The vertical profiles of aerosol size distributions measured during the spring and summer campaigns are shown in Fig. 4. Note that the size distributions between 3 and 10 nm are calculated based on the difference in concentrations monitored by the two CPCs, and the size distributions above 10 nm are measured by the FIMS. The size distributions show strong vertical heterogeneity and strong seasonal variability. For both seasons, the aerosol concentrations at the surface are the highest due to surface emissions. However, the dominant size of these aerosols during summer (∼ 100 nm) is larger than that during spring (∼ 30 nm), which is likely a result of enhanced condensational growth due to stronger volatile organic compound (VOC) emissions and secondary organic aerosol (SOA) formation in summer (Liu et al., 2021). Additionally, the more abundant gaseous precursors can promote aqueous chemistry in cloud droplets, which accumulates the mass and increases the size of the accumulation-mode aerosols (Wang et al., 2007). The larger size of the aerosols in the summer also leads to a larger condensation and coagulation sink for nucleating vapors and newly formed particles, suppressing the occurrence of NPF in the summer. As a result, the concentration of aerosols below 30 nm in the BL (e.g., below 1000 m) during spring is higher than that during summer. In the summer BL, the mode size of the aerosols gradually increases with altitude due to aerosol dynamics such as cloud processing (aqueous reaction and coalescence scavenging) and condensational growth (Wang et al., 2006). Due to the larger size of the aerosols that facilitates cloud droplet activation, the summertime size distributions show a Hoppel minimum, which is the average particle size at which particles become cloud condensation nuclei (CCN) of around 90 nm in the upper BL (600 to 1500 m) because of cloud processing (Fig. 4b). The in-cloud scavenging also removes aerosols and smaller cloud droplets, leading to a lower aerosol concentration at the higher altitudes of the BL. In the FT, aerosols are distributed in layers through long-range transport events. However, the concentration of the transported aerosols is higher in spring, likely due to the more dominant air mass arriving from the west that is associated with wildfires, as also indicated by the higher CO mixing ratio in Fig. S5a. The influence of long-range transport aerosols is also discussed in the case study in Sect. 3.4.4.

To quantify the vertical heterogeneity of the aerosols, we integrated the FIMS-measured aerosol size distributions in different size ranges. We calculated the concentrations of aerosols in size ranges from 10 to 20, 20 to 100, and above 100 nm to qualitatively study the properties of nucleation-, Aitken-, and accumulation-mode aerosols. Strictly speaking, the Aitken- and accumulation-mode aerosols are separated by the Hoppel minimum (not 100 nm), which is dependent on aerosol properties and meteorological conditions. However, due to the high concentration of aerosols at the surface, a clear Hoppel minimum could not be observed for most of the flights. The nucleation mode should also contain aerosols with sizes below 10 nm. Since the FIMS is limited to the measurement of particles above 10 nm, we also analyzed N3–10 derived by the CPC measurements. For simplicity, we use N10–20, N20–100, and N>100 interchangeably with nucleation-mode, Aitken-mode, and accumulation-mode aerosols. We also use N>3 to represent the total particle concentration.

Figure 5 shows the vertical profiles of particle number concentrations under normalized conditions (STP, 273.15 K and 101.325 kPa). Overall, a higher total aerosol concentration (N>3) can be observed in the spring campaign at all the altitudes (Fig. 5a), and the difference in concentrations between the spring and summer is larger in the BL. Moreover, the variation of N>3 in the same altitude bin during spring is larger compared to summer, resulting from the intense but sporadic NPF events during spring, especially in the upper BL. This can also be indicated by the mean values of N3–10 and N10–20 being outside the 25th to 75th percentile box in the altitude range between 500 and 1000 m (Fig. 5b and c). In the spring campaign, there were three NPF events where N3–10 and N10–20 values were above 2000 cm-3 in the altitude range of 500 and 1000 m, which is significantly higher than the median values below 600 cm-3.

During summer, there is a larger variation of N>100 in the same altitude bin compared to that of spring (Fig. 5e); however, the influence of this variation on the variation of N>3 is small due to the relatively small contribution of N>100 to N>3. At the same time, the relatively small variation of N>3 during summer is caused by the weaker NPF during summer. The more consistent source of the air mass from southeast (Fig. 3) during summer also leads to a more consistent background aerosol condition. A study at the SGP shows that the summertime air mass pathways were primarily from the same direction (southerly), potentially resulting in lower variability in observed aerosol properties (Andrews et al., 2011). Moreover, a number of studies documented episodically high concentrations of aerosols at the SGP during the spring season due to agricultural and wildfire sources in the area as well as the influx of biomass burning aerosols from different regions across North America (Peppler et al., 2000; Wang et al., 2006). Therefore, these results also suggest the importance of seasonal synoptic-scale variability for aerosol properties and potential aerosol–cloud interactions by influencing the CCN activity. A recent study analyzing the long-term CCN activity at the SGP shows that aerosols associated with air masses passing over various locations (e.g., Central America, the Gulf of Mexico, and the west) have very different accumulation-mode aerosol concentrations and hygroscopicity (Patel and Jiang, 2021). However, as indicated by the study, particle size is still the most critical factor influencing the ability of aerosols to activate, whereas the effect of chemical composition was secondary.

The breakdown of the total aerosol concentration into different size modes shows that, in the BL, N3–10 and N10–20 are higher during the spring (Fig. 5b and c), while N>100 is higher during the summer (Fig. 5d). Springtime and summertime N20–100 are comparable in the BL, but springtime N20–100 in the FT is higher than that in the summer, most likely resulting from long-range transport events (Sect. 3.3.4). Therefore, the larger BL total aerosol concentration in spring mainly results from higher values of N3–10 and N10–20 caused by the NPF events. The high N3–10 and N10–20 extend to the upper BL (∼ 1000 m), which could be explained by the upward mixing of the surface air but at the same time suggesting that NPF events may occur directly in the upper BL. Indeed, NPF events at different altitudes were observed in multiple research flights, and a few example cases are discussed in Sect. 3.3.1. The higher N>100 in summer is partly due to growth by condensable vapors transported from anthropogenic sources in the south and biogenic sources east of the SGP (Liu et al., 2021). At the same time, because the summertime size distribution shows a Hoppel minimum (Fig. 4b) and most particles above 100 nm are in the accumulation mode, the aqueous-phase chemistry plays a role in increasing N>100 as well. These accumulation-mode aerosols also serve as condensation and coagulation sinks for nucleating vapors and newly formed particles, reducing the concentration of nucleation-mode aerosols.

The aerosol concentrations in different modes show a vertical gradient where the concentration at the surface is highest. This may be due to the stronger emission at the surface, but the low BL heights observed during several research flights can also contribute to a higher average concentration of aerosols near the surface. To exclude the influence of BL height in analyzing the vertical gradient of aerosol concentrations, we examined the particle concentrations as a function of normalized altitude, z/z0, where z is the altitude and z0 is the BL height. The particle concentrations are first binned and averaged using the specific BL height for each flight and are then combined for the entire campaign (Fig. S6). Comparing Fig. S6 against Fig. 5, we can find that the vertical gradient of aerosol concentrations, especially near the surface, is greatly reduced when concentrations are plotted against normalized altitude, suggesting that BL height and atmospheric stability strongly influence the vertical distribution of aerosols. Overall, N>3 and N20–100 are relatively constant in the lower BL (z/z0 < 0.75), while N>100 is higher near the surface, potentially due to the faster growth by the more available condensing vapors at the surface or more efficient removal by clouds at lower altitudes in the BL. Aerosol concentration in all size ranges decreased with increasing altitude in the upper BL (z/z0 > 0.75) due to cloud scavenging. N3–10 and N10–20 have larger values in the middle to upper BL, especially near z/z0≈0.65, again highlighting the occurrence of NPF in the upper BL and their potential contribution to the aerosol population over the SGP.

The chemical composition of the non-refractory aerosols measured by the HR-ToF-MS is shown in Fig. 6. The aerosol mass is dominated by organics and sulfates for both seasons (> 75 %). Except for nitrate, all other components have higher concentrations during summer. Since nitrate aerosols are more stable under low temperatures, the higher nitrate concentrations during spring are likely driven by the lower temperature in the spring (Liu et al., 2021; Parworth et al., 2015). Ammonium is likely associated with agricultural emissions. The increased levels of organics during the summer season can be attributed to biogenic emissions originating from the east of the SGP, while the higher sulfate concentration is a result of the transportation of anthropogenic emissions from the south of the SGP (Figs. 3 and S1). Moreover, the previous study by Xu et al. (2015) showed that the formation of isoprene SOA is mediated by the abundance of sulfate through chemical reactions among sulfate, organics, and water, which allows the increase in both organics and sulfates in the summer. Although the organics and sulfate aerosol mass ratios are similar in spring and summer at the SGP, the detailed chemical compositions are different. The study by Liu et al. (2021) showed that, despite the higher organic aerosol mass in summer, these organics are less oxidized than in spring, as indicated by the O:C and H:C ratios. These summertime organics are not generated by biomass burning plumes as the study by Liu et al. (2021) analyzing the aerosol mass spectrometer data showed that the biomass burning organic aerosol (BBOA) factor was not identified during the summer season. This result is also consistent with the low concentrations of BBOA observed in summer at the SGP in a previous study (Parworth et al., 2015). According to Liu et al. (2021), the different oxidation levels in the spring and summer are likely due to (1) aerosols in the spring being more aged due to a longer residence time in the atmosphere, potentially different oxidant concentrations, or a combination of both; (2) different VOC species potentially contributing to aerosol formation in spring and summer (e.g., higher monoterpene and isoprene concentrations were observed in the summer); and (3) the more abundant biogenic VOCs in the summer not being transformed into a higher-oxygenated form in the aerosol phase, due either to radical chemistry, oxidants, or their residence time in the atmosphere. The chemical composition analysis also showed that the impact of biomass burning on the organic aerosols is low during summer. The higher oxidation level in spring may also explain the stronger NPF events since highly oxygenated molecules can contribute to NPF due to their low volatility (Liu et al., 2021). The more abundant fresh less-oxidized VOCs (such as isoprene) in the summer are reported to suppress NPF (Lee et al., 2016; Heinritzi et al., 2020), which agrees with our observations in this study (Fig. 5b).

Regarding the vertical distribution of the aerosol chemical composition, all species show lower concentrations at altitudes above 2000 m due to the weaker impact from surface emission and reactions. Similarly to the analysis of aerosol concentrations (Figs. 5 and S6), we plotted the aerosol chemical composition against normalized altitude in Fig. S7, which shows a relatively uniform distribution of aerosol mass in the lower BL (z/z0 < 0.75). The in-cloud scavenging and the subsequent precipitation or drizzle process can reduce aerosol mass concentrations in the upper BL. In spring, sulfate and organic concentrations are moderately elevated at higher altitudes, likely due to long-range transport events. Both ammonium and nitrate concentrations increase from the surface to ∼ 1000 m, likely resulting from the lower temperature that stabilizes the ammonium nitrate aerosols. At the same time, NOx is more abundant farther from the surface because there are few ground-level sources at the SGP, and NOx are mostly transported from other regions.

As discussed above, the aerosol properties at the SGP are largely influenced by aerosol dynamics and transport processes. Aerosol dynamics include NPF (Sect. 3.3.1), particle growth (Sect. 3.3.2), and cloud processing (Sect. 3.3.3), which alter the aerosol size distributions within the BL. Vertical and horizontal transport affects the vertical profiles of aerosols. The vertical profile of aerosols is affected by the boundary layer stability (Sect. 3.4.1), where the mixing of the residual layer and the nocturnal boundary layer results in the change in aerosol properties. At the same time, entrainment (Sect. 3.4.2) is a slow process that exchanges aerosols between the upper BL and the lower FT. During the HI-SCALE campaign, we also observed more rapid vertical transport due to deep-convection downdrafts (Sect. 3.4.3), which bring high concentrations of ultrafine particles from the FT to the surface. Long-range transport (Sect. 3.4.4) is an example of horizontal transport that can bring aged aerosol layers to specific altitudes of the atmosphere. Examples of these aerosol dynamics and transport processes are introduced and discussed as follows.

Similar to previous studies at the SGP (Marinescu et al., 2019; Wang et al., 2006; Chen et al., 2018), several NPF events were observed on board G-1 during the HI-SCALE campaign. These newly formed particles and their subsequent growth may be a significant contributor to CCN. Generally, the occurrence of NPF requires sufficient gaseous precursors, a high actinic flux for photo-oxidation reactions, a low temperature to stabilize the newly formed particles, and a low concentration of pre-existing aerosols (i.e., a low condensation and coagulation sink) (Kuang et al., 2009; Kuang et al., 2008; Cai and Jiang, 2017; Zheng et al., 2021; Corral et al., 2022). There is a weak positive correlation between the SO2 mixing ratio and the concentration of nucleation-mode aerosols (N10–20) for most of the flights, showing that the NPF events at the SGP are possibly linked to sulfuric acid (Fig. S8a). However, the total surface area concentration (Stot, which is approximately equivalent to the condensation sink Fig. S8b), and ambient temperature (Tamb, Fig. S8c) also play roles in promoting NPF. Since all the HI-SCALE campaign flight measurements were conducted during the daytime, we could not directly compare the NPF occurrence as a function of actinic flux. However, ground observations during the HI-SCALE campaign show that NPF occurrence is a strong function of solar radiation and has a clear diurnal cycle (Marinescu et al., 2019).

Although the concentration of the nucleation-mode aerosols is highest near the surface of the SGP (Fig. 5), NPF does not necessarily occur only near the surface. By measuring the vertical profiles of the aerosol concentration and size distributions, we found that, out of the 38 research flights, 9 flights observed NPF near the surface (below 600 m a.g.l., 1 flight in the spring campaign and 8 flights in the summer campaign), 7 flights observed NPF in the upper BL (above 600 m a.g.l., 4 flights in the spring campaign and 3 flights in the summer campaign), and 2 flights observed NPF in the lower FT (above the BL, 1 flight in the spring campaign and 1 flight in the summer campaign) over the SGP. Since these NPF events generate high concentrations of nucleation-mode aerosols on a short timescale (sometimes shorter than the mixing time in the BL), they can significantly influence the vertical distribution of aerosol properties. For example, Chen et al. (2018) found that it took ∼ 0.5 to 1.0 h to vertically mix newly formed particles from ∼ 400 m above the ground to the surface. This mixing time will be longer for NPF events in the upper BL, and it also depends on the stability of the BL (Marinescu et al., 2019). In this study, we used the criteria of N>3/N>10 > 1.2 (monitored by the two CPCs with different cut-off sizes) and N10–20 > 500 cm-3 (monitored by the FIMS) to identify NPF events within each altitude bin. Note that here we included the criterion regarding N10–20 to show the NPF's influence on the measured aerosol size distributions and their evolution in the vertical profiles. However, in some NPF cases, the newly formed particles may not grow sufficiently quickly to be detected in the 10 to 20 nm range. Therefore, our criteria may be relatively strict for identifying NPF events. Cases of NPF events occurring at different locations of the atmosphere are discussed below.

An example of a NPF event near the surface of the SGP on 11 September 2016 is shown in Fig. 7. A relatively stable BL with a height of ∼ 900 m is indicated by the LWC and the change in aerosol concentration (Fig. 7a). The potential temperature profile also shows a gradual temperature inversion below 800 m, limiting the transport of aerosols in the BL. Elevated levels of N3–10 and N10–20 are observed at an altitude below 400 m (Fig. 7b). In the NPF altitude range, there is a higher mixing ratio of SO2 near the surface (∼ 0.3 ppb) compared to the higher altitudes (< 0.1 ppb), suggesting that the NPF events may be initiated by sulfuric acid. However, a relatively high concentration of pre-existing accumulation-mode aerosols is observed (∼ 600 cm-3 in the BL), which led to a high total surface area concentration of the aerosols (Stot) reaching around 75 µm2cm-3. A low surface area of pre-existing aerosols is generally required for NPF in clean environments. For example, over the eastern North Atlantic, NPF does not occur when Stot is above 10 µm2cm-3 (Zheng et al., 2021). The NPF occurrence under a high Stot is likely a result of the abundance of nucleating precursors. We further examined the spatial scale of this NPF event and compared flight measurements to ground observations at the SGP (Fig. S9). As shown in Fig. S9a, this NPF event covered a horizontal scale of approximately 20 km × 20 km. The SGP observatory also detected this NPF event, where aerosols below 20 nm emerged at a local time of ∼ 12:00 and the size of the nucleation-mode aerosols continued to grow into the Aitken mode. However, the concentration of the nucleation-mode aerosols below 20 nm is lower than that of the Aitken mode. This is likely because the NPF event did not occur directly at the SGP observatory, and many of the newly formed particles grew to sizes above 20 nm before they were transported to the SGP observatory.

During the HI-SCALE campaign, several NPF events occurred in the upper BL, which could not be captured by the ground-based measurements. An example case on 14 May 2016 is shown in Fig. 8. The vertical profiles of the potential temperature, LWC (Fig. 8a), and CO mixing ratio (Fig. 8c) show a relatively well-mixed BL with a height of ∼ 1000 m. Strong NPF occurred in the upper BL between 600 and 1000 m, as shown by the high concentration of nucleation-mode aerosols (Fig. 8d) and N3–10 (Fig. 8b). The NPF occurred in the upper BL, where the pre-existing aerosols may be removed through cloud processing and precipitation (which occurred at 19:20 LT on the previous day), where a Stot of around 50 µm2cm-3 is observed. The spatial scale of this NPF event is shown in Fig. S10a, where NPF occurred in the upper BL, while the N>3/N>10 near the surface is relatively low. This is indicated by the surface measurements of aerosol size distributions, which did not observe the formation of particles below 30 nm. Instead, particles of ∼ 30 nm appeared at 16:00 LT (the flight measurement was conducted between 12:40 and 15:30 LT), suggesting that 30 nm particles were transported vertically due to the mixing of the upper BL air with the surface.

Since NPF events generally occur during the daytime, the upper BL may be the residual layer that developed above the nocturnal boundary layer. Due to the formation of the stable nocturnal boundary layer, the emission from the surface, including the gaseous precursors for NPF, will have less impact on the residual layer. As the BL gradually mixes, the gaseous precursors are transported to the upper BL, where the lower temperature and the higher actinic flux at the cloud level may promote the NPF, similar to the findings in the recent work modeling the vertical profile of NPF at the SGP (O'Donnell et al., 2023). However, it should be noted that, in this case, the SO2 concentration at the cloud level is lower than that near the surface. This may be for the following reasons: (1) the low Stot plays a dominating role in promoting NPF, and a SO2 mixing ratio of 0.5 ppb is already sufficient for initiating NPF; (2) NPF events that occurred in the residual layer above the BL and the newly formed particles were further mixed down to the upper BL (Lampilahti et al., 2021; Beck et al., 2022), where the SO2 in the upper BL is not related to the NPF events; (3) SO2 was depleted in the upper BL due to its oxidation to sulfate that promoted the NPF event in the upper BL; (4) different nucleating precursors, such as amines and extremely low volatility organic compounds (ELVOCs), were involved in this NPF event.

As shown in Fig. 8d, after the particles are formed in the upper BL, they grow to larger sizes as they are transported to the surface. These newly formed particles can grow to sizes capable of acting as CCN and contributing to cloud formation. Previous balloon-borne measurements over the SGP observed similar NPF events in the upper BL, reaching an altitude of around 400 m (Chen et al., 2018). Our measurements show that NPF events may occur at even higher altitudes in the BL and influence the measurement at the surface. Depending on the atmospheric stability, these newly formed particles may take 0.5 h to several hours to be transported to the surface (Chen et al., 2018). The study on the long-term monitoring of aerosol size distributions at the SGP site examined the correlation between the BL height and the occurrence of high concentrations of particles between 7 and 30 nm (N7–30nm), which is an indicator of NPF (Marinescu et al., 2019). The study found that weekly periods with more consistent diurnal cycles in N7–30nm were associated with deeper BLs that extended into the late afternoon and evening hours, suggesting that BL development may play an important role in NPF events observed at the surface. These results concerning the NPF and BL structure highlight the importance of directly measuring the vertical profiles of aerosols and trace gas properties for a better understanding of NPF mechanisms. Since most climate simulations use data collected from ground observation, NPF occurring in the upper atmosphere and its influence on the surface measurements need to be examined in more detail.

NPF occurs regularly in the FT due to the low surface area of pre-existing aerosols, colder temperatures, and higher actinic flux (Bianchi et al., 2016; Williamson et al., 2019). During the HI-SCALE campaign, limited by the altitude of G-1 (generally below 2500 m), observations were made in the lower FT, where NPF was not as frequent as that in the BL. The intensity of the NPF was also lower, as shown in the example case on 13 May 2016 (Fig. 9). In this event, NPF was observed between 1800 and 2200 m (Fig. 9b and d), which is above the BL with a height of around 1000 m determined based on the potential temperature profile (Fig. 9a). The NPF event coincided with a minor increase in the SO2 mixing ratio (∼ 0.3 ppb, Fig. 9c), showing that the NPF may be initiated by sulfuric acid. Stot at the NPF layer (∼ 55 µm2cm-3) was of a similar order of magnitude compared to the surface (∼ 80 µm2cm-3), suggesting that the colder temperature and the higher actinic flux may play a more significant role in this NPF event. In addition, compared to the NPF events in the BL (Figs. 7 and 8), the concentration of the newly formed particles in the lower FT is lower. The spatial scale of this NPF event can be shown in Fig. S11a, where the region of NPF is relatively localized above the BL. Surface measurement of aerosol size distributions also shows good agreement with the flight measurement (Fig. S11b). In order to have an impact on cloud formation, these newly formed aerosols in the lower FT need to be first transported into the BL via entrainment or convective downdraft (Sect. 3.4.2 and 3.4.3) and further grow to CCN-relevant sizes, which are more complex compared to the processes involving newly formed particles in the BL that can readily grow and contribute to CCN. Therefore, the climate impact of these particles formed in the lower FT over the SGP may be small. We should note that it is also possible that these newly formed particles may be generated from NPF in the residual layer developed overnight (a mechanism similar to the discussion regarding NPF in the upper BL), as the BL height reached 1700 m on the previous day before the development of the nocturnal boundary layer. If so, the NPF cases shown in Figs. 8 and 9 may both occur in the residual layer but are observed before (Fig. 9) and after (Fig. 8) entrainment into the BL. Given that these NPF events occur at the higher altitudes and that there are limited observations regarding NPF in these locations, the mechanisms associated with NPF events require more investigation.

The growth of nucleation- and Aitken-mode aerosols is commonly observed at the SGP (Hodshire et al., 2016; Marinescu et al., 2019; Wang et al., 2006). The growth of these aerosols is also shown in Fig. 8d, where recently formed particles in the upper BL of below 10 nm grow to ∼ 20 nm as they are transported to the surface of the SGP. The particles can grow through different dynamic approaches, including coagulation, condensation, and volume-controlled processes. The growth of these particles by coagulation is unlikely since the characteristic coagulation time (τ = 2/βN; β is the coagulation coefficient, and N is the total aerosol concentration) is more than 80 h (Friedlander, 2000), much longer than the mixing time in the BL. A previous study at the SGP using size-dependent growth law analysis shows that volume-controlled processes dominated the aerosol growth above 60 nm, and sulfate accounts for 50 % to 72 % of the increase in aerosol volume concentration, with the rest likely due to secondary organic species (Wang et al., 2006). The size-dependent growth rate analysis shows that particle growth below 60 nm is likely influenced by condensation (Wang et al., 2006). A more recent study by Hodshire et al. (2016) measured the gas and particle compositions during NPF events at the SGP in spring 2013 and suggested that the growth of the newly formed particles from 10 to 20 nm can be explained by either sulfuric-acid-forming salts with atmospheric bases (either ammonia or amines), the condensation of primarily nonreactive organics, or a combination of these two. The study highlights that, in addition to sulfuric acid, ELVOCs are very important for growth in the rural setting of the SGP. The aerosol measurement in a southern African grazed savannah grassland environment showed that, depending on the gaseous precursors and the size of the newly formed particles, the aerosol growth can be dominated by either sulfuric acid accompanied by ammonium or by organic compounds originating in either biogenic emissions or savannah fires (Vakkari et al., 2015). The contribution of sulfuric acid was larger during the early phases of the growth, but under clean conditions, organic compounds dominated the growth from 1.5 nm up to climatically relevant sizes.

Cloud processing converts CCN to cloud droplets, reducing the number of aerosols through scavenging and enlarging the size of aerosols after cloud droplets are evaporated. At the same time, aqueous reactions in the cloud droplets accumulate mass in the droplet, also enlarging aerosol size upon evaporation (Wang et al., 2007). An example of how cloud processing influences the vertical profiles of aerosols is shown in Fig. 10. The research flight on 1 May 2016 observed a decoupled BL structure with a lower surface layer up to 800 m and an upper layer reaching 1500 m. Inside the surface layer, the aerosol size distribution is bimodal and independent of altitude. Inside the upper layer, cumulus clouds are distributed between the altitudes of 1000 and 1500 m, and the concentration of accumulation-mode aerosols is significantly reduced because of coalescence scavenging by the cumulus clouds. The lower concentration of accumulation-mode aerosol indicates that surface measurements may overestimate the concentrations of CCN that are relevant for cumulus cloud formation when the BL is decoupled. The concentration of Aitken-mode aerosols also exhibits a decreasing trend with altitude due to interstitial coagulation with cloud droplets. We notice that the size of the accumulation-mode aerosols increased inside the upper BL (165 nm) compared to that near the surface (150 nm), as shown by the black line in Fig. 10b. This increase in the accumulation-mode aerosol size is likely due to the aqueous-phase reaction (sulfate production) in cloud droplets, which was also observed in a flight measurement campaign at the SGP site in 2003 (Wang et al., 2007). Measurements and model calculations in that study show that the aqueous-phase reaction is sufficiently fast to overcome the homogenization resulting from turbulent mixing, leading to a vertical heterogeneity in aerosol size distributions. Coalescence scavenging of aerosols in the cloud layer also contributes to particle mass accumulation and particle size increase upon evaporation of the cloud droplets. The accumulation-mode aerosol size increase is shown in the mean size distribution over summer but not in spring (Fig. 4), which could be due to the higher SO2 concentration (Fig. S5c) and stronger aqueous-phase oxidation in summer. Note that the cloud droplet measurement during the HI-SCALE campaign also shows that droplet size during summer is approximately 2 µm larger than that during spring, although the cloud frequency was less during summer (Fast et al., 2019). Therefore, it is likely that the difference in the type of clouds can also lead to different properties of the accumulation-mode aerosols.

Since the aerosols travel with the air mass, the vertical distribution of aerosols is largely dependent on atmospheric stability. The HI-SCALE campaign contains several same-day flights during summer to examine the temporal variation of aerosols over the SGP, where the influence of the BL structure on the vertical distribution of aerosols can be examined. Figure 11 shows the measurement on 4 September 2016 conducted from 09:00 to 12:00 LT (Fig. 11a) and then from 14:00 to 15:00 LT (Fig. 11b). A stable BL can be observed during the morning measurements, as indicated by the inversion of the potential temperature (Fig. 11a). Cumulus clouds can be observed at the altitudes of 1700 to 2100 m, which separates the BL and FT. At this time, the nocturnal boundary layer that developed overnight has not been affected significantly by surface warming, and therefore the surface emission is largely constrained near the ground, leading to a high concentration of aerosols at around 100 nm (∼ 5100 cm-3, Fig. 11b). The upper BL (1000 to 2000 m), which corresponds to the altitude range of the residual layer that developed overnight, has a significantly lower concentration of aerosols (∼ 1500 cm-3) and a larger accumulation-mode aerosol size compared to the surface, likely due to the lack of an emission source and cloud processing that removes and conditions pre-existing aerosols. This difference in concentration leads to a large gradient in aerosol concentration.

The flight measurement in the afternoon shows an evolved BL structure where the air below an altitude of 1000 m is relatively well mixed, as indicated by the relatively constant potential temperature (Fig. 11c). This mixing in the BL led to a more uniform distribution of the aerosol size distribution in the lower BL, and the total concentration of aerosols varied between 4000 and 5200 cm-3 within the BL at different altitudes. In addition to the accumulation-mode aerosols, nucleation-mode and Aitken-mode aerosols emerge in the lower BL, likely due to NPF events or air mixing from another location. The mode size of the accumulation-mode aerosols changed from 110 nm in the morning to 130 nm in the afternoon, potentially due to particle growth in the BL, although the influence of air-mass change cannot be ruled out. The slow growth of the accumulation-mode aerosols compared to the nucleation-mode and Aitken-mode aerosols is caused by their larger surface area, which requires a larger mass of condensing vapors to grow the particle size. Surface measurements conducted on the same day agreed in general with the flight observations (Fig. S12), showing a gradual growth in the size of the accumulation-mode aerosols during the day and the emergence of aerosols below 30 nm between 14:00 and 15:00 LT. In the upper BL, where cumulus clouds are located (> 1200 m), the aerosol size distributions measured in the afternoon are similar to those measured in the morning (Fig. 11b and d), indicating that the BL is still not completely well mixed at this time. The results show that, due to the daily evolution of the BL structure, the aerosols at the cloud level over the SGP may not be strongly related to the aerosols monitored at the surface. This time-dependent mixing of aerosols may further complicate our understanding of aerosol–cloud interactions.

Entrainment at the top of the BL is a slow process that exchanges the air between the BL and FT, affecting the vertical distribution of aerosols in the upper BL and lower FT. Generally, due to the higher concentration of aerosols in the BL at the SGP (especially in summer, as shown in Fig. 4b), entrainment brings cleaner FT air to the BL and dilutes the aerosols in the BL. Such an entrainment process can be observed in Fig. 11a and b (case also shown as an example of mixing in the BL), where the higher concentration of accumulation-mode aerosols emitted from the surface was transported across the cloud top into the lower FT. The continuous-mode sizes of the accumulation-mode aerosols through 1250 to 2500 m show that the aerosols in the upper BL and lower FT are related. The aerosols transported into the FT are less influenced by cloud scavenging and other sinks and, therefore, have a longer lifetime in the atmosphere and may influence other locations through long-range transport. During the HI-SCALE campaign, there are also events where high concentrations of FT aerosols are entrained into the cleaner BL, and an example case on 25 April 2016 is shown in Fig. 12a and b. In this event, the aerosols at around 50 nm in the lower FT above 1200 m are entrained into the cleaner upper BL, and the entrained aerosols grow further as they are transported to the surface.

The flux of aerosols transported through entrainment can be estimated based on the concentration gradient built between the lower FT and upper BL and the entrainment velocity. The entrainment flux Fx,ent can be estimated as 1Fx,ent=we(NFT-NBL), where we is the entrainment velocity, and NFT and NBL are the average concentrations of aerosols in the lower FT and upper BL, respectively (Wang et al., 2021). The value of we largely varies by location, and a recent study at the SGP reported a mean entrainment velocity of 0.74 ± 0.15 cms-1 (Albrecht et al., 2016). Taking into account the seasonal variation of the average BL height (Sect. 3.1), we further calculated the NFT and NBL based on the average aerosol concentration 200 m above and below the top of the BL. The influx of N>10 from the FT to the BL due to entrainment is estimated as -3.25 ± 0.66 × 107 and -7.56 ± 1.53 × 107 cm-2d-1 for the spring campaign and summer campaign, respectively, based on the entrainment velocity determined by Albrecht et al. (2016). Overall, entrainment serves as a sink of aerosols in the BL, which dilutes the aerosols in the BL of the SGP. Entrainment dilutes BL aerosols more effectively in the summer G-1 sampling periods than in the spring G-1 sampling periods due to the larger difference in aerosol concentrations across the top of the BL.

Mid-latitude mesoscale convective systems (MCSs) are often associated with frontal intrusions that can transport air mass horizontally. Transport within convective clouds can also exchange aerosols between the FT and BL locally and episodically, in contrast to more prevalent entrainment. Data collected from the HI-SCALE campaign show that deep-convection events affected aerosol properties at the SGP, potentially through both horizontal and vertical transport processes. Here, deep convection is characterized by a strong equivalent radar reflectivity factor (Z > 50 dBZ) and intense mean Doppler velocity values that are indicative of strong vertical air velocities (> 5 ms-1) in the convective core, as estimated by the ARM Radar Wind Profiler (RWP). A recent study on the characteristics of midlatitude MCSs suggested that the squall lines associated with mature convection over the SGP are often characterized by strong precipitation-driven downdrafts originating at altitudes higher than 6 km (Wang et al., 2019). An example MCS from 9 September to 10 September 2016 is shown in Figs. 13 and 14. On this day, a squall line passage with cloud (echo) top heights exceeding 12 km is observed between 21:30 and 22:00 LT on 9 September, followed by stratiform precipitation from 22:30 local time (LT; the time zone for all instances in the text is LT) on 9 September to 02:00 on 10 September (Fig. 13a). Figure 13b shows the mean Doppler velocity during this MCS, where a positive value indicates downward motion and a negative value indicates upward motion. These velocity estimates combine contributions from the air motions and hydrometeor fall speeds. For this example, one marker of the intensity of drafts in these core regions is evident in mean Doppler velocity values near and above the melting layer that exceed 10 ms-1, well beyond expected hydrometeor fall speeds for those altitudes and indicative of coherent downdrafts.

This deep-convection event is accompanied by significant decreases in aerosol concentrations (Fig. 13c and d) and equivalent potential temperature (θE, Fig. 13e) and changes in surface wind speed and direction (Fig. 13f). The decrease in θE is a separate indicator of the convective downdraft, which vertically transports air mass from the FT to the BL. Its decrease by 15 K takes place ∼ 15 min before precipitation, which is associated with the advancing cold pool–downdraft outflow developed with this squall line and precedes its arrival at the SGP site. Similarly, the change in the wind direction (from south to northwest and then to the north) and wind speed (from ∼ 5 to ∼ 15 ms-1) also indicates the horizontal transport of the air mass associated with the cold pool. This combined transport process potentially leads to a change in aerosol size distributions. Before the deep-convection event (10 min before the decrease in θE), the aerosols have an average total concentration of ∼ 1.4 × 104 cm-3 and a mode diameter of ∼ 91 nm. Aerosols during the deep convection (between 22:10 and 02:00) had a reduced average total concentration of ∼ 2600 cm-3 and a smaller mode diameter of ∼ 45 nm. We further calculated the concentrations of aerosols above 100 nm (N>100) and below 80 nm (N<80) during this event. There was an increase in N<80 between 21:10 and 23:00 (increasing from ∼ 600 to ∼ 1000 cm-3, Fig. 13d), while N>100 remained relatively constant during this period. This change in aerosol properties may be explained by the horizontal transport of air mass, but at the same time the vertical transport of FT air to the BL may also contribute to the change in aerosol properties. Generally, the FT is a source of ultrafine particles in the nucleation mode and Aitken mode due to the prevalent NPF events (Bianchi et al., 2016; Williamson et al., 2019). A study conducted in the Amazon rainforest showed that strong convective downdrafts and subsequent weaker downward motions in the trailing stratiform regions of MCSs are capable of transporting ultrafine aerosols from the FT to the BL (Wang et al., 2016). These vertically transported ultrafine aerosols can grow to CCN sizes and may affect subsequent cloud formation.

In this deep-convection event at the SGP site, a phenomenon similar to that of the Amazon rainforest is observed, where the Aitken mode at ∼ 45 nm emerged during the deep convection (Fig. 13c). The change in aerosol size distribution between 22:00 and 01:00 may be due to the weaker air downward transport in the stratiform regions from different altitudes, but at the same time it may be a result of condensational growth in the BL, where precipitation removed larger particles serving a condensation sink. Since no flights were conducted closely before and after the deep-convection event, we examined the research flight conducted between 11:00 and 13:00 on 10 September, around 10 h after the deep-convection event. Back-trajectory analysis shows that the air masses arriving at 500, 1000, and 1500 m over the SGP site during the flight sampling period agree well with the path of the MCS (Fig. S13) and wind direction (Fig. 13f), suggesting that the air mass sampled during the research flight represents the air mass impacted by the deep convection. Moreover, the back-trajectories indicate vertical air motion from the FT to the BL. The vertical profile of aerosol size distributions (Fig. 14) shows the impacts of the downward transport of aerosols, where the lower FT (1000 to 1800 m) and BL (below 1000 m) aerosols are dominated by the Aitken-mode aerosols and the aerosol size distributions from the surface to the lower FT are continuous, suggesting that the aerosols in the BL are transported from the FT.

Assuming the vertical profile of aerosol size distributions measured on 10 September is the same as that during the deep-convection event, we can estimate the influx of aerosols from the FT to the BL between 21:30 and 22:00 by 2Fx,con=wd(NFT-NBL), and wd is the downdraft velocity measured by the RWP. We use the mean concentration of aerosols between 1000 and 1800 m to represent the FT aerosol concentration (NFT), which is 1790 ± 280 cm-3. The concentrations of aerosols below 1000 m are used to calculate the BL aerosol concentration (NBL), which is 1540 ± 540 cm-3. wd can be estimated from the mean Doppler velocity in Fig. 13b, which is a summation of hydrometeor fall speed and downdraft velocity. In general, hydrometeor fall speed is between 6 and 8 ms-1. Considering this effect, since the mean Doppler velocity at the altitude of 1000 m (through the boundary between FT and BL) between 21:30 and 22:00 is approximately 8.4 ms-1, wd is on the order of 1 ms-1. Therefore, the influx of N>10 from the FT to the BL due to the convective downdraft is approximately 2.50 × 104 cm-2s-1 for this event. Considering that the period of strong convective downdraft lasted 0.5 h (21:30 to 22:00), the total exchange of N>10 from the FT to the BL during this period is approximately 4.50 × 107 cm-2, which is comparable to the absolute exchange of N>10 due to entrainment for a single day. Note that, here, we ignored the subsequent weaker downward motions in the trailing stratiform regions of MCS between 22:00 and 02:00, which may further contribute to the flux of aerosols. Assuming that the BL is well mixed and that there are no other sources or sinks of aerosols, we could estimate the impact of this event on the total aerosol concentration measured at the surface. This can be done by dividing the total exchange (4.50 × 107 cm-2) by the BL height (assumed to be 1 km), which yields an increase in the aerosol concentration of 450 cm-3. Note that this value is similar to that observed in the increase in N<80 between 21:10 and 23:00 (from ∼ 600 to 1000 cm-3), suggesting that the increase in the Aitken-mode aerosols during MCSs may be related to the convective downdraft.

Compared to the vertical transport of these particles by entrainment mixing, which is slower but more prevalent, the deep-convection downdraft process is more rapid and efficient in the vertical transport of aerosols, replenishing the BL with aerosols after wet deposition. However, most current climate models do not include this vertical transport as a source of CCN, as climate models use convective transport parameterization that does not reflect cloud-resolved properties at a high resolution. Moreover, there are not many direct measurements of these processes to confirm the impact of convective downdraft on aerosols. The frequency of deep-convection events varies significantly with geographic location, and thus their contributions to CCN are unpredictable. Future work is needed to investigate further this mechanism's importance under wide geographical and seasonal conditions. Also, as shown in this studied case, it is challenging to differentiate between horizontal and vertical transport of aerosols. Research flights conducted adjacent to deep-convection events or flux towers monitoring aerosol and trace gas properties may help isolate the influences of these two processes.

As shown in the vertical profiles of aerosol size distribution, the aerosol concentrations are generally higher in the BL than in the FT due to the strong emission from the surface at the SGP. However, we observed unusually high aerosol concentrations in the FT in several flight observations, which are due to the long-range transport events. Such events are more prevalent in the spring due to the dominant air mass from north to west potentially influenced by wildfire emissions which injected aerosols into the FT. An example event of long-range transport on 28 April 2016 is shown in Fig. 15. The BL height is around 1000 m, based on the vertical profile of CO, and it is not fully mixed, as indicated by the gradual increase in the potential temperature. Due to the insufficient mixing in the BL, a high concentration of Aitken-mode aerosols near the surface of the SGP is observed.

In the FT, the air is stratified, as indicated by the monotonic increase in potential temperature. We could observe elevated concentrations of aerosols at an altitude of 1600 to 2500 m (∼ 3500 cm-3) compared to the upper BL (∼ 2500 cm-3). It also appears that two layers of aerosols are observed, one at an altitude of 1600 to 2200 m and the other at 2200 to 2500 m. The lower layer is characterized by aerosols with a mode size of ∼ 40 nm, and the upper layer has a mode size of ∼ 60 nm (Fig. 15c). The two air layers also show distinct mixing ratios in CO (0.140 ppm in the lower layer and 0.144 ppm in the upper layer, Fig. 15b), as opposed to that in the boundary layer (0.135 to 0.140 ppm). We carried out back-trajectory calculations using the HYSPLIT model (Fig. 15d), and the back-trajectories indicate aerosol layers observed to have originated 10 d earlier from northeastern Asia, where fire activities were reported (Fig. S14). Note that the upper layer between 2200 and 2500 m showed a broad aerosol size distribution with larger sizes, likely caused by the particle coagulation and condensation of gas-phase species during the long-range transport (Mattis et al., 2003). Overall, both the vertical profiles of aerosol size distributions (Fig. 4a) and the back-trajectory analyses (Fig. 3) show that long-range-transported aerosols in spring play a more important role in affecting the vertical distribution of aerosols than in summer. If these transported aerosols are located near the top of the BL (unlike the aerosol layer in Fig. 15), they can be entrained into the BL and further influence the CCN activities over the SGP.