Hao Miao , Xiaocong Wang , Yimin Liu , Guoxiong Wu , ?

a College of Atmospheric Science, Nanjing University of Information Science and Technology, Nanjing, China

b State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences (CAS), Beijing, China

c College of Earth and Planetary Science, University of Chinese Academy of Sciences, Beijing, China

Keywords:Cloud mass Number concentration Effective radius Cloud types CloudSat/CALIPSO Tibetan Plateau

A B S T R A C T By using four-year CloudSat/CALIPSO satellite data, the authors investigated cloud microphysical properties in three representative regions over East Asia, where models commonly suffer from great biases in simulations of cloud radiative effects. This study aims to provide an observational basis of cloud microphysical properties for the modeling community, against which the model simulations can be validated. The analyzed cloud microphysical properties include mass, number concentration, and effective radius for both liquid and ice phases. For liquid clouds, both cloud mass and number concentration gradually decrease with height, leading to the effective radius being nearly uniformly spread in the range of 8—14 μm. For ice clouds, the cloud mass and effective radius decrease with height, whereas the number concentration is nearly uniform in the vertical. The cloud microphysical properties show remarkable differences among different cloud types. Cloud mass and number concentration are larger in cumuliform clouds, whereas smaller in cirrus clouds. By comparing cloud properties among the Tibetan Plateau, East China, and the western North Pacific, results show the values are overall smaller for liquid clouds but larger for ice clouds over the Tibetan Plateau.

1. Introduction

Clouds are important in modulating the global energy budget and hydrological cycle ( Stephens, 2005 ). Numerical models still struggle to accurately simulate the characteristics of clouds and their radiative effects ( Su et al., 2013 ; Wang et al., 2014a , b ), although considerable progress has been made during the last decade. The uncertainties in modeling can be partly attributed to the poor representation of cloud microphysical properties (e.g., the cloud effective radius), in addition to those caused by cloud macrophysical properties (e.g., cloud fraction and cloud vertical structure), which have been widely explored in the literature ( Weare, 2000 ; Wang, 2017 ; Wang et al., 2020 ). Earlier studies by Slingo (1990) found that decreasing the effective radius of cloud droplets from 10 to 8 μm would result in atmospheric cooling that could offset global warming due to doubling CO. The importance of cloud microphysical properties in modulating cloud radiation was underlined in Van Weverberg et al. (2011) and Zhao et al. (2018) . Therefore, it is highly necessary to extend our knowledge of cloud microphysical properties.

Prior to the launch of the CloudSat/CALIPSO satellite, cloud microphysical properties were mainly obtained from passive sensors onboard satellites such as ISCCP, MODIS, CERES, etc. The provided liquid (ice)water path data in these products have been widely used for validating model performance in the past ( Horvath and Davies, 2007 ; Li et al.,2008 ). In-situ measurements and aircraft experiments are another way to acquire cloud microphysical properties ( Yin et al., 2014 ; Cheng et al.,2015 ; Zhao et al., 2016 ). The disadvantage of these observations, however, is that they are sporadic and region dependent. By contrast, the CloudSat/CALIPSO satellite provides cloud retrievals on a global scale and detects clouds layer by layer, which were inaccessible in earlier passive satellites. Over the past decade, CloudSat/CALIPSO have been widely used for cloud investigation and model evaluation ( Barker et al.,2008 ; Yan et al., 2018 ; Miao et al., 2019 ).

In this study, we aim to provide a comprehensive characterization of cloud microphysical properties over East Asia by using datasets from the CloudSat/CALIPSO satellites, where models commonly suffer from great biases in simulations of cloud radiative effects ( Zhang and Li, 2013 ;Wang et al., 2014a , b ; Li and Mao, 2015 ). Given that cloud radiation is partly dependent on cloud microphysical properties, especially cloud effective radius, an analysis of this kind will provide a reference for the modeling community.

The paper is organized as follows: Section 2 gives a brief description of the CloudSat/CALIPSO data and methods. Section 3 shows how cloud microphysical properties vary in the vertical direction and between different regions and different cloud types. Section 4 summarizes our main findings and conclusions.

2. Data and methods

The cloud microphysical properties (cloud mass, number concentration, and effective radius) data employed in this study come from the 2B-CWC-RO product ( Austin et al., 2009 ). The 2B-CWC-RO product contains composite profiles of the microphysical properties of both ice and liquid phases. The cloud types are from the 2B-CLDCLASS-LIDAR product, which was derived using a combined rule-based and fuzzy logic classification approach that incorporated various inputs (e.g., radar and Li-DAR signals, cloud-top height, and precipitation; Wang et al., 2013 ). The 2B-CLDCLASS-LIDAR product classifies clouds into eight types: stratus(St); stratocumulus (Sc); cumulus (Cu); nimbostratus (Ns); altocumulus(Ac); altostratus (As); deep convective (Dc); and cirrus, cirrostratus, and cirrocumulus (Ci). In this study, St and Sc clouds are grouped together because of the difficulties in distinguishing these two similar cloud types( Sassen and Wang, 2008 ). Details of the algorithm and retrieval methods can be found at http://cloudsat.atmos.colostate.edu/ .

The 2B-CWC-RO product provides cloud microphysical properties on a vertical resolution of 240 m with 125 layers in total, whereas the 2BCLDCLASS-LIDAR product divides cloud type information into 10 layers identified by the CloudLayerType flag. We first converted cloud fields in the 2B-CLDCLASS-LIDAR product to the same vertical resolution as in the 2B-CWC-RO product, according to cloud base and top height in 2B-CLDCLASS-LIDAR. These data were then processed to a coarser resolution of 0.5 km by merging cloud fields from every two adjacent layers to further increase the sample size. The pixels with both valid cloud type flags and non-zero values of cloud microphysical properties were identified as cloudy and used in the statistics. The cloud fraction was defined as the ratio of the number of cloudy pixels to the total number of pixels in each layer, whereas the cloud microphysical properties refer to the incloud values. To avoid the uncertainty caused by few cloudy samples at some layers, we excluded layers where the cloud fraction was less than 5%. Only the lowest 80 layers were used for analyzing cloud properties(about 19 km for the top level height). The lowest four layers near the surface were also excluded for the sake of ground-clutter contamination( Mace et al., 2007 ). All cloud fields were processed to seasonal-mean values by averaging data from 2007 to 2010.

3. Results

Fig. 1 (a) shows the terrain height over East Asia, which varies from sea-level over the western North Pacific to up to 6 km over the Tibetan Plateau. Models have significant biases in simulations of clouds and their radiative effects in these regions ( Zhang and Li, 2013 ; Wang et al.,2014a , b ). Fig. 1 (b) shows that the multi-model mean of cloud shortwave cooling effects in CMIP6 are underestimated over land and overestimated over ocean relative to CERES, with the biases reaching as high as 12 W m. The remarkable standard deviation in Fig. 1 (c) further corroborates that simulations of cloud radiative effects over these regions are of great uncertainty. In view of this, three representative regions were selected: the Tibetan Plateau above 3 km (TP, 28°—36°N,83°—100°E); East China (EC, 22°—32°N, 105°—115°E); and the western North Pacific (WNP, 13°—23°N, 138°—148°E).

Fig. 1. The (a) terrain height in East Asia, and the (b) multi-model mean bias(units: W m ? 2 ) and (c) standard deviation (units: W m ? 2 ) of shortwave cloud radiative effects for CMIP6 models. The three regions are marked by the rectangles.

Fig. 2 shows various cloud properties over the three regions, including cloud fraction (CFR), liquid water content (LWC), liquid number concentration (LNC), liquid effective radius (LRE), ice water content(IWC), ice number concentration (INC), and ice effective radius (IRE).The CFR shows clear seasonal variations over all three regions, where high clouds are frequently formed in the summer. There are also distinctive features between the three regions, with more low-level clouds produced over the TP and less over EC and the WNP. To explore the reasons for these differences, we separated CFR into different cloud types,which are shown in Fig. 3 . The main non-precipitating cloud types over the TP are Cu and Sc(St) (stratus and stratocumulus), whereas Dc and Ci are dominant over EC and the WNP in summer. Moreover, Dc extends to a higher altitude over EC and the WNP than over the TP. It is thus implied that deep convection is much restricted over the TP, as it is difficult for convective parcels rooted in dry conditions to develop to a higher altitude. As Ci is mainly composed of ice crystals detrained from deep convection, its distribution is closely related to that of Dc in upper layers.

Fig. 2. Seasonal variations of (a—c) cloud fraction (CFR, units: %), (d—f) liquid water content (LWC, units: mg m ? 3 ), (g—i) liquid number concentration (LNC, units:cm ? 3 ), (j—l) liquid effective radius (LRE, units: μm), (m—o) ice water content (IWC, units: mg m ? 3 ), (p—r) ice number concentration (INC, units: L ? 1 ), and (s—u) ice effective radius (IRE, units: μm) over the TP (Tibetan Plateau), EC (East China), and the WNP (western North Pacific).

Although CFR shows clear seasonal variations, the cloud microphysical properties do not. However, they do show significant variations in the vertical direction ( Fig. 2 (d—u)). This demonstrates that CFR is mainly modulated by large-scale circulation, whereas cloud microphysical properties are intrinsic. For liquid clouds, both LWC and LNC decrease synchronously with altitude and disappear near 8 km, leading to a nearly uniform value of LRE around 8—14 μm. For ice clouds, the IWC peaks at the freezing level (～9 km) and then gradually decreases with altitude. By contrast, the INC does not show a clear maximum near the freezing level but tends to be more uniformly spread. Correspondingly, the IRE shows a gradual decrease with altitude. The values of LWC and IWC over the WNP are generally larger than those over the TP, which mainly results from more intense upward motion and larger surface evaporation ( Gao et al., 2014 ). The LNC and INC are also larger over the WNP and EC than over the TP, which is presumably caused by the higher aerosol concentrations in the former ( Wang et al.,2014a , b ).

To understand how cloud microphysical properties behave in different cloud types, Fig. 4 and Fig. 5 show the vertical profiles of microphysical properties for liquid and ice clouds, respectively. As shown in Fig. 4 ,all cloud types except Ci and Sc(St) show a gradual increase in LWC and LNC from the surface to a height of 4—5 km and then decrease sharply upward over EC and the WNP. The decreasing feature in the upper part is also apparent over the TP. The LWC is largest in Cu and smallest in Ci. To facilitate quantitative comparison, these values were vertically averaged and listed in Table 1 . Both LWC and LNC show remarkable differences among different cloud types. The LWC varies from 19.1 mg min Ci to 418.4 mg min Cu. The LNC varies from 8.3 cmin Ci to 75.7 cmin Sc(St). The vertically averaged LRE does not vary as much among different cloud types, with the smallest value of 8.7 μm in Ci and the largest one of 14.1 μm in Cu. However, the distinctions are significant below the freezing level, with larger values for cumuliform clouds(e.g., Cu) and smaller ones for stratiform clouds (e.g., As). It is therefore desirable to represent cumuliform and stratiform clouds separately in models to further improve the representation of cloud radiation.

Fig. 3. Seasonal variations in cloud fraction (%) for (a—c) deep convective clouds (Dc), (d—f) cirrus, cirrostratus, and cirrocumulus (Ci), (g—i) nimbostratus (Ns), (j—l)altostratus (As), (m—o) altocumulus (Ac), (p—r) cumulus (Cu), and (s—u) stratus and stratocumulus (Sc(St)) clouds over the TP (Tibetan Plateau), EC (East China), and the WNP (western North Pacific).

The IWC shown in Fig. 5 gradually increases to a peak at around 9 km and then decreases with altitude, whereas the INC (IRE) increases(decreases) monotonously with altitude. The magnitude of IWC in Dc is significantly larger than that in other cloud types. It shows a maximum value of 350 mg mnear 9 km over the WNP, which is nine times larger than that in Ac and Ci at the same height. As in liquid clouds, the IWC and INC show remarkable differences among different cloud types. For instance, the averaged IWC varies from 4.9 mg min Sc(St) to 194.1 mg min Dc. The INC ranges from 5.3 Lin Sc(St) to 188.8 Lin Dc.The values of cloud mass and number concentration are larger over the TP than over EC and the WNP for most cloud types, such as Ac, Sc(St),Cu, and Ns, which is contrary to those found for warm clouds.

Fig. 4. Vertical profiles of the average (a—c) liquid water content (LWC, units: mg m ? 3 ), (d—f) liquid number concentration (LNC, units: cm ? 3 ), and (g—i) liquid effective radius (LRE, units: μm) for seven cloud types over the TP (Tibetan Plateau), EC (East China), and the WNP (western North Pacific). The error bar denotes the standard deviation.

Fig. 5. Vertical profiles of the average (a—c) ice water content (IWC, units: mg m ? 3 ), (d—f) ice number concentration (INC, units: L ? 1 ), and (g—i) ice effective radius(IRE, units: μm) for seven cloud types over the TP (Tibetan Plateau), EC (East China), and the WNP (western North Pacific). The error bars denote the standard deviation.

4. Summary

We investigated cloud microphysical properties (cloud mass, number concentration, and effective radius for liquid and ice phases) over three representative regions in East Asia by using CloudSat/CALIPSO products from 2007 to 2010. Results show that, unlike cloud fraction, cloud microphysical properties do not show clear seasonal variation, but do show significant variations in the vertical direction. For warm clouds,both cloud mass and number concentration synchronously decrease with altitude and disappear near 8 km, leading to an almost uniform effective radius of 8—14 μm. The cloud mass also decreases gradually with altitude in ice clouds, but the number concentration does not, which is instead spread uniformly.

It is revealed that cloud microphysical properties are distinct among different cloud types for both liquid and ice phases. For liquid clouds,the maxima and minima of mass and number concentration are observed in Cu and Ci, whereas they are observed in Dc and Sc(St) for ice clouds,respectively. Most cloud types, such as Ac, Sc(St), Cu, and Ns, show larger mass and number concentrations over the TP than over EC and the WNP in ice clouds, which is contrary to the case in liquid clouds where values are smaller over the TP for almost all cloud types.

Table1Averagevaluesofcloudmicrophysicalproperties.

Funding

This work was jointly supported by the Strategic Priority Research Program of the Chinese Academy of Sciences [grant number XDA20060501], the National Basic Research Program of China [grant numbers 2017YFA0604000 and 2016YFB0200800], and the National Natural Science Foundation of China [grant number 41530426].