Cloud parameters are most important factors affecting the global climate since they directly takepart in the radiative process in the earth's atmosphere. The parameters vary for different clouds and their spatiotemporal distributions directly changethe radiative balance between the earth and the atmosphere. Wetherald and Manabe(1988)pointedout that the radiative forcing effect by clouds ismore significant than the known greenhouse gases and aerosols. Moreover, cloud parameters are closelylinked with global water cycle. Cloud droplet size, one of the parameters, is usually used to describe intensity of the aerosol indirect effect because aerosolmakes the cloud droplet smaller with extended lifetime under certain conditions, which will modify thecloud radiative properties and lessen precipitation(Twomey, 1977; Albrecht, 1989; Ramaswamy et al., 2001; Lohmann and Feichter, 2005; Rosenfeld et al., 2007, 2012). Consequently, changing cloud parametersartificially is an important method in weather modification, such as seeding catalytic agent into cloudsto increase rainfall or to mitigate hail(Lei et al., 2008).

Our knowledge on global cloud characteristics hasbeen greatly improved since the start of the International Satellite Cloud Climatology Project(ISCCP)inthe beginning of the 1980s. The ISCCP collected signals of the visible and infrared channels obtained fromradiometers onboard the constellation of geostationary and polar-orbiting satellites and then retrieved cloudamount and cloud types(Schiffer and Rossow, 1983; Minnis et al., 1992; Klein and Hartmann, 1993; Rossow and Garder, 1993; Rossow and Schiffer, 1999; Li et al., 2003; Liu et al., 2003; Liu and Fu, 2009). With the progress in instruments, suchas the advanced visible and infrared sensor, microwave imager, CloudSat, Cloud-Aerosol Lidar, and Infrared Pathfinder Satellite Observation(CALIPSO), the technology for retrieval of cloud parameters hasbeen improved(Arking and Childs, 1985; Huang and Diak, 1992; Rao et al., 1995; MaKague and Evans, 2002; Stephens et al., 2002; Zhao and Weng, 2002; Winker et al., 2010; Wang et al., 2011), which provides a new opportunity to underst and the propertiesof cloud parameters. In particular, the observationsfrom 36 channels of the Moderate-resolution ImagingSpectroradiometer(MODIS)on Aqua and Terra, and 20 channels of the Medium Resolution Spectral Imager(MERSI)on FY-3 have promoted the studies on cloudparameters.

However, most radiometers do not have as manychannels as MODIS or MERSI. Usually, they have onevisible, one near-infrared, one intermediate infrared, and two thermal infrared channels, such as Visible and Infrared Scanner(VIRS)onboard Tropical Rainfall Measurement Mission(TRMM). Making good useof the observations from these satellites together withthe data from other instruments can promote the studies of cloud parameters. Based on this motivation, this paper summarizes the methods for retrieval ofcloud parameters, especially the bispectral reflectance(BSR)method, reviews the advances in recent studies on the aerosol indirect effects related to cloud parameters, and analyzes the relationship between thecloud parameters and precipitation properties(intensity, type, and structure). 2. The BSR algorithm

The spectrum algorithm for retrieval of cloud parameters involves inherent properties of clouds. It isknown that the scale of cloud particles generally variesfrom 1 to 100 µm, typically 10 µm for liquid cloud and 30 µm for ice cloud. These particles generate different electromagnetic radiation extinction from visibleb and s(0. 38-0. 78 µm), near infrared b and s(0. 8-3. 0µm), mid-infrared b and s(3-8 µm), to thermal infraredb and s(8-14 µm). For shortwave b and s(0. 4-4. 0 µm), especially the visible b and s, cloud particles scatter theincoming solar radiation to various degrees dependingon the particle size and phase, the incident angle ofsolar radiation, as well as the structure of cloud, whilethe absorbing effect of the particles could be neglected. For longwave b and s(4-100 µm), especially the thermal infrared b and s, the incoming radiation is affectedby the absorption/emission of liquid and ice cloud particles, which is mostly determined by the cloud temperature(related to the cloud height). For the thincloud, the upward radiance above it is also influencedby surface upwelling radiation.

During the radiative transfer in cloud particles, reflection and transmission of solar radiation by cloudsare determined by the effective radius(Re) and liquid water path(LWP)of cloud particles(Hansen and Travis, 1974). For the simplicity of radiative transfer calculation, Re, a parameter evaluating the sizeof cloud, is defined as the ratio of the cubic sum toquadratic sum of different cloud particle radii. It hasbeen revealed that cloud optical thickness(τ_c)is sensitive to LWP, Re, and the phase of cloud particles, while LWP is determined by cloud number density(i. e., cloud water content) and cloud thickness, and Re is influenced by the size distribution of cloud particles and the phase determined by cloud temperature.

The BSR algorithm is the most representativemethods for cloud parameter retrieval among numerous algorithms. It is first proposed by Twomey and Seton(1980)to evaluate τ_c and Re. Nakajima and King(1990)improved the method to retrieve τ_c and Refrom the MODIS observations. Generally, the BSR algorithm takes advantage of the characteristics of negligible absorbing effect in visible b and s and distinct absorbing effect in near infrared b and s to synchronouslyretrieve cloud τ_c and Re.

According to the radiative transfer theory, fora plane-parallel(one-dimensional)model, the spectral reflectance R_λ at a given wavelength(λ)in visible/infrared b and s is expressed as the function of τ_c, single-scattering albedo ω(varying with wavelength), asymmetry factor g(determining the scattering phasefunction), l and surface reflectance rs, solar directionangle ξ₀, and observation direction angle ξ, i. e., R_λ =f_λ(τ_c; ω; g; r_s; ξ; ξ₀). Hence, R_λ is a function of τ_c(1-g)(van de Hulst, 1980). Since the absorbing effect ofcloud particles in visible b and s is so small that the assumption of ω = 1 is valid, thus the variation of R_λmainly depends on τ_c(1 - g). In infrared b and s, however, R_λ is influenced by both τ_c(1-g) and ω becausecloud particles have both absorption effect and scattering effect in these b and s. Therefore, the first stepof the BSR is to acquire the simulated reflectance atthe two b and s with different Re and τ_c by using theradiative transfer model, and then create a lookup table of Re and τ_c from the reflectance of the two b and s. At last, the true Re and τ_c values are obtained bycomparing the observed reflectance with the simulatedone. Note that only absorption and scattering of cloudparticles to solar radiation are considered in the BSR, while the thermal radiative emission of cloud particles is ignored. If longer wavelength(such as 3. 7 µm)is considered in the retrieval, contributions from thethermal radiation of cloud particles, which can be derived through cloud temperature, should be taken intoaccount and deducted from the observations.

The reflectance relationship of channels at 0. 75 and 2. 16 µm for a given τ_c and Re is simulated by theradiative transfer model(see Fig. 2 in Nakajima and King, 1990). It is shown that the relationship of bothchannels is not independent when the reflectance islower than 0. 4, contrary to the independent relationship of the two channels at large reflectance(> 0. 4, corresponding to cloudy condition), i. e., an orthogonality relationship between Re and τ_c. These indicatethat the reflectance at 0. 75 µm increases significantlywith the increment of τ_c, while the reflectance at 2. 16µm is not sensitive to the τ_c variation. Similarly, withthe increment of Re, the reflectance at 2. 16 µm decreases gradually against relatively stable reflectanceat 0. 75 µm. Consequently, τ_c and Re can be obtainedby interpolating the reflectance observed by the twochannels.

After τ_c and Re are acquired by the above re-trieval method, LWP(in g m^-2 or kg m^-2)is calcu-lated by the function LWP =2/3ρτ_cRe, where ρ is liquid water density(g cm^-3; Arking and Childs, 1985; Han et al., 1994; Nakajima and Nakajima, 1995). It isworth noting that the capability of solar radiation atvisible/infrared b and s to penetrate through the cloudlayers is so poor that the retrieved τ_c and Re only represent the cloud parameters near the cloud tops.

Figures 1a-c show signals observed by visible, near infrared, and thermal infrared channels of Visible and Infra-Red Radiometer(VIRR)onboard FY-3over the Tibetan Plateau at 0645 UTC 3 July 2011. InFig. 1a, a bow cloud b and and many small cloud cellsto its south with over 0. 6 reflectance appeared over theTibetan Plateau. The bow cloud b and has brightnesstemperatures of 245-260 K at thermal infrared channels while the brightness temperature is lower than220 K for those cloud cells(Fig. 1c). If informationon the near and thermal infrared channels is joined together, it can be speculated that the bow cloud b and comprises clouds with mixed ice and water, and thosecells are composed of only ice particles because theyhave higher cloud top altitudes. By using the BSRmethod, the retrieved Re, τ_c, and LWP are plotted inFig. 2, which indicates that the bow cloud b and is puttogether by many cloud blocks with Re exceeding 20µm. Re for those cloud cells is larger than 35 µm. Ifwe compare the spatial distributions of both Re and LWP, τ_c exhibits more continuity. All the above re-sults illustrate a good observational ability of VIRRin detecting cloud structures near cloud top.

Fig. 1. Signals detected by(a)visible, (b)infrared, and (c)thermal infrared channels of VIRR onboard FY-3 at 0645UTC 3 July 2011.

Figure options

Fig. 2. (a)Re, (b)τc, and (c)LWP, retrieved by signals detected by visible and infrared channels of VIRR.

Figure options

On the basis of the BSR algorithm, triple-channel(such as 1. 6, 2. 1, and 3. 7 µm) and multi-channel algorithms are developed to retrieve the cloud parameters(Stone et al., 1990; Wielicki et al., 1990; Nakajima et al., 1991; King et al., 1992; Ou et al., 1993; Han et al., 1994; Rosenfeld et al., 1994, 2004; Nakajima and Nakajima, 1995; Platnick and Valero, 1995; Masunaga et al., 2002; Platnick et al., 2003; Chen Yingying et al., 2007). Using two near infrared channels(1. 6 and 2. 1µm) and a mid-infrared channel(3. 7 µm)from MODISobservations, Chen R. et al. (2007)obtained structureinformation along different height levels near the cloudtop and retrieved Re at different cloud tops, i. e., thevertical profile of Re. Based on MODIS observations, Ye et al. (2009)presented a retrieval scheme of τ_c and Re for multi-layer clouds. In this scheme, the radiativedatabases for τ_c and Re of the multi-layer cloud, watercloud, and ice cloud are established with Santa Barbara DISORT Atmospheric Radiative Transfer Model(SBDART), with consideration of various geometricalconditions, surface types, and atmospheric states. After the identification of clouds, cloud phase recognition, and multilayer cloud detection, τ_c and Re areretrieved by MODIS channel 1(0. 65 µm) and channel 7(2. 13 µm)data via lookup tables. Meyer and Platnick(2010)provided a new technique of pairingMODIS channels at 1. 38 and 1. 24 µm to evaluate theabove/in-cloud water vapor attenuation and retrievethin cirrus τ_c by such corrected attenuation. Nauss and Kokhanovsky(2011)proposed a novel method relying on asymptotic solutions to radiative transfer theory to acquire the information of τ_c, Re, the liquid and ice water paths, and so on.

The investigations on retrieval algorithms of cloudparameters during nighttime started from the 1990s, which are mainly based on measurements in long-wavelength b and s(Baum et al., 1994; Kubota, 1994; Strabala et al., 1994; Key and Intrieri, 2000; Baum et al., 2003). In principle, τ_c, Re, and phase of cloudare obtained by brightness temperature difference between infrared channels, such as one mid-infraredchannel(3. 7 µm; Ch3) and two thermal infrared channels(10. 8 and 12. 0 µm; Ch4 and Ch5)in the Advanced Very High Resolution Radiometer(AVHRR), and three thermal infrared channels(8. 0, 10. 8, and 12. 0 µm)in the High-resolution Infrared RadiationSounder(HIRS). These channels have different behaviors in absorbing and scattering for the same cloudparticles. Due to strong scattering effect at near infrared and mid-infrared channels(Ch3), the absorptivity and emissivity are far below 1. 0. But for Ch4 and Ch5, both absorptivity and emissivity are approximately 1. 0, as a result of small single scattering albedo, and then the scattering extinction can be negligible. Inthe condition of thick clouds with smaller Re, the emissivity at Ch3 is smaller than that at Ch4 so that thebrightness temperature difference between Ch3 and Ch4, i. e., BTD34, is negative and decreases with increasing Re. As for thin clouds, i. e., semitransparentclouds, the brightness temperature at Ch3 is higherthan that at Ch4 because temperature on l and surfaceis usually greater than that in cloud, which leads topositive values at BTD34. This is why BTD34 is usedto identify thin clouds. Generally, the sensitivity ofBTD34 is higher than that of BTD45(the brightnesstemperature difference between Ch4 and Ch5). Butit is convenient to directly use BTD45 in daytime because there is an additional brightness temperature atCh3 caused by reflected solar incident radiation(Inoue and Aonashi, 2000).

At present, the vertical structure of cloud is obtained from active detection onboard satellites, suchas cloud radar on CloudSat and lidar on CALIPSO. If the active detection is combined with spectrum observation, the ability to obtain cloud parameters inthe vertical direction will be enhanced. For example, Wu et al. (2009)revealed the cloud occurrence frequency at different altitudes by using combined observations of MISR, AIRS, MODIS, OMI, CALIPSO, and CloudSat. The multi-sensor combination has a goodability to detect multi-layer clouds. Hu et al. (2010)investigated the liquid water content, occurrence, and fraction of super cool water clouds through combining measurements of CALIPSO, IIR(Infrared Imaging Radiometer), and MODIS. Joiner et al. (2010)developed a relatively simple algorithm to detect multi-layer clouds and their vertical structure, using A-trainconstellation. Delanoö e and Hogan(2010)retrieved theice water content, effective radius, and extinction coefficient of ice clouds, based on the merged data derived from CloudSat, CALIPSO, and MODIS. Comparing four retrieval algorithms for ice cloud properties with data supplied by CloudSat, CALIPSO, and MODIS, Thorwald at al. (2011)concluded that microphysical assumptions in these algorithms need tobe refined. Base on high temporal resolution observations(15 min)of Spinning Enhanced Visible and Infrared Instrument(SEVIRI)aboard the Second Generation Meteosat, Kü uhnlein et al. (2013)proposeda semi-analytical cloud parameter retrieval method. Wang et al. (2013)presented the global distributionof liquid water in snowing clouds using observations ofMODIS and CloudSat. All above mentioned studiesillustrated the superiority of cloud detection by multiple sensors aboard multiple satellites and the improvement of cloud parameter retrieval methods. 3. Analysis of aerosol indirect effect by using cloud parameters

The impact of aerosol on cloud and precipitationis one of the most challenge problems. Among thestudies in this aspect, the Indian Ocean Experiment(INDOEX)in the 1990s is most representative, whichrevealed for the first time the effect of aerosol emissionfrom the urban region of the Indian subcontinent indry season on downwind clouds and precipitation overthe oceanic region. There appeared an obvious gradient in aerosol concentration, more in north and lessin south, along the wind direction from the subcontinent southward to the southern Indian Ocean. Notable gradients in liquid water content, cloud dropletsize, and other cloud parameters were also observedby airborne instruments and retrieval results(Liu et al., 2001). The interaction between the aerosol and clouds formed the most typical balance in atmosphericradiation in this region(Rhoads et al., 1997). Similarexperiments were carried out in East Asia and offshoreregions of Northwest Pacific, such as the APEX(AsianAtmospheric Particle Environmental Change Studies), ACE-Asia(Asia-Pacific regional Aerosol Characterization Experiments), and TRACE-P(Transport and Chemical Evolution over the Pacific)(Huebert et al., 2003; Jacob et al., 2003; Nakajima et al., 2003; Seinfeld et al., 2004).

The results of ACE-Asia have proved that themixture of dust, black carbon, sulfate, and nitrate inthe Asian Pacific area has caused the instability ofregional aerosol optical properties, and the clear skydirect radiative forcing in spring in this region far exceeds the global average(Remer and Kaufman, 2006; Yu et al., 2006). The joint aerosol observation experiment conducted by China and the USA in 2004 alsorevealed the unique role of urban haze aerosol on regional radiative forcing in Asia(Li et al., 2007; Xin et al., 2007).

In fact, the region offshore East China is an idealplace to investigate the interaction between aerosols and cloud parameters because westerly brings theaerosols that are emitted from inl and to this place. Asan example, Fig. 3 displays wind streamlines at 850hPa on 10 July 2001 in Northeast Asia and the distribution of mean aerosol optical depth issued by MODISfor 8-10 July 2001 over North China, Sh and ong, and Jiangsu. The air flows southward from the Sh and ongPeninsula to the Yellow Sea, as shown in Fig. 3. During this period, non-precipitation clouds over the Yellow Sea near south of the Sh and ong Peninsula wereobserved by VIRS(Figs. 4a-c). Figure 4d shows theretrieved Re by the BSR algorithm. The brightnesstemperature of the clouds in regions A and B is about280 K, and the reflectance at visible b and s is greaterthan 0. 5, which indicates that water clouds are popularin this case. The Re in the windward cloud belt(regionA)is 10 µm lower than in the leeward side(region B). It may be speculated that the size of cloud particles isreduced in region A due to impact of aerosols emittedfrom the Sh and ong Peninsula. The studies about theinteraction between aerosols and cloud parameters onthe eastern coast of China are ongoing.

Fig. 3. (a)Wind streamlines at 850 hPa on 10 July 2001 over Northeast Asia and (b)aerosol optical depth averagedfrom 8 to 10 July 2001.

Figure options

Fig. 4. Signals detected by(a)visible, (b)infrared, and (c)thermal infrared channels of VIRS; and (d)retrieved Re.

Figure options

The effect of aerosols on precipitation remains uncertain from the observational point of view. A few investigations suggested that precipitation is inhibitedby aerosols(Albrecht, 1989; Rosenfeld, 1999), whileothers found that precipitation intensity rises becauseof the aerosol effect(Koren et al., 2012). Disagreement remains about the aerosol effect on rainfall locations(Lowenthal and Borys, 2000; Givati and Rosenfeld, 2004; Lynn et al., 2007). Due to the differencesin aerosol hygroscopicity, shortwave absorptivity, and aerosol size, aerosols impact cloud parameters in different ways during the process of precipitation. Thatis why there are many forms of the aerosol effect onprecipitation(Paldor, 2008). It is also revealed thatmore small droplets and richer ice particles exist insidestratiform precipitating clouds in the region surrounded by dense mineral and dust aerosols, comparedwith the case in clean regions(Min et al., 2009).

In monsoon regions, the interactions among cloudparameters, aerosols, and monsoon is worth investigating. For example, how would the cloud parametersvary under the effect of aerosols on the monsoon activity? How will the changed cloud parameters impacton the monsoon activity? Preliminary studies haveshown that significant differences of cloud parameterssuch as cloud optical depth, cloud ratio, cloud height, and so on occurred over the central and northeastern India and the equatorial Indian Ocean, due to different aerosol effects on the Indian monsoon bursts and interruption(Kiran et al., 2009). As pointed outby Koren et al. (2012), our knowledge on the influences of aerosols on clouds(especially precipitation)isnot even close because such influences vary with manyfactors such as geographic location, season, and spatial and temporal scales. Furthermore, it is an antinomy to make simultaneous observations of cloud and aerosol. In cloudy sky, satellite spectrum is unable topenetrate cloud to obtain the aerosol information below the cloud, meaning great challenges to study thecloud-aerosol interactions in cloudy conditions. However, long-term observations from the multiple sensorsaboard satellites have the advantage of wide coverage and coherent measurement st and ards. If they are combined with other meteorological data such as ground-based observations and reanalysis data, it is possiblefor us to acquire useful information on cloud-aerosolinteractions. 4. Relationship between cloud parameters and precipitationIn the process from formation of clouds to occurrence of precipitation, knowledge on the actuallinkage between precipitation and cloud parameters islimited because of the complicated microphysics inside the cloud. Under the ideal hypothesis of cloudphysics, Ronsenfeld et al. (1989)suggested that thefeatures of cumulus cloud and the formation of precipitation can be identified by the relationship betweeninfrared bright temperature(T) and Re(Rosenfeld and Lensky, 1989; Rosenfeld et al., 1994; Lensky and Rosenfeld, 1997, 2003; Woodley et al., 2000). Thismethod provides an approach for us to study the relationship between cloud parameters and precipitation. Restricted by many factors, the hypothesis mentionedabove needs to be further tested by more experimentsas well as observations. Nowadays, the available information of cloud parameters can only be obtained byfinite flight observations(Woodley et al., 2003; Rosenfeld et al., 2006; Zhou et al., 2010), and study of thecloud microphysical processes depends more on thesimulations by numerical models(Hu et al., 1983; Xiao et al., 1988; Reisin et al., 1996; Guo, 1999; Lei et al., 2008).

To better underst and the relationship betweencloud parameters and precipitation, it is necessary toretrieve the cloud parameters together with the rainintensity and type at the same time. However, theconventional ground-based observations cannot provide all these properties. The combined observationsby ground-based precipitation radar and cloud radarmay be helpful to overcome this difficulty for a limited area. Using the almost simultaneous measurements from the precipitation radar(PR) and VIRSonboard TRMM, joined with the cloud parameter retrieval method mentioned above, may be a good wayto reveal the relationship between cloud parameters and precipitation over a larger region. For this purpose, the author merged the visible/infrared signalsmeasured by VIRS(st and ard product 1B01)with precipitation profiles derived from PR(st and ard product2A25), and obtained a new dataset of precipitationprofiles(with a horizontal resolution of 4. 5 km)in parallel with signals of visible/infrared in 5 channels(Fu et al., 2011). Then, Re and τ_c of cloud particles wereretrieved by the BSR method, based on this dataset. As a result, Fig. 5 shows the spatial distribution ofrain rate, Re, τ_c, and LWP in January 1988 in the tropics and subtropics. A good correlation between cloudparameters and precipitation is revealed, because PRcan distinguish effectively the precipitating cloud fromthe non-precipitating one. Figure 5 also clearly showsthe location of rain b and in equator and to its south, i. e., the ITCZ precipitation in winter. The mean intensity of the rain b and varies within 3-10 mm h^-1, with the maximum exceeding 10 mm h^-1. Outsidethe ITCZ, the mean rain intensity is small except forAfrica and South America, which are southward awayfrom the equator. The mean rain rate over East Chinais about 1 mm h^-1. The mean Re retrieved by the visible and infrared channels of VIRS varies between 10 and 50 µm, and the mean Re corresponding to largerprecipitation in the ITCZ can be more than 15 µm. At 36°N south of East China, the mean Re is usuallysmaller than 15 µm while it is greater than 15 µm inAfrica and South America(regions southward awayfrom the equator). The mean τ_c is larger than 50 and 90 in the ITCZ and in equatorial western Pacific, respectively. Although the mean rain intensity and Reat 36°N south of East China are small, the mean τ_c inthis region is very large. This may be related to thehigh aerosol content in this region, which needs to befurther analyzed. Because LWP is proportional to theproduct of Re and τ_c, its spatial distribution is similarto that of Re and τ_c, with more details omitted here.

Fig. 5. Spatial distributions of(a)rain rate, (b)Re, (c)τc, and (d)LWP in January 1998. The data are derived bymerging the precipitation radar(PR) and the visible/infrared scanner(VIRS)together with use of the BSR algorithm.

Figure options

To reveal the relationship between cloud parameters and the precipitation structure, a frontal cycloneobserved by PR and VIRS in the Jianghuai region at1417 BT(Beijing Time)22 June 2003 is taken as anexample(Zheng et al., 2004). Figure 6 shows the rainrate derived from PR, the visible reflectance and thermal infrared radiative temperature observed by VIRS, the Re and LWP retrieved by the BSR algorithm. Theprecipitation intensity near the cyclone center and thecold front reaches over 40 mm h^-1, while in the warmfront it is less than 8 mm h^-1. Corresponding to largerain rate in the center, lower thermal infrared temperature(mostly lower than 220 K)appears there, i. e., the cloud top is high or the cloud is thick around thecenter. The retrieved Re and LWP distributions revealthat some differences exist in cloud particle size and cloud water content between the cyclone center and the frontal area. The probability distribution function(PDF)of Re and LWP near the cloud top for theconvective and stratiform precipitation in the frontalcyclone is presented in Figs. 7a and 7b, respectively. It is indicated that Re varies mainly between 15 and 25µm while the peak of LWP for convective precipitationis 200 g m^-2, higher than that for stratifrom precipitation(Fig. 7b). It is believed that the liquid watercontent in convective precipitating cloud is larger thanthat in stratiform cloud for this frontal cyclone.

Fig. 6. (a)Rain rate, (b)brightness temperature at 10. 8 ¹m, (c)Re, and (d)LWP in the Jianhuai region(from orbit31925).

Figure options

Ideally, accurate descriptions of vertical distribution of cloud parameters and their relationship withthe intensity and type of rain will be helpful for precipitation parameterization in numerical models. However, it is difficult for us to obtain these descriptionsbased on the existing technology. Therefore, it may bea shortcut to examine the relationship between cloudparameters and precipitation vertical structures by analyzing the precipitation profiles associated with different Re and LWP values near the cloud tops. Inlight of PDF for Re and LWP in Fig. 7, we presentthe mean rain rate profiles for convective and stratiform precipitation with different Re and LWP valuesin Fig. 8. For convective precipitation, the mean profile under high Re and LWP displays the deepest precipitating clouds and the highest surface rain rate. Incontrast, the mean profile for low Re and LWP showsrelative shallower precipitating clouds and lower surface rain rate. However, this relationship is not robustfor the stratifrom precipitation. The Re and LWP nearthe top of stratiform precipitating clouds can only indicate the intensity of surface rain rate but not thethickness of cloud.

Fig. 7. Probability distributions of(a)Re and (b)LWP in a frontal area.

Figure options

Fig. 8. Mean profiles of(a, c)convective and (b, d)stratiform clouds under di®erent Re and LWP values in a frontalcyclone.

Figure options

For statistical significance, large samples of bothconvective and stratiform precipitation cases are examined in the following three regions, and the results are plotted in Fig. 9. The three regions include Jianghuai(23°-34°N, 117°-119°E), South China(25°-29°N, 116°-119°E), and the warm pool(WP)area of the western Pacific(0°-2°N, 140°-150°E). Summer cases from 1998 to 2007 are considered. Inview of the differences in the microphysical processesbetween water cloud(temperature higher than 268 Kat 10. 8-µm channel, usually named as shallow precipitating cloud) and non-water cloud(ice cloud and mixed cloud of ice and water, named as deep precipitating cloud here), precipitating clouds of convective and stratiform cases are classified into two sub-types, water cloud and non-water cloud. For water cloud, Re of convective and stratiform precipitation decreaseswith increasing rain rate in the three regions, especially in the WP area. This may result from the limited vertical extent in water cloud, which suppressesthe cloud particle growth. The mechanism of thiskind of precipitation still needs to be further studied.

For deep precipitating cloud, Re of convective precipitation in the WP area seems not varing with increasing precipitation intensity. This may be causedby the process of droplet increase such as coagulationmainly happens in the lower and middle of the cloudlayer, while particles near the cloud top remain almost stable. Re of stratiform precipitation decreaseswith increasing rain rate in the WP region, possiblydue to the weak updraft inside the oceanic stratiformclouds. Ordinarily, oceanic stratiform precipitation isgenerated from dissipating stage of the convective lifecycle. The weakening updraft makes large dropletsmove down quickly with the earth's gravity. Consequently, the surface rain rate increases while Re nearthe cloud top decreases. In Jianghuai region and SouthChina, Re of convective precipitation increases withincreasing surface rain rate, which may be causedby the stronger updraft bringing larger droplets inthe middle or lower cloud layer up to the near-toplayer. In the circumstances of deep stratiform precipitating cloud in both regions, Re in South Chinaalso increases with increasing surface rain rate. Relative large Re near the cloud top may be caused bythe strong updraft forced by mountainous topographythere. A unique relationship between Re and surfacerainfall intensity for stratifrom precipitation occurs inthe Jianghuai region. As surface rain rate is less than2 mm h^-1, Re becomes smaller with increasing rainrate; afterward, it becomes larger slowly with increasing rain rate. The mechanism of this is still unclear. Moreover, Fig. 9 also shows that for the same surfacerain rate, Re of convective and stratifrom precipitationin the WP area is 3-5 m larger than that over inl and areas, the minimum Re of the deep and shallow convection and shallow stratiform precipitation appearsin the Jianghuai region. Whether the difference of Rebetween l and and ocean is caused by the aerosol indirect effect remains to be further studied.

Fig. 9. The relationship between Re and rain rate for(a, b)water cloud and (c, d)non-water cloud in summer overthe Jianghuai(JN), Huanan(HN), and the warm pool(WP)area of western Pacific, respectively. (a, c)Convectiveprecipitation and (b, d)stratiform precipitation.

Figure options

In terms of the relationship between LWP and surface rain intensity, as shown in Fig. 10, the largerthe surface rain intensity is, the enhanced LWP theclouds have, which is reasonable in physics. This relationship prevails in precipitating clouds with waterphase or non-water phase. But there are still regionaldifferences. For the same surface rain intensity, theLWP in water clouds in Jianghuai and South Chinais 0. 2-0. 4 kg m^-2 higher than that in the WP area. In deep precipitating clouds, increased LWP is alsoneeded in clouds over Jianghuai, especially for the convective precipitation, at the same surface rain rate. The underneath mechanism needs to be studied in detail.

Fig. 10. As in Fig. 9, but for the relationship between LWP and rain rate.

Figure options

Due to the inverse relationship between the radiative temperatures at the thermal infrared channel10. 8 µm and at the cloud top height, temperaturesat this channel can be used to represent the heightof cloud top. Figure 11 shows the variation of LWPwith the temperature at this channel for convective and stratiform precipitation, which may be regardedas the LWP distribution with the height of cloud top, i. e., the LWP varies with different thicknesses of theprecipitating cloud. In Jianghuai and South China, the LWP increases with temperature decreasing from280 to 220 K(i. e., rise of cloud top)at the 10. 8-µmchannel. When the temperature is below 220 K, theLWP remains unchanged. In the WP area of the western Pacific, the LWP increases with the rising cloudtop. The above analysis indicates the differences ofcloud water content in vertical direction with the cloudtop height between l and and ocean. Furthermore, asthe temperature of cloud top is higher than 220 K, themaximum and minimum LWPs occur in the Jianghuai and WP area, respectively, and moderate LWP inSouth China, for the same cloud top height. Theseresults suggest that cloud water content varies in different regions, which directly impacts the surface rainintensity. The more detailed studies in this aspect arestill undergoing.

Fig. 11. The relationship between LWP and brightness temperature at 10. 8 ¹m in summer over Jianghuai(JN), Huanan(HN), and the warm pool(WP)area of western Pacific, respectively. (a)Convective precipitation and (b)stratiformprecipitation.

Figure options

It is well known that the first PR together withVIRS and other instruments are onboard the TRMMsatellite. The effective way is to merge the observations measured by these instruments with the cloudparameters retrieved from the algorithms based onspectrum observations. Then, a new dataset containing rain types and precipitation profiles corresponding to their cloud parameters will be set up, whichcan overcome nonsynchronous shortages of precipitation properties and cloud parameters in the past. Thisdataset will help us to solve the problems such as thedifference of cloud parameters between precipitating and non-precipitating clouds, the relationship betweenrain intensity and cloud parameters, and so on. Thisis the author's primitive motivation to write this article. More detailed studies are on going in ruggedways. 5. Conclusions

Using observations of visible and infrared channels to retrieve cloud parameters is and will still be adominant approach to underst and natures of the cloudsystem, since these channels remain to be the mainworking way in spectrum instruments onboard geostationary and polar orbit satellites, especially withconsideration of the advantage of high temporal resolution in geostationary satellites. Consequently, taking full advantage of the data retrieved from visible and infrared measurements, developing the cloud parameter retrieval methods, and integrating the datasupplied by other instruments such as precipitationradar and cloud radar, have great significance to thestudying of cloud properties, their radiative effects, aerosol indirect effects, and so on. These are helpfulto improve numerical models and enhance their abilities in weather forecast and climate prediction.

In this paper, we have reviewed the principle ofthe BSR retrieval algorithm for cloud parameters, and displayed the BSR-retrieved cloud parameters for acase observed by VIRR. The domestic and foreign advances in studies of aerosol indirect effects from theperspective of cloud parameter analysis in recent yearsare summarized. Moreover, a case is used as an example to present the impact of inl and aerosol on cloudparameters in offshore China. The results show thatthe cloud particle effective radius is reduced by 10 µmdue to the aerosol influence.

Moreover, the relationship between cloud parameters and rain intensity, type, and structure is introduced. The PR and VIRS data are merged, togetherwith the cloud parameters retrieved by the BSR. Theresults indicate that Re of precipitating clouds in thetropics and subtropics varies from 10 to 50 µm, and τ_cof heavy precipitating cloud is larger than 50. It is alsofound that the average precipitation intensity and Rein precipitating clouds are small with large τ_c(> 90)at 36°N south of East China. This may be related tothe higher content of aerosols in the region. The caseanalysis shows that cloud effective radius and liquidwater path near the cloud tops are good indicators ofthe thickness and intensity of convective clouds. TheRe of water cloud reduces with the rain rate increasingin the Jianghuai area, South China, and the warm poolarea of western Pacific. In the warm pool area, as rainrate increases, Re of deep convection seems unchangedwhile Re of stratiform precipitation reduces slightly. Re of deep convection becomes larger with the rainrate increasing in the Jianghuai area and South China. As for the stratiform precipitation over Jianghuai, therelationship between Re and rain rate is complicated. These mechanisms still need to be investigated.

Because the space borne spectroradiometer isunable to capture the information inside the cloud, the retrieval algorithm of cloud parameters based onmerged data, which contain the measurements of spaceborne spectroradiometer, passive micorwave radiometer/imager, precipitation radar, cloud radar, and laserradar, needs to be further developed. With such anadvanced algorithm, not only the cloud parametersnear the cloud tops but also the profiles of cloud parameters inside the cloud can be obtained, which isessential to retrieving precipitation intensity and related latent heat release. This is a challenging researchdirection for the next 10 or more years.

Fortunately, the performance of the multiple sensors aboard FY series satellites has been greatly improved(Yang et al., 2012; Zhang et al., 2012a, b). Taking full advantages of these satellite data will enhance our ability in satellite data application. China islocated in the typical monsoon region where the cloudshave remarkable regional differences, and variations ofthe cloud system are controlled by the monsoon activities. To select suitable locations representing typicalweather systems and to establish ground-based observation stations(super stations)with comprehensiveinstruments are the necessary approach to observecloud structures and other cloud properties, and validate the retrieved cloud parameters, based on theremote sensing by multiple sensors aboard the satellites. Meanwhile, to build airborne systems includingspectroradiometer and active and passive microwaveinstruments, is also necessary for verification of theretrieval results, which will signify a nation's capability in development of the atmospheric science.