases a small amount, by�0.08°C. Evidently, this forcing term does not correlate very wellwith the surface temperature change. Precipitation is reduced by�20%, whereas cloudiness and cloud liquid water increase by�0.061% and �31.5%, respectively. These changes in clouds andprecipitation are consistent with the second indirect effect; i.e.,the reduction of cloud droplet size from anthropogenic aerosolslowers the coalescence efficiency between the relatively smallercloud droplets, leading to less precipitation and more cloudinessand cloud liquid water in the atmosphere. Detailed analysis ofaerosol effects on precipitation over East Asia can be found inHuang (ref. 33; available at http:��etd.gatech.edu�theses�available�etd-03222005-120424).

To explore the cause of the temperature increase obtainedin BHIND2, two additional runs of BHIND2 and BHCONTwith more frequent data output (hourly instead of 6-hourly)were carried out for the winter season when the simulatedtemperature increase was the largest. Table 3 lists the wintermean daily maximum and minimum temperature change (Tmaxand Tmin) and the surface solar and long-wave forcing duringthe daytime and nighttime averaged over the subregion cov-ering the Sichuan Basin, southern China, and eastern coastarea of China where the anthropogenic aerosol loadings arehighest (see Fig. 1a). This subregion is also similar to the studyregion of Zhou et al. (12), where nighttime warming and DTR

decrease were observed. The model-simulated temperatureincrease from BHIND2 during nighttime is �0.68°C, whereasduring the daytime it decreases slightly by �0.05°C. Thus, theDTR decreases by 0.73°C averaged over this subregion. Thissimulated temperature change asymmetry arises becausenighttime temperatures are more sensitive to the radiationchanges than the daytime temperature, with a nighttimesensitivity (the mean surface temperature response at night tothe surface radiative forcing) that is 0.08°C�(W�m�2). Becausesurface temperatures at night and during winter are morestrongly decoupled from those of the overlying atmospherethan that of daytime and summer, for the same amount ofradiative change, nighttime�winter temperature should show alarger change (e.g., refs. 34 and 35). Therefore, our modelresults suggest that the observed nighttime warming inferredfrom Zhou et al. (12) and its underestimation by the reanalysisdata could be caused by long-wave warming at the surfaceduring nighttime resulting from anthropogenic aerosols in-creasing the cloud cover and thickness. The net warming seenin Table 2, despite the negative net radiative forcing at the topof the atmosphere, is explained by the larger nighttime sen-sitivity (Table 3) compared with that of daytime to surfaceradiation.

To compare our results with the observed nighttime warmingand DTR decrease of Zhou et al. (12), an observed AOD trendfrom Luo et al. (36) is adopted, which is an AOD increase perdecade of approximately �0.08 during 1979–1990 over the abovesubregion. An estimate of the model-simulated DTR trend canbe obtained by multiplying the observed trend in AOD by themean DTR change per unit perturbation in AOD, assuming aconstant sensitivity of DTR to the surface radiative forcing. Oursimulated DTR change is �0.73°C, obtained from the simulatedmean anthropogenic AOD of 0.42 over the subregion. Thus, themodel-predicted DTR trend is �0.139°C per decade resultingfrom the aerosol second indirect effect. Given the differencesexpected between the simulated areal average and the observedstation average, this simulated DTR trend may adequatelycorrespond to the observed DTR decrease of �0.195°C perdecade (12).

The winter daily mean Tmax decreases over southeastern Chinaand Japan and increases over the northern and western part ofEast Asia (Fig. 3a), and Tmin increases over the entire continentalEast Asia and up to 1.8°C over the Sichuan Basin and north-eastern China (Fig. 3b). Thus, the DTR decreases by up to�1.5°C with the spatial pattern similar to that of Tmin (Fig. 3c).By dividing the land areas into dry and wet surfaces using theBowen ratio (�, a measure of the ratio of sensible heat to thelatent heat at the surface), it is found that Tmax decreases by�0.03°C over wet surfaces as expected by aerosol cooling(generally over southeastern China) but increases at �0.12°Cover dry surfaces (northwestern part), apparently for dynamicalreasons because aerosol and cloud effects are small over there

Table 2. Annual mean cloud fractional cover and cloud liquidwater path changes, net radiative forcing at the top of theatmosphere, and surface air temperature and precipitationchanges averaged over the interior model land region

Experiments � CLD� CLWP,

%NRF,

W�m�2 �T, °C �P, %

INDIR1-CONT �0.008 �2.3 �4.57 �0.35 �9.3BHIND2-BHCONT �0.061 �31.5 �3.23 �0.08 �20.4

CLD, cloud fractional cover; CLWP, cloud liquid water path; NRF, netradiative forcing at the top of the atmosphere. CLWP is weighted by cloudfraction. Surface air temperature was calculated at 2 m above the surface. Therelative change in percent is given for CLWP and precipitation.

Table 3. Second indirect effect on daytime and nighttimetemperature, surface solar and long-wave radiation in winteraveraged over the subregion of largest aerosol loadings forBHIND2 relative to BHCONT

Time �T, °CSWRFsfc,W�m�2

LWRFsfc,W�m�2

Daytime �0.05* �34.2 �9.2Nighttime �0.68† n�a �8.5Mean �0.31 �17.1 �8.8

SWRFsfc, surface solar radiation; LWRFsfc, surface long-wave radiation. n�a,not available.*Daily maximum temperature (Tmax).†Daily minimum temperature (Tmin).

Huang et al. PNAS � March 21, 2006 � vol. 103 � no. 12 � 4373

ENV

IRO

NM

ENTA

LSC

IEN

CES

GEO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

June

29,

202

0

(Fig. 1a). (Grid cells have been binned into ‘‘dry’’ when � � ��and ‘‘wet’’ when � � �� , where mean �� is averaged over theinterior land region.)

The negative solar forcing and positive long-wave forcing atthe surface (see Fig. 8, which is published as supporting infor-mation on the PNAS web site) are generally consistent with thatof the increased overall cloudiness. Fig. 4 shows an increase oflow-level and overall cloudiness averaged during daytime andnighttime in winter. The increase in the low-level cloudiness (Fig.4 a and c) appears generally over continental China and spatiallycorrelates with the aerosol distribution with a slightly largerincrease during nighttime. Conversely, the overall cloudiness(Fig. 4 b and d) increases over the entire domain with the largestincreases over eastern China and the adjacent ocean areas.

Summary and DiscussionThe aerosols simulated in this study produced changes insurface temperatures as a result of changes in the model’sclouds and radiation that were largest in winter and over theindustrialized parts of China. The simulated changes in cloudproperties for this period and region consequent to theinclusion of anthropogenic aerosols not only reduce the day-time solar heating at the surface but add to the nighttimedownward long-wave radiation, giving a nighttime warming of0.7°C and DTR decrease of �0.7°C. In addition to the cloudcover, the increased cloud liquid water from aerosol indirecteffects contributes to the observed decrease of DTR.

The indirect effects of aerosols on solar radiation noted inthis work are qualitatively similar to what have been reportedin studies of aerosol impacts on surface temperatures (10, 22);in particular, increases in cloudiness and cloud liquid waterwere modeled whose effects are to reduce the absorption ofsolar radiation. The consequences of such changes in cloudproperties on global average solar radiation may be uncertainby nearly an order of magnitude ranging from �0.3 to �1.4W�m�2, depending on cloud physics and treatment of aerosolindirect effects in the models (10, 22). Indeed, this aerosol–cloud interaction is recognized as perhaps the most complexand uncertain issue in addressing anthropogenic impacts onclimate change. Moreover, the nucleation of cloud droplets byaerosols, the coalescence of the cloud droplets into raindrops,and aerosol hygroscopic growth all occur on much smallerspatial scales than simulated by the meteorological models andinvolve variability of humidity on these scales (29, 37, 38). Thelast effect could result in a low bias of simulated AODcompared with the observation. It also is noted that thedescription of the linkage between aerosol concentration andcloud optical properties depends on multiple factors, i.e., clouddroplet activation, updraft velocity, etc. (39). However, phys-ically based methods to determine cloud droplet concentra-tions from aerosol concentrations and other relevant param-eters are not yet established to be any more successful thanempirical relationships such as that used in this work and havethe same qualitative behavior (40). In addition, the effect fromabsorbing BC particles could be quite variable in both sign andmagnitude, depending on their vertical location relative to theclouds (41). The entrainment of the dry overlying free tropo-spheric air by the cloud-topped boundary layer may diminishthe response of cloud water increase resulting from theaerosol-induced suppression of precipitation (42).

Overall, the parameterizations of cloud formation and pre-cipitation used in our model, as in all current climate models,may still have many unrealistic aspects. Application of the varietyof parameterizations currently recommended in the literaturewill give a wide range of results. As summarized in SupportingText, which is published as supporting information on the PNASweb site, repeating our simulations with the autoconversionscheme of Tripoli and Cotton (43) reduces the radiative forcingby a factor of �5. Conversely, Rotstayn and Liu (44) argued 60%reduction in their radiative forcing (from �0.71 to �0.28 W�m�2)using their newly developed autoconversion scheme. Thus, theagreement between the model-predicted and observed DTRmay be fortuitous. Further support (or disproof) for the hypoth-esis that the DTR has been changed by an increase of nighttimedownward long-wave radiation at the surface, as all estimates ofthe aerosol second indirect effect, will require advances inunderstanding how to include aerosol–cloud physics interactionsin a climate model. The purpose here has not been to advancethat understanding but to demonstrate a previously undescribedaspect of this complex situation, which could be important forinterpretation of the observed DTR changes. That is, it may benecessary to address not only how changes in aerosol and cloud

Fig. 3. Model-simulated changes in surface temperature and DTR at 2 m inwinter due to only the second indirect effect, daytime maximum temperature(Tmax) (a), nighttime minimum temperature (Tmin) (b), and DTR (c). Valuesare in °C.

4374 � www.pnas.org�cgi�doi�10.1073�pnas.0504428103 Huang et al.

Dow

nloa

ded

by g

uest

on

June

29,

202

0

microphysics affect solar radiation but also how they may changelong-wave radiation.

Model DescriptionThe framework of our coupled regional climate–chemistry–aerosol model is the National Center for Atmospheric Researchregional climate model (RegCM2), a limited area model devel-oped by Giorgi et al. (45, 46) with the augmentations describedby Giorgi and Shields (47) and Giorgi et al. (27). The coupledmodel was enhanced to include: a sulfur module based onKasibhatla et al. (48) with the revisions and enhancements ofQian et al. (49) and Tan et al. (50); a carbonaceous aerosolmodule to simulate their emissions, advection, removal, andchemical conversion processes (see Supporting Text for details);and the algorithms to simulate direct and indirect aerosol effects(33, 51). Qian et al. (49), Tan et al. (50), and Huang (33) providedetails about the aerosol module and its implementation ofradiative�cloud effect in a climate model (see Supporting Text forthe description of aerosol direct radiative effect).

The aerosol–cloud interaction (indirect effect) is estimatedusing the CCM3 (National Center for Atmospheric ResearchCommunity Climate Model) radiation package in our regionalclimate model. It calculates cloud radiation in terms of twoquantities: the cloud liquid water content (wL) and the cloudeffective radius (re). In the absence of anthropogenic aerosols, reis assigned to be 10 �m, a typical value for large-scale stratiformwater clouds with a background cloud condensation nucleipopulation (52). The parameterization for the first indirect effect(aerosols decrease the cloud droplet size and hence increase thecloud brightness) is represented in the model using an empiricalrelationship between the cloud droplet number concentration(Nc) and total hydrophilic aerosol concentration, then relatingNc to the cloud effective radius (re) (53, 54). The second indirect

effect (the reduction of cloud droplet size may increase the cloudlifetimes) was implemented by modifying the precipitation au-toconversion rate (Pautocv) for large-scale clouds to depend oncloud microphysical parameters (Nc or re) that are in turnaffected by aerosols. In the standard RegCM2 model, a Kesslertype of autoconversion rate is used to convert cloud water intorainwater with a conversion threshold (25). The KS69 schemewas replaced with an autoconversion rate of Beheng (26) thatalso includes Nc or re to simulate the second indirect effect (ref.23; see Supporting Text and also Fig. 9, which is published assupporting information on the PNAS web site).

The model domain encompassed 80 103 grid points(�4,800 6,200 km2) with a resolution of 60 km and is centeredat (34N, 120E), covering East Asia and adjacent ocean areas.The lateral boundary conditions to drive the model simula-tion were obtained from the analysis of observations fromthe European Center for Medium-Range Weather Forecast(ECMWF). A standard relaxation technique was used over abuffer zone of 12-grid point width, with greater forcing in themiddle and upper troposphere and weaker forcing in the lowertroposphere (46). Thus, the aerosol effect on a regional scale wasestimated by the spatial average over the interior model domain(see Fig. 10, which is published as supporting information on thePNAS web site), which excludes this buffer zone.

We thank Drs. Liming Zhou, Rong Fu, Athanasios Nenes, Irina Sokolik,and Michael Bergin (all from the Georgia Institute of Technology);Dr. V. Ramanathan (Scripps Institution of Oceanography, La Jolla, CA);and Dr. Filippo Giorgi (International Center for Theoretical Physics,Italy) for discussion and helpful comments. We also thank the reviewersfor helpful comments and suggestions that have much improved thismanuscript. This work was supported by National Aeronautics and SpaceAdministration Grant NNG04GB89G and National Science FoundationGrant ATM-0129495.

Fig. 4. Model-simulated low-level and overall cloudiness changes averaged during daytime and nighttime in winter due to only the second indirect effect,daytime low-level cloudiness (a), daytime overall cloudiness (b), nighttime low-level cloudiness (c), and nighttime overall cloudiness (d).

Huang et al. PNAS � March 21, 2006 � vol. 103 � no. 12 � 4375

ENV

IRO

NM

ENTA

LSC

IEN

CES

GEO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

June

29,

202

0

1. Coakley, J. A., Jr., Cess, R. D. & Yurevich, F. B. (1983) J. Atmos. Sci. 40,116–138.

2. Charlson, R. J., Schwartz, S. E., Hales, J. M., Cess, R. D., Coakley, J. A., Jr.,Hansen, J. E. & Hofmann, D. J. (1992) Science 255, 423–430.

3. Twomey, S. (1974) Atmos. Environ. 8, 1251–1256.4. Albrecht, B. A. (1989) Science 245, 1227–1230.5. Pincus, R. & Baker, M. (1994) Nature 372, 250–252.6. Houghton, J. T., Meira Filho, L. G., Bruce, J., Hoesung, L., Callander, B. A.,

Haites, E., Harris, N. & Maskell, K., eds. (1995) Climate Change 1994, RadiativeForcing of Climate Change and an Evaluation of the IPCC IS92 EmissionScenarios (Cambridge Univ. Press, New York), p. 339.

7. Kiehl, J. T. & Briegleb, B. P. (1993) Science 260, 311–314.8. Andreae, M. O. (1995) in Future Climates of the World, World Survey of

Climatology, ed. Henderson-Sellers, A. (Elsevier, Amsterdam), Vol. 16, pp.341–392.

9. Kiehl, J. T., Schneider, T. L., Rasch, P. J., Barth, M. C. & Wong, J. (2000) J.Geophys. Res. 105, 1441–2739.

10. Intergovernmental Panel on Climate Change (2001) in Climate Change 2001:The Scientific Basis, Contribution of Working Group I to the Third AssessmentReport of the Intergovernmental Panel on Climate Change (IPCC), eds. Hough-ton, J. T., Ding, Y., Griggs, D. J., Noguer, M., van der Linden, P. J., Dai, X.,Maskell, K. & Johnson, C. A. (Cambridge Univ. Press, New York), pp.289–348.

11. Ramanathan, V., Crutzen, P. J., Kiehl, J. & Rosenfeld, D. (2001) Science 292,2119–2124.

12. Zhou, L., Dickinson, R. E., Tian, Y., Fang, J., Li, Q., Kaufmann, R. K., Tucker,C. J. & Myneni, R. B. (2004) Proc. Natl. Acad. Sci. USA 101, 9540–9544.

13. Kalnay, E. & Cai, M. (2003) Nature 423, 528–531.14. Kanamitsu, M., Ebisuzaki, W., Woollen, J., Yang, S.-K., Hnilo, J. J., Fiorino,

M. & Potter, G. L. (2002) Bull. Am. Meteorol. Soc. 83, 1631–1643.15. Karl, T. R., Jones, P. D., Knight, R. W., Kukla, G., Plummer, N., Razuvayev,

V., Gallo, K. P., Lindseay, J., Charlson, R. J. & Peterson, T. C. (1993) Bull. Am.Meteorol. Soc. 74, 1007–1023.

16. Hansen, J., Sato, M. & Ruedy, R. (1995) Atmos. Res. 37, 175–209.17. Dai, A., Trenberth, K. E. & Karl, T. R. (1999) J. Clim. 12, 2451–2473.18. Kaiser, D. P. (1998) Geophys. Res. Lett. 25, 3599–3602.19. United Nations (1996) United Nations Statistical Yearbook (United Nations

Publication, New York), Issue 41, p. 886.20. Streets, D. G. & Waldhoff, S. T. (2000) Atmos. Environ. 34, 363–374.21. Chameides, W. L., Luo, C., Saylor, R., Streets, D., Huang, Y., Bergin M. &

Giorgi, F. (2002) J. Geophys. Res. 107, 10.1029�2000JD000208.22. Lohmann U. & Feichter, J. (2005) Atmos. Chem. Phys. 5, 715–737.23. Lohmann U. & Feichter, J. (1997) J. Geophys. Res. 102, 13685–13700.24. Takemura, T., Nozawa, T., Emori, S., Nakajima, T. Y. & Nakajima, T. (2005)

J. Geophys. Res. 110, 10.1029�2004JD005029.

25. Kessler, E. (1969) Meteorology Monograph Series (Am. Meteorol. Soc., Boston),Vol. 32, p. 84.

26. Beheng, K. D. (1994) Atmos. Res. 33, 193–206.27. Giorgi, F., Huang, Y., Nishizawa, K. & Fu, C. (1999) J. Geophys. Res. 104,

6403–6423.28. Zhou, X., Li, W. & Luo, Y. (1999) Sci. Atmos. Sin. 22, 418–427.29. Petch, J. C. (2001) Q. J. R. Meteor. Soc. 127, 2385–2394.30. Chameides, W. L., Yu, H., Liu, S. C., Bergin, M., Zhou, X., Mearns, L., Wang,

G. Kiang, C. S., Saylor, R. D., Luo, C., et al. (1999) Proc. Natl. Acad. Sci. USA96, 13626–13633.

31. Qian, Y. & Giorgi, F. (2000) Geophys. Res. Lett. 27, 3521–3524.32. Giorgi, F., Bi, X. Q. & Qian, Y. (2003) Climatic Change 58, 345–376.33. Huang, Y. (2005) Ph.D. thesis (Georgia Institute of Technology, Atlanta), p.

143.34. Zhou, L., Dickinson, R. E., Tian, Y., Jin, M., Ogawa, K., Yu, H. & Schmugge,

T. (2003) J. Geophys. Res. 108, 10.1029�2003JD004083.35. Pielke, R. A., Sr., & Matsui, T. (2005) Geophys. Res. Lett. 32, 10.1029�

2005GL024407.36. Luo, Y., Li, W., Zhou, X., He, Q. & Qin, J. (2000) Acta Meteor. Sin. 14, 490–502.37. Khairoutdinov, M. & Kogan, Y. (2000) Mon. Weather Rev. 128, 229–243.38. Rotstayn, L. D. (2000) J. Geophys. Res. 105, 15495–15507.39. Nenes, A. & Seinfeld, J. H. (2003) J. Geophys. Res. 108, 10.1029�2002JD002911.40. Haywood, J. & Boucher, O. (2000) Rev. Geophys. 38, 513–543.41. Menon, S. (2004) Annu. Rev. Environ. Resour. 29, 1–30.42. Ackerman, A. S., Kirkpatrick, M. P., Stevens, D. E. & Toon, O. B. (2004) Nature

432, 1014–1017.43. Tripoli, G. J. & Cotton, W. R. (1980) J. Appl. Meteor. 19, 1037–1063.44. Rotstayn, L. D. & Liu, Y. (2005) Geophys. Res. Lett. 32, 10.1029�

2004GL021922.45. Giorgi, F., Marinucci, M. R. & Bates, G. T. (1993) Mon. Weather Rev. 121,

2794–2813.46. Giorgi, F., Marinucci, M. R., Bates, G. T. & De Canio, G. (1993b) Mon. Weather

Rev. 121, 2814–2832.47. Giorgi, G. & Shields, C. (1999) J. Geophys. Res. 104, 6353–6375.48. Kasibhatla, P., Chameides, W. L. & St. John, J. (1997) J. Geophys. Res. 102,

3737–3759.49. Qian, Y., Giorgi, F., Huang, Y., Chameides, W. L. & Luo, C. (2001) Tellus 53B,

171–191.50. Tan, Q., Huang, Y. & Chameides, W. L. (2002) J. Geophys. Res. 107,

10.1029�2001JD000769.51. Qian, Y. & Giorgi, F. (1999) J. Geophys. Res. 104, 6477–6499.52. Briegleb, B. P. (1992) J. Geophys. Res. 97, 7603–7612.53. Hegg, D. A. (1994) J. Geophys. Res. 99, 25903–25907.54. Martin, G. M., Johnson, D. W. & Spice, A. (1994) J. Atmos. Sci. 51, 1823–1842.

4376 � www.pnas.org�cgi�doi�10.1073�pnas.0504428103 Huang et al.

Dow

nloa

ded

by g

uest

on

June

29,

202

0