Aerosol‐Correlated Cloud Activation for Clean Conditions in the Tropical Atlantic Boundary Layer During LASIC

Aerosol measurements during the DOE ARM Layered Atlantic Smoke Interactions with Clouds (LASIC) campaign were used to quantify the differences between clean and smoky cloud condensation nuclei (CCN) budgets. Accumulation‐mode particles accounted for ∼70% of CCN at supersaturations <0.3% in clean and smoky conditions. Aitken‐mode particles contributed <20% and sea‐spray‐mode particles <10% at supersaturations <0.3%, but at supersaturations >0.3% Aitken particles contributions increased to 30%–40% of clean CCN. For clean conditions, the Hoppel minimum diameter was correlated to the accumulation‐mode number concentration, indicating aerosol‐correlated cloud activation was controlling the lower diameter cutoff for which particles serve as CCN. For smoky conditions, the contributions of Aitken particles increase and the correlation of cloud activation to accumulation‐mode particles is masked by the lower‐hygroscopicity smoke. These results provide the first multi‐month in situ quantitative constraints on the role of aerosol number size distributions in controlling cloud activation in the tropical Atlantic boundary layer.


Introduction
Aerosol-cloud interactions include both aerosol effects on clouds and cloud effects on aerosols, where the latter effects provide a signature of the cloud activation mechanisms that have affected the aerosol size distribution.Changes in aerosol physical and chemical properties have the ability to impact low-level cloud properties (McComiskey et al., 2009;Rissman et al., 2004), but the dependence of cloud changes on aerosols varies by region (Zhang & Feingold, 2023) and are not well captured in global climate models (Forster et al., 2021).A limited understanding of baseline pre-industrial or so-called "natural" aerosol properties has been identified as the leading component of aerosol-cloud interaction uncertainty in models (Carslaw et al., 2013;Lee et al., 2016).Increased observations of natural aerosol properties have been shown to better constrain the pre-industrial baseline and aerosol indirect effects (Regayre et al., 2020), even though there are few "clean" regions that remain in the polluted present-day atmosphere (Hamilton et al., 2014).Direct observations of the aerosol contributions to cloud condensation nuclei (CCN) are needed in these clean regions to constrain the representation of atmospheric processes in models.
In situ observations of atmospheric aerosol particles in the marine boundary layer have quantified the aerosol contributions to CCN over several of the Earth's oceans (Modini et al., 2015;Quinn et al., 2017;Sorooshian et al., 2019;Zheng et al., 2018).The tropical Atlantic provides "clean" conditions during November-May that contrast with very "smoky" conditions from biomass burning in June-October (Lu et al., 2018;Pennypacker et al., 2020;Zuidema et al., 2018).The combination of these diverse types of aerosol populations over the course of each year means that different months have very different clean and smoky particle size distributions.Clean aerosol measured over the open ocean and coastal regions are often characterized by three modes commonly identified by dry diameter size ranges of <100 nm known as the Aitken mode, 100-500 nm known as the accumulation mode, and >500 nm known as the sea-spray mode (Modini et al., 2015;Quinn et al., 2019;Wex et al., 2016;Zheng et al., 2018).In contrast, size distributions measured in smoky air masses in this region are effectively bimodal, hiding the sea-spray mode below the upper diameters of the high-concentration smoky accumulation mode (Che, Segal-Rozenhaimer, et al., 2022;Wu et al., 2020).Previous studies in the tropical Atlantic have focused on measurements during the June-October dry season to characterize the range of polluted and less polluted size distributions (Dobracki et al., 2023;Wu et al., 2020), with limited analysis of the "natural" or clean conditions in that region (Andreae et al., 1995;Hoppel et al., 1996).Consequently, the differences in particle size distributions between clean and smoky conditions, as well as the extent of their monthly variability, have not been well characterized in this region.
CCN in clean and smoky conditions can be sensitive to differences in size distributions (Carslaw et al., 2010;Wex et al., 2016).In smoky conditions at low supersaturations (<0.3%), smoke particles in the accumulation mode can serve as an important source of CCN (Lathem et al., 2013;Royer et al., 2023).CCN can also be formed from biogenic marine secondary sulfate and contribute to Aitken and accumulation modes during clean conditions (Russell et al., 1994;Zheng et al., 2018).Coastal and open ocean measurements suggest only a minor contribution of sea-spray aerosol to CCN (<30%), based on the low sea-spray number concentration relative to the accumulation mode (Andreae et al., 1995;Modini et al., 2015;Quinn et al., 2017;Russell et al., 2023;Sanchez et al., 2021;Zheng et al., 2018).The sparsity of observations of the contributions of size distribution modes to CCN in clean conditions means that model representations are biased toward the regions with measurements (Saliba et al., 2019;Xu et al., 2022).Therefore, characterizing the budget of CCN particles from each size mode for clean and smoky conditions in the tropical Atlantic is important to accurately represent marine aerosol and CCN in models.
The differences in clean and smoky particle chemical composition also change which aerosol particles act as CCN (Petters & Kreidenweis, 2007), although small differences in composition in some regions have negligible effects on CCN (Dusek et al., 2006;Ovadnevaite et al., 2017).Measurements and modeling have shown a primarily sizedependent control on CCN behavior in clean to moderately polluted conditions, but the sensitivity of aerosol acting as CCN is more complex in highly polluted conditions when more particles compete for the available water vapor (Kacarab et al., 2020;Pohlker et al., 2021;Reutter et al., 2009;Rissman et al., 2004).Since direct measurements of particle hygroscopicity are not available across clean and smoky conditions in the tropical Atlantic, the role of chemical composition for aerosol contributions to CCN has not been well constrained.
To attribute CCN to aerosol size modes in clean and smoky conditions, we analyzed measurements of aerosol properties in the remote tropical Atlantic boundary layer over two smoky seasons (June-October 2016 and 2017) as well as the clean conditions in between (November and May).By applying an automated size mode fitting algorithm, we distinguished the contributions of three separate modes of the aerosol size distributions to the measured CCN concentrations.The submicron composition measurements provided mode-specific estimates that constrained chemical effects on hygroscopicity.The mode fitting allowed the identification of the Hoppel minimum diameters, providing a characteristic signature of cloud processing.Together these results provide the first in situ quantification of the CCN budget and the sensitivity of cloud activation processes to the large dynamic range of clean and smoky aerosol particle concentrations in the remote tropical Atlantic.

Clean and Smoky Aerosol Size Distribution Modes
Two-hour averaged aerosol measurements from the Department of Energy Atmospheric Radiation Measurement Layered Atlantic Smoke Interactions with Clouds (LASIC) campaign on Ascension Island (8°S, 14.5°W, 360 m ASL) (Zuidema et al., 2018) provided particle concentration and composition measurements (Text S1, Table S2, Figures S1, S2 in Supporting Information S1), of which we classified ∼40% as "clean" and the remaining 60% as "smoky" using criteria previously developed (Dedrick et al., 2022a) (Text S2,Figures S3,S4  greater than 0.75 for more than 60% of the 2-hr averaged measurements, meaning that the below-cloud air often cycled through the frequent overlying clouds. Clean conditions at Ascension Island had an average aerosol number concentration of condensation nuclei >10 nm (CN 10 ) of 238 ± 84 cm 3 (Table S3 in Supporting Information S1).During the June-October smoky conditions, the average CN 10 was 394 ± 233 cm 3 in 2016 and 2485 ± 197 cm 3 in 2017.For the most polluted smoky conditions, 2-hr CN 10 concentration was greater than the 90th percentile of CN 10 observed during LASIC by >500 cm 3 , with maxima of 1,257 cm 3 in 2016 and 2,988 cm 3 in 2017.
The size distributions during November-May clean conditions were quite consistent, with most size distributions within 10%-20% of the median distribution (Figures 1f-1l).During June-October clean conditions, there was large variability in modal number concentrations and size (Figures 1a-1e).The shape of the size distributions also varied during June-October smoky conditions (Figures 1a-1e), revealing changes in both the diameter and number of the Aitken and accumulation modes, similar to size distributions measured in aged biomass burning plumes from vegetation fires (Lathem et al., 2013;Wu et al., 2020) (Table S1, Text S6 in Supporting Information S1).
The number concentration of Aitken-mode particles (N Ait ) accounted for 28 ± 20% of CN 10 on average and ranged from 2% to 94% (Table S4 in Supporting Information S1).The N Ait contribution to smoky CN 10 (20 ± 13%) was lower than to clean CN 10 (37 ± 23%), although the average N Ait were similar between November-May (82 ± 41 cm 3 ) and June-October (70 ± 35 cm 3 ).The mean Aitken-mode diameter (D g,Ait ) was 10-20 nm larger in smoky conditions at 48 ± 15 nm compared to clean at 35 ± 5 nm (Figures 1a-1e, Figure S6d in Supporting Information S1).November-May clean conditions were characterized by the number concentration of accumulation-mode aerosol (N acc ) accounting for 52 ± 13% of CN 10 with an N acc of 129 ± 59 cm 3 (Table S6 in Supporting Information S1).June-October clean period N acc were the lowest of the seasonal cycle at 89 ± 47 cm 3 and contribution to CN 10 of 41 ± 16%.N acc doubled during smoky conditions and had high standard deviations of 287 ± 200 cm 3 in 2016 and 2,312 ± 175 cm 3 in 2017, contributing 72 ± 20% of CN 10 (Figure S6b in Supporting Information S1).The mean diameter of accumulation-mode particles (D g,acc ) was consistently 30 ± 10 nm larger during smoky conditions at 179 ± 16 nm compared to clean conditions at 149 ± 18 nm (Figures 1a-1e, Figure S6e in Supporting Information S1), consistent with growth and coagulation processes in aged biomass burning plumes that increase D g,acc (Che, Segal-Rozenhaimer, et al., 2022).
Smoky N acc contributed 71 ± 10% to CN 10 in June-August and was slightly lower at 60 ± 10% in September-October (Table S4 in Supporting Information S1).The differences in mean N acc between clean and smoky conditions were statistically significant at the 95% confidence level (p = 0.03; one-way ANOVA) in June-August, while September-October clean and smoky N acc were not statistically different (p = 0.10; one-way ANOVA).The June-August increases in N acc coincided with the climatological peak in regional subsidence that may have led to more entrainment of smoke particles into the Ascension Island boundary layer compared to September-October (Gaetani et al., 2021).Mean sea-spray-mode aerosol number concentration (N ss ) was 5 ± 3 cm 3 during clean conditions of November-May and June-October and contributed only 3 ± 2% of CN 10 (Figures 1f-1l, Table S4 in Supporting Information S1), with very similar mean N ss of 5 ± 3 cm 3 in June-October and 4 ± 2 cm 3 in November-May (Figure S6c in Supporting Information S1).These low sea-spray contributions to N ss and CN 10 were in the range of previously reported remote N ss in the tropical Atlantic (Suhre et al., 1995;Wex et al., 2016).The low dynamic range of wind speeds may explain the weak dependence of N ss on local wind speed (R = 0.19, p < 0.05, two-tailed t test; Figure S7 in Supporting Information S1), suggesting upwind history was more important than local winds for N ss during LASIC (Grythe et al., 2014;Quinn et al., 2000).The diameter of the sea-spray mode (D g,ss ) showed a large range of values each month (0.3-0.7 μm, 5th-95th percentile) (Figure S6f in Supporting Information S1), but the means and standard deviations were similar between the November-May (0.50 ± 0.1 μm) and June-October (0.52 ± 0.1 μm) clean conditions (Table S4 in Supporting Information S1).
Modal number concentration contributions to N CCN were evaluated by calculating the modal hygroscopicity (κ mode ) from the estimated composition and by determining the critical diameter (D crit ) required for particles to serve as CCN in each mode (Petters & Kreidenweis, 2007) (Text S4, S5, Figures S9, S10 in Supporting Information S1).Measured submicron composition provided modal composition estimates of κ Ait = 0.29 and κ acc = 0.38 for smoky conditions and κ Ait = 0.74, κ acc = 0.76, and κ ss = 1.01 for clean conditions.The small differences between Aitken and accumulation-mode hygroscopicity meant that the differences in composition between the modes had a small effect on D crit during both clean and smoky conditions, with mean D crit differences of 4 ± 4 nm during clean conditions and 15 ± 7 nm during smoky conditions at supersaturation <0.3% (Figures 3d-3f).At 0.2% supersaturation, accumulation-mode D crit was 96 ± 3 nm during smoky conditions and 78 ± 5 nm during clean conditions.The larger D crit values retrieved during smoky conditions were because the smoky conditions had higher mass fractions of organic and black carbon than clean conditions, which reduced the hygroscopicity (Text S4, S5 in Supporting Information S1).
Modal number contributions to N CCN (N mode, Dcrit /N CCN ), calculated as the number concentration of each mode integrated from D crit to the maximum mode diameter, showed that the accumulation mode had the largest contribution to N CCN at each supersaturation for clean and smoky conditions (Figure 2b).For clean conditions, N acc, Dcrit /N CCN ranged from 75% at 0.1% supersaturation to 30% at 1% supersaturation.Smoky N acc, Dcrit /N CCN ranged from 100% at 0.1%-59% at 1.0% supersaturation during 2017.Each decrease in N acc, Dcrit /N CCN had a corresponding increase in N Ait, Dcrit /N CCN for supersaturations increasing from 0.1% to 1.0% during clean and smoky conditions (Figure 2a).The increase in N Ait, Dcrit /N CCN is most apparent at >0.3% supersaturation, where contributions begin to exceed 10%.This shift in the contribution to CCN from accumulation-mode to Aitkenmode particles in clean and smoky conditions is explained by the decrease in mean accumulation-mode D crit below D g,acc as supersaturation increased (Figure 2e).The large contribution of N acc during smoky conditions (∼70%) meant that the average N Ait, Dcrit /N CCN did not exceed 20% even at supersaturations >0.3% (Figure 2a).In contrast, during clean conditions, Aitken-mode particles contributed 35 ± 12% of N CCN at supersaturations >0.3%, with the highest average contributions during June-October, when the mean CN 10 and N acc were lower than during November-May.The increased contribution of Aitken-mode particles to the CCN budget has been reported during clean conditions in remote marine regions when cloud-base supersaturation exceeds >0.3% (Hudson et al., 2010;Russotto et al., 2013).
The sea-spray mode contributed a minor fraction (<10% on average) to the CCN budget and decreased with increasing supersaturation from 0.1% to 1.0% (Figure 2c).The average N ss,Dcrit /N CCN at <0.3% supersaturation in November-May and June-October was 6 ± 3%.This result supports similar observations of minor sea-spray contributions to CCN in marine boundary layers (Modini et al., 2015;Quinn et al., 2017Quinn et al., , 2019;;Russell et al., 2023;Sanchez et al., 2021;Wex et al., 2016;Zheng et al., 2018), in agreement with the modeled contribution of sea salt to tropical Atlantic N CCN (Che, Stier, et al., 2022).N ss,Dcrit /N CCN was largest during clean conditions of June-October with an average value of 10 ± 5% and maximum of 50% at 0.1% supersaturation.Overall, N acc provides the majority of N CCN during clean and smoky conditions with up to 30%-40% from Aitken particles at supersaturations >0.3% and up to 10% from sea-spray particles at supersaturations <0.3% and low CN 10 .

Aerosol-Correlated Effects on CCN
The distinct separation or "gap" between the Aitken and accumulation modes (Figure S10 in Supporting Information S1) is a canonical feature of submicron size distributions in clean marine regions, which has been explained as the result of cycling of particles to droplets and back to particles in non-precipitating clouds (Hoppel  et al., 1986).The gap forms because so-called "cloud processing" redistributes mass to larger particles through aqueous reactions or scavenging of interstitial particles (Feingold et al., 1996;Kaufman & Tanre, 1994;O'Dowd et al., 1999) (Table S6 in Supporting Information S1).The diameter at which the gap formed is between the particles that activated and the interstitial ones that did not activate and is named the "Hoppel minimum" (D HM ), which can be interpreted as the average critical diameter of the clouds that processed the particles.The observed D HM provides an effective lower size cutoff for the drops that activated in clouds (Gong et al., 2023), and N acc provides the concentration of aerosol that acted as CCN at the effective supersaturation (s eff ) of clouds, where here the "clouds" are interpreted as an effective average of clouds that most recently activated the aerosol measured (Blot et al., 2013;Gong et al., 2023;Kruger et al., 2014;Miyakawa et al., 2023;Sanchez et al., 2017).Thus, D HM and N acc provide observed properties that reflect the supersaturation (which sets D HM ) and the increased aerosol concentrations (which sets N acc ).To explicitly link the cloud activation processes to aerosol concentration, we can also compare N acc to the s eff , which is estimated from D HM and κ acc .D HM was 70 ± 9 nm and ranged from a minimum of 25 to a maximum of 100 nm (Figure S12 in Supporting Information S1), consistent with similar measured averages in the Atlantic marine boundary layer (Gong et al., 2023;Hoppel et al., 1986).D HM during clean conditions was largest during November-May (75 ± 8 nm) and lower during June-October (67 ± 7 nm).D HM for clean conditions in June-October was similar to the smoky conditions of 2016 at 61 ± 10 nm and 2017 at 64 ± 10 nm.The 2016 smoky conditions included the largest range of D HM, extending down to 25 nm.Smaller D HM observed during smoky conditions results in part from smoke entrained in the boundary layer, which increased the D g,acc but broadened the mode width, as observed in aircraft measurements of aged smoke plumes in the Southeast Atlantic (Che, Segal-Rozenhaimer, et al., 2022).
During clean conditions, D HM and N acc increased with a moderate positive correlation (R = 0.48) that was statistically significant (p < 0.05, two-tailed t test, Figure 3b).The correlation of low D HM with low N acc during clean conditions indicate that D crit was lower when fewer particles acted as CCN.The lower CCN concentration resulted in more water vapor remaining available to reach higher supersaturations producing smaller D HM (Hudson & Noble, 2014;Rissman et al., 2004).Similarly, the s eff is negatively correlated with N acc (Figure 3c), illustrating a possible feedback effect that decreases cloud supersaturation because of the increase in accumulation-mode particles.Since covariability of cloud activation and aerosol concentration is unlikely in clean marine conditions, the correlation is likely to be causal.Aerosol effects on cloud activation have been shown to result from particle competition for available water vapor controlling the maximum supersaturation reached (Reutter et al., 2009;Russell et al., 1999), so that the higher aerosol number concentrations activated caused the lower supersaturation as observed (Figure 3c, Figure S13 in Supporting Information S1).The reverse causation of cloud activation changing aerosol concentration would tend to activate more Aitken particles, which would increase the accumulation mode at higher supersaturations, but that is not what is observed (Figure 3c).
During smoky conditions, when both the mean CN 10 and N acc were higher than clean conditions, the correlation of D HM to N acc was very weak and negative but statistically significant (R = 0.23, p < 0.05) (Figure 3e).It is possible that the correlation between D HM and N acc in clean conditions is dampened during smoky conditions because of the contribution of N acc from entrained smoke, leading to a negligible correlation between s eff and N acc (R = 0.12) (Figure 3f).
To consider the range of aerosol conditions observed during LASIC, we separated the clean and smoky conditions into bins of N acc and found opposite correlations above 400 cm 3 compared to below (Figure 4).The correlation of D HM with N acc is weak and positive for N acc < 400 cm 3 (statistically significant p < 0.05) but becomes weak and negative (and not statistically significant p > 0.05) for N acc > 400 cm 3 .When N acc < 400 cm 3 , the aerosolcorrelated behavior is effectively driven by N acc (even though there is some non-monotonic variation in κ acc with N acc ), while for smoky conditions with N acc > 400 cm 3 , the κ acc behavior is nearly constant.This 400 cm 3 threshold divides the two regimes of aerosol correlation to cloud activation, at a level comparable to the ∼500 cm 3 particle number threshold proposed by Kacarab et al. (2020) for clean conditions over the South Atlantic.Above this value, "intermediate" (CN ∼ 500-800 cm 3 ) or "polluted" (CN > 800 cm 3 ) regimes may exist leading to the droplet activation becoming more sensitive to changes in the aerosol hygroscopicity or cloud updraft than to N acc or CCN (Hudson & Noble, 2014;Pohlker et al., 2021;Reutter et al., 2009) (Table S5 in Supporting Information S1).

Contrasting Clean and Smoky Aerosol Effects
The DOE ARM LASIC aerosol measurements illustrate the differences in aerosol size distributions between conditions of clean air and those impacted by smoke pollution from the South-central African continent.For approximately 40% of the 17-month campaign, the conditions were classified as clean and maintained low particle number concentrations (∼200 cm 3 ) with the aerosol size distribution characterized by three modes.The largest contribution to aerosol number concentrations were accumulation-mode particles that accounted for 50% of CN 10 at ∼100 cm 3 , with variable number concentrations of Aitken-mode particles making up ∼40% of CN 10 at ∼80 cm 3 and small number concentrations from sea-spray particles of only 5 cm 3 (<10% of CN 10 ).The main difference during smoky conditions was that accumulation-mode particles were 30 nm larger than during clean conditions with double to triple the concentrations.This larger accumulation-mode diameter and higher accumulation-mode number concentration during smoky conditions caused the 15%-30% increase in the fraction of particles that serve as CCN since the smoky particles were only a little less hygroscopic.N acc provides most N CCN during clean and smoky conditions (>90%) with 30%-40% contributions from Aitken particles at supersaturations >0.3% and 10% contributions from sea-spray at supersaturations <0.3% and low CN 10 .
Retrieved Hoppel minimum diameters from fitted size distributions had a moderate, positive, and statistically significant correlation with the number of accumulation-mode particles for clean conditions and aerosol concentrations <400 cm 3 .Further, the effective supersaturations that controlled cloud activation processes were correlated weakly and negatively to accumulation-mode particle concentrations for clean conditions, likely indicating an aerosol feedback on cloud activation processes.This result showed that cloud droplet formation was correlated to aerosol concentrations but effectively insensitive to composition in clean conditions (<400 cm 3 ), consistent with idealized modeling and some prior observations (Chen et al., 2016;Hudson & Noble, 2014;Kacarab et al., 2020).In contrast, concentrations >400 cm 3 showed a negative correlation between D HM and N acc , indicating that cloud supersaturations were not aerosol-correlated, possibly due to dampening of feedback effects by low-hygroscopicity smoke particles as well as other meteorological drivers (Reutter et al., 2009;Rissman et al., 2004;Zhang & Feingold, 2023).
These observation-based findings provide specific quantitative constraints for representing the perturbations of increased aerosol concentrations on clouds in the marine boundary layer, highlighting how aerosol size distributions can control not only the magnitude but also the mechanism of aerosol-cloud interactions by altering cloud supersaturation.These direct observations of the aerosol contributions to CCN over multiple months of clean aerosol conditions can improve the representation of the behavior of aerosol-cloud interactions for pre-industrial conditions, better constraining the representation of atmospheric processes in models.
in Supporting Information S1).Submicron aerosol size distribution measurements were fit with three modes comprising of Aitken, accumulation, and sea-spray modes, with the latter constrained by supermicron nephelometer scattering during clean conditions (Text S3, FigureS5in Supporting Information S1).Ascension Island cloud fraction was

Figure 1 .
Figure 1.Fitted number size distributions from 2-hr averaged SMPS, UHSAS, and nephelometer measurements for each month during November-May (black) and June-October (blue) clean conditions and smoky conditions in 2016 (red) and 2017 (magenta).Size distributions are shown as the median (solid line) and 25th and 75th percentiles (shading) of the fitted aerosol number size distributions.

Figure 2 .
Figure 2. Contributions of fitted aerosol modes to cloud condensation nuclei (CCN) as well as the observed critical diameters, including (a-c) Two-hour mean and one standard deviation (whiskers) of modal number concentration contributions to measured CCN (N mode, Dcrit /N CCN ) and (d-f) Critical diameters (D crit ) from estimated seasonal hygroscopicity of each mode.Shaded diameter ranges in panels (d-f) represent the mean ± one standard deviation of mode geometric mean diameters.Data are colored by clean conditions during January-May (black) and June-October (blue) and smoky conditions of 2017 (magenta).Note only 2017 results are shown because ACSM measurements were not available in 2016 to estimate κ.

Figure 3 .
Figure 3. Scatter plots of relationships between N acc and κ acc (a, d), D HM (b, e), and s eff (c, f) during clean and smoky conditions.Pearson correlation and statistical significance provided.

Figure 4 .
Figure 4. Correlation coefficients of D HM versus N acc (left ordinate, black circles), mean D HM (first right ordinate, maroon crosses), and the mean κ acc (second right ordinate, blue squares) at the mean N acc of each N acc bin (abscissa).N acc bin limits are 0, 100, 250, 400, 750, and 1,000 cm 3 shown by horizontal whiskers.Vertical whiskers represent one standard deviation of the mean.Linear correlation coefficients that are not statistically significant ( p < 0.05) are shown with open circles.