Cloud condensation nuclei characteristics at the Southern Great Plains site: role of particle size distribution and aerosol hygroscopicity

The activation ability of aerosols as cloud condensation nuclei (CCN) is crucial in climate and hydrological cycle studies, but their properties are not well known. We investigated the long-term measurements of atmospheric aerosol properties, CCN concentrations (NCCN) at supersaturation (SS = 0.1%–1.0%), and hygroscopicity at the Department of Energy’s Southern Great Plains (SGP) site to illustrate the dependence of NCCN on aerosol properties and transport pathways. Cluster analysis was applied to the back trajectories of air masses to investigate their respective source regions. The results showed that aged biomass burning aerosols from Central America were characterized by higher accumulation mode particles (Naccu; median value 805 cm−3) and relatively high aerosol hygroscopicity (κ; median value ∼0.25) values that result in the higher CCN activation and relatively high NCCN (median value 258–1578 cm−3 at a SS of 0.1%–1.0%). Aerosols from the Gulf of Mexico were characterized by higher Naccu (∼35%), and NCCN (230–1721 cm−3 at a SS of 0.1%–1.0%) with the lowest κ (∼0.17). In contrast, relatively high nucleation mode particles (Nnucl; ∼20%) and low NCCN (128–1553 cm−3 at a SS of 0.1%–1.0%) with higher κ (∼0.30) values were observed on the aerosols associated with a westerly wind. The results indicate particle size as the most critical factor influencing the ability of aerosols to activate, whereas the effect of chemical composition was secondary. Our CCN closure analysis suggests that chemical composition and mixing state information are more crucial at lower SS, whereas at higher SS, most particles become activated regardless of their chemical composition and size. This study affirms that soluble organic fraction information is required at higher SS for better NCCN prediction, but both the soluble organics fraction and mixing state are vital to reduce the NCCN prediction uncertainty at lower SS.


Introduction
Suspended atmospheric aerosols allow for the water vapor condensation under certain supersaturation (SS) conditions and subsequently evolve into cloud droplets by serving as cloud condensation nuclei (CCN). Changes in the amount or the CCN properties will indirectly affect the climate by perturbing the cloud development and precipitation (Rosenfeld et al 2008, Li et al 2011. In this sense, the CCN concentration is an important parameter affecting aerosol-cloud interaction. The radiative forcing induced by aerosol-cloud interaction (−0.55±0.63 Wm −2 ) is larger than the one induced by aerosol-radiation interaction (−0.27±0.50 Wm −2 ) (IPCC 2014). Moreover, the uncertainty in the radiative forcing associated with the aerosol-cloud interaction remains high and is significantly larger than that associated with the aerosol-radiation interaction (IPCC 2014). Despite the considerable efforts to better understand aerosol-cloud interaction during the last decade, the uncertainty associated with the radiative forcing due to aerosol-cloud interaction has not decreased significantly . Reducing this uncertainty is crucial for increasing our confidence in predictions of global and regional climate models (IPCC 2014). The fundamental parameter relevant for understanding the aerosol-cloud interaction is the CCN (Rosenfeld et al 2014). Hence, the significant uncertainty in aerosol-cloud interaction points to the necessity of dedicated observational and modelling efforts to improve the scientific understanding of CCN activation and accurately quantify the aerosol perturbed change in cloud microphysics (Rosenfeld et al 2014).
The ability of particles to act as CCN is mainly controlled by aerosol particle size following by chemical composition (Dusek et al 2006) and meteorological conditions (i.e., supersaturation (SS) and uplift force of air parcels; Seinfeld & Pandis 2016). CCN are particles directly emitted (as primary particles) into the atmosphere from natural and anthropogenic sources (Després et al 2012, Duan et al 2018 or aerosol particles that have undergone growth processes and possible chemical transformation in the atmosphere. These fine particles may originate from atmospheric new particle formation (NPF) events, anthropogenic combustion, or other various emission sources (Paasonen et al 2018). With the development and dissemination of techniques, measurements of ambient aerosol size distribution and chemical composition have increased in the last few decades. However, the global coverage of such data is still far from sufficient (Fan et al 2016), and the effort to systematically combine surface measurement sites is only at its earliest stages. Earlier studies suggested that measuring the aerosol hygroscopicity under subsaturated water vapor conditions has been proposed as a way to estimate the CCN activity of aerosols (Brechtel & Kreidenweis 2000, Kreidenweis et al 2005. In those studies, particle hygroscopicity was used to predict the critical supersaturation or critical diameter of particles, above which the thermodynamic equilibrium between the aerosols and the surrounding vapor collapses and the vapor condensation rate exceeds the evaporation rate. This leads to the continuous growth of the particles, which are thus solution droplets. The advancement of a single parameter κ that incorporates Raoult's law and the Kelvin effect with the given value of surface tension of water made the quantitative comparison between hygroscopicity at subsaturated condition and CCN activation more feasible (Petters & Kreidenweis, 2007). Information on CCN number concentration (N CCN ) at specific SS values is needed for the present-day climate models. To fulfill this requirement, many attempts have been made to retrieve N CCN from aerosol hygroscopicity and size measurements at various regions such as the Amazon rainforest (Gunthe et al 2009, Pöhlker et al 2016, rural continental sites (Dusek et al 2006, Cho Cheung et al 2020, large cities (Lance et al 2009, Rose et al 2010, coastal locations (Dusek et al 2003, Gong et al 2020 and subarctic (Kammermann et al 2010). Several investigations have raised attention to the effect of aerosol mixing state on N CCN (Ervens et al 2010, Wang et al 2010, Wex et al 2010. However, hygroscopicity data for ambient aerosols is still far from sufficient (Swietlicki et al 2008).
The Atmospheric Radiation Measurement (ARM) program initiated by the US Department of Energy (DOE) aims to improve the parameterization of clouds in global climate models (Stokes & Schwartz 1994). The Southern Great Plains (SGP) site under the ARM program is one of the world's largest and most extensive climate research facilities, which has over 20 years of long-term ground-based measurements of aerosol and cloud properties (Ackerman & Stokes 2003, Dong et al 2005. While the SGP site is located in a rural environment, around 40 km away from the nearest population centers, it is influenced by a mixture of anthropogenic, biogenic, and biomass burning aerosol sources along with long-range transported aerosols. Indeed, earlier studies showed a distinct seasonal variability in size and composition of aerosols, aided by favorable air masses at the SGP (Parworth et al 2015, Logan et al 2018. Therefore, it provides a unique platform to investigate the role of aerosol in CCN activity extensively. The prevailing aerosols at the ARM-SGP site typically contain organic and black carbon associated with biomass burning and inorganic aerosols composed of sulfate and nitrate species (Parworth et al 2015, Logan et al 2018. None of the prior studies at the SGP site demonstrated the differences in intrinsic hygroscopicity among those aerosol species and their roles in aerosol activation processes. Thus, this study aimed to characterize the variation in aerosol hygroscopicity and CCN activity under the influence of different air mass histories by utilizing a long, continuous record (January-December, 2019) of aerosol properties and CCN concentration measured at the SGP site. Section-1 outlines the methodology, instrumentation, and observations used in this study. Section-2 provides the present study results, which cover identifying the aerosol source regions and associated variation in aerosol characteristics and their activation processes. Section-3 summarizes the key findings of the present study and discusses areas of focus for future research.

Site description and measurements
The present study utilizes comprehensive in situ measurements conducted at the ARM-SGP extended central facility (E13) (36.605°N 97.486°W) site located in a mixed land-use area of cattle pastures and agricultural fields in Lamont, Oklahoma, U.S. (https://www.arm.gov/capabilities/observatories/sgp). The central facility employs surface-based instruments and remote sensing equipment that can provide continuous measurements of the physical and chemical properties of atmospheric constituents and local meteorological environments. The climate at the site is continental, with cold winters and hot summers. The site is impacted by various air masses originating from Central America, the Gulf of Mexico, and influence from long-range transport (figure 1 & Figure S1 (available online at stacks.iop.org/ERC/3/075002/mmedia)) with accompanying diversity in aerosol concentrations and properties. The site experiences complex and highly variable aerosol chemical compositions with time and particle size, as reflected in that of growth factor (GF) described by Mahish & Collins, (2017). The ARM-based Aerosol Observation System (AOS) has several ground-based instruments that can retrieve aerosol physical and chemical properties at the lowest atmospheric levels. Table 1 lists the routine aerosol measurements at the site that were used for the analysis presented here. All datasets used for this analysis are available for download from the ARM archive (https://adc.arm.gov/). The scanning mobility particle sizer (SMPS) measured dry particle number size distribution (PNSD) with a TSI 3936 SMPS at 5 min interval. The SMPS system consisted of an electrostatic classifier with a differential mobility analyzer (DMA) and a TSI 3772 CPC. A total of 109 bins were used to measure a diameter range from 10.6 nm to 512 nm. A sheath to aerosol flow ratio of 5:1 was used for the DMA. An additional diffusion correction for the inlet tube was applied, assuming a laminar flow (Hinds 1999). Before deployment, the sizing accuracy of the SMPS was confirmed using the National Institute of Standards and Technology (NIST)-certified polystyrene latex spheres. Total particle number concentrations for particles larger than 10 nm in diameter were obtained from the condensation particle counter (CPC). A CPC, which has a ∼10% detection efficiency for particles of 10 nm diameter, was connected to the same inlet as the SMPS. The additional details about the calibration, operation, accuracy, and uncertainty of both instruments are discussed elsewhere (Kuang 2016a(Kuang , 2016b. N CCN was measured using a continuous-flow, streamwise thermal gradient dual-column cloud condensation nuclei counter (CCNC). The sampled aerosol particles are guided within a sheath flow through this chamber and can become activated to droplets, depending on the supersaturation conditions and particles ability to act as CCN. Dual-column CCNC has two columns to measure different samples at different supersaturations (SS) simultaneously. During the experiment, one column measures CCN concentration for fixed value of supersaturation (ss=0.4% in the present study), whereas second column measures CCN concentration at 5 different supersaturation conditions (0.1%, 0.2% 0.4%, 0.8% and 1.0%). The additional description about the operation, calibration, data quality, and uncertainty is discussed elsewhere (Uin 2016).
The Aerosol Chemical Speciation Monitort (ACSM) employs thermal vaporization, electron impact ionization mass spectrometer that can measure ground-level species, such as organics (carbonaceous compounds), ions of nitrate, sulfate, ammonium, and chloride, in units of μg m −3 (Ng et al 2011). ACSM, in conjunction with the other measurement platforms and air mass trajectory analysis, provides crucial information about the air mass pathways and potential aerosol sources over regions. More detail about ACSM measurements are given elsewhere (Watson et al 2018(Watson et al , 2020. 2.2. Derivation of the particle hygroscopicity parameter (κ) Firstly, the N CCN and total particle number concentration (N CN ) data were synchronized into 5-minute averages, which corresponded to the time interval for particle size distribution data measured by SMPS. According to Köhler theory (Köhler 1936), whether or not an aerosol particle can act as a CCN is primarily controlled by its size, chemical composition, and maximum supersaturation in its vicinity. The aerosol hygroscopicity (κ) values in the present study were derived using κ-Köhler theory (Petters & Kreidenweis 2007) using CCN activity data (κ CCN ). For κ CCN >0.1, the following approximate expressions can be used, assuming the surface tension of the examined solution droplets (σ s/a ) is that of pure water: where d crit is the critical diameter above which all particles are activated into droplets for a certain supersaturation ratio, S (= supersaturation+1). M w and ρ w are the molecular weight and water density, while R The median of particle number size distribution associated with derived four airmass clusters during the study period, where shaded area shows the first and third quartiles. The total number of size distribution spectra used in each cluster throughout the study period is denoted as N in the figure. The gray dashed line separates out the nucleation, Aitken and accumulation modes. (c) frequency distribution of the geometric mean diameter of the aerosol system during all four airmass clusters along with mean value of GMD for each cluster has mentioned. and T are the ideal gas constant and the absolute temperature, respectively. Simultaneously measured N CCN and PNSDs were used to derive d crit using the approach described by Rose et al (2008) and Mei et al (2013). Additionally, a simple mixing rule on chemical volume fractions for an assumed internal mixture proposed by Petters & Kreidenweis (2007) is used to calculated κ chem : Where ε i and κ i are the volume fraction and hygroscopicity parameter, respectively, for the individual (dry) chemical components, and i is the number of components in the mixture. The ACSM measured bulk composition is used to calculate κ chem in the present analysis. The ACSM-measured aerosol components mainly consisted of organics, (NH 4 ) 2 SO 4 , and NH 4 NO 3 , Zhang et al 2016. The κ values for (NH 4 ) 2 SO 4 , and NH 4 NO 3 are 0.67 and 0.61, respectively, which are derived from previous laboratory experiments (Petters & Kreidenweis 2007). The linear function derived by Mei et al (2013) was used to estimate κ org (= 2.10 × f 44 -0.11) in our study, where f44 is the fraction of m/z=44 in total organics. The particle hygroscopicity is thus the volume average of the participating species. Volume fractions of species were derived from mass concentrations and densities of the participating species. The densities of NH 4 NO 3 and (NH 4 ) 2 SO 4 are 1720 kg m −3 and 1770 kg m −3 , respectively. The density of organics is assumed to be 1200 kg m −3 (Turpin & Lim, 2001). More detailed descriptions of the method to derive κ chem can be found elsewhere . Additionally, to remove the outliers in κ data, we defined an outlier by values larger or smaller than 1.5 times the interquartile range (IQR) as follows: Where Q1 and Q3 are the first and third quarters of kappa data and IQR is Q3 minus Q1. About 7% of the data has been removed, according to equation (4).

Results & discussions
3.1. Identification of air mass origins and potential source regions In order to assess the potential origin of air masses that affect aerosols characteristics during the study period, seven-days air mass back trajectories over study location were calculated every 4 h using the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) for the entire sampling period (Stein et al 2015). For the analysis, a model is initiated with the six-hourly Global Data Assimilation System (GDAS) archived data with a resolution of 1°in longitude and latitude. The endpoint of the trajectories was 200 m a.g.l. (above ground level) at the ARM-SGP Table 3. Parameters for each mode of the fitted lognormal distributions for the number size distributions for all four clusters shown in figure 2. N 0 represents the total number concentration within the mode (cm −3 ), the value in bracket shows the median value of particle number concentrations for respective mode and cluster, D m is the median diameters (μm) and σ g represents the geometric standard deviation. site. These back trajectories from various sources ( Figure S1) were classified into clusters using the cluster analysis technique (Dorling et al 1992). This analysis resulted in four distinct air mass back trajectory clusters (see figure 1; hereafter referred to as C1, C2, C3, and C4). Month wise occurrence frequency of each cluster is listed in table 2. Synoptic and mesoscale dynamics play essential roles in governing the movement of the air masses that transport aerosols from their sources to the SGP site (Parworth et al 2015, Logan et al 2018. The high-elevated air mass cluster C1 is associated with a westerly wind, contributing ∼27% of the total back trajectories, bringing air masses from the arid and mixed forest regions in the west of the SGP site. The intercontinental transport from Asia (mainly from eastern China) has also been seen during C1, mainly containing land-based anthropogenic aerosols . Like C1, C3 also shows long-range transport from Canada, mainly influenced by continental emissions, contributing the least (∼11%) to total back trajectories. This set of trajectories also originates from the northern plain, passed over grassland and cropland (Trishchenko et al 2004), and is influenced by weak biogenic emissions (Liu et al 2020). Both C1 and C3 mostly dominate during late Fall and Winter. Unlike C1 and C3, most back trajectories during C2 and C4 are low-elevated and transport from shorter distances, depending on the position of high-and low-pressure systems passing through the region. During C2, the air masses arriving at the SGP site primarily originated from Central America, dominated during spring, contributing ∼26% of the total back trajectories. This set of trajectories are primarily attributed to agricultural burns and wildfires in Central America (Peppler et al 2000, Wang et al 2009. During C4, air mass arriving at the SGP site originated primarily from the southern region, contributing the highest (∼32%) to the total back trajectories. During summer, these trajectories were passed by urban/industrial areas such as Oklahoma City and eastern Texas, which are influenced mainly by anthropogenic emissions. A fraction of trajectories during C4 pass over the biogenic-rich emissions regions southeast of the SGP site. This indicates that biogenic emissions likely contribute to secondary organic aerosol mass at SGP (Parworth et al 2015, Liu et al 2020. Previous studies reported that a set of back trajectories associated with air mass cluster C4, having a signature of biomass burning smoke aerosols (especially during spring) and transport of marine air masses from the Gulf of Mexico

Aerosol physical and chemical characteristics
Particles of different sizes have different formation routes, sources, and behaviors. Therefore, the simultaneous observations of aerosol particle number size distributions are co-located to various air mass clusters and further divided into the subsets to examine their dependence on the air mass origin. Figure 2(b) represents the corresponding median particle number size distributions for all four air mass clusters. The shaded area indicates the range between the 25th and 75th percentiles. To better understand the particle size modes, we fitted the particle number size distribution (PNSD) to three lognormal functions. The parameterization function for lognormal distribution is as follows (Seinfeld & Pandis 2016): N 0 is the total particle number concentration (in cm −3 ) of the ith mode, σ g is the geometric standard deviation of the ith mode distribution, and D m is the median diameter (in μm) of the ith mode. Every PNSD was individually parameterized by a trimodal distribution, where the number of I=1,2,3 represents the nucleation, Aitken, and accumulation mode, respectively. Nucleation mode (N nucl ) is that particles observed below 30 nm, Aitken mode (N aitk ) is that from 30 to 100 nm, and accumulation mode (N accu ) is that beyond 100 nm (Ueda et al 2016, Willis et al 2016. For each PNSD, we searched for an optimal fitting function until the coefficient of determination was larger than 0.97. The parameters for the number size distribution (N 0 , D m, σ g ) for each of the three size ranges are shown in table 3. PNSDs resemble those show three modes, i.e., nucleation, Aitken, and accumulation modes, which can be distinguished, as shown in figure 2(b). While similarities are evident in the size distribution's shapes and modes, several differences between the clusters can be seen in figure 2. The average relative contribution of the particle number concentration in the three modes for individual air mass clusters is shown in figure 2(a). Figure 2(b) represents the corresponding median particle number size distributions for all four air mass clusters. The shaded area indicates the range between the 25th and 75th percentiles. The frequency distribution of geometric mean diameter (GMD) for each air mass cluster, along with the mean GMD values, is shown in figure 2(c).
Variations in the contributions of the particle modes during the observational period indicated a potential difference in the particle characteristics over the region during the four distinct air mass histories. In general, the PNSD exhibited a bimodal distribution for all four air mass clusters with varying degrees of particle concentration. The median PNSDs during C1 and C3 have the first peak at below 50 nm and the second at > 100 nm. In these periods, the majority (50%) of the aerosol particles were present in the Aitken mode and least (∼14%) in the accumulation mode. Interestingly, a prominent presence of nucleation mode particles (∼30%) was also observed during both the environment even though the air mass history varies. It also reflects in the frequency distribution of GMD, spreading towards the lower size range with the mean values 52 nm (C1) and 49 nm (C3). Changes in GMD of the PNSD for a given observation site most likely reflect the different particle emission sources and aging histories. Moreover, the relatively lower occurrence frequency of C3 (table 2) contributes significantly lower to total aerosol number concentration with a median value of 705 cm −3 . On the other hand, C1 had a second maximum contribution to total aerosol number concentrations (median value of 1046 cm −3 ). In general, nucleation mode particles are produced by homogeneous and heterogeneous nucleation processes, formed during natural gas-to-particle condensation. Nucleation mode particles are transferred to the Aitken mode through coagulation of nucleation particles, condensation of vapors onto existing particles (Seinfeld & Pandis 2016), or cloud processing (Hoppel et al 1994), during which they grow into that size range.  figure 2(a). We observed that, while Aitken mode particles still dominated the particle number concentration, the accumulation mode exhibited an increase in particle number concentration during C2 (38%) and C4 (35%). The Aitken mode peak was observed at ∼70 nm, and the accumulation mode peaked at >150 nm. The consistent presence of such particles is seen as the spread of the distribution of the GMD in figure 2(c), with the mean GMD values 80 nm (C2) and 71 nm (C4). Additionally, C4 contributes highest to the total aerosol number concentration than the other air mass clusters, with a median value of 1147 cm −3 , followed by C2 with a median value of 903 cm −3 . Accumulation mode particles are mostly emitted to the atmosphere from natural sources, e.g. mineral dust, marine aerosol or bioaerosols, mainly long-range transported or more aged aerosols. These results are similar to the previous studies (Andrews et al 2011, Marinescu et al 2019. Moreover, the new particle formation (NPF) event is the primary source of secondary aerosol particles. This could significantly increase the number concentrations of nucleation mode particles, and those particles are growing into Aitken and/or accumulation mode size ranges and last for a few hours until they disappear into the atmospheric condensation and coagulation sinks (Dal Maso et al 2005). Therefore, the occurrence frequency of NPF has been investigated during each air mass environment ( Figure S2) to find their contribution to total particle concentration. In total, 205 NPF events were visually identified by following the protocol given in Kulmala et al (2012) during the entire study period. The investigation reported that NPF occurred more frequently during C1 (∼36%) may significantly contribute to nucleation mode particles. Interestingly, the occasional NPF occurrence during C3 (∼12%) suggests that nucleation mode particles are associated with long-range transportation via the westerly wind from Asia. The observation shows that NPF occurrence frequency during C2 (20%) and C4 (29%) may contribute to nucleation mode particles. According to Köhler's theory, particle activation depends on their solubility for particles of constant size, which is a function of their composition, including the number of potential solute molecules and their solubility. Therefore, we analyzed the ACSM-measured bulk chemical composition of aerosol particles, reaching the study location via different air mass histories. Figure 3 gives the mass fraction of organics, ammonium, chloride, nitrate, and sulfate as measured by ACSM, and the average mass concentration of these components is summarized in table 4 during April-December, 2019. Organic aerosols (OA) contributed the most considerable fraction to the total mass concentration during C1, C2, and C4 periods, accounting for >60% on average, whereas inorganics were greater than 50% of the total mass during C3. Sulfate was abundant in absolute mass during C4 with an average concentration of 0.7 μg m −3 . In contrast, nitrate is dominated in C3 (mean concentration 0.51 μg m −3 ) and C1 (mean concentration 0.38 μg m −3 ). The nitrate mass was lowest during C4 (mean concentration 0.23 μg m −3 ) than others due to its semi-volatile behavior pushing the equilibrium back to the gas phase with warmer temperatures. Moreover, cations and anions illustrate a good correlation ( Figure S3), but during C4, ammonium is insufficient for full neutralization of the anions, suggesting the aerosols associated with air mass cluster C4 were more acidic. In general, ammonium nitrate will not partition into the condensed phase until particulate sulfate is fully neutralized (Guo et al 2017). Thus, the more acidic behavior of aerosol might be another explanation for the lower nitrate during C4. These given values might have slight variation due to incomplete datasets of aerosol chemical composition, but the variation in chemical composition associate with air mass clusters are almost similar in comparison to previous studies (Parworth et al 2015, Mahish et al 2018, Liu et al 2020. Subsequently, we used these measurements for the computation of the chemical-based hygroscopicity parameter (κ chem ) and the further evaluation of CCN closure analysis (discussed in the following section).   in κ CCN values observed in this study. Furthermore, the median values of κ CCN against their corresponding d crit for different air mass histories is shown in figure 4. The error bars represent the interquartile range of κ CCN . Both κ CCN and d crit decrease with an increase of supersaturation, suggesting that the chemical composition was not uniform among sizes. Smaller particles tended to have lower κ CCN values corresponding to less hygroscopic species, while more hygroscopic species in larger ones. These low κ CCN values in smaller diameter suggest the presence of organic material, which has also been observed in previous studies (Parworth et al 2015, Mahish & Collins 2017, Liu et al 2020. This observed trend was consistent with the assumption that larger particles are activated first. Although the Kelvin effect may cause some decrease of κ with decreasing particle size, this effect is small, less than 5% for particles in the diameter ranged from 50-200 nm (Swietlicki et al 2008. Moreover, κ CCN values are higher during C1 and lowest during C4. Due to incomplete datasets of aerosol chemical composition, we were unable to determine the higher κ CCN values in C1 directly. However, the previous studies reported that the increased hygroscopicity is probably associated with the promotion of condensation of semi-volatile species such as ammonium nitrate and semi-volatile organic species due to a Ratio of activations ratios for all SS over the activation ratio at 0.5% SS for each air mass clusters based on particle size distribution>10 nm. At SS=0.5% (x-axis) ratio is 1. The gray dotted line is the average fit through all curves, whereby y=A * ln (SS %)+b with A=0.55±0.02 and b=1.41±0.02. combination of shallow boundary layers, enhanced emissions, and low temperature (Parworth et al 2015). Moreover, during C1, a relatively high concentration of hydrophobic particles such as dust and organic soil particles from crop harvesting contribute to be depression of κ CCN of larger particles (Mahish & Collins, 2017). Conversely, the highest organic mass concentration from biomass burning and anthropogenic emissions during C4 further reduces the κ CCN values. During C2, the biomass burning aerosol contribute a significant fraction of the organic aerosol that oxidize to a more water-soluble form, likely due to aging occurring during the transport (Parworth et al 2015, Liu et al 2020, resulting in relatively high value of hygroscopicity. Particles in this size range are mostly accumulation mode and have undergone cloud processing and aging. In comparison of C2, κ CCN is relatively low in C4, likely due to aerosol during C2 are more aged than C4 (Liu et al 2020). Interestingly, the presence of high inorganic concentration in comparison to organic during C3 still have lowest κ CCN than other air mass clusters, likely due to coating of hydrophobic particles such as carbonaceous particles associated to intercontinental transportation that suppress the overall hygroscopicity of particle. Furthermore, the κ CCN curve against size is found flatter during C2 and C4 in comparison to C1 and C3, likely due to the presence of more aged and chemically homogenous aerosols.

CCN concentration and aerosol hygroscopicity
Besides, the comparison of mean κ chem with that derived from the N CCN (κ CCN ) measurements against the aerosol particle size is shown in Figure S5. Here, κ chem is computed from the bulk chemical composition, whereas κ CCN is arithmetic based on measurements of particles with a diameter range from 10 nm to 512 nm. As illustrated in Figure S5, the κ chem calculated from equation (3) agreed well with the measured κ CCN and having a similar trend. The difference between κ chem and κ CCN is statistically insignificant at all diameters during C3 and C4, while the one during C1 and C2 became statically significant, particularly at larger particle sizes. This large difference in C1 and C2 (more frequent during winter and spring) can be due to incomplete aerosol chemical composition datasets (only available during April-December). Although the ACSM-based κ chem may overestimate particle hygroscopicity because of its sensitivity to larger particles, the mean κ chem is lower than κ CCN . One possible explanation for the lower κ chem is uncertainty in the hygroscopicity of organic aerosols because it is assumed to be a simple linear function of f44 (Mei et al 2013). Because the coefficients in the linear function κ org (= 2.10 × f 44 -0.11) are based on measurements in different regions, they may not fit the aerosols sampled during the present study. Another cause may be the particle aging/coating process, for instance, condensation of secondary aerosol on preexisting particles. The resulting particle hygroscopicity may depend more on the coating layer than on the preexisting particle composition (Ma et al 2013).

CCN activation
The fraction of aerosol particles acting as CCN at a given supersaturation is known as the CCN activation ration/ fraction (AR) and is a crucial parameter for characterizing the CCN activity (Dusek et al 2006, Andreae, 2009. Figure 5(a) shows a direct comparison of the AR spectra for the given SS segregated to various air mass histories for the period of interest, which reveals characteristics differences in the curve's shape. The observed differences among the AR spectra in figure 5(a) reflect some of the significant trends in the aerosol variability in SGP. In general, the key parameters in the CCN activation behavior are aerosol number size distribution (primary) and, in a secondary role, the chemical composition of particles (Dusek et al 2006). Thus, the cluster-wise averaged PNSD (in figure 2) and κ CCN (in figure 4) have to be considered to explain the different shapes in figure 5(a). Focusing on C2 and C4, it can be stated that with increasing SS, the d crit decreases and is shifted from the accumulation-mode towards the Aitken-mode size range. Thus, comparatively small SS levels can already activate most particles of the pronounced accumulation mode. In contrast, during C1 and C3, while the same SS levels still activate the accumulation mode remains inactivated. It means that the ratio of Aitken and accumulation-mode particles determine the activated fraction as a function of SS and thus also the steepness of the activation spectra in figure 5(a). During C2 and C4, 50% of particles activate at 0.4% SS, while in the cases of C1 and C3, 50% activation occurs at SS=0.8% and 1.0%, respectively, reveling that the ratio of Aitken and accumulation-mode particles determine the activated fraction as a function of SS and thus also the steepness of the activation spectra. While size appears as the dominant parameter in the particle activation behavior, in certain cases variability in chemical composition also matters. In figure 5(a), this can be seen between C2 and C4.
In the presence of more aged aerosol in C2, the 50% activation occurs already at SS=0.32% than to be C4 (SS=0.44%) behavior. While figure 2(a) shows that the relatively higher accumulation mode presence during C4 than C2, the observed difference in figure 5(a) can be explained by the deviations in the corresponding κ CCN size distribution ( figure 4). In other words, the elevated κ CCN during the intrusion of more aged aerosols in C2 allows the activation of particle sizes that remains inactivated at the lower κ CCN levels in the C4 due to relatively larger contribution of organic aerosols. Therefore, the difference in chemical composition can explain the decreased SS in C2 and C4 cases. Furthermore, the lower κ CCN would yield a larger d crit or higher critical SS, and thus a lower CCN and cloud droplet concentration, which will turn lead to uncertainty in evaluating the associated aerosol indirect effects on clouds and climate. Contrary, the higher κ CCN would accordingly result in a smaller d crit or lower critical SS, and consequently higher estimated CCN and cloud droplet concentration.
We calculated the activation ratio (AR) for each measured SS based on the particle number size distribution >10 nm to compare the particle activation behaviors from various air mass clusters. Further, we computed the AR ratio at each SS (AR X ) to AR at SS=0.5% (AR 0.5 ) to assess the changes in AR with respect to SS. In this study, N CCN at SS=0.5% were not measured directly therefore, the value was linearly interpolated. The result of the analysis is shown in figure 5(b). The dashed gray line represents a logarithmic fit through all four curves. A steep slope represents that the aerosol particle population activation is sensitive to small changes in SS, while a flat slope describes that a further increase in SS would not have a significant influence on the AR. The curves in figure 5(b) suggest that particles at all four clusters have distinct activation properties with changing SS, reflecting the results shown in figure5(a). Particles observed in the C2 & C4 have a high steeper slope up to SS=0.5% thereafter, the curves flatten, indicating that the aerosol particle population is relatively insensitive to higher SS and that most particles activate at SS=0.5%. The frequency distribution of GMD for C2 & C4 in figure 2(c) suggests that most particles are larger than 70 nm which will already activate at SS lower than 0.5%. The previous study associated with the lower AR at higher SS link it to the influence of nearby biomass burning and hence smaller less hygroscopic particles (Pöhlker et al 2016). Furthermore, Gunthe et al (2009) have shown in their findings that particles with a smaller electrical mobility diameter are less hygroscopic than larger particles due to the difference in composition. The mass fraction of inorganic constituents is higher in larger  particles (as seen in figure 3). Surprisingly, particle activation behavior in C2 (associated with pure continental signature) portrays that most particles active already at SS<0.4%, which is in line with the measured large particle sizes. It suggests that the mixing between water-solute organics and the natural (biogenic) sources leads to size distribution, which has hygroscopic behavior. Likewise, the mixing between biomass burning from the Gulf of Mexico and natural (marine) sources associated with C4 leads to a complex particle activation behavior and it is sensitive to SS<0.5%. Conversely, particle activation behavior in C1 & C3 is sensitive to higher SS, indicating the influence from long-range transported air pollution at the site. Regarding the C1, findings from a previous study  showed that growing international trade in China exports a significant amount of air pollution to the USA, which highly contains land-based anthropogenic particles such as black carbon, carbon monoxide, sulfate, etc . These smaller anthropogenic particles (GMD=57 nm; figure 2(c)) activate at higher SS greater than 0.5% due to less hygroscopic behavior of particles. Similarly, particles in C3, associated with transboundary transportation of land-based polluted particles, are sensitive to higher SS.

CCN Closure Analysis
Cloud resolving models require simple and efficient parameterizations of the complex microphysical basis to adequately reflect the spatiotemporal CCN cycling (Cohard et al 1998, Andreae 2009). The previous literature (Andreae 2009, Cai et al 2018, Deng et al 2013, Gunthe et al 2009, Jurányi et al 2011, Pöhlker et al 2016, Rose et al 2010 reported several different schemes for the prediction of N CCN to understand the involvement of size and chemical composition of aerosol particles in CCN activity under different environments, to improve the knowledge of aerosol-induced CCN activation further. Based on this fact, the present study used three different assumptions regarding particle diameter, chemical composition, and mixing state (internally & externally) for a CCN closure analysis.
(1) Using experimental average hygroscopicity (M1): in this scheme, the average hygroscopicity estimated using equation (1) represents the particle chemical composition and mixing state. However, it only represents the regional CCN activation due to aerosol hygroscopicity in SGP.
(2) Using bulk chemical composition and internally mixed (M2): in this scheme, the ACSM-based averaged chemical composition was assumed to be size-independent and internally mixed. All particles have an identical chemical composition in the entire size range.
(3) Using bulk chemical composition and externally mixed (M3): in this scheme, the aerosol chemical composition was assumed to be size-independent and externally mixed. There were three types of particles at each size: NH4NO3, (NH4)2SO4, and organics, and the concentrations of these three types of particles at each size were identical.
The κ-Köhler theory, as indicated in appendix-1, was used to calculate the critical diameter for the prediction of total N CCN . Due to limited aerosol chemical composition datasets, a CCN closure analysis was performed during April-December, 2019.  Figure 6 shows the normalized mean bias of closure analysis for different airmass histories under low (SS<0.5) and high (SS>0.5) supersaturation conditions.
The fitting results depicted that M3 was more accurate than M2 at all SS values in all air mass histories, likely due to the internally mixed particles with homogenous composition had already grown into larger particles or activated as cloud droplets under high moist environments due to their hygroscopic nature. Simultaneously, heterogeneous reactions may have existed in the nucleation particle formation and the remaining particles were mainly externally mixed. The best CCN closure results were achieved in C2 and C4 compared to C3 and C1 in all the schemes, particularly for higher SS, possibly due to aged aerosols, which were less affected by assumptions of chemical composition and mixing state. A previous study conducted in Mexico City reported a similar pattern to that of C3 and C1, in which the presence of primary organic aerosols and black carbon in the form of an external mixture deteriorated the closure ratio (Wang et al 2010). The low concentration and simple species in C3 made the M3 scheme better for CCN prediction. Figure 6 illustrates that the N CCN prediction accuracy using chemical methods (M2 & M3) depends on SS, typically underpredicting N CCN at low SS but always overpredicting at high SS. This reflects the mean hygroscopicity derived from bulk chemical composition, usually higher than the hygroscopicity of smaller particles but lower than larger particles. In general, the mean hygroscopicity in chemical methods is closest to the hygroscopicity of particles with a d crit of 100-130 nm, corresponding to SS of 0.1%-0.2%. However, the weak size-dependency of aerosol hygroscopicity due to particle aging neglected such an effect in C2 and C4 but expected to be prominent in C1 and C3 due to the strong size-dependency (as seen in figure 4). These results further confirm that the prediction of N CCN is less sensitive to κ at high SS than at low SS and that the impact of hygroscopicity on the N CCN prediction decreases with increasing SS. In addition, CCN closure is sensitive to solubility & surface tension of organics can lead to a poor closure ratio with a low-soluble inorganics fraction (Chang et al 2007). It is worth noting that the solubility of organics significantly affects CCN prediction in both external and internal mixing cases. The previous study in a region influenced by urban and industrial sources also suggested that knowledge of the water-soluble organic compound fraction combined with the assumptions of internally mixed aerosols can significantly improve CCN prediction (Asa-Awuku et al 2011). Notable underprediction of N CCN is reduced (10%-15%) from C2 to C4 to C1 at lower SS when assuming external mixing, in agreement with the aerosol aging, which should lead to both oxidized organic aerosol and a significant amount of internal mixing. This pattern indicates that at lower SS, the solubility of organics plays a crucial role in the CCN activity of aerosols, whereas at higher SS values, most particles become activated regardless of their chemical composition and size. In general, CCN closure error increases (15%-20%) as κ CCN decreases (from C1 to C2 to C4) at lower SS, suggesting that detailed size-resolved composition and mixing states information is crucial for accurate prediction of N CCN . The study conducted over the Northern Pacific Ocean (Schulze et al 2020) suggested that the environment with intense organic aerosol intrusion into the marine boundary layer similar to C4 are least likely to be precisely reproduced by regional models that need a detailed investigation of their particle characteristics, frequency, and the resulting impact on cloud properties is warranted. Figure 7 illustrates the comparison of different prediction methods under low supersaturation (SS<0.5) and high supersaturation (SS>0.5) conditions irrespective of any airmass cluster. It was more reliable to assume that the external mixing method was more accurate than the internal mixing method at both low and high supersaturation environments. It is worth noting that the soluble organic fraction is required at higher SS, but both the soluble organics fraction and mixing state are substantial at lower SS for CCN prediction. As illustrated in figure 7, the predictions obtained from M1 were slightly poor than the chemical methods (M2 & M3), while at higher supersaturation, the M1 method was more robust. It suggests that the chemical method was still unreliable under high supersaturation due to the critical diameters of particles at high supersaturation being too small. The CCN closure error increases as SS decreases, particularly for M1, suggesting that detailed mixing state and size-resolved compositional information or both is critical for accurate CCN prediction. As the aerosol hygroscopicity calculation used in this study relies on an assumption of internal mixing of organic and inorganic aerosol components, it is difficult to determine whether CCN closure error results from external mixing of organic and inorganic aerosols or a result of variable composition with size. In general, the particle mixing state in the CCN closure studies has been considered size-independent, which may not be correct for all cases because the mixing state varies with the time of the day, aging of aerosols, and distance from the potential source. The atmospheric processes, such as coagulation, condensation, and photochemical transformation, shifts the externally mixed aerosols to an internally mixed state. Therefore, the better CCN closure results are unexpected by considering the long lifetime of externally mixed aerosols. Aerosol aging processes such as condensation, mass transfer processes, or reactive uptake cause the production of secondary species and reduce the overall contribution of primary species that further reduces error in CCN closure. It suggests that implementing an external mixing state, particularly for aged aerosols, is still valid in global climate models. However, it does not represent actual ambient conditions. Therefore, a knowledge of size-resolved chemical composition and the mixing state assumption is required to reduce the CCN closure error.

Summary and Discussion
The present study investigates aerosol characteristics and their ability to activate as CCN using measurements obtained over the ARM-SGP site. Measurements of aerosol properties were combined with cluster analysis of the back trajectories to gain insight into aerosol characteristics and their influence on CCN under various airmass environments. Trajectory analysis demonstrated that aerosols that prevail at SGP have major source regions: The Gulf of Mexico, Mexico, and Central America, along with long-range transportation. The characteristics of these airmass clusters and their contribution to CCN activation are as follows: • Airmass transported from Central America (C2) was characterized by biomass burning aerosols, contributing highest to accumulation mode particles (∼38%) with relatively high aerosol hygroscopicity (∼0.25), likely due to the aging of aerosols. Aged aerosols with a GMD>80 nm and relatively high aerosol hygroscopicity (∼0.25) demonstrates that ∼50% of particles activate at SS<0.4%. It portrays both particle size and chemical composition together play a crucial role in the CCN activation.
• Airmass transported from the Gulf of Mexico (C4) has the signature of urban pollution, biomass burning smoke, and marine aerosols, contributing the highest to total particle concentration (∼median value 1147 cm-3) and Naccu (∼35%). The significant contribution from anthropogenic emissions reduces the aerosol hygroscopicity to the lowest (∼0.17). Interestingly, even in the poor aerosol hygroscopicity condition, the relatively lower ratio of Aitken and accumulation mode particles increase the CCN activation ratio, and ∼50% of particles activate at SS<0.5%. This demonstrates that particle size matters more than the chemical composition.
• High-elevated airmass associated with a westerly wind (C1) has a prominent presence of nucleation mode particles (∼30%) and least contributes to accumulation mode particles (∼19%) with a geometric mean diameter (GMD)∼50 nm. Although the condensation of semi-volatile species under favorable meteorology increases the aerosol hygroscopicity (0.30), the ∼50% particle activation occurs at 0.8% SS. It suggests that CCN activation mainly depends on particle size followed by chemical composition.
• The minimal contribution from elevated airmass clusters (C3) from Canada to total have the lowest contribution to total particle contribution (median value 705 cm −3 ) and relatively prominent contribution in fine mode particles (∼20%). It mostly has a signature of continental aerosols, which are hygroscopically inactive, and∼50% activation occurs at 1.0% SS.
• The better association of biomass burning aerosols with CCN and accumulation mode aerosols during C2 indicates aerosols' aging process possibly enhanced the CCN activity prior to the marine airmass (C4). It suggests that particle size is responsible first, followed by chemical composition for the CCN activity.
• The CCN closure analysis indicates that the prediction method using bulk chemical composition is more reliable at lower SS. In contrast, the experimental average hygroscopicity method provides a robust prediction of NCCN at higher SS. Additionally, the external mixing method was more accurate than the internal mixing method at both low and high supersaturation environments. The solubility of organics plays a crucial role in the CCN activity of aerosols. The present closure analysis suggests that the soluble organic fraction is required at higher SS, but both the soluble organics fraction and mixing state are substantial at lower SS for better NCCN prediction. Finally, the closure error analysis suggests that detailed size-resolved compositional information and mixing state are critical to reducing the NCCN prediction uncertainty. This information will help establish the empirical hygroscopicity relationship for the climate models.