Hygroscopic Growth Factors of Sub-micrometer Atmospheric Aerosols at Four Selected Sites in Thailand

Hygroscopic growth factor (Gf) of aerosols is related to water loading on particles in sub-saturated conditions. It is an essential parameter for assessing the role of atmospheric particles in the radiative transfer and cloud formation process. Therefore, the temporal variation in Gf of size-selected atmospheric particles from the Aitken mode (D p ≤ 100 nm) and accumulation mode (D p > 100 nm) was measured using a humidified tandem differential mobility analyser (H-TDMA) and the relationship between particle size and Gf for various locations and meteorological conditions was determined. The origin of ambient particles primarily defines their properties and governs their participation in atmospheric processes. Thus, the measurements were performed in locations with different land-use types: urban, rural, coastal-industrial, and landlocked industrial areas. The data showed site-dependent patterns of temporal and spatial changes in Gf. The results indicated that the number-weighted Gf averaged over the investigated particle size range (30–250 nm) was highest in rural areas (Gf = 1.27), followed by coastal-industrial (Gf = 1.19), urban (Gf = 1.11), and landlocked industrial areas (Gf = 1.06). Particles in the urban and landlocked industrial areas had relatively low Gf values, suggesting that they originated mainly from fossil fuel combustion, in contrast to particles at other sites which can be attributed to coastal proximity


INTRODUCTION
Atmospheric aerosols can be viewed as one of meteorological parameters.Their distributions are continuous and usually multi-modal, subject to the size range being investigated (e.g., Willeke and Whitby, 1975;Salma et al., 2011;Hansson and Bhend, 2015).Size dependent they participate in different physico-chemical processes, ranging from interaction with solar radiation of dry particles to changes in size due to ambient humidity, which in turn modifies the optical properties of the particles, affecting radiative transfer and climate.Due to the particles' material and ambient conditions, atmospheric water vapour can be adsorbed creating under sub-saturated conditions water films of varying thicknesses.This level of interaction with water vapour and indicates that particles with a high hygroscopic growth factor (Gf) favourably participate in droplet nucleation and growth.If the particle composition is known, the multicomponent hygroscopic behaviour can be computed by weighing the component hygroscopicity by volume fractions in the mixture.In the absence of information on chemical particle composition, experimental data regarding the water affinity of an aerosol can be used to calculate Gf, which is the ratio of the particle size exposed to high relative humidity (RH) to the same particle size exposed to low RH (Svenningsson et al., 1997;Swietlicki et al., 1999Swietlicki et al., , 2008;;Massling et al., 2005Massling et al., , 2009;;Ye et al., 2011).The values of Gf are commonly used to classify the water affinity of aerosols into following groups: near-hydrophobic (NH), less-hygroscopic (LH), more-hygroscopic (MH), and sea-salt (SS) particles, whereas the actual boundaries for these Gf groups are not uniquely defined (e.g., Vu et al., 2015Vu et al., , 2021)).A summary of pertinent data from selected past studies and from this research is shown in Table 1.
Depending on particle size and chemical composition, a certain number of fine aerosol particles under given meteorological conditions can activate to a cloud droplet serving as cloud condensation nuclei (CCN).The effectiveness of an aerosol particle to serve as a CCN depends on its size, affinity to water represented by Gf, and available atmospheric water vapour yielding in the process of heterogeneous nucleation and condensation (Petters and Kreidenweis, 2007;Swietlicki et al., 2008;Coggon et al., 2012).This process is governed not only by the particle size and its elemental composition but also depends on the level of external and internal mixing status of the included elements (e.g., Riemer et al., 2019).Atmospheric particles are emitted from a variety of natural and anthropogenic sources, with different origins depending on human activity and land use.Consequently, the physical and chemical properties of these aerosols may significantly vary.Particles containing such ions as Na + , K + , Cl -, SO4 2-, and organic carbon are thought as good water adsorbers (Decesari et al., 2002;Svenningsson et al., 1994;Zhang et al., 2007).Rural areas are often a source of particles containing inorganic compounds and tend to be more hygroscopic, whereas aerosols resulting from crop burning and combustion of fossil fuels are usually a source of carbonaceous particles with pronounced hydrophobic properties (Gu et al., 2010;Keuken et al., 2012).Some may be so hydrophobic that water adsorption does not occur, even at a relative humidity of 100% (Sumner, 1988;Raes et al., 2000).However, the ability of these particles to attract water and act as CCN may change over time (atmospheric aging) and depend on the nature and amount of possible coating material on these particles.As these processes are quite complex and with varying environmental impacts, numerous studies on the hygroscopicity of aerosol particles have been conducted in recent years using the humidified-tandem differential mobility analyser (H-TDMA) system (e.g., Svenningson et al., 1994;Swietlicki et al., 2008;Massling et al., 2003;Rungratanaubon et al., 2017;Wu et al., 2013).
In this study, particles in the Aitken mode (Dp < 100 nm) and accumulation mode (Dp ≥ 100 nm) were of interest because of their important role in the processes of cloud formation and precipitation and their impact on aerosol optical properties with implications for climate (Meier et al., 2009;Jiang et al., 2016;Wang et al., 2018;Xu et al., 2020).Commonly, the types and proportions of particle distributions have different characteristics resulting from land use, human activities, and environmental conditions.Usually, it is expected that particles from the Aitken mode will dominate the size distribution in typical urban areas, whereas in marine aerosols containing sea salt, most particles are considered to belong to the accumulation mode (Heintzenberg et al., 2000).
In large urban areas, such as Bangkok, which is also true for other cities in Southeast Asia, fine particles mainly originate from anthropogenic sources, such as local traffic, local industrial plants, or transported biomass burning (Karagulian et al., 2015;Choomanee et al., 2020;Wang et al., 2014).If they are associated with black carbon (BC) due to combustion they would likely exhibit hydrophobic behaviour.This frequently causes particles in urban areas to have low CCN potential, with Gf approximately equal to 1 (Xu et al., 2020).In contrast, marine aerosols may have Gf >> 1 even at relatively low RH values (Zieger et al., 2013).However, marine-type aerosols exhibit also complex size and morphology-dependent hygroscopicities (Laskina et al., 2015).It is a fact of utmost significance as these particles are an important component of natural aerosols with climate implications because they act as a source of CCN (O'Dowd and de Leeuw, 2007).
A deep understanding of the climatic role of atmospheric aerosols on regional and global scales is still limited; however, it is known that they influence regional and global climate conditions (IPCC, 2013;Chakraborty et al., 2021).The presence, abundance, and type of fine atmospheric particles depend on seasonal environmental conditions, geographic location, and anthropogenic activities.Moreover, time series and diurnal patterns of Gf provide insight into the possible health and environmental impacts of atmospheric aerosols (Hsiao et al., 2016;Vu et al., 2015;Won et al., 2021).A recent large-scale study (Chen et al., 2022) regarding hygroscopic growth of ambient aerosols revealed that Gf-values might increase with an aggravated air pollution problem.There, it was pointed out towards an interesting relationship of particle mass concentrations and numberweighted mean growth factors, a connection not yet entirely resolved.This work focused on the study of the hygroscopic growth factor of fine particles in areas with varying land use and human activities at four selected sites in Thailand.Temporal variations of particle fractions were linked to their Gfs and to the precipitable water content modelled with ERA-Interim Reanalysis software (https://confluence.ecmwf.int).These aspects are important because a quantitative understanding of local-scale aerosol properties and the affinity to water is a crucial to assess the atmospheric fate and transformation of aerosols, as well as their impact on local-to-regional precipitation patterns.That also would allow the examination of the role of urbanisation, land use, and agricultural practices on the regional environment and climate.

Measurement Sites and Sequence of Data Collection
For this study, four different locations were selected: an urban area (Bangkok, S1), a rural area with a mangrove forest (Laemphakbia, S2), a coastal industrial area (Maptaphut industrial estate, S3), and a landlocked industrial area (Mae Moh, S4).The locations of the measuring stations and surrounding land use are shown in Fig. 1.
Site S1 was located on the roadside of the Chatuchak District, Bangkok (in the centre of an urban office building environment), including shopping malls and road traffic.The southern part of the site is a major road with heavy traffic, approximately 0.15 km.
Site S2 was located in the Phetchaburi Province in the western part of Thailand adjacent to the Gulf of Thailand, where the surrounding environment consists of farmland and mangrove forest at a distance from the sea coast of approximately 1.1 km in the SE direction relative to the measuring system location.
Site S3 was located within the petrochemical processing plant industrial estate area in Rayong Province in the coastal zone close to the Gulf of Thailand.The distance from the seashore in the SW direction was approximately 4.9 km.
Site S4 is located in the valley of Lampang Province, northern Thailand.This area is surrounded by high mountains, with lines of valleys oriented generally from SW to NE. Site S4 was situated about 10 km N-NE of the Mae Moh coal-fired power plant with lignite mines located about 3 km away in the WSW direction.
Aerosol measurements and a collection of meteorological data were performed in the period from 5 th April to 26 st May 2014.Relevant atmospheric parameters such as air temperature,  pressure, wind direction, wind speed, and relative humidity were determined at 10 m above the ground.The details of the location, sampling period, and environmental conditions are summarised in Table 2.All data shown in this contribution were obtained throughout 48-h per stations and averaged for three hours.
At the time of measurements, the wind speed and direction in all four study areas (S1, S2, S3, and S4) originated mainly from the SW direction.In all study areas, S1, S2, S3, and S4, the average wind speed during the measurements was 1.56, 2.39, 1.52, and 0.76 m s -1 , respectively.

The Measuring System: H-TDMA
Hygroscopic Tandem Differential Mobility Analyzer (H-TDMA) is a measuring system which has been applied in recent years to determine the water uptake on sub-micrometer particles at subsaturated conditions.The H-TDMA size measurement is based on particle number concentration with unique advantages in investigating the characteristics of fine particles; thus, it has been widely used (e.g., McMurry et al., 1996;Liu et al., 2011;Wu et al., 2016;Zhang et al., 2017;Fan et al., 2020).The operational parameters of the H-TDMA determine the actual measurement size range.In this work, the particle number concentrations and hygroscopic properties of fine aerosolswere investigated at particle sizes of 30, 50, 100, 150, 200, and 250 nm which is comparable with other studies (Massling et al., 2005;Ye et al., 2009;Massling et al., 2009).A schematic diagram of the measurement system is shown in Fig. 2. Ambient aerosols enter through the sample inlet into a diffusion dryer, which defines the internal ambient low RH and dries the sample.Subsequently, particle charge equilibrium was obtained using a soft X-ray charger (Model 3081, TSI Inc.).The first differential mobility analyser (DMA-1, TSI, Model 3081L) selected quasi-monodisperse particles with a given dry size and geometric standard deviation typically < 1.2.These particles were exposed to well-defined humidity conditions using a Nafion humidifier (Perma Pure, PD-50T Gas Dryers) before entering the second DMA-2 (TSI, Model 3081L).The possible size changes of these particles can be analysed with a subsequent DMA-2, which, in combination with the nano-water-based condensation particle counter (N-WCPC, TSI, Mod.3788), allows the number concentrations of the particles in question to be determined.The measured changes in particle size indicate the extent of hygroscopic growth, thereby elucidating the particle size-dependent Gf-values (e.g., Svenningson et al., 1994;Tan et al., 2013;Zhang et al., 2018;Wang et al., 2020).
Measurements of particle number concentrations at given sizes were performed with a time resolution of 15 min, whereas hygroscopic growth measurements of the selected dry particle diameters were repeated every 40 min.Absolute number concentration values were determined by the stability of the flowrate, which was typically of the order of ± 5%.The H-TDMA intake flow rate was fixed at 1.5 L min -1 .The diffusion dryer provided dry conditions at 27% ± 1% RH.Quasimonodisperse, dry particle sizes of 30, 50, 100, 150, 200, and 250 nm were selected by applying a given fixed voltage to DMA-1 and subsequently measured by N-WCPC.Wet conditions (relative humidity (RH) = 87 ± 1%) were achieved using a Nafion humidifier.The humidified particles entered DMA-2, where the wet particle sizes (Dp,RH87) were determined, and their number concentrations measured (Chen et al., 2018;Xu et al., 2020).The H-TDMA system was calibrated using ammonium sulphate ((NH4)2SO4), based on the extended aerosol inorganic model (E-AIM, model II) (Clegg et al., 1998;Wexler and Clegg, 2002).The experimental uncertainties were comparable to those of previous studies that used the same measurement system (Mochida et al., 2006;Aggarwal et al., 2007;Mochida et al., 2011;Boreddy et al., 2014;Jing et al., 2018).Further details, especially regarding the calibration, operation and principal measuring accuracy can be found elsewhere (Tan et al., 2013).

Hygroscopic Growth of Atmospheric Particles
The affinity of particle surface towards water and its adsorption on particles at ambient temperatures and pressures may result in particle size growth at sub-saturated conditions.The primary particle parameters derived from measurements with the H-TDMA as a function of particle size are the hygroscopic growth factor (Gf) and number fraction (Nf) of particles belonging to the observed groups of hygroscopic growth.Nf represents the ratio of particles belonging to a given hygroscopic mode to the total particle count.

Particle Number Concentrations and Growth Factors
The concentrations of ambient aerosols were measured at particle diameters of 30, 50, 100, 150, 200, and 250 nm, as shown in Fig. 3(A).There are distinctive differences between the urban area (S1) and other measuring sites.The number concentrations at S1 were approximately five-fold higher than those at S2 and S3 for all measured particle sizes.This fact was traffic-related, mirrored by the high amount of Aitken particles (Meier et al., 2009).The measuring sites S2 and S3 exhibited relatively low numbers of aerosols experiencing no traffic (S2) or limited traffic (S3).Because both sites are in the coastal region, differential heating of the atmosphere usually results in dilution of aerosols due to the horizontal air movement.The landlocked S4 measuring site was exposed to aerosols originating from the coal-fired power plant and from biomass burning in this area.This was also inferred from the different shapes shown by the particle number size distribution at this location, with a maximum in the range of 150-200 nm.The corresponding hygroscopic growth factors are shown in Fig. 3(B).At site S1 there is a domination of traffic-generated particles which is confimed by the obtained Gf ≈ 1.1.This is because particles related to vehicle exhaust typically contain carbonaceous compounds, predominatly black carbon (BC) (e.g., Watson et al., 2005;Tunguraiwan, 2009;Sakamoto et al., 2015).In particular, the site S1 was located near roads with substantial traffic, resulting in ambient particles at this site containing a high concentration of BC as was reported elsewhere (Choomanee et al., 2020).Aerosol particles at the site S4 also exhibited relatively low Gf values.This is understandable because coal combustion at this site must produce a considerable amount of hydrophobic soot and metallic trace elements.
The average Gf at a particle size of 30 nm was the highest in all study areas and decreased as the particle size increased within the Aitken mode.Gf increased again for the accumulation-mode particles (Dp = 150-250 nm), similar to that observed in earlier studies (Massling et al., 2009;Ye et al., 2011).The data at locations S2 and S3 show that the particles at these sites exhibited higher Gf values than those at the other two sites.Sites S2 and S3 are located next to the sea; hence, particles at these locations may contain ions such as Na + , Mg 2+ , and Cl -which have high water affinity (McMurry, 2000).The Gfs measured at all stations as a function of the particle number fraction for each particle size are shown in Fig. 4 and Table 3. Usually, all particles with Gf < 1.11 are considered nearly-hydrophobic (NH) (Swietlicki et al., 2008).In this study, we split the measured Gf values from the customary NH group into two groups: strongly-hydrophobic (SH, Gf ≤ 1.01) and nearly-hydrophobic (NH, Gf = 1.01-1.10),followed by the less-hygroscopic group (LH, 1.10 ≤ Gf ≤ 1.34).The objective of the insertion of Gf ≤ 1.01, representing SH particles, was to obtain somewhat finer gradation of the measured water affinity of aerosols in question.This reasoning is based on the hygroscopic growth theory developed for undersaturated conditions (Petters and Kreidenweis, 2007).From this model, the number of water monolayers on the surface of a particle can be obtained by assuming a single water molecule with a size of 0.277 nm.Consequently, for a particle size of 50 nm in diameter, representing the centre of the Aitken mode, Gf = 1.01 results in slightly less than one monolayer of water surface coverage.Gf = 1.11 for a particle size of 150 nm (which is about the centre of the accumulation mode) results in a particle surface coverage with approximately 30 monolayers of water, allowing the postulation that such a particle could likely serve as a CCN under favourable ambient conditions.Based on the hygroscopicity boundaries stated above, it was apparent that particles at each site possessed a specific hygroscopicity-related fingerprint (Fig. 4).In all four study areas, the dominant fraction were hydrophobic particles, although the relative amounts varied significantly with the particle size and location.The boundary introduced at Gf ≤ 1.01 showed that at location S2, approximately 30% of the Aitken particles were strongly hydrophobic (Fig. 4).This seemingly surprising result for a measurement site (S2) practically surrounded by the mangrove area is probably due to the farmland in the SW direction at a distance of approximately 3 km, where biomass was burned.Nevertheless, owing to the proximity of the coastal area, over 55% of all measured particles at this site exhibited Gf > 1.11.In contrast, at the S4 site with not too distant coal-burning facility and wind direction toward the measuring location the hydrophobicity of particles was very pronounced in the accumulation mode, with nearly 40% of particles with Gf ≤ 1.01.At the urban site (S1), more than half of all measured particles had Gf < 1.1, with approximately 40% of the particles having Gf > 1.11.In contrast, data from the industrial coastal measuring site (S3) displayed the dominance of Aitken particles, with nearly 70% having Gf > 1.11.This can be attributed to the fact that the prevailing winds at this site were from the SW direction, transferring maritime air towards the measuring site.

Temporal Variation of Particle Fractions and their Growth Factors
Temporal variations in the hygroscopic growth factor and changes in the relative particle number concentrations were measured over a 48-h period at all four sites and are summarised in Fig. 5.It can be seen that each site possesses a specific pattern regarding the changes in particle growth factor and size-related concentrations with daytime.Each bar represents the data averaged over 3 hrs and indicates the relative abundance of the particle number concentration for a given size.Measurements were categorised into the Aitken mode group (Dp = 30, 50, and 100 nm) and the accumulation mode group (Dp = 150,200,and 250 nm).The results confirmed the strong sitedependent influence of land use on the fine ambient particles.At the S1 location (urban) during the daytime period, the Aitken mode dominated the accumulation mode, with 60-70% of all particles belonging to this mode.The night-time period shows the dominance of the accumulation mode.Approximately 70% of the particles were from the accumulation mode, which was likely caused by the atmospheric ageing of aerosols.The observed Gf values indicate that daytime particles had lower hygroscopicity (Gf ≈ 1.07) which increased at night (Gf ≈ 1.15).Daytime ultrafine particles at this site mainly originated from heavy traffic and consequently coagulated and agglomerated with time, along with photochemical transformations (Salma et al., 2011).As mentioned before, the Aitken mode particles at this site have likely high BC content with low water adsorption potential (Choomanee et al., 2020;Xu et al., 2020).The observed particle size shift to the accumulation mode with higher average Gf-values at night indicates aerosol ageing and aggregation processes (Wang et al., 2014).At site S2 (rural), Aitken mode particles dominated during the day, peaking in the early afternoon hours.These particles originated from agricultural activities, including crop burning, as revealed by Gf < 1.1.In contrast, at night, the growth factor continuously increased, reaching Gf > 1.3 during the 1:00-3:00 am phase, likely due to the influx of marine aerosols (Notario et al., 2013) from the nearby coastal area.Consequently, this caused favourable horizontal air parcel transport.The petrochemical industrial site (S3) is located approximately 4.9 km from the coast in a southwestern direction.This site exhibited the dominance of Aitken mode particles over the entire measuring period; however, their hygroscopicity had a diurnal pattern, increasing in the morning hours up to Gf ≈ 1.3, then falling continuously from the afternoon to the minimum (Gf = 1.03) (at approximately 1:00 am) before increasing again.The observed pattern was related to the influence of terrestrial and marine winds.During the daytime, the sea breeze carries usually marine aerosols towards the land, resulting in a mixture of industrial particles and, consequently, higher average Gf values.The night-time fine particles with low Gf values were most likely emissions from industrial actions, site-related, and local traffic activities and weredominated by the particles in the Aitken mode (Zappoli et al., 1999).Gfs showed a decreasing trend, reaching a minimum value at approximately 1:00 am.Site S4 (industrial, land-locked) is located in an area surrounded by mountains in the vicinity of the Mae Moh power plant.Particles in this location mainly originated from local sources, such as coal-fired power generation, lignite mining activities, and biomass burning, resulting in the dominance of particles from the accumulation mode having pronounced stable hydrophobic properties with a number-weighted average Gf = 1.06 and practically without showing any temporal pattern.

Assessing Precipitable Water Vapour Content
The interplay between fine particles and the amount of water vapour in the atmosphere results in a phase state of atmospheric aerosols.The phase state is often difficult to quantify; however, it is important to consider cloud formation via the transition of ambient particles to CCNs.Precipitable water vapor (PW) can used to diagnose the atmospheric humidity over a specific location.It is the vertically integrated amount of water vapor in the atmosphere, which is equivalent to the depth of liquid water that would result if all the water vapor in the atmospheric column would condense.Increasing amounts of ambient precipitable water (PW) could leqad to the rising number of ambient CCN.Therefore, we linked the Gf values at given particle sizes with the modelled amount of precipitable water at all measurement sites.
Many atmospheric models are not pertinent for description of small-scale processes.Therefore, reanalysis of datasets considering atmospheric temperature and precipitable water content are important for appraising the thermodynamic profiles as they could contribute to the formation of CCNs.
To obtain a better insight into the atmospheric situation at the measurement sites the meteorological reanalysis software (ERA-Interim Reanalysis) offered by the European Centre for Medium-Range Weather Forecasts (ECMWF) was used in this study.It is a widely employed modelling product applied in numerous climatic and atmospheric trend analysis studies (Rienecker et al., 2008;Bock and Nuret, 2009;Berrisford et al., 2011;Dee et al., 2011;Zhang et al., 2018;Buntoung et al., 2020) and is available online (https://confluence.ecmwf.int).
For the appraisal of precipitable water (PW) content at the measuring sites (S1-S4) the spatial resolution of the relevant meteorological data was set to approximately 80 km (corresponding to T255 spectral resolution) on 60 levels in the vertical from the surface up to a pressure of 0.1 hPa (Berrisford et al., 2011).By definition, reanalysis data are developed by revisiting the forecasts issued by regional and global climate models.Predictions for a specific time (Ti + ∆t) obtained from Ti were compared with the measured data from various sources and then assimilated into the model output at a time (Ti + ∆t).The model parameters were then adjusted to provide the best possible fit to the measured data at Ti + ∆t, thereby providing a range of model variables that best matched the observations.Fig. 6 shows the temporal variation in the number fraction of particles from the Aitken and accumulation modes at all four sites (S1-S4).The corresponding Gf values are presented along with the PW vapour content.The PW amount is similar at sites S1, S2, and S3, typically ranging from 5.5-6 cm, except for site S4 (industrial, landlocked), which has a PW of ~3.5 cm.This fact, together Fig. 6.Averaged temporal variation of mode-related particle fractions showing the modelled associated precipitable water (PW, blue squares) amount and the hygroscopic growth factor (Gf, red dots).

Fig. 2 .
Fig. 2. Schematic diagram of the experimental setup of the H-TDMA.

Fig. 5 .
Fig. 5. Mode-depending temporal changes in the relative abundance of particle number concentration (3 h-averages, left-side y-axis) and corresponding hygroscopic growth factors (Gf) values shown on the right-side y-axis.

Table 1 .
Variability of hygroscopic growth factors of sub-micrometer atmospheric aerosols from selected studies.

Table 2 .
Details of the measuring sites, environmental data and time periods of the campaign.

Table 3 .
Hygroscopic growth factor and number fractions (shown in brackets) of atmospheric particles for selected particle sizes at the four measuring sites.