Volume 49, Issue 19 e2022GL099478
Research Letter
Free Access

Effects of Different Types of Aerosols on Deep Convective Clouds and Anvil Cirrus

Jinming Zhang

Jinming Zhang

International Center for Climate and Environment Sciences, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

University of Chinese Academy of Sciences, Beijing, China

Contribution: Conceptualization, Methodology, Validation, Formal analysis, ​Investigation, Data curation, Writing - original draft, Visualization

Search for more papers by this author
Bin Zhao

Corresponding Author

Bin Zhao

State Key Joint Laboratory of Environmental Simulation and Pollution Control, School of Environment, Tsinghua University, Beijing, China

State Environmental Protection Key Laboratory of Sources and Control of Air Pollution Complex, Beijing, China

Correspondence to:

B. Zhao and X. Liu,

[email protected];

[email protected]

Contribution: Conceptualization, Methodology, Formal analysis, ​Investigation, Resources, Writing - review & editing, Project administration, Supervision

Search for more papers by this author
Xiaohong Liu

Corresponding Author

Xiaohong Liu

Department of Atmospheric Sciences, Texas A&M University, College Station, TX, USA

Correspondence to:

B. Zhao and X. Liu,

[email protected];

[email protected]

Contribution: Conceptualization, Resources, Writing - review & editing, Project administration

Search for more papers by this author
Guangxing Lin

Guangxing Lin

International Center for Climate and Environment Sciences, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

Contribution: Writing - review & editing

Search for more papers by this author
Zhe Jiang

Zhe Jiang

Carbon Neutrality Research Center, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

Contribution: Writing - review & editing

Search for more papers by this author
Chenglai Wu

Chenglai Wu

International Center for Climate and Environment Sciences, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

Contribution: Writing - review & editing, Supervision, Funding acquisition

Search for more papers by this author
Xi Zhao

Xi Zhao

Department of Atmospheric Sciences, Texas A&M University, College Station, TX, USA

Contribution: ​Investigation, Writing - review & editing, Supervision

Search for more papers by this author
First published: 20 September 2022

Abstract

Deep convective clouds and associated anvils exert opposite radiative effects. The impact of different aerosol types on these two categories of clouds remains a major challenge in understanding aerosol-cloud interactions. Using 11-year satellite retrievals, we find that cloud top height (CTH) and ice cloud fraction of deep convective clouds and anvil cirrus identified by Cloud-Aerosol Lidar with Orthogonal Polarization increase with small aerosol loadings and level off or even decrease with further aerosol increase. Compared with continental aerosols, CTH affected by marine aerosols starts to decrease at smaller aerosol loadings. Moreover, cloud optical depth (COD) of deep convective clouds decreases with aerosol loadings. COD of anvil cirrus increases with increased loadings of most aerosol types but decreases with smoke aerosol. These relationships are mainly attributed to the aerosol effect rather than the meteorological effects. Our findings contribute to the development of models and better assessment of aerosol-cloud radiative forcing.

Key Points

  • The height and amount of Cloud-Aerosol Lidar with Orthogonal Polarization-identified deep convective clouds and anvil cirrus first increase and then level off with aerosol increase

  • Marine aerosols start to decrease cloud top height when aerosol optical depth is relatively smaller than continental aerosols

  • With increased aerosol loadings, deep convective clouds have a decreased cloud optical depth (COD) while anvils have an increased COD. One exception is for smoke

Plain Language Summary

By acting as the seeds of clouds, aerosols affect the formation and development of clouds, thereby affecting climate. Deep convective clouds and associated anvil cirrus are often accompanied with severe weather events. These two types of clouds have opposite climate effects: the former generally cools the Earth system while the latter warms the Earth system. Using 11 years of satellite data, we find that with the increase of aerosol loadings, the cloud top height (CTH) and cloud fraction of deep convective clouds and anvil cirrus identified by Cloud-Aerosol Lidar with Orthogonal Polarization first increase and then remain virtually unchanged or even decrease. We also analyze the effects of different types of aerosols on deep convective clouds and anvil cirrus. We find that, compared with continental aerosols, CTH affected by marine aerosols starts to decrease at smaller aerosol loadings. As the aerosol loadings increase, the cloud optical depth of deep convective clouds decreases while the optical thickness of the anvil cirrus increases. Therefore, these two categories of clouds as well as the effects from various aerosol types should be carefully considered when quantifying the aerosol effects on deep convective cloud systems.

1 Introduction

Deep convective clouds contribute significantly to global precipitation, radiation budget, and large-scale circulations (Arakawa, 2004; Fan, Leung, et al., 2012; Fan, Rosenfeld, et al., 2012; Fan et al., 2016; Han et al., 2022; Houze, 2014; Peng et al., 2016; Tao et al., 2012). In general, deep convective clouds develop until they hit the inversion layer, such as the tropopause, and further spread out, forming anvils (Fan et al., 2013; Koren, Remer, et al., 2010). Aerosol is an important factor influencing the characteristics and the radiative effects of deep convective clouds and detrained anvil cirrus. By serving as cloud condensation nuclei and ice nucleating particles (INPs), aerosols influence cloud microphysics, and thus change cloud macro-physical and radiative properties. This is known as aerosol-cloud interaction (ACI), which is a major source of uncertainties in the assessment of aerosol radiative forcing on climate (Boucher et al., 2013; Masson-Delmotte et al., 2021).

Many studies have investigated the impact of aerosols on deep convective clouds and found that aerosols can invigorate the convection by the so-called “aerosol invigoration effect”: increasing aerosol loadings produce more but smaller cloud droplets that suppress warm rain formation. Those smaller droplets are transported to supercooled levels to freeze, releasing larger amount of latent heating that adds to the local buoyancy, which invigorates the updrafts of cumuli for a deepening of cloud layer and higher rain rate (Andreae et al., 2004; Chakraborty et al., 2018; I. Koren et al., 2005, Koren, Feingold, et al., 2010; Pan et al., 2021; Rosenfeld et al., 2008). In addition, heat uptake of ice particles falling at lower levels and the subsequent evaporative cooling of cloud water enhance cold pools near the surface, which further affects the convection (Rosenfeld et al., 2008; Tao et al., 2007). Moreover, the radiative effects of aerosols may suppress or enhance convection by altering the stability and heating rate of the atmosphere (Fan et al., 2015; Lee et al., 2016; Li et al., 2019; M. Jiang et al., 2016; Y. Wang et al., 2013).

However, our understanding of aerosol effects on deep convective clouds and anvil cirrus is rather inadequate due to the complexity of aerosol-clouds interaction processes and mechanisms. The complexity partly stems from the opposite radiative effects of deep convective clouds versus anvil cirrus. That is, deep convective clouds are likely to induce a cooling effect while anvil clouds are likely to produce a warming effect on the Earth's radiation budget. For example, the cloud radiative effects calculated by Feng et al. (2011) at the top of the atmosphere over the Southern Great Plains region are −0.2 W m−2 for convective clouds and 0.8 W m−2 for anvils during June–August 2009–2010. Therefore, it is imperative to examine the aerosol effects on deep convective clouds and anvil cirrus separately. Nevertheless, only limited studies (Ekman et al., 2007; Fan et al., 2013; Koren, Remer, et al., 2010; Sarangi et al., 2018) thus far have investigated the aerosol effects on these two categories of clouds simultaneously. Koren, Remer, et al. (2010) found that increased aerosol loadings are associated with deeper convective clouds as well as larger and thinner anvils over the tropical Atlantic and Pacific in summer. Using long-term satellite data and model simulations, Sarangi et al. (2018) found that the aerosol invigoration effect leads to deeper convective clouds and enhanced formation of thicker anvil clouds at higher altitudes with smaller ice particles. Besides, the larger number of smaller ice particles and longer lifetime of anvil cirrus caused a net cooling effect over the Indian summer monsoon regions. Yan et al. (2014) showed the thickening of deep convective clouds and the expansion and thinning of anvils as aerosol loadings increase over the Southern Great Plains. These inconsistent changes of anvils with aerosol loadings call for further investigations.

The effects of aerosols on deep convective clouds and anvil cirrus through ACIs vary among aerosol types. I. Koren et al. (2005) found a systematic invigoration of convective clouds by pollution, dust, and biomass burning aerosols over the North Atlantic Ocean. J. H. Jiang et al. (2018) analyzed the effect of polluted continental aerosols, dust, and smoke over South America, Central Africa and Southeast Asia and found that the polluted continental aerosols and smoke tend to invigorate and suppress the convection, respectively, while the dust aerosol effect of convection is regionally dependent. Zhao, Gu, et al. (2018) found that the cloud optical depth (COD) of anvil clouds increases with increasing concentrations of dust and polluted continental aerosols but decreases with increasing concentrations of smoke over East Asia. W. Wang et al. (2010) and Huang et al. (2006), however, showed that dust aerosols, by serving as INP, decrease the COD of ice clouds over Northwestern China and downwind regions. Since previous studies on the effects of different aerosol types are limited and often show contradictory results, further research is much needed. To our knowledge, none of the previous studies have investigated the effect of different types of aerosols on deep convective clouds and detrained anvils simultaneously.

In this study, the interactions of various aerosol types with deep convective clouds and anvil cirrus are studied based on analyses of 11-year satellite observational data. The region of interest is over East Asia (70°–135°E, 15°–55°N), where there are abundant aerosol loadings as well as various aerosol types.

2 Data and Methodology

2.1 Sources and Collocation of Satellite Retrievals

The observations of aerosol and cloud properties from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard Aqua, the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) aboard Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) and the Cloud Profile Radar (CPR) aboard CloudSat Satellite are collocated and used in our analysis, as summarized in Table S1 in Supporting Information S1.

First, the profiles defined as deep convective clouds or anvil cirrus are selected as the reference profiles according to the CALIPSO level 2 merged aerosol and cloud layer product (05kmMLay, version 4.10) at a 5 km along-track resolution. It is noted that the deep convective clouds observed by CALIOP is high and opaque from a lidar perspective. These clouds are presumably associated with convection, but due to the limited ability of CALIOP to penetrate clouds, they may not necessarily penetrate through the atmospheric column (Liu et al., 2005). In this study, “deep convective clouds” refer to the clouds identified by CALIOP to be the “deep convective” type, which differs to some extent from the commonly used definition of deep convective cores (Takahashi et al., 2017). Only single-layer clouds with valid quality assurance (QA) flags (20 ≤ CAD score ≤ 100, Extinction QC = 0/1) are considered in our study. More details about the identification of deep convective clouds and anvil cirrus are given in Section 2.2. Then we collocate the MODIS/Aqua and CPR/CloudSat observations to the reference profiles. Aerosol optical depth (AOD) from the MODIS MYD04 (Level 2, Collection 6) product at 10 × 10 km resolution is averaged within a radius of 50 km around the reference profiles to present the loadings of aerosols interacting with clouds.

Considering the difference between radar and lidar and that the anvil cirrus are much thinner than convective clouds, COD from the CloudSat 2B-TAU product (Polonsky et al., 2008) and layer COD from the CALIPSO 05kmMLay product are used for convective clouds and anvil cirrus, respectively. We average the COD from 2B-TAU within a 3-km radius of the reference profile to represent the COD of deep convective clouds. As for ice cloud fraction (Cfi), following Hubanks et al. (2016), we calculate the ratio of the number of 1 × 1 km MODIS overcast cloud pixels (MYD06 product, collection 6) at which the “primary cloud retrieval outcome” is successful and the “cloud phase” is ice to the number of all pixels within a 20-km radius of the reference profile. For studying the interactions between aerosols and deep convective clouds, one of the most important cloud properties is cloud top height (CTH), which represents the strength of convection to some extent (Takahashi & Luo, 2014). CTH used here is from the 2B-CLDCLASS-LIDAR (version P1_R05), a combined CloudSat-CALIOP product that detects more cirrus cloud layers. The closest CTH from 2B-CLDCLASS-LIDAR is selected to represent the reference profile. More detailed explanations about the selection of satellite retrieval products are provided in Text S1 in Supporting Information S1.

To exclude the influence of meteorology on ACIs, a dozen of meteorological variables are obtained from the Final Analysis reanalysis data product of the National Centers for Environmental Prediction (NCEP) with 1° × 1° and 6-hr resolutions. The meteorological variables considered here include relative humidity, horizontal and vertical wind speed, vertical wind shear, and the energy related to convection (see a full list in Table S2 in Supporting Information S1). Since CALIPSO passes East Asia around 5:00 to 8:00 UTC, the NCEP data at 06:00 UTC is used to match the CALIPSO reference profiles.

2.2 Determination of Deep Convective Clouds, Anvil Cirrus, and Aerosol Types

The CALIPSO 05kmMLay product classifies each cloud layer into one of eight cloud types including deep convective clouds and cirrus. The determination of cirrus detrained from deep convective clouds, that is, anvil cirrus, is much more complex. G. G. Mace et al. (20012006) divided cirrus into two major types based on the distinct formation mechanisms: anvil cirrus and cirrus generated in situ by frontal systems or gravity waves. According to previous work (Riihimaki & McFarlane, 2010; Zhao, Liou, et al., 2018), a reference profile is defined as anvil cirrus if the CALIOP-classified cirrus is connected to a deep convective cloud. If two neighboring ice cloud profiles vertically overlap and are horizontally away from each other no more than 5 km, the two profiles are considered to be physically connected. Moreover, Zhao, Liou, et al. (2018) have proved that the classification of ice clouds is reasonable.

The CALIOP 05kmMLay product classifies each aerosol layer into one of seven aerosol types including dust, polluted dust, clean continental aerosols, polluted continental aerosols, clean marine, dusty marine, and elevated smoke. For each reference profile, if all vertical layers of the CALIOP profiles within the 50-km radius possess the same aerosol type, we will consider the reference profile to be primarily affected by that aerosol type. We adopt such a strict aerosol type identification criterion to decrease the possibility of contamination by other aerosol types.

3 Results and Discussion

3.1 Response of Cloud Top Height and Ice Cloud Fraction to Different Types of Aerosols

Samples of reference profiles are divided into five AOD bins with almost the same sample size in each bin. Each bin is averaged for both AOD and cloud properties, indicating the changes in CTH and Cfi as a function of AOD for CALIOP-identified deep convective clouds and anvil cirrus. For deep convective clouds, as shown in Figures 1a and 1b, CTH and Cfi increase with AOD under the condition of small aerosol loadings (AOD < 0.2–0.3) for all types of aerosols. Under large aerosol loading conditions (AOD > 0.2–0.3), CTH and Cfi either slightly decrease or barely change with AOD increase for all types of aerosols. Similar changes appear for anvil cirrus (Figures 1c and 1d) as for deep convective clouds. Under the lower aerosol loadings (AOD < 0.1–0.2), CTH and Cfi of anvil cirrus increase sharply with AOD increase. As AOD further increases, the CTH decreases either significantly or slightly depending on aerosol types, while Cfi slightly increases or barely changes.

Details are in the caption following the image

The response of cloud properties to aerosol loadings. The changes of cloud top height (CTH, a) and ice cloud fraction (Cfi, b) with the aerosol optical depth of various types for Cloud-Aerosol Lidar with Orthogonal Polarization-identified deep convective clouds. (c and d) are similar to (a and b), but for anvil cirrus. The error bars represent the standard errors (urn:x-wiley:00948276:media:grl64870:grl64870-math-0001), where σ is the standard deviation and N is the sample number. The sample numbers of CTH and Cfi are urn:x-wiley:00948276:media:grl64870:grl64870-math-0002 and urn:x-wiley:00948276:media:grl64870:grl64870-math-0003 respectively for deep convective clouds, and urn:x-wiley:00948276:media:grl64870:grl64870-math-0004 and urn:x-wiley:00948276:media:grl64870:grl64870-math-0005 respectively for anvil cirrus.

The increasing tendency of CTH and Cfi for small aerosol loadings is consistent with the invigoration hypothesis which indicates that more aerosols in the air suppress the warm rain processes and cause more cloud water to be lifted above the freezing level; as a result, more latent heat released owing to the freezing of the water droplets will invigorate the convection (Rosenfeld et al., 2008). At larger aerosol loadings, more cloud droplets form in the air and the lifting of the air parcels consumes more buoyancy energy, which is not conducive to the development of convective clouds in the later stage (Rosenfeld et al., 2008). Moreover, large amounts of aerosols attenuate the solar radiation reaching the surface through aerosol-radiation interactions, which weakens the convection (J. H. Jiang et al., 2018). Another possible reason for the decreasing tendency of CTH and Cfi at large aerosol loadings is that some absorbing aerosols, such as smoke, dust and polluted dust, promote the evaporation of cloud droplets and increase the atmospheric stability due to heat absorption (Y. Wang et al., 2013), which also inhibits the development of convection. Interestingly, CTH for deep convective clouds and anvil cirrus affected by marine aerosols decreases earlier than that affected by continental aerosols as AOD increases. A possible reason is that the more humid marine environment leads to the formation of more cloud water. That is, when the warm-phase rain is suppressed due to high aerosol loadings over the ocean, more buoyancy energy is consumed due to the larger cloud water amount over the ocean than that over the land. As a result, marine aerosols come upon the stage of inhibiting the convection sooner than continental aerosols.

There is a difference between deep convective clouds and anvil cirrus in the variation of CTH (Figures 1a and 1c). That is, with the increase of aerosol loadings, the CTH of anvil cirrus starts to decrease at smaller aerosol loadings than that of deep convective clouds, and once it starts to decrease, it also decreases faster. A possible reason for this difference is that anvil cirrus is more sensitive to convective strength than deep convective clouds. Therefore, when the convection strength is suppressed at large aerosol loadings, the CTH of anvil cirrus is inhibited more than that of deep convective clouds. Moreover, Figures 1c and 1d show the contrast between the aerosol effects on CTH and Cfi of anvil cirrus. That is, under polluted conditions, Cfi slightly increases or barely changes while CTH decreases as the convection is inhibited, probably because CTH is primarily dependent on convective strength while Cfi depends on both convective strength and cloud droplet activation/ice nucleation. The more frequent cloud droplet activation/ice nucleation owing to increased aerosols produces a larger number of smaller cloud droplets and ice crystals, resulting in the slight increase of Cfi with lower sedimentation velocities of smaller ice crystals in anvil cirrus.

3.2 Response of Cloud Optical Depth to Different Types of Aerosols

Figure 2 shows the variations of COD of deep convective clouds and anvil cirrus with the change of AOD for different aerosol types. Under smaller aerosol loadings (AOD < 0.3), for each aerosol type, the COD of deep convective clouds decreases with increasing aerosols. For anvil cirrus, COD increases with increasing aerosol loading for all aerosol types except that COD shows a decreasing tendency with smoke aerosols. With the further increase of aerosol loadings (AOD > 0.3), the variations of COD for both deep convective clouds and anvil cirrus become small, while the COD of anvil cirrus still decreases significantly with smoke aerosol. The distinct changes of COD for deep convective clouds and detrained anvils under small AOD (AOD < 0.3) are possibly because more water droplets or ice crystals are detrained to the anvils under stronger convection due to the aerosol invigoration effect, which decreases the COD of deep convective clouds while increasing that of anvil cirrus. Fan et al. (2013) also found increasing COD of anvil cirrus in polluted conditions when examining the impact of aerosols on the lifecycle of deep convective clouds. They suggested that in polluted conditions, convective clouds detrain larger amounts of smaller cloud hydrometeors with smaller fall velocities, leading to larger and thicker anvils. As aerosol loadings increase (AOD > 0.3), the detrainment of ice crystals from the deep convective clouds to the anvils is weakened along with the inhibited convection and thus the COD changes become smaller. As mentioned above, there is a significant difference in COD variations for anvil cirrus between smoke and other types of aerosols (Figure 2b). Possibly, this is because smoke aerosols, as a strong absorber of solar radiation, heat up the ice crystals and the surrounding atmosphere (Y. Wang et al., 2013), which can accelerate the sublimation of ice crystals, known as the semi-direct effect. This effect has also been shown to be applied to liquid clouds (B. T. Johnson et al., 2004; Lu et al., 2018). Therefore, the semi-direct effect of smoke aerosol exerts a more important role than the detrainment of ice crystals by its microphysical effect in this case.

Details are in the caption following the image

The changes of cloud optical depth (COD) with the aerosol optical depth of various types for Cloud-Aerosol Lidar with Orthogonal Polarization-identified deep convective clouds (a) and anvil cirrus (b). The error bars represent the standard errors (urn:x-wiley:00948276:media:grl64870:grl64870-math-0006), where σ is the standard deviation and N is the sample number. The sample numbers of COD are urn:x-wiley:00948276:media:grl64870:grl64870-math-0007 and urn:x-wiley:00948276:media:grl64870:grl64870-math-0008 for deep convective clouds and anvil cirrus, respectively.

3.3 Disentangling the Influence of Meteorological Factors

The relationships between AOD and cloud properties described above do not always mean causality. That is, aerosols and cloud properties may vary simultaneously due to the variations of certain meteorological factors. As a result, we may mistakenly think that clouds are affected by aerosols. However, the fact is that the correlations between aerosols and clouds should be attributed to the covariation of meteorological factors. Therefore, we must check whether meteorological variations explain the aerosol-cloud correlations. Here, two methods are adopted to disentangle the influence of meteorological factors from the aerosol effects.

The first method is to restrict the meteorological variability into smaller ranges and compare the relationships between aerosols and deep convective clouds for different ranges. In our study, 13 meteorological variables that possibly influence ACIs are considered, including the U-component of horizontal wind speed at 300 hPa (U300) and 1,000 hPa (U1000), the V-component of horizontal wind speed at 300 hPa (V300) and 1,000 hPa (V1000), the vertical wind velocity at 300 hPa (VV300) and 500 hPa (VV500), the relative humidity at 300 hPa (RH300), 500 hPa (RH500) and 850 hPa (RH850), convective available potential energy (CAPE), convective inhibition (CIN) and the U- (VWSH_U) and V- component of vertical wind shear (VWSH_V). Here we calculate the correlation coefficients between cloud properties and meteorological variables (Table S2 in Supporting Information S1) and find that U300, CAPE, and VWSH_U correlate best with cloud properties among all of the meteorological variables. Therefore, the relationships between AOD and cloud properties of deep convective clouds and anvil cirrus are examined in Figures S2 and S3 in Supporting Information S1 by restricting these three parameters. We find that the relationship patterns between AOD and CTH, Cfi and COD are similar under different ranges of U300. The same is also true for CAPE and VWSH_U. Therefore, the meteorological covariations are not the primary reason for the observed aerosol-cloud relationships.

The other method of disentangling the impacts of meteorological factors is to calculate the partial correlations that define the linear correlation between two statistical variables (AOD and cloud properties here) under the condition that all other relevant variables (i.e., 13 meteorological variables mentioned above) are restricted (Engström & Ekman, 2010; PSU, 2017; R. A. Johnson & Wichern, 2007). If the sign of the total correlations between AOD and cloud properties is the same as that of the partial correlations eliminating the effect of certain meteorological variables, the correlations between AOD and cloud properties are not attributable to these meteorological variables. If the total and the partial correlations show opposite signs, the impact of meteorological variables should not be ruled out when analyzing the aerosol-cloud relationships. Figure 3 shows the total and the partial correlations of CTH, Cfi and COD for deep convective clouds and anvil cirrus with AOD considering 13 meteorological variables mentioned above. Specific values of the correlations can be found in Tables S3 and S4 in Supporting Information S1. Over 85% of correlations are statistically significant at the 99% level. Except for these, most total and partial correlations between AOD of dust, marine and “all aerosol types” and CTH of deep convective clouds, as well as the correlations between the dusty marine AOD and Cfi, do not pass the significance test. As for the anvil cirrus, the total and partial correlations of marine AOD and CTH are not statistically significant. Notably, these insignificant correlations are also very small. The possible reason is that the sample number of marine aerosols is rather small while the responses of CTH to dust and “all aerosol types” are not monotonical. Moreover, the fact that the total and the partial correlations tend to pass or not pass the significance test at the same time suggests that the meteorological factors have limited effect on the relationship between AOD and cloud properties. Remarkably, only the partial correlation between AOD and CTH excluding the effect of RH300 shows the opposite sign to the total correlation, but both the correlations are negligibly small (see Table S3 in Supporting Information S1). In other words, the sign of total correlation for each aerosol type is almost always the same as that of the corresponding partial correlations excluding the influence of one or all meteorological factors, implying that the aerosol-cloud relationships for deep convective clouds and anvil cirrus are significantly attributed to the aerosol effects. Remarkably, the COD of deep convective clouds shows negative correlations with AOD for all types of aerosols, whereas that of anvil cirrus shows positive correlations with AOD for all aerosol types except for smoke aerosol.

Details are in the caption following the image

Total and partial correlations between aerosol optical depth (AOD) and cloud top height (CTH) (a and d), AOD and ice cloud fraction (Cfi, b and e), and AOD and cloud optical depth (COD) (c and f) for Cloud-Aerosol Lidar with Orthogonal Polarization-identified deep convective clouds (a–c) and anvil cirrus (d–f) with the effect of meteorological factors eliminated individually and simultaneously. For the cases of CTH and Cfi, the AOD range is [0, 0.8] for all types of aerosols except for clean marine and dusty marine for which the AOD range is [0, 0.2]. For the cases of COD, the AOD range is [0, 0.8] for all types of aerosols. The AOD ranges are determined to be the interval in which cloud properties change monotonically with AOD. Over 85% of correlations are statistically significant at the 99% level (bold font in Tables S3 and S4 in Supporting Information S1).

4 Conclusions and Implications

In this study, the effects of different types of aerosols on both deep convective clouds and anvil cirrus identified by CALIOP were investigated using 11-year of satellite data. We found that cloud top height (CTH) and Cfi of deep convective clouds and anvil cirrus increase with increasing aerosol loadings under AOD < ∼0.2 for all aerosol types possibly because of the aerosol invigoration effect. Under more polluted conditions with higher AOD, Cfi slightly increases or even decreases, while CTH slightly decreases for deep convective clouds and noticeably decreases for anvil cirrus, likely due to the consumption of buoyancy energy as well as the block of solar radiation by aerosols. Besides, we found that the CTH affected by marine aerosols decreases at lower AOD than that affected by continental aerosols which may be attributable to the more efficient formation of cloud droplets on the marine aerosols in the strong updrafts that consume the buoyancy energy. We also recognized some distinctions in the variations of CTH that the CTH of anvil cirrus starts to decrease when AOD is relatively smaller than the case for deep convective clouds. For anvil cirrus, compared with the decreasing tendency of CTH at large aerosol loadings, Cfi shows a slightly increasing tendency likely because aerosol activation is enhanced with the presence of more aerosols. As for cloud optical properties, COD of deep convective clouds decreases as the aerosol loading increases. This is because with increased loading of aerosols, the convection is enhanced due to the aerosol invigoration effect. Therefore, more water droplets or ice crystals are detrained to the anvils by stronger horizontal winds in the upper atmosphere, which brings down the COD of deep convective clouds. This effect is much stronger than the increased COD due to reduced size of cloud droplets with increased aerosol loading. In contrast, COD of anvil cirrus increases with increasing AOD for all types of aerosols except for elevated smoke. This is because smoke aerosols can heat the atmosphere, which will decrease cloud amount and increase atmospheric stability, resulting in a reduced COD with increased smoke aerosol.

Many previous studies investigated the aerosol effect on deep convective clouds but the deep convective clouds and detrained anvils were seldom considered separately. Given the distinct optical properties of deep convective clouds and anvil cirrus, the former produces the cooling effect while the latter induces the warming effect. Therefore, these two types of clouds should both be considered when estimating the overall radiative effects of deep convective cloud systems. Our results show the observed variations of cloud physical and optical properties with aerosol changes especially the opposite trends of COD for deep convective clouds and anvil cirrus, which helps to better quantify the aerosol effects on deep convective cloud systems. In addition, our study includes different aerosol types in the investigated East Asia regions with anthropogenic aerosols from intensive human activities mixed with natural dust and long-range transported aerosols. In this study, natural and anthropogenic aerosols are distinguished to examine their respective effects, which is conducive to a more accurate assessment of human activities on deep convective clouds, extreme weather, and regional climate (Wei et al., 2022). This long-term observational analysis separating deep convective clouds and anvil cirrus and taking various aerosol types into account can be used to perform a systematic and refined evaluation of numerical models and to improve model physical representations of deep convective clouds and ACIs. Finally, the mechanisms underlying the interactions of aerosols and deep convective cloud systems proposed in this study should be validated by models and the relative contributions of different mechanisms should be quantified in future studies.

Acknowledgments

J. Zhang and C. Wu were supported by the National Natural Science Foundation of China (Grant No. 41830966). Z. Jiang was supported by the National Natural Science Foundation of China (Grant No. 41975178). B. Zhao is supported by the Tencent Foundation through the XPLORER PRIZE.

    Data Availability Statement

    The satellite and meteorology data products used in this study are publicly available at the following sites. MODIS/Aqua Aerosol 5-Min L2 Swath 10 km Product (MYD04_L2, Collection 6) and MODIS/Aqua Clouds 5-Min L2 Swath 1 and 5 km Product (MYD06_L2, Collection 6) are available from https://ladsweb.modaps.eosdis.nasa.gov/search/. The Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) Level 2 Cloud, Aerosol and Merged Layer (05kmMLay) V4.20 Products can be directly downloaded from the NASA Langley Research Center Atmospheric Science Data Center (NASA/LARC/SD/ASDC), available at https://asdc.larc.nasa.gov/data/CALIPSO/. The CloudSat Level 2 Combined Radar and Lidar Cloud Scenario Classification Product (2B-CLDCLASS-LIDAR) and Level 2 Cloud Optical Depth Product (2B-TAU) used in this analysis can be downloaded from the online open repository at the Colorado State University website (https://www.cloudsat.cira.colostate.edu/). The National Center for Environmental Prediction-Final (NCEP FNL) can be downloaded from the Research Data Archive (https://doi.org/10.5065/D6M043C6) managed by the National Center for Atmospheric Research (NCAR).