Clouds play a crucial role in various physical and chemical processes occurring in the lower troposphere and hence affect the Earth's radiation budget (Seinfeld et al., 2016; Tilgner et al., 2014). Aerosol particles, including primary and secondary ones generated from natural and anthropogenic sources, either directly alter radiative forcing or act as cloud condensation nuclei (CCN) to indirectly influence it. At present, aerosol–cloud interactions unquestionably affect radiative forcing and global climate (McFiggans et al., 2006; Rosenfeld et al., 2014; Seinfeld and Pandis, 2006). CCN become cloud droplets through the condensation of water vapor when the relative humidity (RH) of an air parcel increases above saturation (Farmer et al., 2015). Size, chemical composition, and mixing state are main factors affecting the ability of a particle to act as CCN (Dusek et al., 2006; Fan et al., 2016; Hudson, 2007; Li et al., 2011a; Rosenfeld, 2000). In addition, aerosol particles incorporated into cloud droplets can be easily lifted into the free troposphere during cloud development and further extend their influence on cloud precipitation and regional climate (Fan et al., 2016).

Owing to rapid industrialization and urbanization in Asia, large quantities of aerosol particles from anthropogenic sources are released into the atmosphere, which can dramatically affect the chemical composition of CCN, and furthermore change the properties of clouds such as radiative forcing, lifetime, and precipitation patterns (Drewnick et al., 2007; Ervens, 2015; Li et al., 2011b; Twohy and Anderson, 2008). High concentrations of aerosol particles increase the number of cloud droplets and reduce their size, which further results in the reduction of precipitation efficiency and in extending the lifetime of clouds (Fan et al., 2016; J. Li et al., 2017; McFiggans et al., 2006; Qian et al., 2009; Rosenfeld, 2000). Moreover, anthropogenic aerosol particles – especially fly ash, metal, and soot particles – are incorporated into cloud droplets, and are transported long distances to affect ecosystems, human health, and radiative forcing (Li et al., 2013; Rosenfeld et al., 2014). In particular, the toxic and bioaccumulative metals can deposit into the ecosystem following fog or precipitation and further cause severe health problems to human beings (Liu et al., 2012). Moreover, transition metals such as iron (Fe) and manganese (Mn) can enhance the in-cloud oxidation of sulfur dioxide to sulfate (Harris et al., 2013).

Recently, many studies have been performed worldwide to investigate aerosol–cloud interactions and the composition of cloud droplets. Schroder et al. (2015) investigated the activation of refractory black carbon particles in stratocumulus clouds at a marine boundary layer site using a counterflow virtual impactor and single-particle soot photometer. Ueda et al. (2014) reported the effects of in-cloud processes on the compositional changes of sea salt particles by collecting individual aerosol particles in and below clouds, respectively. Pierce et al. (2015) calculated size distribution changes and radiative forcing effects due to the scavenging of interstitial particles by cloud droplets in a clean remote region. Roth et al. (2016) analyzed the composition and mixing state of cloud residues and out-of-cloud aerosol particles by single-particle aerosol mass spectrometry (SPAMS) on a mountain site and found that soot particles internally mixed with sulfate and nitrate were the dominant ones in cloud residues. All of the above studies were carried out in the clean atmosphere; they could not observe the clear interactions between abundant anthropogenic particles and cloud droplets. However, the latest satellite observations indicated that large numbers of anthropogenic fine particles assembled in the cloud base and might modify cloud properties in heavily polluted air influenced by industrial and urban emissions (Eck et al., 2018). Field observations are needed to confirm this and understand the interactions of aerosol–cloud over polluted areas, especially in North and South Asia.

Mt. Tai, the highest mountain in the NCP, is surrounded by several medium-sized industrial cities. The altitude of Mt. Tai is close to the top of the planetary boundary layer (PBL) above the NCP. Therefore, Mt. Tai is an ideal site to study the effects of regional transport and local emissions of anthropogenic aerosols on cloud properties. Numerous studies have been conducted on Mt. Tai, but virtually all the researchers mainly focus on the variation of chemical composition and size distribution of aerosol particles (Zhang et al., 2014) and the chemical composition of cloud water (J. Li et al., 2017; Wang et al., 2011). Because of the limitation of sampling and analyzing techniques, these studies did not consider the aerosol–cloud interactions at the top of Mt. Tai in North China.

Transmission electron microscopy (TEM) has become a powerful technique to characterize the particle morphology, size, and composition and mixing properties of different components in individual particles in recent years (Li et al., 2016a; Ueda et al., 2014). Many studies used SPAMS to characterize the composition of residual particles of individual cloud droplets (Lin et al., 2017; Pratt et al., 2010; Zhang et al., 2017). Compared to SPAMS, TEM can directly observe the morphology and different components of individual cloud droplet residual (cloud RES) and interstitial particles (cloud INT) (Kojima et al., 2004; Li et al., 2011a; Twohy and Anderson, 2008; Ueda et al., 2014). Therefore, the TEM technique can not only be used to identify cloud RES and cloud INT collected in a same cloud event but also reveal interactions between aerosol and clouds based on the different components within individual particles. In this study, we collected individual particles during cloud events at the summit of Mt. Tai and applied TEM to obtain and compare the morphology, size, and composition of cloud RES and cloud INT. This is helpful to understand the influence of anthropogenic sources on cloud properties above the heavily polluted region.

TEM observations in this study reveal that cloud RES contained large numbers of refractory-containing particles primarily emitted from various anthropogenic sources in the heavily polluted NCP. As much as 76 % of cloud RES were identified as S-refractory particles (Fig. 7a). We found that soot-containing particles (i.e., S-soot and S-fly ash/metal-soot) were the most abundant (62 %) among the cloud RES, followed by the relatively abundant fly ash/metal-containing particles (i.e., S-fly ash/metal and S-fly ash/metal-soot, 37 %) compared with 18 % and 7 % in the cloud INT, respectively (Fig. 7a). Although these refractory particles such as soot with hydrophobic properties could not be CCN directly, they can be easily accumulated by the existing cloud droplets as inclusions (Zuberi et al., 2005). In the heavily polluted NCP, large numbers of soot and fly ash/metal particles are released from anthropogenic sources (e.g., industrial activities and vehicular exhaust). During cloud events, abundant refractory particles can be efficiently entrained into existing liquid cloud droplets by wet scavenging. Li et al. (2011a) reported that particle number decreased dramatically during cloud formation at Mt. Tai with a scavenging ratio of 0.54, which demonstrated that aerosol particles could efficiently be incorporated into cloud droplets. Physical coagulation of interstitial particles with cloud droplets is an important process in developing clouds, which can lead to the reduction of cloud INT number and a size increase of cloud RES after the clouds dry (Pierce et al., 2015). As we know so far, the extremely complicated mixture of secondary and primary particles observed in the present study has seldom been found in cloud droplets in clean air over developed countries (Hao et al., 2013; Kojima et al., 2004; Schneider et al., 2017; Ueda et al., 2014), remote areas (Hiranuma et al., 2013) and the ocean (Hopkins et al., 2008; Twohy and Anderson, 2008; Zhang et al., 2006). For example, Zhang et al. (2006) reported that S-rich particles were predominant in the cloud RES with a small number fraction of sea salt particles over the Sea of Japan and soot or fly ash/metal particles were not observed. The present study reveals that individual cloud droplets are a far more complicated system in polluted air in North China than in the pristine continental and clean ocean air of the world.

Cloud properties such as albedo and lifetimes could be largely modified by the aerosol–cloud interactions, especially in heavily polluted regions (Wang et al., 2010, 2013). The model simulation revealed that black carbon aerosols had a noticeable impact (up to nearly 20 %) on cloud droplet number concentration in polluted black carbon source regions (Cherian et al., 2017). Especially, abundant black carbon particles incorporated into cloud droplets could lead to a decrease in cloud albedo by absorbing radiation and an increase of temperature in the troposphere, then an acceleration of the evaporation of the cloud droplets (Ackerman et al., 2000; Adachi et al., 2010; Wang et al., 2013; Zuberi et al., 2005). In the past few decades, precipitation was significantly reduced over east central China due to the large amount of anthropogenic aerosols (Qian et al., 2009; Zhao et al., 2006). Because an excess of aerosols in clouds could reduce precipitation, the nonprecipitating clouds in the NCP tend to evaporate back to aerosol particles by solar radiation. It is highly probable that abundant black carbon particles present in the cloud droplets in the heavily polluted NCP in this study (e.g., particles 1 and 2 in Fig. 8) significantly affected the cloud properties and regional climate.

Fly ash and metal particles are a typical “fingerprint” pointing to the coal-fired power plants and boilers in factories and heavy industries (e.g., steel plant and smelting factory) (Chen et al., 2012; Li et al., 2016a; Moffet et al., 2008). Indeed, the most intense emissions from various industries in the world occur in Hebei and Shandong provinces in the NCP (Qi et al., 2017). It is well known that these industrial activities continuously release anthropogenic pollutants via high stacks into the upper air. Liu et al. (2012) reported that the concentration of Zn reached 249.1 µgL-1 in the cloud and fog water samples at Mt. Tai, followed by Al (157.3 µgL-1), Fe (105.8 µgL-1), Pb (46.2 µgL-1), and Mn (42.8 µgL-1), which were much higher than those values reported at Mt. Schmücke in Germany (Fomba et al., 2015). Combining these results with our present study, we infer that fine primary particles emitted from these industrial activities might spread and be lifted to the upper air more easily than at ground level. These metal particles, especially Pb and Zn of nanometer size, can harm ecosystems and human health (Roberts et al., 2004). The fly ash and metal particles incorporated into cloud droplets (e.g., particle 1 in Fig. 8) could go through the atmospheric acid Fe dissolution processes during long-range transport reported by W. Li et al. (2017). If they were further transported to remote oceanic regions, soluble Fe species in the aerosol particles can fertilize plankton on the surface of the ocean (Ito and Shi, 2016; W. Li et al., 2017). Therefore, these anthropogenic fly ash/metal particles in polluted air contaminate cloud droplets and further amplify potential impacts of fine metal particles on the biogeochemical cycle in the troposphere.

Some studies suggested that the aqueous oxidation of SO2 to sulfate by H2O2 and O3 in cloud droplets was dominant at Mt. Tai (Shen et al., 2012). However, cloud water collected on Mt. Tai contained high concentrations of soluble Fe, Mn, Zn, and Pb (Liu et al., 2012). These soluble metals in cloud droplets are released from aqueous reactions between metal particles and acidic sulfates in cloud droplets (W. Li et al., 2017). Harris et al. (2013) estimated that the oxidation of SO2 in cloud droplets catalyzed by natural transition metal ions (TMIs) in mineral dust was dominant at Mt. Schmücke in Germany. For this study, how the soluble anthropogenic TMIs drive sulfate formation through TMI catalysis in micro-cloud droplets is still unresolved in polluted air. We propose that anthropogenic TMI catalysis contributing to sulfate production should be further studied in cloud droplets in the polluted NCP.

The nonprecipitating cloud processes over the polluted air of NCP quickly change the composition of aerosol particles and cloud droplets in the upper air, potentially causing various effects such as human health, regional climate, and biogeochemical cycle at the larger regional scale. To better illustrate the aerosol–cloud interactions in this study, we created the conceptual model of Fig. 9. The tall stacks of plants can emit smoke plumes that contain fine refractory particles and gaseous pollutants to the upper air. A portion of particles in urban areas can also be lifted to the mountaintop by prevailing valley winds. Once the clouds form on a mountaintop (or are later transported in other directions), these cloud droplets act as collectors to scavenge the refractory particles. These refractory particles as inclusions might complicate the cloud chemistry in micro-cloud droplets. Gaseous pollutants such as SO2, NOx, and VOCs may have enhanced aqueous oxidation potential in the complex cloud droplets. Our study is designed to better understand the aerosol–cloud interactions on the mountaintop in polluted industrial and urban areas. Recently, a study showed that the major aerosol pollution events with very high fine-mode aerosol optical depth (>1.0 in mid-visible) in the China–Korean Peninsula–Japan region are often observed to be associated with significant cloud cover (Eck et al., 2018). Therefore, we expect that large numbers of fine refractory particles from polluted areas scavenging in clouds have important impacts, not only at a local scale, but also at a large regional scale.

Individual aerosol particles were collected during cloud events on Mt. Tai from 22 July to 1 August 2014. Cloud RES and INT particles were separated by their distribution on TEM grids, and their composition was identified by TEM–EDS. Individual particles were classified into S-rich, refractory (i.e., mineral, soot, fly ash/metal), and S-refractory (i.e., S-mineral, S-soot, S-fly ash/metal, and S-fly ash/metal-soot). According to air mass backward trajectories and PM2.5 concentrations on Mt. Tai, the entire sampling period was divided into three classes: a clean period (PM2.5<15 µgm-3), a lightly polluted period (15 µgm-3 < PM2.5 < 35 µgm-3), and a polluted period (PM2.5 > 35 µgm-3). In the clean period, individual particles were dominated by S-rich particles (78 %), whereas the fraction of refractory particles and S-refractory particles increased significantly and dominated during the polluted periods. This suggested that anthropogenic pollutants from tall stacks of coal-fired power plants and heavy industries and vehicular exhaust in cities could be lifted to the summit of Mt. Tai under the prevailing southerly winds in summer.

TEM observations showed that 76 % of cloud RES were S-refractory particles contaminated by anthropogenic refractory particles, compared with only 22 % of cloud INT. Cloud RES displayed a larger size than cloud INT, which indicates that particle size decidedly affects CCN ability. Our study reveals that large numbers of anthropogenic refractory particles were incorporated into cloud droplets through in-cloud processes. It is especially important that abundant black carbon particles in cloud droplets could alter the radiative forcing of clouds and accelerate the evaporation of cloud droplets. The high concentrations of transition metal ions might favor the aqueous-phase oxidation of SO2 by O2 in cloud droplets under the heavily polluted conditions in the NCP. Fly ash/metal-containing cloud droplets could be transported long distances and harm ecosystems and human health through wet deposition. We propose a conceptual model to show the aerosol–cloud interactions on mountaintops influenced by heavily polluted air.