Air mass physio-chemical characteristics over New Delhi: Impacts on aerosol hygroscopicity and CCN formation

This work presents for the first time long term and time-resolved estimates of hygroscopicty parameter (к) and CCN for Delhi, a megacity that is subjected to high local anthropogenic emissions and long-range transport of pollutants. As a part of the Delhi Aerosol Supersite (DAS) campaign, characterization of aerosol composition and size distribution were conducted from January 2017March 2018. Air masses originating from the Arabian Sea 15 (AS), Bay of Bengal (BB) and South Asia (SA) exhibited distinct characteristics of time-resolved sub-micron non-refractory PM1 (NRPM1) species, size distributions, and CCN number concentrations. SA air mass had the highest NRPM1 loading with high chloride and organics followed by BB air mass which was relatively more contaminated than AS with a higher organic fraction and nitrate. The primary sources were identified as biomassburning, thermal power plant emissions, industrial and vehicular emissions. The average hygroscopicty parameter 20 (к), calculated by the mixing rule was ~ 0.3 (varying between 0.13 and 0.77) for all the air masses (0.32±0.06 for AS, 0.31±0.06 for BB and 0.32±0.10 for SA). The diurnal variations of к were impacted by the chemical properties and thus source activities. The total, Aitken, and Accumulation mode number concentrations were higher for SA, followed by BB and AS. The mean values of estimated CCN number concentration (NCCN, 3669-28926 cm) and the activated fraction (af, 0.19-0.87) for supersaturations varying from 0.1-0.8% also showed the same trend (SA 25 > BB > AS). The size turned out to be more important than chemical composition directly, and the NCCN was governed by either the Aitken or Accumulation modes depending upon the supersaturation (SS) and critical https://doi.org/10.5194/acp-2019-1044 Preprint. Discussion started: 3 December 2019 c © Author(s) 2019. CC BY 4.0 License.

important than chemical composition directly, and the NCCN was governed by, either the Aitken or Accumulation modes depending upon the supersaturation (SS), and critical diameter (Dc). The af was governed mainly by the Geometric Mean Diameter (GMD), and such a high af (0.71±0.14 for the most dominant sub-branch of SA air mass (R1) at 0.4% SS) has not been seen anywhere in the world for a continental site. The high af was a consequence of very low Dc  nm, for SS ranging from 0.1%-0.8%) observed for Delhi. Indirectly, the 5 chemical properties also impacted CCN and af by impacting the diurnal patterns of Aitken and accumulation modes, к and Dc. The high hygroscopic nature of aerosols, high NCCN, and high af can severely impact the precipitation patterns of the Indian Monsoon in Delhi, the radiation budget, and the indirect effect, and need to be investigated to quantify the impacts.

Introduction 10
High aerosol loading can have huge climatic repercussions on precipitation including land surface feedback through rainfall, surface energy budget, and variation in latent heat atmospheric influx (Tao et al., 2012). Added CCN may nucleate a larger number of smaller droplets, which then take a longer time to coalesce into raindrops (Gunn and Phillips, 1957;Squires, 1958). A greater cloud depth indicating higher rain initiation occurs in more polluted clouds. Complete suppression of warm rain might also occur, and get aggravated due to additional CCN 15 activation above the cloud base (Braga et al., 2017). While rain suppression was observed in case of polluted urban and industrial plumes (Rosenfeld, 2000), and smoke arising from forest fires (Rosenfeld, 1999), the precipitation tendency increases due to influx of giant CCN consisting of sea salt  and salt playas (Yinon et al., 2002) due to acceleration of the auto-conversion rate . To understand the impact of pollution on indirect radiative forcing and precipitation in highly polluted regions, the information 20 on CCN number concentration is essential in Global Climate Models (GCMs) and Regional Climate Models (RCMs).
As per the fifth IPCC report (Boucher et al., 2013), the two most important factors governing CCN activation and number concentration are, size followed by composition. The aerosol chemical composition impacts the aerosol hygroscopicity which impacts the critical diameter and hence CCN activation. The hygroscopicity parameter (к) is 25 defined as the total water uptake ability of aerosols (Petters and Kredenweis, 2007). Further, the increase in relative humidity (RH) due to water uptake by aerosol can impact visibility (Lee et al., 2016;Liu et al., 2012), secondary particle formation (Ervens et al., 2011), and measurements of remote sensing (Wang et al., 2007;Brock et al., 2016), aerosol loading, and it's chemical composition (Chen et al., 2018). Hence, it is essential to determine the hygroscopicity of aerosols, especially in the polluted regions of the world, where these impacts are expected 30 to be highly significant.
Although the recent precipitation data during 1950-2011 averaged over July and August for Delhi, reveals a significant decreasing trend, there has been an increasing trend in the frequency of heavy rainfall events, and a decrease in the frequency of wet and rainy days when it rains for a shorter period (Guhathakurta et al., 2015).
These occurrences are most likely signatures of aerosols impacting the cloud nucleating properties, which calls for detailed CCN data examination. The high uncertainties associated with radiative forcings, both direct and 5 indirect, especially at the regional level, are a result of poor representation of the aerosol distributions in GCMs. This is critical for the Indian sub-continent, where the variability in aerosol microphysical properties is very high, at various spatial and temporal scales. These necessitate the measurement of long term aerosol physiochemical properties, hygroscopicity parameter, and CCN estimates. Detailed CCN and к measurements have been carried out in different parts of the world (Rissler et al., 2004;Bougiatioti et al., 2012;Engelhart et al., 2011), and in 10 India, at places like Kanpur Bhattu et al., 2014;Ram et al., 2014), Mahabaleshwar (Leena et al.,. 2016), and eastern Himalayas (Roy et al., 2017). However, no CCN measurements or estimates have been developed so far for Delhi. There is only one study that has estimated aerosol hygroscopicty based on PM2.5 mass, RH, and visibility data (Wang et al., 2019).
In this work, for the first time for Delhi, time-resolved size distribution and chemical speciation 15 measurements were carried out from 15 th January 2017-31 st March 2018 as a part of the DAS campaign (Gani et al., 2019). The time-resolved hygroscopicity parameter and CCN estimates were derived using chemical speciation data from Aerosol Chemical Speciation Monitor (ACSM), and number concentration data from Scanning Mobility Particle Sizer (SMPS), measured during January 2017 to March 2018. Data were analyzed to investigate the hypotheses: (a) the precursors to SOA formation critically impact the chemical composition over 20 Delhi; (b) the emission sources significantly impact CCN formation by governing the size distributions and chemical composition, and thus hygroscopicity; (c) physical properties impact CCN more compared to chemical properties directly; however, the physical properties are in turn, shaped by the chemical properties. sampling site at New Delhi was free from any source activity, except for a road 150 m away. The IIT campus is relatively cleaner than the rest of the city. However, it lies in the heart of the city, and the outskirts of the campus experience fresh traffic influx. The IIT campus allows only limited access to vehicles and therefore has relatively less traffic compared to the city in general. A temperature controlled room was used to carry out the measurements.

Instrumentation
Two separate and thermally insulated sampling lines (3/4 inch outer diameter stainless steel tubes) with flows of 3 LPM and 2 LPM equipped with PM1 cyclone, in line with a water trap and a Nafion membrane diffusion dryer (Magee Scientific sample streamdryer, Berkeley, CA) were used, for 1) SMPS and ACSM in conjunction with a 5 flow controller and 2) Aethalometer, respectively. A brief description of the instruments is given below. A detailed description of the instruments is given in Gani et al. (2019) and Bhandari et al. (2019).
The SMPS comprised of a differential mobility analyzer (DMA, TSI 3081), an electrostatic classifier (TSI, 3080), X-ray aerosol neutralizer (TSI, 3088), and a water-based condensation particle counter (CPC, TSI 3785). The ambient air was sampled in the size range 12 nm-560 nm, with a time difference of 135 s between two scans. The 10 sheath to aerosol flow was 4:1, and the total flow drawn by the CPC was 1 LPM. The two dominant modes (Aitken and Accumulation) are well captured within this size range. ACSM sampled the inlet air at a flow rate of 0.1 LPM with a time resolution of ~1 min. The calibration procedures and data processing are discussed in Gani et al. (2019). ACSM collected time-resolved NRPM1 (Non-refractory Particulate Matter) based on species that volatilize by 600 ○ C and included NO3 -, Cl -, SO4 2-, NH4 + , and organics. 15

Qualitative separation of organic aerosols as BBOA, HOA, and OOA
The composition data presented in this work were collected in the Delhi Aerosol Supersite (DAS) study. PMF analysis was conducted on the 15 months in the dataset. As a result, BBOA could be resolved as a separate factor only in spring 2018. This inability to resolve primary organic aerosol (POA) to separate factors, namely hydrocarbon like organic aerosol (HOA) and biomass burning organic aerosol (BBOA), was attributed to the unit 20 mass resolution of the instrument (Bhandari et al. (2019)  together with those usually employed as reference profiles for PMF analysis likely accounts for this variability.
As a part of the analysis conducted here, the mean strength at the relevant m/z(s) (m/z 57 and 60) as well as the standard deviation (S.D.) of the profiles at these m/z(s) were utilized in the analysis.
Organic aerosols were qualitatively segregated by comparing the m/z ratios of f57 and f60 with the reference profiles of BBOA (f57: 0.0337±0.00884, f60: 0.025±0.00521), HOA (f57: 0.0838±0.00378, f60: 30 0.00227±0.00214) and OOA (f57: 0.00997±0.00786, f60: 0.00571±0.00349), as reported by Ng et al. (2010). This was done by first calculating the cluster means of f57 and f60 for each cluster (Table 2). This was followed by an evaluation of residuals. Residuals represent the deviation of the cluster means from the reference profiles. The HOA, BBOA and OOA residuals ( , and ) were then calculated based on cluster means of f57 and f60 ( 57 and 60 ), with respect to the corresponding means of reference profiles ( 57 60 ), as given below for HOA in Eq 1: ), as given below for HOA in Eq 2: The residuals of the cluster means were then compared with the reference residuals as per six conditions described in detail in Text S1 and classified as HOA, BBOA, OOA or mixed. 10

Estimation of к and CCN:
The ACSM data was used to calculate к as per the following mixing rule in Eq. 3 (Petters and Kredenweis, 2007): where, є and к represent the volume fractions and individual hygroscopicity parameters of the various 15 components. The inorganics were represented by (NH4)2SO4, NH4Cl, and NH4NO3. The organic к was taken as 0.1 (Gunthe et al., 2010;Gunthe et al., 2011;Dusek et al., 2010;Rose et al., 201. кi values were taken as 0.61 for (NH4)2SO4, 1.02 for NH4Cl, and 0.67 for NH4NO3 (Sullivan et al., 2009;Petters and Kredenweis, 2007). The density values to estimate the volume fraction of the inorganic constituents were taken as 1770kg m -3 for (NH4)2SO4, 1519 kg m -3 for NH4Cl, and 1720 kg m -3 for NH4NO3 (CRC Handbook of Physics and Chemistry, 20 95th Edition). The density of organics was taken as 1500 kg m -3 (Bougiatioti et al., 2009). к for BC was taken as zero, as reported in several studies (Hong et al., 2014, Leng et al., 2014, Wu et al., 2013. It should be noted that we assumed that the к calculated from NRPM1 data of ACSM represents the bulk hygroscopicity parameter in the absence of size-resolved measurements, and is a limitation of this work. The difference due to the assumption cannot be accounted for and should be investigated in the future. However, it is 25 reported that for к >0.1, CCN closures within 20% can be achieved assuming bulk composition and internal mixing (Wang et al., 2010). Temperature, relative humidity (RH), and the calculated к were then used to calculate the critical diameter (Dc) from the multi-component к -Kohler theory . The temperature and RH data are available from the RK Puram site (~ 3-4 km aerial distance from the measurement site) maintained by the Central Pollution Control Board, India. NCCN was then estimated by integrating the size distribution 30 obtained from SMPS above Dc. The CCN estimates were obtained for supersaturations (SS) = 0.1%, 0.15%, 0.2%, 0.35%, 0.4%, 0.5%, 0.6%, 0.7%, 0.75%, 0.8%, 0.85%, and 1%. However, for the sake of detailed analysis, 0.1%, 0.4%, and 0.8% were chosen. 0.1% represents the condition when the effect of chemical composition is expected to be the highest, 0.4% represents the condition for convective clouds, and 0.8% represents a high supersaturation state when almost all aerosols tend to get activated as CCN.

Airmass characterization 5
To characterize the air masses, the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model was used (Draxler and Rolph, 2003) to determine the major pathways of aerosols reaching Delhi. The 5-day back trajectory analysis was done at the receptor site at a height of 500 m. The cluster analysis was then performed seasonally to identify the cluster mean trajectories per season. These mean trajectories were then again re-clustered to identify three main clusters based on the directions of the mean cluster trajectories as the Arabian Sea (AS) 10 branch (16.5% of total trajectories), the Bay of Bengal (BB) branch (13% of total trajectories), and the South-Asian (SA) branch (70.5% of total trajectories). The BB branch was further classified as B (54%) and B.reg (45%), where B represents the air masses reaching the sea, while B.reg represents the air masses aligned towards reaching the Bay of Bengal but did not hit the sea. The SA branch was partitioned into L (17.5%), R1 (54%), R2 (18%), and R3 (11%). L represents the local trajectories originating within India, and mainly from Delhi,Punjab and,15 Haryana. R1 represents trajectories coming from Pakistan and Afghanistan. R2 represents trajectories originating from Iran. R3 is representative of all trajectories beyond including a portion of South Africa, the Mediterranean Sea, and Turkey. The seasonal clusters for winter, spring, summer, and monsoon of the year 2017 and winter and spring of the year 2018 are shown in Figure S1. The re-clustering is shown in Figure 1. All the chemical speciation data from ACSM and size distribution data from SMPS were then categorized as per the classification discussed 20 above and used in the following discussion.

Aerosol aging estimation
The NOx, toluene and benzene data inventory for the entire campaign was taken from CPCB for RK Puram, whenever available. The data was not available for branch B. To determine the photochemical aging of aerosols, toluene and benzene concentrations were used to calculate the life (in hours), as per Nault et al. (2018) in Eq (4): 25 but due to the unavailability of data of its variation for Delhi, it was assumed constant for aging calculation. 30

Results and Discussion
The HYSPLIT analysis revealed that the north-west direction is the most dominant direction, which is representative of SA air masses, and within it, R1 is the most dominant, indicating that on an overall basis, the emissions from Pakistan and Afghanistan regions and the sources en route govern Delhi's aerosol characteristics.
However, the chemical signatures were potentially different for the various clusters, which explains the variation 5 of aerosol properties with time. Due to the different nature of sources and pathways, aerosol properties vary resulting in different hygroscopic properties and CCN forming potential. These aspects are discussed in the following sections.

Introduction to characteristics and sources of air masses
Out of the three main branches, the SA branch was the most anthropogenically contaminated, followed by BB 10 and AS branches, as indicated by the mean NRPM1 mass concentrations: 125.2±91.6, 45.9±23.3, and 32.5±20.6 g m -3 respectively ( Figure 2a). The total NRPM1 loading for the SA branch followed the sequence: NRPM1(L) < NRPM1(R3) < NRPM1(R1)< NRPM1(R2). Amongst the SA branches, L was associated with the lowest organic non-ideal burning of (NH4)2SO4 occurs (Chakraborty et al., 2015).
On comparing the BB branches, total NRPM1 for B (41.9±20.8 g m -3 ) was slightly less than that for B.reg 20 (47.3±25.4 g m -3 ) and can be attributed to the fact that B.reg air mass does not travel over water (originates adjacent to the coastline) butis subject to its influence, while B air mass travels over water and is therefore relatively cleaner. B.reg had a slightly higher norganic and organic content than that of B.
The relatively higher abundance of aerosols of BB over AS can be attributed to both the sources and the pathways of air masses. In terms of source, the Bay of Bengal is more anthropogenically impacted than the Arabian Sea, as 25 concluded by the ICARB campaign (Kalapureddy et al., 2009). Previous studies (Nair et al., 2008a(Nair et al., , 2008bMoorthy et al., 2008) reported higher aerosol number concentration (NCN), as well as Black Carbon (BC) concentration over the Bay of Bengal than that of Arabian Sea, in all size ranges within the marine boundary layer as well as the vertical column. The BB air mass traveled over the Indo-Gangetic Plains and the AS air-mass traveled across Western India and the desert region of Rajasthan. Based on previous emission estimates (Habib et al., 2006), the emission fluxes from fossil fuel dominate the aerosol burden over the Indo-Gangetic plains (IGP).
The aerosol over IGP is largely composed of Inorganic oxidized matter (IOM), including fly ash from coal-fired power plants and mineral matter from open crop waste burning (Habib et al., 2006). The AS air mass travels over 5 western India and brings pollution from both fossil fuel combustion and desert dust (Habib et al., 2006).

PM1 chemical composition of different air masses
Mass closure between SMPS size distribution data and the sum of ACSM species together with BC was achieved (R 2 = 0.83) as detailed in our parallel manuscript Gani et al. (2019). The NRPM1 species (NH4 + , Cl -, NO3 -, SO4 2-, POA, and OOA) and BC varied significantly for the different air-masses, both in terms of the mass of species 10 ( Figure 2c) and the diurnal patterns ( Figure 3), leading to different aerosol chemistry and chemical reactions. A summary of the average mass of each species for all air masses is detailed in Table 1. In brief, both POA and OOA followed by NO3 -, SO4 2and Cldominated the PM composition for the SA air mass, while OOA followed by SO4 2and OOA was dominant for BB and AS air masses. High chloride was a special feature of the SA air mass, which was not apparent in the other two branches.
Detailed ANR values are given in Table S1. ANR values revealed that while AS, B, and L branches were completely neutralized, B.reg, R1, R2, and R3 were only partly neutralized, indicating that minor components of sulphate, chloride, and nitrate may be bound to non-volatile salts such as NaNO3 or NaCl or Na2SO4 or, are associated with organics as organosulphates, organochlorides, or organonitrates, evidence for which is shown in 25 a previous DAS study (Bhandari et al., 2019).
The organic speciation revealed that AS organics were BBOA, BB (B and B.reg both) organics were threemixed 15 ,and SA organics were BBOA wherein L, R1, and R2 organics were BBOA, while R3 organics ere both HOA and BBOA.
The NOx emissions (in g m -3 ) for SA air mass (96.88±127.22) were the highest, followed by that in BB (38.30±64.79), and then in AS (36.71±68.13) air mass. BB isrepresentative of B.reg only, as NOx data for B air mass was not available. NOx data for AS was also scarce. The SA aerosols exhibited less aging (4.38±4.49 h) 20 compared to B.reg. (11.58± 3.45 h), but both were representative of aged aerosols. Aging was not calculated for AS due to very little data availability.
To determine the presence of biomass burning and traffic emissions, BCwb (wood-burning component) and BCff  Table S4. Since fossil fuel sources are active the 25 year-round, there was a strong presence of BCff ranging from 70% to 86%. However, biomass burning is only active during certain specific times for short durations and is very prominent in the north-west direction for the SA air masses. It was observed that the more distant air masses exhibited a higher BCwb contribution compared to those originating within proximity. Hence, while L was associated with 13.9% BCwb, R3 exhibited 29.2% BCwb. The BCwb contribution for A and BB air masses was 21%. It can thus be concluded that both biomass burning and traffic emissions are important sources contributing to the chemical composition of the various air masses.

3.3.1The South-Asian air mass 5
This air mass ranked highest in NH4 + concentration compared to other branches. Locally (i.e. for L), the source for NH4 + may be attributed to NH3 gas from the nearby agricultural fields of the Indian Agricultural Research Institute (IARI) (Sharma et al., 2014). For R1, R2, R3, sharp spikes in early morning hours seen in the diurnal patterns of NH4 + indicate its formation from ammonia, as a result of industrial exhausts of untreated ammonia. This is because its diurnal variation is very similar to the diurnal of NH3 emissions of industrial origin (Wang et 10 al., 2015). A very prominent feature of the SA that made it distinct from the other two air-masses was the presence Locally, plastic burning, refuse burning, and soil dispersion; (c) Biomass Burning, which is a very prominent feature of the SA branch as indicated by f57 and f60values and is also indicated by alarge number of fire counts 15 from MODIS fire-count data , dominantly in Punjab,Haryana, and a few places in Pakistan; (e) Coal-based thermal power plants in Delhi, Punjab, Haryana, and Pakistan; (f) Small and medium scale metal processing industries in Delhi, Punjab, and Haryana where HCl is used in pickling process of hot and cold rolling of steel sheets, and acid recovery from fume generation is not practiced (Jaiprakash et al., 2017).
As far as the increase in chloride with the increasing length of trajectories is concerned, the most plausible 20 explanation is biomass burning. It is pointed out in Section 3.2 earlier, that BCwb contribution increases as the air mass trajectories become distant, a feature similar to chloride emissions. There was a marked similarity in the diurnal patterns of NH4 + and Clions, such that the sequence of [Cl -] for SA sub-branches was also valid for[NH4 + ], indicating the formation of NH4Cl. NH4Cl may also be emitted directly from cement plants (Cheney et al., 1983) in Punjab. The equilibrium constant for Clis more sensitive to ambient temperature than NO3 -, as a result of 25 which, during the daytime, a large amount of NH4Cl dissociates to form NH3 and HCl, if the temperature exceeds 10 o C (Kaneyasu et al., 1998). The diurnal patterns of both [NH4 + ] and [Cl -] exhibited a sharp decrease after 08:00 in the morning which is obvious since NH4 + for SA air mass is mostly associated with Cl -. At the same time ] showed an increase between 9:00-10:00 and then started decreasing, but the rate of decrease was relatively lower than NH4Cl. This is expected as NH4NO3 is relatively more stable than NH4Cl (Kaneyasu et al., 1998). During winter, the ambient temperature drops slightly below 10 ○ C in the morning hours, and then increases sharply after 08:00 to reach a maximum at 14:00, then again starts decreasing and reaches around 10 ○ C at mid-night (Gani et al., 2019). Since during winters, the air mass comes majorly from the north-west direction of SA air mass, it is evident that the formation and dissociation of NH4Cl were governed by the ambient temperature at the receptor 5 site.
The reduction in chloride concentrations during mid-day can also be attributed to sulphate substitution mechanism, when sulphate formation enhances, and was also marked by the ratio R[SO4 2 -] ratiogreater than two for L, R1, R2, and R3. This is valid especially for L, wherein [SO4 2 -] increased significantly. However, the diurnal patterns of [SO4 2-] and [NH4 + ] did not resemble each other, indicating that (NH4)2SO4 may be present in small 10 amounts but primarily SO4 2is associated elsewhere. Hence, SO4 2in combined form can be expressed in two ways: (a) small amounts of (NH4)2SO4 and (b) majorly in combination with K + . Thus, the sharp jump in  in locally originated air masses in the late morning and afternoon hours may be attributed to SO2 emissions. SO2 emissions in India are primarily attributed to power generation plants that make use of coal combustion as the chief source (Reddy and Venkataraman, 2002), followed by transportation. Such coal-based power plants are 15 located in the IGP, with a high concentration in Haryana. SO2 dissolves readily in water and can form sulphite ion, which in the presence of ozone can form sulphate ion (Erickson et al., 1976). H2SO4 formed from the reaction of SO2 and ozone, can react with NH3 to form NH4HSO4, which combines with NH3 again to form (NH4)2SO4 (Stelson et al., 1982;Seinfeld 1986). Since the ozone spiked during the daytime (10:00-16:00), more sulphate formation was seen when ozone was maximum. The diurnal variation for ozone is explained in Gaur et al. (2014) 20 for Kanpur where a spike in ozone levels was seen during 10:00-16:00. A peak in sulphate concentration was also previously observed for the foggy period in Kanpur at 10:00, due to the resumption of photochemical activity after fog dissipation (Chakraborty et al., 2015). SO4 2in SA branches may also combine with K + , as K + is produced in biomass burning. Evidence for the presence of K + along with SO4 2in accumulation mode is reported in Fuzzi et al. (2007). SO4 2and NO3emissions may also be associated with secondary formation for R1, R2, and R3, due 25 to industrial emissions from metal product manufacturing industries in Punjab and Haryana, large-scale manufacturing of porcelain insulators, switchgear in Islamabad (Jaiprakash et al., 2017), and steel rolling mills in Iran, Iraq, and Turkey and Punjab.
The NO3levels wer very high for SA. The high NO3in SA can be explained due to the non-ideal burning of NH4NO3 and NOx emissions due to mining equipment in the coal mines in Pakistan, leading to high NO3 -30 formation (Chakraborty et al., 2015). It is mentioned in Section 3.2 that NOx emissions in the SA branch are very high. NH4 + neutralizes NO3simultaneously with Cl -, however, the correlation of [NO3 -] with [ns-NH4 + ] is moderate for SA. Therefore, it is expected that NO3might be associated with K + and Na + , since biomass burning results in K + and Na + emissions (Fuzzi et al., 2007). K + and Na + exhibit a high affinity for nitrate during neutralization reactions, thus aiding in particulate nitrate formation (Bi et al., 2011). This is in addition to other 5 nitrate sources that are discussed above along with SO4 2sources for SA.
The BC concentrations were highest for SA, followed by that in AS, and then in BB. The biomass burning in SA air mass could be a major source of BC besides power plants, cement plants, local traffic, and industries. The POA emissions for SA followed the order BC(L) < BC(R3) < BC(R1)< BC(R2). The spikes during early morning hours and night time of the POA diurnal profile may be attributed to lower boundary layer heights during the two periods. 10 BC and POA were well correlated (r 2 = 0.77) for R3, indicating that they are coming from primary emissions. The diurnal profiles for all branches were similar, which shows a decline as the day proceeds, followed by an increase as the night proceeds. OOA was present significantly in all the three branches, but is maximum for SA. Its diurnal variation resembled that of NO3 -(r 2 = 0.78), indicative of its semi-volatile nature. OOA and NO3correlations were strongest for L (r 2 = 0.91), followed by R2 (r 2 = 0.85), R3 (r 2 = 0.81), and R1 (r 2 = 0.75). 15

The Bay of Bengal air mass
[Cl -] was lower in this air mass compared to that in SA. For both of the BB branches, fossil fuel combustion was the most likely source of Clas fossil fuel emissions dominate IGP. Clfor BB was not correlated with ns-NH4 + , and may probably be present in the form of methyl chloride, methylene chloride, carbon tetrachloride and tetrachloroethene (Ho et al., 2004). 20 Fossil fuel combustion from coal plants along IGP can be explained as a common source for both SO4 2and NH4 + ions, leading to (NH4)2SO4 formation (r 2 NH4+/SO42-= 0.78). This was also seen in the diurnal profiles of [SO4 2-] and [NH4 + ], both of which exhibited a sharp spike in the early morning hours between 10:00 and 16:00. SO2 emissions, as explained for SA via photochemical oxidation by O3 in combination with NH3, can lead to (NH4)2SO4 formation. For B.reg branch, the diurnal profiles of NH4 + and SO4 2exhibited double spikes (M-pattern, which is 25 a typical feature of NOx profile for traffic emissions) during heavy traffic hours (06:00-08:00 and around 16:00-19:00), indicating (NH4)2SO4 formation. NO from automobile exhausts can also form NH3 in three ways catalytic convertors (Gandhi and Shelf, 1991), which in combination with SO2 formed due to pyrolysis of sulphide fuels and subsequent oxidation, can lead to (NH4)2SO4 formation.
The correlation of [NO3 -] with [ns-NH4 + ] was appreciably high for both B and B.reg. For both these branches, the fossil fuel combustion resulting in NO3emissions in combination with NH4 + (Rajput et al., 2015;Pan et al., 2019) can lead to NH4NO3 formation. This is also evident from the diurnal profile of NO3that shows a very similar pattern to NH4 + , and is expected to be in the form of NH4NO3. The diurnal profiles of NO3show a decline as the temperature increases during the day and NO3converts back to HNO3, due to its semi-volatile nature. 5 The BC concentration in BB air masses was considerably lower than in the SA air masses. The missing points in the diurnal variability of BC for B region are on account of the unavailability of Aethalometer data. The BC in IGP can be emitted from industries (as for B), traffic (as for B.reg) and natural sources (Derwent et al., 2001). For B and B.reg, the B.reg branch was subjected to a longer duration of anthropogenic influence compared to B which also spent considerable time on the water, hence after the early morning hours, when the various fresh emissions 10 start increasing, the magnitude of POA for B.reg exceeds B. However, POA for BB was very low compared to SA. For B.reg, the spike in OOA during daytime hours was very similar to that of odd oxygen (O3+NO2) for Delhi.
The OX profile for Delhi is given in Tiwari et al. (2015), indicating its production by local photochemistry, despite the increase in boundary layer height in the afternoon.

The Arabian Sea air mass 15
Chloride amounts were relatively very low for AS compared to that in SA. Biomass burning as indicated by f57 and f60 measurements seems to be the main Clcontributor to AS and might be associated with K + which is also emitted along with it.
Similar to the cases of the L branch in the case of SA air mass and the BB air mass, the power stations in Gujarat and Rajasthan lead to SO2 emissions. Since the power plants in this region over which the AS air mass traverses 20 are relatively less in number, the SO2concentration is much lower compared to that in the Land BB air masses.
SO2 emissions subsequently lead to (NH4)2SO4 formation which was the main salt present in this branch and was also evident from the high correlation between the two ions. (NH4)2SO4 may be formed both due to emissions from power plants and traffic (similar to B.reg). Traffic emissions can be understood from the M-pattern in diurnal profiles of NH4 + and SO4 2-, though the variation was not very pronounced, and might be suppressed due to power 25 plant emissions. The traffic signal was more clearly implied by the diurnal profile of (NH4)2SO4 for AS as seen in Figure S2. The correlation of [NO3 -] with [ns-NH4 + ] was very poor for AS indicating that NO3might be associated with K + and Na + , similar to B.reg.
Both BC and POA for AS air mass were less than in the SA air mass. However, compared to BB, BC was slightly higher and POA was comparable. BC was likely of industrial origin. The POA diurnal profile was similar to the other air masses. Similar to B.reg, the OOA diurnal pattern resembled that of odd oxygen, where the odd oxygen profile is reported in Tiwari et al. (2015).
Thus, the direct emission sources and the pre-cursors (SOX, NOX, NH3, O3, and Ox) that lead to particulate matter 5 formation, strongly impacted the chemical properties of aerosols. The chemical properties of aerosol also impacted the hygroscopicty of aerosols, as will be discussed in the following section.

Impact of chemical composition on the hygroscopicity of air masses
This study provides the first long term estimation of aerosol hygroscopicity in the PM1 regime. The mean к was approximately the same for all the air masses which is ~ 0.3 (0.32±0.06 for AS, 0.31±0.06 for BB, and 0.32±0. 10 10 for SA), and in line with the global average value of 0.27± 0.21 for continental aerosols Petters and Kreidenweis, 2007;Pöschl et al., 2009;Pringle et al., 2010). Including BC in к calculations lead to a difference of 10% in к on an average, shifting the mean к of 0.32 to 0.29. The BC mass fraction and volume fractions were 10% and 9% respectively. Thus, the change in к due to the introduction of BC was not significant. к varied from 0.13-0.77, and there was a difference in the diurnal variation of the hygroscopicity parameter for 15 the various air masses (Figure 4). A similar finding was observed in China with a mean к of 0.3 and varying in the range 0.1-0.5 (Rose et al., 2010). Recently, к of 0.42±0.07was also reported for PM2.5 for Delhi based on Beta Attenuation Monitor (BAM) measurements of PM2.5 (Wang et al., 2019), using an indirect method in the absence of direct measurements. Thus, the dependence of к on size cannot be underestimated for Delhi and should be dealt with in the future. 20 Variation of к with size has been seen at several places in the world. E.g., at 97% RH, mass growth factors of 6.95 and 9.78 were reported for the size ranges, 0.53-1.6 μm and 1.6-5.1 μm respectively, at the Slovenian coast (Tursic et al., 2005). Aitken mode к was 0.25 while the accumulation mode к was 0.45 for Beijing ).
An increase in к for the higher size regime points out to the fact that organic fraction is higher in the smaller size range, while the inorganic fraction increases substantially with size 25 In the PM1 size range, while к of 0.1 indicates secondary organic aerosol, к varied from 0.01-0.8 for biomass burning aerosols in lab studies (Peters et al., 2009). к varied from 0.15 to 0.25 with lower values (around 0.16) be observed during the night when biomass burning particles prevailed during wintertime at Athens, Greece (Psichoudaki et al., 2018). Thus,к values for Delhi can represent both secondary formation and biomass burning. This is true for Delhi which had both POA and OOA in all the air masses, while BBOA was present in AS and SA air masses, as detailed in the preceding sections on chemical properties.
An important observation for all branches is that when the inorganic volume fraction (of dominant salt) increased ( Figure S2), or during the times when к was high, (Figure 4) or when the organic volume fraction decreased 5 ( Figure S2), a dip in DC ( Figure 6) was seen, implying that a larger size regime was available for activation. The diurnal variation of к ( Figure 4) followed more strongly the diurnal pattern of the dominant inorganic salts for a cluster ( Figure S2) since the hygroscopicity parameters for inorganic salts are considerably higher than that of organics. Pearson correlation coefficient (r) values between к and the salt volume fractions revealed that the diurnal patterns of к were governed dominantly by volume fractions of (NH4)2SO4 (r: 0.85) for AS, moderately 10 by (NH4)2SO4 (r: 0.55) and NH4NO3 (r: 0.49) for BB, and dominantly by NH4Cl for SA air masses. For the two BB branches, к of B branch was governed dominantly by (NH4)2SO4 (r: 0.78) and moderately by (NH4)2SO4 (r: 0.57) and NH4NO3 (r: 0.54) for B.reg. For SA air masses, к of R1, R2, and R3 was governed dominantly by NH4Cl (r values of 0.71, 0.89, 0.95), and jointly by NH4Cl (r: 0.65) and NH4NO3 (r: 0.73) for L.
High volume fractions of (NH4)2SO4 and NH4NO3 may be attributed to SOx, NOx, and NH3 emissions due to power 15 plant emissions and traffic. In the SA sub-branches (Figure 4), the spike in к during the early morning (07:00-08:00) exhibited the sequence: к(R3) > к(R2) > к(R1) >= к(L), and the lower spike in late evening (18:00-22:00) exhibited R3 < R2 <R1 < L, and is attributed to NH4Cl formation. This implies that during the morning, R3 aerosols were most hygroscopic, while L aerosols were least hygroscopic, while after 09:00, L aerosols were most hygroscopic and R3 aerosols were the least hygroscopic. The flatter curve of к can be attributed to two factors: 20 (a) the chloride contribution of distant trajectories decreased very steeply with time compared to the local emissions, and (b) к of L was also supplemented substantially by NH4NO3. Thus, the source activities due to which the chemical properties of aerosols are shaped impacted the hygroscopicity parameter tremendously. This consequently impactedthe size regime of aerosols available for activation and is discussed in the following section.  Table S2.

Impact of governing parameters on CCN estimates of air masses
High values of NCCN for Delhi are consistent with other polluted regions in the world. The relevant statistics for two highly polluted sites, namely Beijing and Kanpur, are presented from Gunthe et al. (2011) and Bhattu et al. 10 (2015) respectively. NCCN was 7660± 3460 and 900-27000 (in cm -3 ), at 0.46% SS and in the range 0.18-0.6% SS, respectively. The high NCCN is consistent with the high NCN. Correspondingly, NCN was 16800±9100 and ~50000 (cm -3 ). Even though high number concentrations of CCN and CN have been reported, the af was not seen to be so high. The af in Beijing was 0.54±0.23and 0.66±0.23, at 0.46% SS and 0.86% SS, respectively. For Kanpur, af was reported as ~0.018-0.54 for 0.18-0.60% SS. However for Delhi, the af ranged from 0.19 for AS, at 0.1% SS to 15 0.86 for R1at 0.8% SS, implying that even at low SS, a considerably large number of particles were activated and at high SS, almost all particles reached the activated state. It should be noted here that the statistics for Beijing and Kanpur correspond to the range (3-900 nm) and (14.6-680 nm), while the estimates for Delhi are given in the (10-560 nm) range. This finding is also consistent with Wang et al. (2019) which states that for Delhi, activation of a 0.1 μm particle requires SS ~0.18± 0.015% compared to ~0.3% for Beijing, 0.28-0.31% for Asia, Africa, and 20 South America and ~0.22% for Europe and North America. The high activated fractions of aerosol can impact the precipitation patterns in Delhiand may be responsible for the short, intense precipitation events, and decrease in overall rainfall. However, no study to date has validated this growing trend with CCN measurements or estimates and this needs to be investigated in the future.
The af and NCCN for all air masses as expected increased with an increase in supersaturation. The variation of CCN 25 and activated fraction with SS are shown in Figures 5and S3. The figures clearly show that even though NCCN for SA was far greater compared to BB and AS, the activated fractions were fairly close. B.reg had higher NCCN, but close af compared to B. Similarly, L had the lowest NCCN among all SA branches but highest af. These features elucidate that many factors are at play and impact NCCN and af differently. To determine the governing parameters impacting both NCCN and af, the diurnal patterns of NCCN, af, Dc at 0.1%, 0.4% and 0.8% SS, NCN, NAitken, NAccumulation, к, and GMD are shown in Figures 6, S4 and S5.
The NCCN curve for SA showed a sharp diurnal feature which was not as prominent for the other two. However, at noontime, at 0.4% and 0.8% SS, while NCCN dipped for SA, it rose for AS and BB ( Figure 6). Further, with the increase in supersaturation, the dip in CCN of SA increased such that (a) at 0.1% SS, SACCN > ASCCN, BBCCN, (b) 5 at 0.4% SS, SACCN ~= ASCCN, BBCCN, and (c) at 0.8% SS, SACCN < BBCCN and ~= ASCCN. The explanation for this observation is that at 0.1% SS, the CCN was governed by the accumulation mode, but at 0.4% and 0.8% SS, it was governed more by the Aitken mode. It is the Aitken mode that dominantly governs the total number concentration, and hence it can be said that CCN is governed by CN at higher SS, and by the accumulation mode at low SS. This is because as supersaturation increases, Dc increases. At low SS, Dc is high (Dc > 100 nm), almost 10 always at 0.1% SS, for all branches as shown in Figure 6d, hence the size distribution that is integrated to get the CCN involves the accumulation mode only. At 0.4% SS, the Dc was around 40-47 nm, therefore a considerable fraction of Aitken mode and accumulation mode was available for activation. At 0.8% SS, the Dc was ~25-30 nm, therefore, the contribution of Aitken mode further increases and accumulation mode is also available as usual.
These findings are also true for BB branches (B and B.reg) and SA branches (L, R1, R2, and R3) as shown in 15 Figure S4. It is the low value of DC relative to other places that is responsible for high CCN. As explained in Section 3.4, Dc is largely associated with к. The DC at other places such as Kanpur varied from 50-200nm, for SS ranging from 0.18-0.60 , compared to which Dc for Delhi is lower (17-142 nm, for SS ranging from 0.1%-0.8%), implying a larger regime is available for activation. It is pertinent to mention here that the dip in CCN for SA during mid-day and the peak during the same for AS and BB can be attributed to the following. 20 (a) The dip in NAccumulation for SA during mid-day was much more prominent compared to AS and BB. (b) During mid-day, the NAitken for SA also decreased while it increased for AS and BB. Thus, the dip in NCCN was strengthened by the simultaneous dip of both Aitken and Accumulation modes, while the peak in NCCN for AS and BB was a manifestation of the dominant peak in Aitken modes. Similar features were also exhibited for subbranches of BB and SA. A deeper insight reveals that the dip in number concentration during mid-day for SA was 25 most aptly seen in the diurnal pattern of POA (which is the most dominant NRPM1 species) and to quite a good extent, in other NRPM1 species barring SO4 2ion. Similarly, the peak in Aitken mode for AS and BB can be attributed to NH4 + , SO4 2-, and OOA concentrations (the dominating species in the respective branches). Thus, the source activities and trajectory pathways impact CCN concentration at the receptor site.
The diurnal pattern for af showed a dip in the mid-day hours for all the air-masses, even though NCCN during midday for AS and BB peaked during mid-day hours. The time of dip in af (more prominent at 0.4% and 0.8% SS) occurred earlier for AS, compared to that for BB and SA ( Figure 6). This was governed by the time of dip in the Geometric Mean Diameter (GMD) of the three branches. The GMD diurnal variation was very similar to af, and the r 2 values between GMD and af enlisted in Table S3 also point out the same. Thus, even though a dip in DC 5 should correspond with an increase in af and vice-versa, this does not usually happen as the change in Dc is less compared to the shift in size distribution such that not only the highest number concentration values changes, the diameter at which this occurs also changes, thereby changing the number available for activation. For example, in Figure 7 two different size distributions for AS air mass at 02:00 and 11:00 are compared. At 02:00 and 11:00, the following characteristics were noted: (a) GMD02:00: 83.78±16.58 nm which is considerably higher than 10 GMD11:00: 47.18±13.01 nm, (b) GSD02:00: 1.69±0.13 nm which is nearly the same as GSD11:00: 1.62±0.14 nm, (c) DC02:00: 44.45±3.83 nm which is slightly higher than DC11:00: 43.48±2.28 nm (at 0.4% SS), (d) CN02:00: 15849±9269 cm -3 which is lower than CN11:00: 25873±9840 cm -3 , (e) CN_Aitken02:00: 9819±5945 cm -3 which is lower than CN_Aitken11:00: 22376±9693 cm -3 , (f) CN_Accu02:00: 5729±3684 cm -3 which is higher than CN_Accu11:00: 3311±1569 cm -3 ,(g) af02:00: 0.75±0.12 which is considerably higher than af11:00:0.38± 0.17, at 0.4% SS, and (h) 15 CCN02:00: 11595±6710 cm -3 which is higher than CCN11:00: 8946±3899 cm -3 , at 0.4% SS. The af at 02:00 is higher than that at 11:00. Dc decrease should correspond to CCN increase, but the magnitude of ∆DC = 0.97 nm is very small. The decrease in GMD (∆GMD = 36.6 nm), on the other hand, is very high with negligible changes in GSD.
At 02:00, since DC was considerably less than GSD, most of the particles are counted for activation. At 11:00, since DC and GMD were very close, nearly 50% of particles are not available for activation. The very high Aitken 20 mode concentration at 11:00 (higher than that at 02:00) was not available for activation at both the times. Thus, NCCN over here is governed by the accumulation mode, which was higher at 02:00, thus making NCCN higher. In this scenario, NCCN and af go hand in hand. However, there also exists a second possibility. E.g., for the SA branch, NCCN, at 0.4% SS, at 08:00 is higher than that at 05:00, while the af at 08:00 is lower than that at 05:00 ( Figure 6).
NCCN was governed at these times by Aitken mode which was higher at 08:00, while af is governed by GMD 25 which was lower at 08:00. Thereby, it is established that CCN is governed by CN (dominantly by either, Aitken or Accumulation mode as the case may be), while the af is governed by GMD.
At this junction, it is also pertinent to mention how chemical dispersion and parameters governing CCN are interconnected. The standard deviation of к (σ(к)) around к is often used as an estimate of the degree of heterogeneity (chemical dispersion) of particles (Psichoudaki et al., 2018;Lance et al., 2013). The chemical dispersion for all 30 air masses and their sub-branches is shown in Figures 6, S4 and S5. The chemical dispersion for SA air mass during the early hours (6:00-8:00) coincided with chloride emissions, and during late-night, after 20:00, with POA and OOA emissions. During the time of high chloride emissions, к also peaked, since inorganics are associated with high hygroscopicity, while during the late hours, к dropped due to an increase in organics associated with low hygroscopcity. The diurnal patterns of activated fraction, GMD and chemical dispersion were also similar. 5 This implies that higher heterogeneity shifts GMD to a high value, thereby increasing the available regime for activation and vice-versa. There was no discernible pattern noted for the other air masses.
It is hereby established that aerosol physical and chemical properties, and their time evolution, are tightly linked with each other. The indirect impact of chemical composition on CCN and af manifests in two ways: (a) NRPM1 species impact the diurnal patterns of Aitken and Accumulation modes which in turn impact CCN, and (b) NRPM1 10 species impact к, hence subsequently CCN, by impacting the size regime available for activation.

Conclusion
Long term measurements of NRPM1 species and size distribution data were carried out at New Delhi. The air masses originated from SA (L, R1, R2, and R3), BB (B and B.reg), and AS. к was estimated using mixing rule and the bulk к was assumed for the entire size distribution. Using к and size distribution data, CCN estimates were 15 obtained. The SA air mass was the most contaminated air-mass followed by BB and then AS. This resulted in higher NRPM1, NCN (both Aitken and Accumulation modes), NCCN, and af for SA, followed by those in BB, and then in AS. The most dominant salts turned out to be (NH4)2SO4 for AS, (NH4)2SO4 and NH4NO3 for BB, and NH4Cl for SA. AS, B and L branches were completely neutralized while B.reg, R1, R2, and R3 were partially neutralized. The diurnal variations of NRPM1 species were governed by source activities aerosol pre-cursors like 20 SOx, NOx, NH3, O3, and Ox. The high PM1 concentration for Delhi, which exceeds the National Ambient Air Quality Standards, can be mitigated only by controlling both the primary emissions and precursors. To address the situation justly, the following measures are lacking. a) Data enlisting measurements of PM1 emissions from various industries in India and Asia, b) Description of the chemical constituents of aerosol that are emitted, both qualitatively and quantitatively, and c) Defining emission limits and complying with the same. 25 The mean к ~0.3 was the same for all air-masses, with the diurnal variation of к governed by chemical species and thus, source activities. The к diurnal trends impacted the DC diurnal trend which in turn affected the available regime for activation. The NCCN diurnal patterns were driven by Accumulation mode at lower SS and Aitken mode with an increase in SS, depending upon DC, which decreases with an increase in SS. The DC obtained for Delhi was lower than that seen at other places in IGP, for example, Kanpur. The activated fraction for Delhi was very high (0.71±0.15, at 0.4% SS for R1), with the means of activated fractions varying between 0.19-0.87, for SS varying from 0.1-0.8%, whereby their diurnal patterns were governed by GMD. A CCN measurement study with a CCN counter in the future can help verify the estimates, and a closure ratio may be determined. However, in the absence of long-term CCN counter measurements, the importance of these findings cannot be neglected. These 5 results can serve as valuable inputs to GCMs to better quantify precipitation. The high NRPM1 loading and activated fractions are bound to significantly impact precipitation over Delhi, aerosol radiation budget and indirect effects, and need to be investigated thoroughly in the future. These investigations may answer the short intense precipitation events occurring over Delhi and the decrease in the overall rainfall over the past half-century.
Author contributions. LHR, JSP, GH, and ZA designed the study. ZA, SG, and SB out the data collection. ZA 10 carried out data processing and analysis. ZA and GH carried out the interpretation of results. ZA wrote the manuscript and was assisted by SB, LHR, and GH in reviewing the manuscript.
Competing interests. The authors declare that they have no conflict of interest.     Cluster mean trajectories were obtained for all seasons and clubbed as per directions. AS branch originated from 5 the Arabian Sea, BB branch with sub-branches B and B.reg originated from Bay of Bengal and SA branch with sub-branches L, R1, R2 and R3 from the north-west direction originated mainly in the South Asian landmass. The map layer used is from "World Countries (Generalized)", by Esri, Garmin International, 2010.

List of figures
(https://www.arcgis.com/home/item.html?id=170b5e6529064b8d9275168687880359). Copyright by Esri, Garmin. All rights reserved. This map is an intellectual property of Esri, Garmin and used under license. Further 10 details may be found on www.esri.com.