The effects of aerosols on water cloud microphysics and macrophysics based on satellite-retrieved data over East Asia and the North Pacific

This study examines the characteristics of the microphysics and macrophysics of water clouds from East Asia to the North Pacific, using data from active CloudSat radar measurements and passive MODerate-resolution Imaging Spectroradiometer (MODIS) retrievals. Our goals are to clarify differences in microphysics and macrophysics between land and oceanic clouds, seasonal differences unique to the midlatitudes, characteristics of the drizzling process, and cloud vertical structure. In pristine oceanic areas, fractional occurrences of cloud optical thickness (COT) and cloud droplet effective radius (CDR) increase systematically with an increase in drizzle intensity, but these characteristics of the COT and CDR transition are less evident in polluted land areas. In addition, regional and seasonal differences are identified in terms of drizzle intensity as a function of the liquid water path (LWP) and cloud droplet number concentration (Nc). The correlations between drizzle intensity and LWP, and between drizzle intensity andNc, are both more robust over oceanic areas than over land areas. We also demonstrate regional and seasonal characteristics of the cloud vertical structure. Our results suggest that aerosol–cloud interaction mainly occurs around the cloud base in polluted land areas during the winter season. In addition, a difference between polluted and pristine areas in the efficiency of cloud droplet growth is confirmed. These results suggest that water clouds over the midlatitudes exhibit a different drizzle system to those over the tropics.


Introduction
The Earth's radiation budget is some affected by the scattering and absorption properties of aerosol, which are referred 30 to as aerosol-radiation interactions.In addition, aerosol particles play an important role in the climate system by serving as cloud condensation nuclei (aerosol-cloud interaction).This affects cloud optical thickness (COT) and cloud particle size (e.g., Twomey, 1977) as well as cloud lifetime (e.g., 35 Albrecht, 1989).However, accurate and quantitative evaluation of these indirect aerosol effects is required to address the considerable uncertainty related to the heterogeneous nature of the spatial and temporal distributions of aerosols.With respect to numerical models, many climate models have been 40 developed and improved for accurate estimation of the global radiation balance.Practically all of the climate models, however, have uncertainty in their cloud precipitation parameterization schemes (e.g., Suzuki et al., 2013a) due to the difficulty of representing the complex aerosol-cloud interactions.

45
The cloud profiling radar (CPR) of CloudSat, whose mission began in 2006, may help clarify the details of cloud physical properties (Stephens et al., 2002), including vertical information that cannot be obtained from conventional satellite passive sensors, and is important to clarify indirect 50 aerosol effects.Research on the physical properties of water clouds has advanced significantly in the last few years.Haynes and Stephens (2007) studied the relationships between cloud thickness and precipitation in the marine tropics, and found regional differences in the cloud vertical structure 55 (shallow, middle, and deep modes) of precipitating clouds.Lebsock et al. (2008) investigated mainly aerosol-cloud interactions based on multi-sensor satellite observations, and found a relationship between variations in the cloud liquid water path (LWP) and the thermodynamic conditions.Kubar et al. (2009) compared the physical properties of water clouds in regions over tropical and subtropical oceans and stressed the importance of cloud macrophysics and microphysics for drizzle frequency and intensity.They also investigated which parameters were important for drizzle processes, focusing on macrophysics (cloud thickness and LWP) and microphysics (cloud droplet effective radius (CDR) and cloud droplet number concentration (N c )). Sorooshian et al. (2009) performed a binning study of LWP to clarify the effects of aerosol perturbation (e.g., precipitation susceptibility, aerosol cloud interactions), and suggested that intermediate LWP (∼ 500-1000 g m −2 ) cloud tends to be more susceptible to aerosol than shallow cloud with low LWP.Furthermore, they expanded the study of Stephens and Haynes (2007) who introduced a method of estimating conversion (from cloud water to rain water) rates from CloudSat-CPR and MODerate-resolution Imaging Spectroradiometer (MODIS) retrieved data, and discussed 20 the relationships between conversion rate and aerosol types, associated with the category of lower tropospheric static stability (LTSS) and LWP (Sorooshian et al., 2013).Nakajima et al. (2010) and Suzuki et al. (2010) attempted to visualize the vertical structure of cloud on a global scale 25 using a method that they termed "contoured frequency by optical-depth diagram" (CFODD).Kawamoto and Suzuki (2012) applied CFODD to investigate precipitation process, and demonstrated that precipitation over the Amazon occurrs in optically thicker locations than is the case over China.
Many researchers have investigated the physical structures and precipitation characteristics of low-level water clouds based on satellite data, as mentioned above.However, most of these studies were limited to the tropics/subtropics or areas over oceans; only a few have compared clouds over land and ocean.Very few have focused on East Asia, where some areas have significant levels of air pollution (e.g., Kawamoto and Suzuki, 2013).Therefore, clouds in these regions may exhibit drizzle characteristics that differ from those of clouds over tropical oceanic areas.
This study focuses on seasonal differences in water clouds that are characteristic of the mid-latitudes, and compares the characteristics of clouds over China (a region with considerable anthropogenic aerosols) with those over the North Pacific (a clean/pristine environment).We also analyze the transition processes of drizzle over both land and ocean (e.g., Nakajima et al., 2010) in the mid-latitudes, which have been evaluated in only a few other studies.
2 Data and methodology 2.1 CloudSat and MODIS 50 CloudSat, launched by the National Aeronautics and Space Administration (NASA) in 2006, was the first project to include a spaceborne millimeter-wavelength (3 mm, frequency = 95 GHz) radar (Stephens et al., 2008) to help resolve the vertical structure of cloud droplets.The vertical and 55 spatial resolutions of the CloudSat data products are approximately 480 m and 1.4 × 1.8 km (across and along tracks), respectively.However, the data are vertically 2× oversampled, and therefore ∼ 240 m sampled data are available (Stephens et al., 2008).We obtained information about cloud proper-60 ties, including the visible COT and CDR near the cloud top from the 2B-TAU product (Polonsky, 2008), and also radar reflectivity and the cloud mask from the 2B-GEOPROF product (e.g., Mace et al., 2007;Marchand et al., 2008).We used temperature and pressure data for each altitude from the Eu-65 ropean Center for Medium-Range Weather Forecasts Auxiliary (ECMWF-AUX) objective analysis (Partain, 2007).The analysis periods were June, July, and August (JJA) from 2007 to 2009, and December, January, and February (DJF) from 2006to 2009(i.e., December 2006-February 2009).
We used the following Eq.(1) to estimate N c (e.g., Brenguier et al., 2000;Wood, 2006;Kubar et al., 2009), where B = (3πρ w /4) 1/3 = 0.0620, ρ w is the density of liquid water, and Γ eff is the adiabatic rate of increase in the liquid water content with height, which is a function of two variables, profile of temperature and pressure, as shown in Fig. 85 1 of Wood (2006).The difference in CDR retrieval error between land and ocean, e.g., due to the differences in cloud type (e.g., Zhang et al., 2012), may also cause uncertainty in estimation of N c .However, we apply a CDR uncertainty threshold of < 1 µm, as mentioned above, which reduces N c 90 uncertainty as much as possible.Other possible errors due to the assumption of deriving N c (e.g., adiabaticity, vertical homogeneity) are documented elsewhere (e.g., Grandey and Stier, 2010;Kubar et al., 2009).In addition, we calculated LWP by the following Eq.( 2) (Brenguier et al., 2000), where τ c and r e were obtained from MODIS retrieval, matched along the CloudSat footprint (i.e., CloudSat 2B-TAU product, mentioned earlier).

T. Michibata et al.:
The effects of aerosols on water cloud

Regions and methods
Figure 1 shows maps of the regions investigated in this study.Inland includes the Gobi Desert.We select an area of Northeastern China (NE China) to study the effects of soil dust aerosols transported from the Gobi and Taklamakan Deserts.
Human activity generates many anthropogenic aerosols in the Industrial area, and this region is one of the most airpolluted areas in the world (upper panel of Fig. 1).Some areas of Japan region also discharge anthropogenic aerosols, but the main reason for selecting this region is to compare it 10 with the Industrial area.We refer to the outflow regions of anthropogenic aerosols as North Pacific 1, 2, and 3 in order of their distance from East Asia.We investigated how large amounts of aerosols transported from East Asia affect cloud properties in these areas.
This study focuses only on low-level water clouds, because most aerosols remain in the lower troposphere.We define water clouds as those with a cloud mask value greater than 30 (good/strong echo), which means high-confidence detection (estimated false detection < 4.3%; see Marchand et al. (2008), Table 1), and a temperature above 273 K for the entire cloud layer.Furthermore, we use only the data with uncertainty values of less than 3 and 1 µm for COT and CDR, respectively.Multilayered clouds are excluded from the analyses to avoid ambiguous statistics.

25
LTSS is defined as the difference in potential temperatures between 700 hPa and the surface (Klein and Hartmann, 1993).This index was calculated from the ECMWF-AUX product (vertical temperature and pressure profiles).

Cloud physical properties for each area
Table 1 lists the physical properties of clouds over each of the seven areas.DJF values are given in parentheses.The landsea mask is not applied in our analysis, and therefore the data for the Japan, NE China, and Industrial area, including the 35 ocean part, do not necessarily represent data only over the continent.The results suggest that the precipitation occurrence is related to LWP (e.g., North Pacific 1 where higher LWP region is accompanied with high '[%] with rain'; Table 1), except for the Industrial area (i.e., high LWP but lower '[%] with rain', and vice versa).It is noteworthy that there are large seasonal differences of more than 7 K in LTSS in the Industrial area.Therefore, there is a possibility of different cloud types over the Industrial area; i.e., cumulus cloud in JJA (unstable lower LTSS environment) than over the oceanic area.The passive MODIS sensor tends to retrieve errors on cumulus inhomogeneous cloud (e.g., Zhang et al., 2012;Zhang and Platnick, 2011;Zinner et al., 2010) because of its simplifying assumptions; i.e., clouds are plane-parallel and homogeneous, any effects of drizzle/rain drops are ig-nored (Zinner et al., 2010), etc.These assumptions may lead to retrieval bias of CDR; e.g., shadowing effects can lead to underestimation of COT and overestimation of CDR (Marshak et al., 2006).The smaller COT and larger CDR are estimated with increasing cloud inhomogeneity, which results in 55 underestimation of LWP for cloudy scenes (Painemal et al., 2013).Therefore, care should be taken with regard to this background of CDR retrieval error and underestimation of LWP, especially over the Industrial area in JJA.
Figure 2 shows the probability distribution function (PDF) of each cloud physical variable.The distribution of maximum radar reflectivity in the cloud layer (Z max ) (Fig. 2a) is similar for both the Industrial area and North Pacific 3, although we observed a slight shift to weaker Z max for the Industrial area.We confirmed the tendency that smaller CDR 65 values, larger N c values, and optically thicker clouds were observed over land areas than over the oceanic regions in Fig. 2 and Table 1, supporting the findings of previous studies (e.g., Kawamoto et al., 2001).However, these results are not as obvious in the region over Japan as in other land areas, 70 Inland, NE China, or the Industrial area.It is possible that the properties of clouds over NE China are affected in a complex manner by dust aerosols from the adjacent western deserts and emissions of anthropogenic aerosols from highly populated areas, such as Beijing.The North Pacific 1 area has 75 slightly larger values for COT, LWP, and N c compared with the other oceanic areas, and the values of CDR are almost the same for all oceanic areas.Small seasonal differences are observed during JJA and DJF over the three oceanic areas; these differences are more obvious over the four land areas, which may be due to the high levels of aerosols in DJF, when atmospheric conditions are most stable.
The mode radii are approximately 15 µm over the three oceanic areas, whereas they are approximately 9 µm over the Industrial area in DJF, which may result in less efficient 85 precipitation.The following subsections discuss how differences in the physical properties of clouds over land and ocean regions affect the rainfall characteristics.

COT-CDR diagram
COT and CDR are commonly considered cloud physical 90 variables.The fact that the correlation between these parameters reflects cloud growth (only liquid phase warm cloud) and precipitation processes has been well documented in previous studies based on satellite observations (e.g., Nakajima et al., 1991;Nakajima and Nakajima, 1995).That is, both 95 COT and CDR increase early in the growth process of cloud droplets, resulting in a positive correlation between them.The cloud particles grow to almost 15 µm, and precipitation begins.With precipitation, COT decreases and CDR increases due to coalescence.This precipitation process leads 100 to a negative correlation pattern.Suzuki et al. (2006) extended these analyses, and successfully simulated the pattern using a spectral-bin microphysics model.Suzuki et al. (2011) documented fractional occurrences as a function of COT and CDR for each rain category (no precipitation, drizzle, and rain), and compared A-Train observations with model simulations.
Figure 3 shows fractional occurrences on COT-CDR diagrams for each rain category ([A] no precipitation; Z max < −15, [B] drizzle; −15 ≤ Z max < 0 and [C] rain; 0 ≤ Z max ) (Comstock et al., 2004;Stephens and Haynes, 2007).The diagrams in the pristine remote ocean (North Pacific 3, Fig. 3gl) reveal that the main group systematically shifts from the 10 lower COT-CDR region to the higher COT-CDR region with an increase in the rain category (i.e., from no precipitation to rain, with a monotonous increase in LWP and a slight decrease in LTSS) during both seasons.This tendency was also reported by Suzuki et al. (2011) and Kawamoto and Suzuki (2013).The fact that JJA (Fig. 3g-i) and DJF (Fig. 3j-l) have similar distributions suggests that the relation between COT and CDR has considerable universality with the rain categories over oceanic areas.However, in the Industrial area where air pollution by anthropogenic aerosols is severe, the transition pattern is not as clear as over the ocean, and the variations of LWP are relatively small.The category Rain in JJA (Fig. 3c) has relatively high values of fractional occurrence (approximately 0.2-0.5) in the small COT-CDR region (COT < 15, CDR < 15 µm), while most values in this region (see Fig. 3i, l) are less than 0.2.Furthermore, we found that large numbers of samples are concentrated in this region and that the cloud-top height in the Industrial area is much higher (3.3 km) than that in the North Pacific 3 area (2.4 km).These findings suggest the existence of other predominant factors that affect drizzle intensity in the Industrial area during JJA, in addition to COT and CDR. Matsui et al. (2004) reported that not only the amount of aerosol but also the static stability was important for growth from cloud droplets into drizzle.
The vertical inhomogeneity of CDR (larger particles appear in the lower part of clouds) is one possible reason for this observation.Further analyses are required to clarify this issue.

Transition pattern of precipitation
Some researchers have considered how the properties of clouds over land and ocean differently affect precipitation They found that the drizzle frequency increased with LWP when N c was constant and decreased with increasing N c and constant LWP.We focused on the mid-latitudes in the North-ern Hemisphere, but more detailed analyses of mid-latitude regions would be valuable.
Figures 4 and 5 show the Z max distribution as a function of LWP and N c , because we focused on the transition process of drizzle intensity rather than its frequency.Over three ocean regions (Figs.4e-g and 5e-g), the drizzle intensity increased with increasing LWP under a constant N c , and increased with 60 decreasing N c under a constant LWP.It is important to clarify the physical parameters of clouds to understand the behavior of drizzle over the mid-latitudes as well as over the tropics/subtropics.As the correlation coefficient r 1 between LWP and Z max (∼ 0.6) is greater than r 2 between N c and 65 Z max (∼ −0.3) in these areas, LWP has a stronger correlation than N c with drizzle intensity.This correlation is less clear over land areas than over oceanic areas, as shown in Figs.4a-d and 5c-d.In particular, high values of Z max over the Industrial area are scattered during JJA because parame-70 ters other than LWP and N c have strong effects on the drizzle transition process (from cloud droplet to drizzle and precipitation).This is consistent with our hypothesis that there is a more important dominant factor than cloud physical properties, such as COT, CDR, LWP, and N c over the Industrial 75 area in JJA.The seasonal difference is more obvious over the land areas than over the oceanic areas, with the magnitudes of the correlation coefficients r 1 and r 2 being higher in DJF than in JJA.The land areas in JJA are in the unstable lower LTSS environment, with the exception of Japan.The 80 low specific heat of the land surface would yield unstable conditions due to heating by stronger shortwave radiation in the JJA season.Such local heating may result in forced precipitation.This is responsible for the scattered distribution of high Z max values.In addition, variations in the dynamics 85 over land areas (e.g., vertical velocity) would also be associated with this seasonal difference.
Values of Z max greater than 0 dBZ e (orange and red in Fig. 4) are uncommon in the Inland and NE China areas during JJA, which indicates very few precipitating clouds.

90
Over these regions in DJF, generally only few water clouds are observed due to low temperature and/or low water vapor levels.In the Industrial area, there are some occasions when N c is larger than 500 cm −3 , and Z max values are lower as N c becomes larger during DJF.Even LWP values, which 95 are more strongly correlated with drizzle intensity, are larger.These findings suggest that the cloud lifetime increases due to storage of water within the cloud layer.These findings are also observed in Japan (Fig. 5d), where a significant transition pattern appear as follows: LWP of 300 g m −2 and N c 100 of 250 cm −3 , to LWP of 450 g m −2 and N c of 100 cm −3 , to LWP of 300 g m −2 and N c of 15 cm −3 , as shown by the black arrows in Figs.4d and 5d.LWP values increase to 400-500 g m −2 as N c values decrease because drizzle occurs only inside the cloud layer with no loss of water.At the same time,

105
CDR values increase slowly within the range of 10-15 µm and then rapidly to larger values (15-25 µm), which leads to precipitation.The conditions in Japan are not as clean as in the three oceanic regions, but are not as polluted as in the Industrial area, which is likely the reason for this V-shaped transition pattern.

Cloud vertical structure
Cloud geometrical thickness is a cloud macrophysical variable, in addition to the cloud-top height and LWP.Over the tropical ocean, cloud-top height is offset by a constant from the cloud geometric thickness, because the cloud base height is almost constant (e.g., Kubar et al., 2009).Cloud base height is, however, not always constant over mid-latitudes, particularly over the land.Therefore, we use cloud geometrical thickness as a representative macrophysical variable.In fact, cloud geometrical thickness has a robust correlation with Z max (0.28-0.83; shown in Fig. 6), which is an index of precipitation intensity, stronger than the relationship between 15 cloud-top height and Z max (0.04-0.63).However, it should be noted that the "cloud geometrical thickness" mentioned here does not always accurately represent the cloud thickness.Specifically, in some cases of non-precipitating cloud, determination of the cloud base is difficult because the re-20 flectivity at this point is too weak to be observed.However, in the case of precipitating cloud, the detected value would include not only the cloud but also some of the precipitating layer.Thus, the "cloud geometrical thickness" represents the detected hydrometer thickness.

25
The PDFs of cloud geometrical thickness are shown in Fig. 6.Solid and dotted lines represent drizzling/precipitating and non-precipitating cloud, respectively.The correlations between cloud geometrical thickness and Z max for JJA and DJF are denoted as r jja and r djf , respectively.Almost all of the non-precipitating clouds have geometrical thickness less than 1000 m, and the clouds with precipitation are ∼ 500-1000 m thicker.This trend and strong correlation between cloud geometrical thickness and Z max suggest the importance of cloud geometrical thickness for the occurrence of precipitation.The modal cloud geometrical thickness of the nonprecipitating category is ∼ 500 m for all seven regions during both seasons.On the other hand, the precipitating clouds have large seasonal variability.For example, oceanic clouds (Fig. 6e-g) become thicker in DJF. Figure 7 shows a histogram of cloud geometrical thickness for thin (< 800 m; red), middle (800-2000 m; green), and thick (≥ 2000 m; blue) clouds, which correspond roughly to non-precipitating, drizzling, and precipitating clouds, respectively.The LTSS values listed in Table 1, which represent the air stability, tend to be consistent with the cloud geometrical thickness.More specifically, middle or thicker clouds exist predominantly in the unstable environment over the Industrial area in JJA (i.e., LTSS = 12.2 K).Conversely, in the stable environment in DJF (i.e., LTSS = 19.6K), thinner clouds are predominant.Similar to this tendency, the cloud geometrical thickness, which reflects the seasonal difference in LTSS, is also seen among other regions.Lebsock et al. (2008) confirmed that high-aerosol conditions tend to decrease LWP in non-precipitating clouds, and 55 the magnitude of the reduction in LWP is greater under the unstable low LTSS environment.These findings suggest the importance of LWP and thermodynamics to understanding aerosol-cloud interactions (L'Ecuyer et al., 2009).We further investigated the cloud vertical structure based on a com-60 parison with the atmospheric conditions (pristine or polluted) associated with LWP and LTSS.Use of the CFODD to illustrate cloud vertical structure facilitates identification of associations with cloud optical properties, particularly for singlelayered water clouds (e.g., Nakajima et al., 2010;Suzuki 65 et al., 2010).In general, the vertical and horizontal axes are allocated to geometrical height and radar reflectivity, respectively, when illustrating the frequency of the vertical radar profile.CFODD visualization methods apply the in-cloud optical depth (ICOD) as the vertical axis instead of altitude.

70
In this way, normalization of the vertical coordinate by ICOD facilitates interpretation focusing on optical properties using composited clouds of different geometrical thicknesses.We obtained information on the layered optical depth from the 2B-TAU product.sents evaporation and condensation processes, and CDR bins [B] and [C] represent mainly collision and coalescence processes.Therefore, an increase in LWP with an increase in CDR is expected.However, the rate of increase of LWP differs significantly between the Industrial area and North Pa-90 cific 3, as shown in Table 2.That is, the rate of increase over North Pacific 3 is greater than that over the Industrial area.This result implies that the clouds over North Pacific 3 are more efficient than those over the Industrial area in terms of cloud droplet growth.Over the Industrial area in DJF, which 95 is in the stable and high LTSS environment, non-precipitating clouds are dominant (61.5 %; see Table 1) and contain much cloud water, as depicted in Fig. 8d.This may suggests the occurrence of the second indirect effect (Albrecht, 1989).Under such high-LWP and small-CDR conditions, cloud albedo 100 can also increase, as can be seen in the following Eq.(3), which is another form of Eq. (1), In fact, the COT in DJF (τ c = 35.9) is much higher than that 105 in JJA (τ c = 19.5).
We can see non-precipitating clouds mainly in the smallest CDR bin (CDR < 12 µm), and an obvious transition of the CFODD to drizzle (12 µm ≤ CDR < 18 µm) and rain (18 µm ≤ CDR) phases.In addition, there is a clear difference between the CFODDs of the Industrial area and North Pacific 3, with regard to the transition process for drizzling clouds.More specifically, the CFODDs over the polluted land area transit ICOD mainly from near the cloud-top to the cloud-base, while those over the ocean transit mainly in the deeper ICOD region (approximately over 30).This feature is consistent with some previous reports (e.g., Nakajima et al., 2010;Suzuki et al., 2010Suzuki et al., , 2011)).We would like to interpret this characteristic of CFODDs as a result of suppression of precipitation due to high concentrated aerosols around the cloud base (large part of ICOD) over the Industrial area.DJF is a dry season over mid-latitudes in the Northern Hemisphere, and the stable and high LTSS environment results in a high aerosol concentration near the surface.Therefore, an aerosol-cloud interaction may occur that results in weaker radar reflectivity in the larger ICOD region.This may be one of the hypotheses, and further analysis (e.g., sensitivity experiments using numerical modeling) is required in order to enhance the credibility.It is also possible that the difference in cloud vertical structure between land and ocean is caused by the difference in updraft strength (Nakajima et al., 2010), or other meteorological factors as well.The mission of "Earth Clouds, Aerosols and Radiation Explorer (Earth-CARE)," which will start in 2016, is helpful because it will equip the CPR with Doppler speed sensor functions (e.g., Sy et al., 2013;Nakatsuka et al., 2012;Schutgens, 2008) that can detect vertical velocity.In addition, numerical modeling experiments are required for further understanding of the aerosol-cloud-radiation interaction.Lebsock et al. (2008) emphasized the importance of performing investigations on regional and seasonal scales in both numerical modeling and observational studies to gain a more detailed understanding of cloud dynamics.Suzuki et al. (2013b) also suggested that the complex behavior of CFODDs at different latitudes (see their Fig.S3) and models could not reproduce the satellite-observed CFODDs due to a lack of knowledge of parameterization of cloud dynamics at different latitudes.The results of the present study based on regional and seasonal analysis associated with the aerosolcloud interaction will contribute to the improvement of cloud physical parameterization in numerical models.
The effects of spatial difference of meteorology on aerosol-cloud interaction were not considered in our study; therefore, further analyses are necessary.We must be careful about following two ideas: one is the fact that genuine aerosol-cloud interactions may behave differently under different meteorological conditions; and the other is the fact that meteorology may drive aerosol-cloud relationships (even in the absence of any aerosol-cloud interactions).Such the meteorological gradients sometimes cause spurious correlations (Grandey and Stier, 2010).
For example, the difference in the autoconversion rate over land and ocean, or in JJA and DJF, may provide some insight into the indirect aerosols effects (e.g., Stephens and Haynes, 2007;Sorooshian et al., 2013).Although the data presented here are insufficient to link the ocean versus land differences 60 to aerosol effects, further studies to determine the effects of atmospheric conditions (i.e., aerosol concentration, static stability) on cloud physical structure would be valuable.

Conclusions
We conducted a comparative study of the physical proper-65 ties of water clouds over the region from East Asia to the North Pacific in the mid-latitudes based on CloudSat/CPR and Aqua/MODIS retrievals.In addition to confirming several known characteristics regarding cloud physical properties, such as larger N c , smaller CDR, and larger COT values 70 over land, we found that the cloud differences over land versus the ocean are more obvious during DJF than JJA.
In the pristine area, we found a clear tendency for lower to higher COT-CDR with rising precipitation categories during both JJA and DJF.However, this transition pattern does not 75 appear clearly in the polluted area during JJA, and precipitation occurs even in the lower COT-CDR region.
An investigation of the transition process of precipitation reveals that during DJF the polluted areas have larger N c values, and the clouds could contain much more LWP with 80 larger N c values than during JJA.Oceanic cloud properties over the mid-latitudes do not change significantly between the two seasons, and their behavior is similar to that of oceanic clouds over the tropics/subtropics.However, we observe considerable seasonal differences over land.

85
Such differences also appear in the LTSS.Although the LTSS is correlated with cloud geometrical thickness, it is less important for the cloud growth process.On the other hand, LWP increases monotonically with growing CDR.However, we confirmed a smaller rate of increase in LWP over pol-90 luted land.In addition, we found a difference in "contoured frequency by optical-depth diagram" (CFODD) between the pristine oceanic area and the polluted land area, implying the aerosol-cloud interaction.However, we cannot completely exclude the possibility that other meteorological factors may 95 be responsible for the differences between land and ocean.
To clarify these differences in cloud properties and drizzle characteristics between land and ocean, and between the tropics/subtropics and mid-latitudes, it is important to estimate the radiation budget accurately.We determined some 100 of the characteristics of aerosol-cloud interaction based only on satellite data.However, composite studies with numerical modeling (e.g., sensitivity experiments for the influence of aerosol and atmospheric stability to cloud physics) are required to gain a detailed understanding of the aerosol-cloud 105 interaction.This study does not preclude the possible effect of spatial gradient changes in the meteorology on aerosol-cloud interaction, and further analyses taking such environmental conditions into consideration are required.Chem. Phys., 10, 9535-9549, doi:http://dx.doi.org/10.5194/acp-10-9535-201010.5194/acp-10-9535-2010, 2010.

40efficiency.
Leon et al. (2008) analyzed CloudSat and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) data, and illustrated the global distribution of drizzle frequency as a function of LWP and CDR.We used N c instead of CDR because we focused on differences in the amount of aerosol between land (polluted) and ocean (cleaner) regions.Kubar et al. (2009) also investigated the drizzle frequency of water clouds over oceanic areas in the tropics and subtropics as a function of a typical macrophysical variable (LWP) and a typical microphysical variable (N c ).

75
CFODDs of each CDR bin ([A] 5-12 µm, [B] 12-18 µm, [C] 18-35 µm) over the Industrial area and North Pacific 3 are presented in Fig. 8.Although LTSS is correlated with cloud geometrical thickness, as mentioned earlier, LTSS seems to have little relation with the cloud growth process, 80 because the values are almost identical among the three CDR bins.The CFODDs show that the LWP increases monotonically with increasing CDR, which corresponds to the transition from cloud particle (category [A]) to drizzle (category [B]), and raindrop (category [C]).That is, CDR bin [A] repre-85

Fig. 1 .Fig. 2 .
Fig. 1.Whole (top) and individual (bottom) regions in this study.Spatial distribution of aerosol optical thickness τa (550 nm) for the 3 year mean derived from monthly Aqua/MODIS level 3 products are illustrated in the top panel.Missing values are shown in white.figure

Table 1 .
Cloud physical parameters in each area.JJA and DJF values are 3 year means.DJF values are shown in parentheses.Maximum values are shown in bold and minimum values are underlined.Maximum radar reflectivity in the cloud layer (Zmax) is used for precipitation categories (no precipitation; Zmax < −15, drizzle; −15 ≤ Zmax < 0, rain; 0 ≤ Zmax).The Inland and NE China regions in DJF, where none or few samples met the criteria, are shown by "not available (N/A)".

Table 2 .
LWP and its rate of increase for each CFODD.