Aerosols, which are an important component ofthe earth atmosphere(Ramanathan et al., 2001a;Solomon et al., 2007), comprise a mixture of mainlysulfate, soil dust, carbonaceous material, and sea salt. Atmospheric aerosols are dispersed worldwide. TheIndo-Asian haze that was documented by the IndianOcean Experiment(INDOEX)spread widely acrossmost of the North Indian Ocean, South and SoutheastAsia(MÄuller et al., 2003; Ramanathan et al., 2001b). Biomass burning and dust aerosols from North Africa(Sahara Desert and Sahel regions)are distributed overmost of the subtropics(Swap et al., 1992; Chiapello et al., 1997; Goudie and Middleton, 2001). Asian dust(Zhang et al., 1998; Wang et al., 2000; Nakajima et al., 2003; Zhang et al., 2003; Arimoto et al., 2006; Mikami et al., 2006; Hsu et al., 2013; Xu and Ma, 2013) and anthropogenic aerosols can travel across the Pacificfollowing the jet stream into the North American continent(Uno et al., 2001; Gong et al., 2006; Zhao et al., 2006; Jiang et al., 2007). Recently, some studies havealso reported the properties of Arctic aerosols(e. g., Breider et al., 2014) and their relationship with Arcticwarming. The impact of these aerosols is becoming animportant environmental and climatic problem(Penner et al., 1992; Jaffe et al., 1999), given of their globaldistribution.

Aerosols affect the earth's climate system by altering the radiative properties of the atmosphere. Atmospheric aerosols influence the earth's radiation budget directly through scattering and absorbing solarradiation(Charlson et al., 1991, 1992; Miller and Tegen, 1998; Hayasaka et al., 2007; Liu et al., 2011) and indirectly through affecting cloud properties(e. g., Twomey, 1974; Twomey et al., 1984; Ackerman et al., 2000; Huang et al., 2006a, c, 2009, 2010). Studies ofthe optical properties of aerosols are crucial to fullyunderst and their radiative effects.

Different aerosols scatter or absorb sunlight tovarying degrees, depending on their optical properties. Although cooling(negative forcing)is a consequenceof the total direct effect of aerosols, different aerosolshave different effects. Absorbing aerosols such as dust(e. g., Overpeck et al., 1996) and black carbon(BC)(e. g., Gu et al., 2010)are able to heat the atmosphereby absorbing solar and thermal radiation(Jacobson, 2002; Menon et al., 2002a; Andreae and Gelencsér, 2006). In contrast, non-absorbing aerosols such as sulfate scatter solar radiation(Charlson et al., 1991, 1992;Kiehl and Briegleb, 1993; Posfai et al., 1999) and generate relatively weaker atmospheric heating effect thanthe absorbing aerosols.

In addition to the profound direct impact ofaerosols on the radiation budget of the earth-atmosphere system, the development of clouds in apolluted environment can also significantly affect theradiation budget, and changes in cloud properties canhave an influence on precipitation(e. g., Flossmann et al., 1985; Andreae et al., 2004; Kaufman et al., 2005a; Andreae and Rosenfeld, 2008; Jiang et al., 2011; Gu et al., 2012). Aerosols play a critical rolein the process of cloud formation, although the absorbing and non-absorbing aerosols affect clouds differently(Kaufman and Koren, 2006). Whereas absorbing aerosols prevent clouds from forming, nonabsorbing aerosols extend cloud lifetimes and are associated with enhanced cloud cover(Coakley et al., 1987; Platnick et al., 2000). Absorbing aerosols suspended near clouds are also believed to contribute tocloud evaporation(Hansen et al., 1997; Huang et al., 2010). However, numerous studies describe additionalmechanisms whereby absorbing aerosols may either reduce or increase cloud cover(Chen et al., 2000; Ackerman et al., 2000; Jacobson, 2002; Small et al., 2011). In contrast, non-absorbing aerosols, such as sulfate, are largely derived from anthropogenic activities and are concentrated mainly in the Northern Hemisphere. Their rapid increase in the atmosphere could have contributed significantly to the cross-equatorial sea surface temperature gradient, which substantially alterslow-latitude cloud, circulation, and rainfall(Hulme and Kelly, 1993; Williams et al., 2001; Rotstayn and Lohmann, 2002; Ackerley et al., 2011).

The effects of aerosols on radiation and climatethrough their interaction with clouds are complex and incompletely captured by climate models(e. g., Takemura et al., 2005, 2007; Suzuki et al., 2008). As suggested in the 5th Assessment Report(AR5)of the Intergovernmental Panel on Climate Change(IPCC; Stocker et al., 2013), scientific underst and ingof aerosol radiative effects is still at mid-low or lowlevels.

This review summarizes the progress made instudies of aerosol optical properties and their radiativeeffects. The review has concentrated on the literatureregarding aerosol optical properties, and radiative effects of the main types of aerosols(dust, carbonaceous, sulfate, and sea salt). Based on the latest literatureon aerosol optical properties and /or their radiative effects, previous studies have been traced and major results are taken into consideration. Section 2 focuseson the aerosol optical properties, and the aerosol radiative effects are discussed in Section 3. Summary and discussion are provided in Section 4. 2. Optical properties of aerosols

The interaction of aerosols with radiation is usually measured by aerosols' optical properties, e. g., thescattering coe±cient, absorption coe±cient, aerosoloptical depth(AOD), single scattering albedo(SSA), and AngstrÄom exponent(AE). The AOD is a parameter used to measure the magnitude of aerosol extinction due to scattering and absorption, integrated in thevertical column. It represents the e-folding length ofthe decrease in a direct beam when traveling throughthe aerosol layer. The SSA is a ratio of the scatteringcoe±cient to the extinction coe±cient and measuresthe relative importance of scattering and absorption. The aerosol effects on the radiation budget at the topof the atmosphere(TOA)switch from net cooling towarming at a certain value of the SSA, depending onthe local surface albedo(Hansen et al., 1997). The AErepresents the wavelength dependence of AOD, withhigh values of AE indicating small particles and lowvalues representing large particles.

As the uncertainties associated with aerosol radiative effects stem mainly from optical properties withhigh inhomogeneous horizontal and vertical distributions(Liu et al., 2009; van Donkelaar et al., 2010;Zhang and Tang, 2012), attempts to better underst and aerosol optical properties are underway(e. g., Zhou et al., 2011; Gao et al., 2012; Wang Na et al., 2013). Theinvestigation of aerosol optical properties has encompassed a range of complementary approaches, including in-situ monitoring(Nakajima et al., 1996; Dubovik et al., 2002; Anderson et al., 2003; Yan et al., 2010;Yang et al., 2012; Zheng et al., 2013; Che et al., 2014), satellite remote sensing(Huang et al., 2008a, b; Chen et al., 2010; Wang W. C. et al., 2010; Cai et al., 2011;Wang Zhao et al., 2013; Ahn et al., 2014), and numerical modeling(Takemura et al., 2002, 2003; Kinne et al., 2006; Textor et al., 2006; Han et al., 2010).

Table 1 summarizes some major ground-basedaerosol observation networks. Ground-based measurements can provide databases to validate satellite retrievals and model simulations. Among the groundbased aerosol networks, Aerosol Robotic Network(AERONET) and Skyradiometer Network(SKYNET)are two major long-term, continuous, and readily accessible public domain databases that have been usedin studies of aerosol optical properties. AERONETincludes information for 897 sites around the world, covering almost all major tropospheric aerosol regimes(Holben et al., 1998). SKYNET includes informationfor about 25 sites around the world(Nakajima et al., 1996; Campanelli et al., 2004). Both AERONET and SKYNET observations are collected via passive remotesensing depending on sunlight, and provide information on aerosols at the surface. As some compensation, Raman lidar and micro-pulse lidar, which areused in lidar networks such as Micro-Pulse Lidar Network(MPLNET), have been used to retrieve profilesof aerosol backscattering and extinction during bothday and night.

Table 1. List of main ground-based networks for aerosol observation

Table options

Satellite remote sensing is the best, and indeed, the only way to observe aerosols on the global scaledue to the short lifetime of aerosols and their complex chemical composition. To effectively determinethe properties of aerosols, some sensors such as Advanced Very High Resolution Radiometer(AVHRR)(Heidinger et al., 2002; Zhao et al., 2008) and Total Ozone Meteorological Satellite(TOMS)(Ginoux and Torres, 2003)have been designed to monitoraerosols from space. Satellite remote sensing nowincludes new and enhanced sensors such as Polarization and Directionality of the Earth's Reflectance(POLDER; Fougnie et al., 1999), Moderate Resolution Imaging Spectroradiometer(MODIS; Salomonson et al., 1989), and Multi-angle Imaging SpectroRadiometer(MISR; Fisher et al., 2014). The launch ofsatellite-borne lidars such as Geoscience Laser Altimeter System(GLAS; Zwally et al., 2002) and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation(CALIPSO; Winker et al., 2006)has further enabled profiling of vertical aerosol distribution. Thesesensors allow quantitative analysis of aerosol opticalproperties, especially AOD, and provide additional information regarding aerosol size, SSA, and refractiveindex. Such advanced sensors will also provide aerosolglobal distribution information, seasonal and interannual variations in the sources, optical properties, and the direct and indirect effects of aerosols.

Combining ground-based measurements and satellite remote sensing makes it possible to obtainrelatively reliable information regarding global aerosoldistribution. Data from AERONET and satellite remote sensing(Fig. 1)indicate that high aerosol loadsover the continents are found in and downwind of theregions with specific sources. For example, high AODvalues have been recorded in the east of China dueto the transport of Asian dust. Such observationaldata provide a basis for underst and ing the properties.

Fig. 1. Global distribution of AOD. The image was ob-tained by merging data from several satellite sensors withground-based sun photometer measurements(Andreae and Rosenfeld, 2008).

Figure options

It is extremely di±cult to observe aerosols byusing ground or satellite remote sensing techniques, when the aerosols are near or inside clouds. One wayto deal with this observational di±culty is to analyzean extended record of satellite measurements in a statistical sense(e. g., Kaufman et al., 2005b). The presence of aerosols inside clouds can be inferred by using other observable gaseous pollutants as a proxy forthe aerosol. Carbon monoxide(CO)is a good aerosolproxy because incomplete combustion produces bothCO and aerosols, with possible sources including forest fires, coal burning power plants, and fossil fuelpowered automobiles(e. g., Jiang et al., 2008; Su et al., 2011). Additionally, aerosol models can be usedas an effective method to offset the limitations of observations. From the early models with only simpleschemes for physical and chemical processes(Cooke and Wilson, 1996), a series of aerosol models havebeen developed, e. g., the dynamic aerosol(Remer and Kaufman, 1998), GLObal Model of Aerosol Processes(GLOMAP; Spracklen et al., 2005), and Spectral Radiation-Transport Model for Aerosol Species(SPRINTARS)models(Takemura et al., 2002; Takemura, 2012; Dai et al., 2014). Recently, the BeijingClimate Center atmospheric general circulation model(BCC-AGCM)has been used to produce simulationsof aerosol properties and radiative effects. For example, Zhang et al. (2012b)simulated the aerosol(including sulfate, dust, carbonaceous material, and sea salt)optical properties over globe(Fig. 2). Thehighest simulated AOD values(0. 4-0. 7)occurred overthe Sahara Desert, followed by Arabia in West Asia(0. 2-0. 4). The simulated pattern of AOD distribution is generally in good agreement with the satellitedata in Fig. 1 for all areas of the globe. For SouthAmerica, the AOD simulated by BBC-AGCM waslower than observation. For Africa, high AOD values were observed from the Sahara Desert to southernAfrica, whereas the simulated high AOD values onlyoccurred around the Sahara Desert. The pattern ofsimulated AOD over East Asia was in agreement withobservations, although the simulated values were lowerthan the observed. This discrepancy may arise fromthe parameterization of aerosols in the model. Thekey factors affecting the aerosol simulation include thel and surface properties, soil moisture, wind speed, and parameterization of the aerosol's emission, transport, and deposition.

Fig. 2. Global annual mean distributions of simulated(a)total AOD and (b)single scattering at 550 nm(Zhang et al., 2012b).

Figure options

In addition to satellite remote sensing, model simulation could also provide a global distribution ofaerosol properties, which enables the validation using the aerosol observational data, especially for theaerosol global distribution.

Unfortunately, both ground- and satellite-basedobservations and model simulations contain uncertainties. For example, satellite aerosol retrievals can becontaminated by thin cirrus, resulting in an overestimate of AOD by about 0. 02 ± 0. 005(mean ± st and ard deviation)(Kaufman et al., 2005b). Combination of observation and model simulation(i. e., dataassimilation)is a good way to reduce the uncertainties to some degree. Aerosol data assimilation(Schutgens et al., 2010a, b)may overcome the respectiveweaknesses of observation and simulation, minimizethe misfit between them, and produce an optimal estimate of aerosol optical properties.

Under whole-sky conditions, the annual and global average AOD was reported(Yu et al., 2006)to be 0. 191 ± 0. 017 over l and and 0. 126 ± 0. 046 overocean. Many studies have focused on aerosol optical properties over Asia(Xin et al., 2005; Huang et al., 2008a, b; Che et al., 2009, 2013; Ge et al., 2010;Wang X. et al., 2010; Bi et al., 2011; Liu et al., 2011;Zhang et al., 2012a; Qi et al., 2013; Alam et al., 2014;Gong et al., 2014), Africa(Diner et al., 2001; Eck et al., 2003; Kim et al., 2011; Queface et al., 2011), and South America(Zhang Y. et al., 2012; Rosáario et al., 2013). Among these studies, the AODs show different values depending on the local conditions, e. g., the distance between aerosol sources, characteristics ofl and surface, and meteorological conditions. The retrieval of AOD is relatively reliable by far. However, big discrepancy on the absorption property of aerosolsis presented, even for the same region. The basic reason for the difference in the absorption property maystem from the huge uncertainty on the single scattering albedo retrieval.

The SSA, which reflects the scattering and absorbing properties of aerosols, is related to the mixingof different chemical species in atmospheric aerosols. This involves the same type of particles(internal mixing)or different types(external mixing). The mixingstate has little effect on scattering(Chylek and Wong, 1995; McMurry et al., 1996); however, it can have agreat effect on the absorption e±ciency. For example, the absorption of a mixture of BC and transparentparticles is significantly higher with an internal ratherthan an external mixture, resulting in a smaller SSA(Jacobson, 2000, 2001). As seen in the BBC-AGCMmodel simulation(Fig. 2b), the highest SSA valueswere observed over Europe, reflecting the high sulfateloading, and the lowest values were over East Asia, South Asia, and South America, where carbonaceousaerosols are the main component. Determining aerosolabsorption properties is crucial for reducing the uncertainties associated with aerosol effects. The SSA values vary in different regions, and they are influencedby different aerosol components. As suggested by Xiaet al. (2006), the characteristics of SSA are differentin different urban locations, and therefore special urban aerosol models should be created and used withsatellite remote sensing in different urban regions.

The trends in aerosol optical properties, such asAOD and AE, have also been investigated in variousstudies(e. g., Kishcha et al., 2007; Meij et al., 2010;Hu et al., 2011; Yoon et al., 2012; Dong et al., 2013). A method using total direct solar radiation to retrieveAOD has been proposed and used to analyze its longterm trends(e. g., Qiu and Yang, 2000; Luo et al., 2001; Wang and Shi, 2010). Atmospheric aerosols areinfluenced to some degree by the development of local economies and production techniques, and they, inturn, affect local economic development and environmental conditions. Variations in atmospheric aerosolsare also partially induced by changes in atmosphericcirculation. Studies of AOD trends may also identifythe reasons for global dimming and brightening.

Despite the large number of studies, the opticalproperties of aerosols are still far from being understood(Sokolik and Toon, 1996), especially the absorption properties. With the increase in human activities, the components of aerosols over different regions havebecome more complex. In regions influenced to a different extent by human activities, the components ofaerosols and their optical properties are remarkablydifferent, which influence the climate in diverse ways. 3. Radiative properties of aerosols

Aerosols may influence the radiation budget ofthe earth-climate system. Three types of radiative effects have been postulated: direct, indirect, and semidirect; they are reviewed in Sections 3. 1, 3. 2, and 3. 3, respectively. 3. 1 Direct effect of aerosols

The growing awareness of the potential climateimpact of aerosols has resulted in a large research effort that has significantly improved our underst and ingof the role of aerosols in the earth's radiation balance. With a full set of aerosol optical properties available, the direct radiative effect(DRE)of aerosols can be calculated by using satellite-based(Huang et al., 2006c, 2010; Deng et al., 2010), model-based(Takemura et al., 2003, 2005; Wang et al., 2004; Gu et al., 2006;Wang Hong et al., 2010; Zhang et al., 2009, 2010;Wu and Han, 2011; Rap et al., 2013), satellite-modelintegrated(Yu et al., 2006; Huang et al., 2009), orground-model integrated(Ge et al., 2010; Liu et al., 2011; Wang W. C. et al., 2013)approaches. Table 2lists recent estimates of the aerosol direct solar effect and direct climate forcing(DCF). A brief descriptionof each is given in the table, and readers are encouraged to refer to the relevant reference for more details.

Table 2. Summary of seasonal and annual(ANN)average clear-sky DRE or DCF(W m^-2)at the top of theatmosphere(TOA) and surface(SRF)over the globe derived with different methods and data

Table options

Aerosols are derived from a range of sources, bothnatural and anthropogenic. Natural sources includevolcanic emissions, plant vapors, and chemicals released by sea creatures. Anthropogenic aerosols arecomposed of a variety of aerosol types and components including water-soluble inorganic species(e. g., sulfate, nitrate, and ammonium), condensed organicspecies, elemental or black carbon, and mineral dust(Penner et al., 1994).

As listed in Table 2, Yu et al. (2006)suggested that the annual mean DRE of natural and anthropogenic aerosols over l and at the TOA and surface(SRF)could reach -5. 5 and -13. 5 W m^-2, respectively(based on the integration of MODIS and Goddard Chemistry Aerosol Radiation and Transport(GOCART)databases). The annual mean DRE ofnatural and anthropogenic aerosols over ocean at theTOA and SRF could reach -6. 5 and -11. 1 W m^-2, respectively(based on the integration of MISR and GOCART databases). Given that the radiative forcing in the atmosphere is the difference between thevalues at the TOA and those at the SRF(e. g., Liu et al., 2011), the annual mean heating of natural and anthropogenic aerosols in the atmosphere over l and canbe as high as 8. 0 W m^-2, which is much higher thanthat over ocean(4. 6 W m^-2).

Since the early 1970s, anthropogenic aerosols havebeen assumed to offset a large portion of the warming resulting from anthropogenic greenhouse gases(GHGs). In the early 1980s, studies based on numerical models estimated that anthropogenic aerosolspartially offset more than half of greenhouse warming and this dominated the uncertainty associated withthe anthropogenic driving of climate change. As reported by the IPCC, despite the large uncertaintyranges regarding aerosol forcing, there is a high confidence that aerosols have offset a substantial portionof the forcing due to GHGs(Stocker et al., 2013). Thebest estimate of anthropogenic DCF is -0. 35 [-0. 85 to+0. 15] W m^-2(high confidence)according to the evidence from aerosol models with some constraints fromobservations(Stocker et al., 2013). Using SPRINTARS, a radiation-transport model for aerosols, Takemura(2012)simulated the distribution of the adjustedforcing due to aerosol direct effect at the SRF in 2000(Fig. 3), in which the adjusted forcing was defined asa change in the net irradiance after allowing the atmospheric and l and temperature, water vapor, clouds, and l and albedo to adjust to the prescribed sea surfacetemperatures and sea ice cover. The adjusted forcingdue to the aerosol direct effect at the SRF is currentlynegative over most regions, with the maximum negative adjusted forcing over Asia being greater than -5W m^-2. The negative effect due to aerosols at theSRF suggests the influence of some absorbing aerosols(e. g., BC)on solar and thermal radiation.

Fig. 3. Global distribution of adjusted forcing(W m-2)due to the aerosol direct effect at the surface in 2000(aver-aged for the period 1998{2002), as simulated by SPRINTARS(Takemura, 2012).

Figure options

Although the net effect of anthropogenic aerosolsis cooling(negative forcing)(e. g., Chen et al., 2011;Liu and Zhang, 2012), some components may contribute a warming effect(positive forcing). Jacobson(2000)showed high positive forcing from BC, with themagnitude exceeding that of forcing due to CH4, suggesting that BC may be the second most importantcontributor to global warming after CO2 in terms ofdirect forcing. Chung and Seinfeld(2005)showed(see Table 2)that different mixtures of anthropogenic BCcould generate different DCFs, with the annual meanDCFs of externally mixed anthropogenic BC at theTOA over the NH, SH, and the whole global beingpositive and reaching 0. 52, 0. 15, and 0. 33 W m^-2, respectively. Internally mixed anthropogenic BC couldinduce greater warming(with annual DCF values atthe TOA over the NH, SH, and the whole globe being 0. 93, 0. 28, and 0. 60 W m^-2, respectively). Asnoted in Section 2, the absorption property of internally mixed BC is significantly higher than that of externally mixed BC, resulting in lower SSA(Jacobson, 2000, 2001). The strong warming effect of BC derivesfrom its strong absorption property. Although heatingeffects are postulated, the warming effects of carbonaceous aerosols are still within a wide range(e. g., Zhang et al., 2009; Tian et al., 2013), which contributes to theuncertainties regarding the overall effects of aerosols.

Mineral dust, another absorbing aerosol, can havea significant effect on the earth's radiation budget, because it can both scatter sunlight back to space(leading to negative radiative forcing) and absorb solar and infrared radiation(leading to positive forcing)(Sokolik and Toon, 1996). It has been suggested that theDCF due to mineral dust may be comparable to theforcing by other anthropogenic aerosols, and on a regional scale, it can greatly exceed that due to sulfate. Dust aerosol may exert a negative(positive)DCF atthe SRF(TOA)by reducing the amount of incomingsolar radiation reaching the ground and locally heatingthe atmospheric column(Islam and Almazroui, 2012). As reported by Highwood et al. (2003), the effect onirradiance due to a dust outbreak over the Sahara wasa decrease of 6. 5 W m^-2 at the TOA and an increaseof 11. 5 W m^-2 at the surface. Over the TaklimakanDesert, Huang et al. (2009)found that the averagedaily mean net radiative effect of atmospheric dust was44. 4, 41. 9, and 86. 3 W m^-2, respectively, at the TOA, SRF, and in the atmosphere. As also shown in Fig. 4, over the Loess Plateau near the Taklimakan Desertin Northwest China, the average shortwave DRE ofdust for spring 2009 was 0. 68, -70. 02, and 70. 70 Wm^-2, respectively, at the TOA, SRF, and in the atmosphere. The dust aerosols heat the atmosphere duringdust events by about 2 K day^-1(daily average), whichis 60% larger than the heating(1. 25 K day^-1)inducedby background aerosols(Liu et al., 2011). With theFig. 4. Daily mean direct radiative forcing due to aerosolsat the surface(SRF), top of the atmosphere(TOA), and in the atmosphere(ATM)in spring 2009 over NorthwestChina(Liu et al., 2011). outbreak of dust events in Northwest China, especially in early spring, the mixture of local pollution(especially carbonaceous aerosols from heating)couldstrengthen the absorption properties of dust aerosols. The uncertainties in the DRE due to dust aerosols(e. g., Su et al., 2011; Ansell et al., 2014)are mainlydue to inaccuracies in measurement of dust aerosol optical properties, the SSA.

Fig. 4. Daily mean direct radiative forcing due to aerosolsat the surface(SRF), top of the atmosphere(TOA), and in the atmosphere(ATM)in spring 2009 over NorthwestChina(Liu et al., 2011).

Figure options

Additionally, over much of the world, biogenic organic aerosols have been shown to dominate the massof fine aerosols(Zhang et al., 2007; Jimenez et al., 2009). The properties of biogenic secondary organicaerosols have also been the focus of recent studies(e. g., Mentel et al., 2013; Han et al., 2014). The presence of biogenic organic aerosols can affect the earth'sradiative balance and cloud properties. As Scott etal. (2014)suggested, the annual global mean DREdue to biogenic secondary organic aerosols is between-0. 08 and -0. 78 W m^-2. However, the properties ofbiogenic secondary organic aerosols are still not wellunderstood.

Despite many advances in underst and ing the direct effect of aerosols, as summarized Table 2, largeuncertainties remain regarding several key variablessuch as surface albedo, aerosol particle size, verticaldistribution, optical depth, and the imaginary part ofthe refractive index(Liao and Seinfeld, 1998). 3. 2 Aerosol-cloud interactions

Clouds themselves are important regulators of theearth's radiation budget. Overall, clouds cool theearth-atmosphere system at the TOA. Losses of 48W m^-2 at TOA in the solar spectrum by clouds areonly partially compensated for 30 W m^-2 by cloudtrapped infrared radiation. Small changes to macrophysical(coverage, structure, and altitude) and microphysical properties(droplet size and phase)of cloudshave significant effects on climate. For example, a 5%increase in shortwave cloud forcing would compensatefor the increase in greenhouse gases during 1750-2000(Ramaswamy et al., 2001).

Aerosol particles serve as condensation nuclei forthe formation of cloud droplets and atmospheric iceparticles. Recent advances in this field have revealed amuch more complex interaction of aerosols with clouds(Rosenfeld et al., 2014). Aerosols affect cloud properties in various ways(see Fig. 5). Smaller clouddroplets, which are polluted by aerosols, when thereis a constant cloud water path, evaporate faster and cause more mixing of ambient air into the cloud top, which could further enhance the evaporation of clouddroplets(process A in Fig. 5). These small clouddroplets polluted by aerosols could induce an increasein cloud albedo, causing a huge cooling effect throughthe suppression of precipitation and extending of thecloud lifetime(process B in Fig. 5). Additionally, the aerosols may transport large quantities of smallice particles to the anvils of deep convective clouds, leading to a warming effect through reduction of thethermal radiation emitted to space(process C in Fig. 5). However, the contributions of individual processesto the overall aerosol-cloud interaction cannot be easily separated.

Fig. 5. Schematic diagram of the interaction of aerosols with clouds(Rosenfeld et al., 2014).

Figure options

Aerosol-cloud interaction includes the indirect radiative effect(IRE) and the semi-direct radiative effect. The IRE consists of two components: the albedoeffect(Twomey effect or the first indirect effect) and the cloud lifetime effect(the second indirect effect). Each affects the size distribution and chemical natureof atmospheric aerosols as well as the chemical composition of clouds and precipitation.

As shown in Table 3, almost all IRE values arenegative. The global annual IRE of anthropogenicaerosols at the TOA and SRF could reach -2. 9(Menon et al., 2002a) and -1. 8Wm^-2(Lohmann and Feichter, 2001), respectively. The annual IRE of anthropogenicaerosols over l and and ocean at the TOA could reach-4. 9(Menon et al., 2002a) and -2. 2Wm^-2(Rotstayn, 1999), respectively. Shen et al. (2011)modeled themonthly average IRE at the tropopause due to dustaerosols over China, and produced values of -1. 26, -2. 0, and -2. 69 W m^-2 in March, April, and May, respectively. In some regions, a value of -7 W m^-2 wassimulated. Su et al. (2008)postulated that the combination of indirect and semi-direct shortwave radiativeforcing for Asian dust was 82. 2 W m^-2, 78. 4% of thetotal dust effect. Huang et al. (2006a)suggested thatduring the growth of clouds in the presence of dustover East Asia, the instantaneous TOA net radiativeforcing for the no-dust region was about -208. 2 Wm^-2, 42. 1% larger than the value in cloud-over-dustregions. The presence of dust aerosols under cloudscould significantly reduce the cooling effect of clouds.

Table 3. Summary of the aerosol-cloud indirect radiative effect IRE(W m^-2)

Table options

Despite considerable advances in our underst and ing of the interactions of aerosols with clouds, furtherprogress has been hampered by limited observationalcapabilities and coarse-resolution climate models. Differences in the cloud microphysics scheme, especiallyin the auto conversion rate, have led to uncertainties inestimates of the indirect aerosol effect(Lohmann and Feichter, 1997; Jones et al., 2001; Menon et al., 2002a). For the indirect effects due to aerosols, most studies have only considered the cloud albedo effect(e. g., Jones et al., 1994; Lohmann and Feichter, 1997). Evenwhen both the cloud albedo effect and the cloud lifetime effect are considered, some discrepancy in theirimportance still exists. For example, Williams et al. (2001) and Kristjansson(2002)concluded that thealbedo effect at the TOA was four times as importantas the cloud lifetime effect, whereas Ghan et al. (2001)simulated a cloud lifetime effect that was larger thanthe albedo effect. In the following section, the albedoeffect and cloud lifetime effect are discussed separatelyin review of the aerosol indirect effect. The semi-directeffect is not included in this section but is discussedseparately in Section 3. 3. 3. 2. 1 Albedo effect

Because solar radiation is mainly scattered butonly minimally absorbed by cloud droplets, an increasein cloud condensation nuclei(CCN)at constant liquidwater content leads to a large concentration of smallradius cloud droplets. This enhances cloud reflectivity, rendering the radiative forcing negative. This aspectof the aerosol indirect effect is referred to as the firstindirect(or albedo)effect(Twomey, 1974; Twomey et al., 1984).

The Third Assessment Report(TAR)of the IPCCconcluded that the first indirect albedo effect of anthropogenic aerosol particles amounts to between 0 and -2 W m^-2 as a global mean(Ramaswamy et al., 2001). The IPCC later refined this range from -0. 5 to-1. 9 W m^-2(Lohmann and Feichter, 2005). The cloud albedo depends on both the clouddroplet size and the cloud thickness. Many studieshave investigated cloud droplet size(Brenguier et al., 2000, 2003; Schwartz et al., 2002)to study the impactof polluted clouds on albedo.

Most studies(e. g., Langner and Rodhe, 1991;Jones et al., 1994; Boucher and Lohmann, 1995)relate cloud droplet number concentration to the massof sulfate aerosols, assuming that the dominant sourceof CCN is sulfate aerosols, and estimate the change incloud albedo by calculating shortwave forcing with and without anthropogenic sulfur emissions. In the studies referred to above, the simulated indirect aerosoleffect including the first indirect albedo effect rangesbetween -0. 5 and -1. 6 W m^-2. Lohmann and Feichter(1997)fully coupled a sulfur cycle module to a cloudmicrophysics scheme and estimated the albedo effectof sulfate aerosols. Le Treut et al. (1998)obtained theindirect forcing by considering the albedo effect onlyin the range of -0. 4 to -1. 6 W m^-2. Lohmann et al. (2000)found that the change in radiative cooling bythe cloud albedo effect between present-day and preindustrial conditions ranged from near zero to -0. 4 Wm^-2 due to anthropogenic sulfate and -0. 9 to -1. 3 Wm^-2 due to anthropogenic carbonaceous aerosol.

Many studies have suggested that the cloudalbedo effect contributes most to the indirect effectdue to aerosols. For example, Lohmann and Feichter(1997)showed that 60% of the indirect forcing of -1. 4W m^-2 was due to the cloud albedo effect. However, there are very large uncertainties in the albedo effectdue to the presence of different types of aerosols inthe atmosphere. Lohmann et al. (2000) and Chuanget al. (2002)reported that the impact of carbonaceous aerosols on cloud albedo was 3-6 times of thatdue to sulfate aerosols. Conversely, Ghan et al. (2001)suggested that the cloud albedo effect was dominatedby sulfate. In addition to the model simulations, somestudies have considered the impact of aerosols on clouddroplet size(e. g., Huang et al., 2006c).

In future studies of the albedo effect of aerosols, the issues of(i)albedo effects of aerosols on differentcloud types, (ii)albedo effects of different aerosols, and (iii)a validation of the albedo effect between satelliteanalysis and modeling studies are key to underst and ing and reducing the uncertainties associated with theaerosol indirect effect. 3. 2. 2 Lifetime effect

The lifetime effect(Squires, 1958; Hudson and Frisbie, 1991)is an aerosol-cloud interaction mediatedby precipitation. It is based on the hypothesis thatchanges in the atmospheric aerosol lead to changes incloud droplet size, which then influence precipitationformation and the residence time of clouds. The distinguishing quality of the lifetime effect hypothesis isthe idea that the macrostructure of the cloud(such asits spatial extension or liquid water content)is determined by the e±ciency with which precipitation develops, which is in turn regulated(at least in part)bythe aerosol. This differentiates the cloud lifetime effectfrom the cloud albedo effect(Twomey, 1977).

Two approaches have been taken to studies ofthe lifetime effect hypothesis(Albrecht, 1989; Stevens and Feingold, 2009). The first considers statisticalmeasurements of the lifetime effect, either in observations or by enforcing relationships among the aerosol, clouds, and precipitation in large-scale models. Thesecond considers the effect to be hypothetical in order to test it or the various assumptions upon whichit is based, e. g., through dedicated field or modelingstudies that explore how individual clouds or fields ofclouds respond to changes in the ambient aerosol.

In 1966, a study of the relationship between cloudnuclei produced by cane fires and the cloud dropletconcentration was performed by using a cloud dropletsampler and thermal diffusion chamber installed inan aircraft(Warner and Twomey, 1967). It hadbeen suggested that aerosols suppressed the precipitation in warm(ice-free)boundary layer clouds(Albrecht, 1989). Therefore, many studies were initiatedto determine the cloud lifetime effect due to aerosols. Lohmann and Feichter(1997)included the cloud lifetime effect in their estimate of indirect forcing, and estimated that 40% of the forcing of -1. 4 W m^-2 wasdue to the cloud lifetime effect. Rotstayn(1999)suggested that the cloud lifetime effect of anthropogenicaerosols was -1. 0 W m^-2. Le Treut et al. (1998)included the cloud lifetime effect of aerosols by regardingit as a feedback.

Recently, many studies have investigated thecloud lifetime effect with high-resolution models and satellite data analyses(e. g., Suzuki et al., 2008; Small et al., 2009; Wang et al., 2012). Although large uncertainties still remain in estimations of the cloud lifetimeeffect, it should not be ignored in future calculationsof the indirect forcing due to aerosols. 3. 3 Semi-direct effect(cloud evaporation)

The term "semi-direct effect" was postulated byHansen et al. (1997)to describe the impact of absorbing aerosols on clouds, but relatively little attentionwas devoted to it until the INDOEX field campaign, which highlighted its importance(Ackerman et al., 2000). The absorption of solar radiation by aerosolsleads to a heating effect in the atmosphere, which canresult in the evaporation of cloud droplets. Such heating can partially offset the cooling due to the aerosolindirect effect. Conversely, Penner et al. (2003) and Johnson et al. (2004)indicated that the semi-direct effect could result in a cooling depending on the locationof the absorbing aerosols with respect to the clouds. Thus, there is a lack of consensus regarding the sign and magnitude of the global semi-direct effect(Morgan et al., 2006), as shown in Table 4, whereas estimatesof the direct and indirect effects are relatively moreconsistent. However, reducing the uncertainties associated with the semi-direct effect is important, as somestudies have found that the magnitude of the semidirect effect could exceed the direct forcing(Johnson et al., 2004).

Table 4. Summary of the semi-direct radiative forcing(W m^-2)

Table options

The semi-direct effect of absorbing aerosols hasbeen studied not only in small regional cloud simulations but also in global models, because aerosol heating can drive large-scale motions that affect clouds(Ackerman et al., 2000; Allen and Sherwood, 2010). Lohmann and Feichter(2001)provided the first assessment in the global annual mean semi-direct forcing and revealed significant reductions of liquid waterpath in regions where the absorbing aerosols burdenwas high. Huang et al. (2006b)postulated that thecloud water path of dust-contaminated clouds was considerably smaller than that of dust-free clouds becauseof the semi-direct effect of dust aerosols over East Asia(Fig. 6). Their results indicated that the mean ice water path(IWP) and liquid water path(LWP)of dustyclouds were less than those of their dust-free counterparts by 23. 7% and 49. 8%, respectively. Sakaeda etal. (2011)indicated that the total aerosol radiativeeffect of biomass burning aerosols over the southernAfrican/Atlantic region at the TOA was significantlyaltered by inclusion of the semi-direct effect. Theirresults also indicated that the semi-direct effect wasprimarily driven by changes in cloud cover and to alesser extent by changes in liquid water path over theocean and l and (Sakaeda et al., 2011).

Fig. 6. Comparison of the cloud water path over the pure cloud(CLD)region and the cloud over dust(COD)regionas a function of effective cloud top temperature(Te)for the(a)ice water path(IWP) and (b)liquid water path(LWP)(Huang et al., 2006b).

Figure options

When considering the effect of cloud evaporation, a 0. 17-K warming in the global annual mean surfacetemperature is expected due to the reduction in lowclouds, whereas the warming without cloud feedback is-0. 41 K(Hansen et al., 1997). Although many studies(e. g., Chylek et al., 1996; Hansen et al., 1997; Cook and Highwood, 2004; Johnson et al., 2004; Huang et al., 2006b, c)have identified a positive semi-direct effect, negative semi-direct forcing results have also beenreported(e. g., Penner et al., 2003).

For the absorbing aerosols, the semi-direct effectis an important aspect to consider alongside the directeffects. The semi-direct effects of aerosols in different regions are significantly different, and their signsmainly depend on the vertical profile of aerosols. Indifferent regions, the impact of aerosols on precipitation through the semi-direct effect varies, which is important for future studies of the hydrological cycle onboth global and regional scales. 4. Summary and discussion

In this review, we have summarized the majorprogress so far achieved in studies of aerosol opticalproperties and radiative effects. Main conclusions arepresented in the following two subsections, and important issues for future studies are highlighted respectively. 4. 1 Optical properties

Aerosol optical properties have been studiedthrough ground-based observation, satellite remotesensing, and numerical modeling. These approachescan also be used to validate each other. With ground- and satellite-based observations and model simulations, data assimilation is expected to reduce the uncertainties regarding aerosol optical properties. However, large uncertainties still remain, especially concerning the absorption properties of different aerosolsin different regions. Some important considerationsinclude:

(1)For in-situ measurement of aerosols, the representativeness of the extent of the observed opticalproperties at each station is an important issue.

(2)When using satellite data, validation withground-based observational data is required.

(3)Despite significant improvements in cloudscreening from satellite retrievals, it remains difficultto observe aerosols when they are near or inside clouds. Alternative methods, such as the use of an aerosolproxy and the use of aerosol models, have been developed to compensate for this observational weakness.

(4)Determining aerosol absorption propertieswhen the aerosols are polluted by absorbing aerosols, such as BC, is important to reduce the uncertaintiesassociated with aerosol radiative effects. 4. 2 Radiative effects

The aerosol radiative effects can be divided intothe direct effect on the radiative balance and the indirect effect through interaction with clouds. Aerosolradiative effects can be further divided into those thatexert a positive perturbation on the radiation budget and those that exert a negative perturbation.

Many studies have attempted to calculate the direct effect due to aerosols. Although the net directeffect of anthropogenic aerosols is cooling(negativeforcing), some components may contribute a warmingeffect(positive forcing). For simulation of the aerosoldirect effect, the uncertainties in the type of aerosols and their precursors, parameterizations of a variety ofsub-grid aerosol processes(e. g., wet, dry, and gravitational deposition, cloud convection), and assumptionsof aerosol size, absorption, mixture, and humidification of particles, should be further reduced.

Some studies have attempted to determine the indirect effect due to aerosols, whereas some have ignored the cloud lifetime effect(e. g., Jones et al., 1994;Lohmann and Feichter, 1997). A large number ofstudies have indicated that both the albedo effect and cloud lifetime effect act to cool the earth-atmospheresystem by increasing cloud optical depth and cloudcover, respectively.

Some important considerations for future studiesinclude:

(1)Determination of the cloud lifetime effect and the aerosol effect on mixed-phase and ice clouds is currently insu±cient and should not be neglected whenattempting to calculate the total radiative effect.

(2)When assessing the aerosol indirect and semidirect effects from data, the shift in emissions ofaerosols and their precursors may be a critical problem. In the future, due to climatic and environmentalchanges across the globe, determining the main emission sources will be crucial for predicting the aerosolindirect and semi-direct effects.

(3)To fully underst and the interaction of aerosolswith clouds, better observational capabilities and high-resolution climate models are needed.

Acknowledgments. We would like to thank the two anonymous reviewers for their valuable comments and suggestions.