Chemical composition and source attribution of submicron aerosol particles in the summertime Arctic lower troposphere

We use airborne measurements of aerosol particle composition to demonstrate the strong contrast between particle sources and composition within and above the summertime Arctic boundary layer. In-situ measurements from two complementary aerosol mass spectrometers, the ALABAMA and the HR-ToF-AMS, with black carbon measurements from an SP2 are presented. Particle composition analysis was complemented by trace gas measurements, satellite data, and air mass history modeling to attribute particle properties to particle origin and air mass source regions. Particle composition above the sum5 mertime Arctic boundary layer was dominated by chemically aged particles, containing elemental carbon, nitrate, ammonium, sulfate, and organic matter. From our analysis, we conclude that the presence of these particles was driven by transport of aerosol and precursor gases from mid-latitudes to Arctic regions. Particularly, elevated concentrations of nitrate, ammonium, and organic matter coincided with time spent over vegetation fires in northern Canada. In parallel, those particles were largely present in high CO environments (> 90 ppbv). Additionally, we observed that the organic-to-sulfate ratio was enhanced with 10 increasing influence from these fires. Besides vegetation fires, particle sources in mid-latitudes further include anthropogenic emissions in Europe, North America, and East Asia. The presence of particles in the Arctic lower free troposphere correlated with time spent over populated and industrial areas in these regions. Further, the size distribution of free tropospheric particles containing elemental carbon and nitrate was shifter to larger diameters compared to particles present within the boundary layer. Moreover, our analysis suggests that organic matter when present in the Arctic free troposphere can partly be identi15 fied as low-molecular weight dicarboxylic acids (oxalic, malonic, and succinic acid). Particles containing dicarboxylic acids were largely present when the residence time of air masses outside Arctic regions was high. In contrast, particle composition within the marine boundary layer was largely driven by Arctic regional processes. Air mass history modeling demonstrated that 1 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c © Author(s) 2020. CC BY 4.0 License.

However, one important aspect concerning the Arctic climate system pertains to the coupling between aerosol, clouds, and radiation. Interaction between aerosol particles, clouds, and radiation strongly depends on number, size, and chemical compo-30 sition of the aerosol particles (e.g., Haywood and Boucher, 2000;Boucher et al., 2013), which are in turn influenced by particle sources, formation, and atmospheric processing.
In Arctic summer, organic matter significantly contributes to submicron aerosol mass (e.g., Schmale et al., 2011;Chang et al., 2011;Lathem et al., 2013;Breider et al., 2014;Willis et al., 2017;Leaitch et al., 2018;Lange et al., 2018;Tremblay et al., 2019). Organic aerosol encompasses a large variety of chemical compounds that vary significantly across Arctic sites, ow-35 ing to differences in sources and chemical processing (e.g., Shaw et al., 2010;Fu et al., 2013;Hansen et al., 2014;Leaitch et al., 2018). Boreal fires and to a lesser extent anthropogenic activities in North America and northern Eurasia can strongly influence the organic aerosol burden in the summer Arctic free troposphere (FT) (Hirdman et al., 2010;Schmale et al., 2011;Lathem et al., 2013;Breider et al., 2014). Whereas in the summertime Arctic boundary layer (BL), organic matter can also be influenced by both primary emissions and secondary chemical processes from Arctic marine and terrestrial sources (e.g., 40 Willis Köllner et al., 2017;Croft et al., 2019). Sources and identities of secondary organic aerosol (SOA) are poorly characterized (Willis et al., 2018). However, tracers of SOA, including a variety of dicarboxylic acids (DCA), such as oxalic, malonic, and succinic acid, have been detected at Arctic sites (e.g., Kawamura et al., 1996;Kerminen et al., 1999;Fu et al., 2009;Kawamura et al., 2012;Fu et al., 2013;Hansen et al., 2014). Several studies demonstrated that particulate DCA are less abundant in summer than in spring, driven by diminished long range transport of precursors and efficient aerosol above 3.5 km to avoid damages under low pressure conditions in the cabin.

Single Particle Soot Photometer -SP2
Concentration of particles containing refractory black carbon (rBC) was acquired with a Single Particle Soot Photometer (SP2) manufactured by DMT (Schwarz et al., 2006;Gao et al., 2007). The instrument includes a continuous intra-cavity Nd:YAG laser (λ = 1064 nm) to classify individual particles as incandescent. The peak incandescence signal is linearly related to rBC 120 mass. For further details see Schulz et al. (2019) and Table S3.

Complementary experimental methods
Meteorological and state parameters were measured by the AIMMS-20 by Aventech Research Inc., integrated on the Polar 6 aircraft. The module Air-Data Probe provides temperature, pressure, relative humidity, wind direction, wind speed, and the three-dimensional aircraft-relative flow vector (true air speed, angle of attack, and side slip). The second module, a GPS, pro-125 cesses satellite navigation signals and determines aircraft three-dimensional position and altitude (Aventech, 2018).
Carbon monoxide (CO) mixing ratios were measured by an Aero-Laser ultra-fast CO monitor (model AL 5002), based on the fluorescence of CO in the vacuum ultraviolet (VUV) at 150 nm (Scharffe et al., 2012;Aero-Laser GmbH, 2013;Wandel, 2015;Bozem et al., 2019). Here, CO is used as an indicator of air mass influenced by combustion sources, including fossil fuel, and biomass burning combustion (Andreae and Merlet, 2001). Further details can be found in Bozem et al. (2019) and Table S3. 130 Number concentrations of particles between 85 nm and 1000 nm were measured with the Ultra-High Sensitivity Aerosol Spectrometer (UHSAS), manufactured by DMT. The instrument is based on particle light scattering technique. Particles that cross the beam of a laser (λ = 1054 nm) are counted and sized by their scattering signals. The intensity of the scattered light provides information on the optical diameter (d opt ) (e.g., Cai et al., 2008;Schulz et al., 2019). For further details see Schulz et al. (2019) and Table S3. Hereafter, only number concentration in a size range between 320 nm and 870 nm (N >320 ) are used (see Sect.

Analysis of the ALABAMA single particle mass spectrometer data
The marker ion method is used to classify particles measured by the ALABAMA (see Köllner et al., 2017). The ion marker 140 method classifies particles based on the presence of pre-selected species that are of interest. This approach is not dependent on the absolute value of the ion peak areas within one spectrum, but only on the presence of a certain ion peak area above a defined threshold value. In principle, four analysis steps are required. First, a pre-selection of chemical compounds that are of interest for the study is necessary. Second, ion markers of laboratory-generated particles with known composition are compared to ions from ambient particles detected during the field experiment. Third, a threshold for the ion peak area is determined to distinguish 145 5 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.  Angelino et al. (2001) Brands et al. (2011), (11) Silva and Prather (2000), (12) Lee et al. (2003), (13) Sullivan and Prather (2007), (14) Yang et al. (2009). the ion signal from background noise (for details see Köllner et al., 2017). Finally, internal and external mixtures of chemical compounds are distinguished by checking the respective ions occur in the same particle mass spectrum or in separate particle mass spectra. Both types of mixtures can be presented in a decision tree. The decision tree is structured as follows: upper/lower branches refer to positive/negative response for whether ion markers of the respective substance are above/below the ion peak threshold. This study focuses on five particle types based on the presence of the following substances (see Table 1 Figure 1. Example for the three-step scaling procedure. (a) The ALABAMA number fraction of the specific particle type (P F ), (b) the averaged UHSAS number concentration in a size range between 320 nm and 870 nm (N>320), (c) and the scaled ALABAMA number concentration (P F ·N>320) are compared with variables that are simultaneously measured. In this study, we used the following variables: FLEXPART PES fractions, altitude, H/C ratio, O/C ratio, and CO mixing ratio. and/or chloride and/or nitrate (Na/Cl/Nitrate), elemental carbon (EC), nitrate (NO 3 ), trimethylamine (TMA), and dicarboxylic acids (DCA). We choose these substances because they are well suited to study aerosol sources as well as aerosol processing along transport that are the subject of this work, such as marine, vegetation fires, flaring, and other anthropogenic sources. We further analyzed ion peaks that might correspond to levoglucosan, but the marker ions are ambiguous (see Supplement Sect. 4). As a result, we will not discuss levoglucosan in this analysis.

155
The conversion of unscaled ALABAMA measurements into quantitative particle number concentrations allows for assigning particle types to different sources (Sullivan and Prather, 2005). Several studies using single particle mass spectrometry applied a similar scaling procedure (e.g., Qin et al., 2006;Healy et al., 2013;Gunsch et al., 2018;Froyd et al., 2019). Based on particle classification, the fraction of the ALABAMA particle types is calculated (Fig. 1a). To determine particle fractions, total analyzed (i.e., detected+ablated/ionized) particles and specific particle types that are of interest are categorized into bins of a 160 simultaneously measured variable (see horizontal axis in Fig. 1). For each bin, the particle fraction is defined as the number of particles of the specific type divided by the number of all analyzed particles (P F ; see Fig. 1a). Bin widths are chosen to guarantee an acceptable level of both statistical significance and resolution. Bins containing less than 20 spectra in total 7 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.
were excluded from the analysis. Further, the particle fraction is scaled to number concentrations, using measurements of a co-located quantitative reference instrument, here the UHSAS onboard the Polar 6 (see Sect. 2.3). The UHSAS number con-165 centration in a size range between 320 nm and 870 nm (N >320 ; Fig. 1b) is used. The selected UHSAS lower and upper size cut offs refer to the ALABAMA size range measured during NETCARE 2014 (see Sect. 2.2). Conceptually, scaled number concentrations of the ALABAMA particle types are determined by multiplying the particle fraction with the averaged UHSAS number concentration in each bin (P F · N >320 ; Fig. 1c).
By using this scaling method, it is assumed that the d opt measured by the UHSAS and the d va measured by the ALABAMA are 170 equal. It is further assumed that the ALABAMA transmits, detects, and ablates/ionizes all particles with the same efficiency.
These assumptions introduce uncertainties in the scaled number concentrations, as the detection efficiency (DE) and the hit rate are observed to be dependent on several factors. The DE is size-dependent, mainly caused by the transmission efficiency of the aerodynamic lens and the optical detection (Zhang et al., 2004;Liu et al., 2007;Brands et al., 2011). However, a sizedependent scaling factor was not deployed in this study, owing to the low particle counting statistics per bin. As a result, the 175 scaling method may lead to uncertainties when determining the concentration of particles smaller than approximately 400 nm.
Likewise, the hit rate in laser ablation/ionization technique is dependent on various particle characteristics, such as size, composition, and shape (Thomson et al., 1997;Kane et al., 2001;Moffet and Prather, 2009;Brands et al., 2011). The hit rate can thus vary during ambient measurements, which may cause a problem when interpreting the fraction of particle types. Further details on the hit rate variability during NETCARE 2014 are given in the Supplement Sect. 1.

180
Given that the scaled number concentration is estimated under several assumptions, we solely discuss trends in number concentrations in comparison with other parameters to identify sources of different particle types, rather than discussing the absolute number concentration. In addition, we will only discuss trends that are clearly visible in both the number fraction and scaled concentration. For clarity, we focus on the scaled number concentration (P F · N >320 ) in the following sections. Additional figures showing particle fraction and averaged UHSAS number concentration are in given in the Supplement Sect. 6. 185 2.5 Air mass history modeling 2.5.1 FLEXPART Air mass history analysis is crucial for understanding aerosol sources and transport processes, both of which control aerosol properties in Arctic regions. In this study, we use the Lagrangian FLEXible PARTicle dispersion model (FLEXPART, version 10.0). FLEXPART is a comprehensive tool for modeling atmospheric transport (e.g., Stohl, 1998;Stohl et al., 2003, 2005) and190 is widely used in Arctic studies (e.g., Stohl, 2006;Paris et al., 2009;Schmale et al., 2011;Evangeliou et al., 2019;Zhu et al., 2020). In this study, FLEXPART was operated backward in time with operational data from the ECMWF with 0.25 • horizontal and 3 hours temporal resolution as well as with 137 vertical hybrid sigma-pressure levels. Tracer air parcels (usually called tracer particles) are released from a certain receptor location and followed for 15 days backward in time (Stohl et al., 2003;Seibert and Frank, 2004;Stohl et al., 2005). The receptor location depends on airborne sampling location. In detail, releases 195 were initialized each 10-min sampling interval along the flight track. By this approach, between 16 and 29 releases were created 8 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.
for each flight, depending on flight duration. The release volume of the tracer particles for each 10 min interval was set by the maximum and minimum geographic locations during this time segment. The main variable calculated by FLEXPART is the so-called Potential Emission Sensitivity (PES) that describes the sensitivity of air sampled at the receptor to source emission locations (Seibert and Frank, 2004;Stohl et al., 2005;Hirdman et al., 2010). The PES value has the dimension of seconds 200 and essentially is proportional to the parcel's residence time in a particular output grid cell (Seibert and Frank, 2004;Hirdman et al., 2010). It is assumed that each source contributes with unit strength in the respective cell. Further, the transported tracer is inert, which means no impact of precipitation and no chemical reactions are considered in FLEXPART (Stohl et al., 2003;Hirdman et al., 2010). The PES function is provided every six hours (i.e., 60 values for a 15 day backtrajectory), with uniform horizontal grid spacing of 0.25 • , and with five vertical levels (400, 1000, 2000, 5000, and 15000 m).

205
FLEXPART output combined with geographical locations of emission sources may provide a more detailed analysis of the relationship between Arctic aerosol composition and sources. This analysis requires the following data processing steps: First, the total residence time of air masses in each grid cell is obtained by integrating PES data in time for a given vertical layer. It is assumed that surface emissions into this vertical layer, the so-called footprint layer, are mixed instantaneously by turbulence in the BL. Therefore, the lowest vertical level (0 -400 m) is applied for most ground sources (e.g., Stohl et al., 2013), except 210 for vegetation fires related to buoyant effects (see details in the Supplement Sect. 2.1). Second, we show maps that specify geolocations of source sectors and source regions (see Sect. 2.5.2). Third, the prior obtained two-dimensional (latitude and longitude) PES map is folded with the two-dimensional map of source regions and sectors. Finally, each grid cell is summed to obtain the total residence time of sampled air mass spent over a specific source within the 15 days prior to sampling and within the vertical footprint layer. In general for this study, the PES fraction is defined as the ratio of the PES of sampled air above 215 source regions/sectors within the footprint layer of the model domain over a 15-days integrated backward simulation compared to the PES of sampled air in the total vertical column of the model domain over a 15-days integrated backward simulation.

Source regions and sectors
Geolocations of selected source regions (e.g., Europe) and sectors (e.g., vegetation fires) are visualized by using two-dimensional maps with horizontal grid spacing of 0.25 • . Geographical regions that likely contribute to Arctic aerosol in summer comprise 220 primarily northern Eurasia, North America, East Asia, and the Arctic itself. Based on several Arctic studies (Stohl, 2006;Shindell et al., 2008;Bourgeois and Bey, 2011;Liu et al., 2015;Yang et al., 2018), selected source regions are shown in Fig. 2.
We consider Europe and Siberia as well as Alaska, northern Canada (nCa), and southern North America (sNA) separately due to differences in vegetation fire occurrence, industrial sources, flaring spots, and population density (Figs. 3a-d, respectively). Arctic open water areas (Fig. 3e). Data on sea ice coverage are based on detected brightness temperature (Cavalieri et al., 1996, updated yearly). In general, these approaches classify whether the source sector is present or absent in the grid cell irrespective from source strengths, rather than providing quantitative source-specified emission data.

245
3 Results and discussions

Meteorological overview and air mass history
Meteorological conditions combined with air mass history play a major role when understanding Arctic aerosol composition influenced by local as well as distant sources and transport. The synoptic situations during NETCARE 2014 have been discussed in previous publications (see Bozem et al., 2019;Burkart et al., 2017;Köllner et al., 2017;Leaitch et al., 2016). Air  changed over the course of NETCARE 2014 from an initial Arctic air mass period (July 4-12) to a southern air mass period (July 17-21) with a transition in between Bozem et al., 2019). Since this has a significant impact on the distributions of the observed quantities, we analyze the measurements for each period individually.

255
During the Arctic air mass period, the measurement region was under the influence of a prevailing high-pressure system generally clear skies, some low-level stratocumulus clouds , low wind speeds, and a stable stratified  shallow BL (Bozem et al., 2019). Inferred from the vertical profile of pseudoequivalent potential temperature, the local upper BL height was on average between 250 m and 450 m, including a surface-based temperature inversion .
Air mass history during the first period reflects the concept of isentropic transport to Arctic regions. Near-surface regions are 260 largely isolated from the rest of the atmosphere and regions aloft are episodically influenced by air originating from southern latitudes (Fig. 4a). The distinct change in air mass history at around 3 km altitude depicts the transition to measurement regions 12 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. located outside of the polar dome ( Fig. 4a; Bozem et al., 2019).
The period is referred to as Arctic air mass period since air mass history 15 days prior to sampling shows a dominant influence of air masses from the high Arctic. Inside the BL, we predominantly sampled air masses that had resided in the Arctic regions 265 (up to 90 % in Fig. 4a). Particularly, a significant fraction of the high Arctic air masses has resided over open water (up to 20 % in Fig. 5a). Thus, air masses inside the BL originated to a large degree from pristine Arctic regions with mainly natural aerosol sources affecting composition. Above the BL, influences of Arctic air masses were still pronounced (Fig. 4a).
However, contributions from southern latitude regions, i.e. Europe, Siberia, northern Canada, Greenland, and the Atlantic Ocean, increased with altitude due to enhanced quasi-isentropic transport from these regions into the high Arctic ( Fig. 4a; e.g., 270 Klonecki et al., 2003). An example of such a transport from Europe towards the Arctic is shown in Figure S35. Air masses at low levels over Europe ascended along the upward sloping isentropes when they were transported polewards. Thus, the impact of the emissions from these source regions on our measurements becomes apparent at higher altitudes (Fig. 5a). Air mass history above 3 km altitude, above the polar dome (Bozem et al., 2019), indicates a sharp transition to a dominating influence of mid-latitude air masses (Fig. 4a). Consistent with several modeling studies, contributions of regions south of 50 -60°N (i.e.,

275
East Asia, Pacific, and southern North America) became significantly important only above 3 km, whereas sources in Europe and northern Canada frequently affected measurements in lower altitudes (e.g., Bourgeois and Bey, 2011;Sobhani et al., 2018;Yang et al., 2018). Thus, anthropogenic pollution in altitudes above 3 km can be attributed to sources in East Asia and southern North America (compare Figs. 4a and 5a). In general during the Arctic air mass period, the contribution of vegetation fires was negligible (Fig. 5a). This finding is in line with low CO mixing ratios (< 90 ppb v ) measured during the first period (Bozem 280 et al., 2019).

Southern air mass period
Meteorological conditions changed after 12 July due to a low-pressure system that was initially centered to the west above the Beaufort Sea and finally passed through Resolute Bay on 15 July. Until 17 July, flying was impeded by the local weather situation with low visibility, fog, and precipitation along with the low-pressure system . However, the 285 measurement region was increasingly impacted by an intense low-pressure system located south of Resolute Bay that led to a different meteorological situation during the period from 17 -21 July (Bozem et al., 2019). Meteorological conditions were characterized by overcast sky, occasional precipitation, moist air throughout the lower troposphere, higher wind speeds, and warmer temperatures compared to the Arctic air mass period Bozem et al., 2019). The BL was more difficult to identify compared to the prior period due to a less well defined temperature inversion. According to the detailed BL 290 study by Aliabadi et al. (2016a), the BL height reached altitudes between approximately 200 m and 400 m. However, vertical mixing between the BL and FT was likely promoted caused by a combination of less intense temperature inversions, higher wind speeds, warmer temperatures, and a higher abundance of clouds compared to the first period (e.g., Stull, 1997;Fuelberg et al., 2010;Tjernström et al., 2012).
This period is referred to as southern air mass period, since air mass history shows the prevalence of air masses originating Upper/lower branches refer to the presence/absence of substances in the particle mass spectra. Numbers are given as relative abundances in % normalized to 10137 particles. The initial classification provides three groups of particle mass spectra with dual-, single positive, and single negative polarity. Particle types, indicated by colors, contain: Na/Cl/Nitrate (dark blue), EC (black), non-sea-spray-(nss)-NO3 (light blue), trimethylamine (TMA) (yellow), dicarboxylic acids (DCA) (brown), and "Others" (gray). As an example, 4 % of 10137 particles (indicated by the asterisk) contain DCA and nss-nitrate (from right to left).
Arctic composition (Figs. 4b and 5b). First, the cumulative contribution of all regions outside the Arctic dominated air mass history within the lower troposphere, making up to 97 % of the PES (Fig. 4b). Along with this, northern Canada and Alaska exerted the largest influences and the European influence was smaller than during the first period, consistent with the prevailing cyclonic activity south of Resolute Bay. Transport of air masses from northern Canada was accompanied by emissions from 300 vegetation fires in the Great Slave Lake region (compare Figs. 3a and S36). Contributions from industrial and flaring activities were likely accompanied by transport of air masses from Alaska, Siberia, southern North America, and Europe (Figs. 4b and   5b). Also, the contribution of air masses that had resided above Arctic open water was significantly lower during the southern air mass period (Fig. 5b), compared to the Arctic air mass period. The representativeness of both July 2014 measurement periods with respect to climatological conditions is discussed in the Supplement Sect. 3. Briefly, synoptic situations during the 305 Arctic air mass period were generally consistent with the climatology, whereas the southern air mass period with influences of the low-pressure system were significantly anomalous compared to climatological mean. 15 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.

Single particle chemical composition
By means of the ALABAMA, 10137 particles were analyzed during the NETCARE 2014 study when sampling outside clouds; 94 % of the spectra include size information; 78 % of the spectra have dual polarity (Fig. 6). Potential reasons for the lack of 310 negative ions are discussed in Köllner et al. (2017). It might be that single-polarity spectra are preferably produced in high relative humidity environments (Neubauer et al., 1998;Spencer et al., 2008), in particular in marine environments (Guasco et al., 2014). Particle classification by the ion marker method introduced in Sect. 2.4 is illustrated here by means of a decision tree in Fig. 6. Particle groups that include less than 300 particles (∼ 3 %) are not further sub-classified due to the low counting statistics. This means EC-and Na/Cl/Nitrate-containing particles are excluded from other types due to their low number. In 315 contrast, particles that contain substances such as nitrate, TMA, and/or DCA and that are externally mixed from EC and Na/Cl can be assigned to several types. TMA-and DCA-containing particles are the most prominent types with relative contributions of 23 % and 31 % to the total number of analyzed particles, respectively (indicated by yellow and brown filling in Fig. 6). In contrast, particles containing Na/Cl/Nitrate, EC, and nitrate (externally mixed from Na and Cl) account for 2 %, 2 %, and 7 %, respectively. In this analysis, 46 % of all mass spectra are not included in the particle classification (gray filling in Fig. 6). The 320 mean spectrum of those "Others" primarily shows peaks of potassium and sulfate (Fig. S9). In the following, mean spectra of the main particle types (Fig. 7) combined with additional ion signals (Table S2) provide an overview of the averaged chemical composition.
The term particulate DCA implies the presence of oxalic, malonic, and/or succininc acid (see Table 1), with oxalic acid as most abundant (86 %), followed by succinic acid with 41 % and malonic acid with 38 % (not shown). In addition, the majority of the 325 DCA spectra show the concurrent presence of oxidized organics at m/z -45, -59, -71, and -73 ( Fig. 7e and Table S2). From our ALABAMA analysis, we can infer that particulate DCA was partly internally mixed with particles containing sodium, chloride, and/or nitrate (Na/Cl/Nitrate), which are interpreted as sea spray particles (Table S2 and Fig. S8). In detail, the presence of DCA was found in 60 % of those sea spray particles. The presence of DCA together with sea spray had been observed earlier during measurement campaigns with focus on marine air (Sullivan and Prather, 2007), also in Arctic regions (Kerminen et al.,330 1999). However, the majority of DCA were externally mixed from sea spray (Figs. 6 and 7e). Moreover, 14 % of those DCAcontaining particles (without sea spray) were internally mixed with TMA and MSA (Figs. 6 and 7e). In our prior study, we concluded that particulate TMA in the Arctic summer results from secondary formation, owing to the external mixture of TMA and primary components, such as sodium . Thus, together with the secondary nature of TMA and MSA (Hoffmann et al., 2016), the internal mixture of DCA with TMA and MSA suggests that DCA results from the secondary 335 (potentially aqueous) conversion of precursor gases (e.g., Kawamura and Bikkina, 2016).
Nitrate was also externally mixed with sea spray particles, hereafter referred to as non-sea-spray (nss)-nitrate. The averaged chemical composition of EC-and nss-nitrate-containing types provide evidence for internal mixing with secondary substances such as sulfate, MSA, and oxygen-containing organic matter (OM; Figs. 7b-c and Table S2). In addition, 20 % of the EC particles were internally mixed with nitrate ( Fig. 7b and Table S2). More than 50 % of the nss-nitrate-and EC-containing 340 particle spectra include ion signals of DCA (Fig. 6), particularly oxalic acid at m/z -89 (Fig. 7b-c).

Arctic marine influences on particle composition
The clean and pristine Arctic background conditions during the first period (Fig. 4)  was high (Fig. 8), linking those particles to locally emitted sea spray. In addition, both particle types are predominantly present in the Arctic marine BL (Fig. 9). Taken together, air mass history demonstrates that Arctic marine sources contributed to the abundance of particulate TMA internally mixed with MSA, sulfate, and/or other organics as well as sea spray particles in the 355 summertime marine Arctic BL.

Long range transport influences on particle composition
In a next step, we will focus on the particle composition measured in the summertime Arctic lower FT. In contrast to particle composition in the Arctic marine BL, our analysis suggests that particle composition in the Arctic FT was much more influenced by long range transport (see Sect. 3.1). To illustrate this, we show the vertically resolved aerosol composition in Fig. 10  in Fig. 11.
Particularly during the Arctic air mass period, the ALABAMA analysis suggests that organic matter in the FT was partly composed of particulate DCA. The number concentration of DCA-containing particles was increasing with increasing altitude (Fig. 10a). Further, Figs. 12 and 13 show the number concentrations of DCA-containing particles in comparison with the HR-

365
ToF-AMS estimated oxygen-to-carbon (O/C) and hydrogen-to-carbon (H/C) ratios as well as with the FLEXPART derived air mass residence time over Arctic regions, respectively. The abundance of DCA-containing particles coincided with air masses with more oxidized organic matter (Fig. 12). Consistent with this result, the size of DCA particles was larger compared to other particle types (Fig. S11). Finally, air mass history shows that the abundance of DCA-containing particles correlated with time spent outside Arctic regions (Fig. 13). This result is in line with findings of an Arctic airborne study by Schmale et al.  Figure 11. Aerosol composition measured by the ALABAMA, by the HR-ToF-AMS, and by the SP2 (indicated on the right side) as a function of FLEXPART derived air mass residence time over the East Asia (a), northern Canada (b), Europe (c), and southern North America (d) during the Arctic (a,c) and the southern air mass (b,d) periods. Top panel represents each the ALABAMA scaled number concentration (P F · N >320 ) with right axis referring to EC-containing (black) and left axis referring to nss-nitrate-containing (light blue) particles. Lower panels represent each median (solid line with marker) and interquartile ranges (shaded area) of the HR-ToF-AMS mass concentrations of sulfate (red), nitrate (blue), ammonium (orange), organic matter (light green), the sulfate-to-organic ratio (pink), and the organic-to-sulfate ratio (dark green) as well as the SP2 mass concentration of refractory black carbon (rBC) (black). Uncertainty analyses are given in the Supplement Sect. 7. 20 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.
anthropogenic precursor gases as well as primary emissions. However, direct observations of DCA in Arctic regions were up to now confined to measurements in the BL (e.g., Shaw et al., 2010;Kawamura et al., 2012;Leaitch et al., 2018), suggesting summer minimum concentrations due to a combination of diminished transport of precursors and more efficient aerosol wet 375 removal compared to other seasons. Along with results in our study and the stable stratified nature of the Arctic BL, it is likely that that these previous BL studies were not sensitive to the presence of organic acids in the Arctic FT influenced by long range transport. Besides the presence of particulate DCA in the FT, DCA-containing particles internally mixed with TMA and MSA (see Sect. 3.2) were largely present in the stable stratified BL (Fig. S12) and when the residence time within the Arctic was high (Fig. S13), indicating sources of particulate DCA in the summertime Arctic. This finding is consistent with previous 380 ground-based and shipborne studies, linking the abundance of DCA in the summertime Arctic to regional Arctic sources (e.g., Kawamura et al., 1996;Kerminen et al., 1999;Kawamura et al., 2010Kawamura et al., , 2012. Aerosol transport from lower latitudes further influenced Arctic composition by particles that contained nitrate and EC. Both species were more pronounced in the FT compared to the BL (Fig. 10). Comparing both periods, the nitrate and organic concentrations measured by the ALABAMA and HR-ToF-AMS were higher during the southern air mass period (compare Figs. 385 10a and b), reflecting the pronounced influence of vegetation fires during this time (see Sect. 3.1.2). Consistently, particles containing nitrate were predominantly present when the residence time over northern Canada was high (Fig. 11b). However, also during the Arctic air mass period, the abundance of nss-nitrate measured by the ALABAMA can be assigned to aerosol transport from lower latitudes (Figs. 11a and c). Air mass history shows contributions from East Asia and Europe to aerosol composition during the Arctic air mass period (Figs. 11a and c, respectively). Figure S33 presents an example of the FLEX-390 PART derived transport pathway from East Asia (particularly Japan) to the Arctic. The advection of those air masses to Arctic regions took place mainly at altitudes above 1 km. In contrast, observed EC measured by the ALABAMA and rBC from the SP2 have their sources in northern Canada, Europe, and southern North America respectively). However, rBC mass concentrations were generally low during the NETCARE 2014 study , reflecting the largely isolated nature of the summertime Arctic from the rest of the atmosphere. Finally, nss-nitrate-and EC-containing particles were larger 395 in diameter compared to TMA-containing particles (Fig. S11), consistent with pronounced influences of aerosol processing during long range transport.
Measurements by the HR-ToF-AMS combined with air mass history analysis suggest that long range transport from sources outside the Arctic contributed to the abundance of particles containing sulfate and ammonium. Both species show slightly higher mass concentrations in the Arctic FT during both periods (Fig. 10). In addition, sulfate was predominantly abundant 400 when the residence time prior to sampling was high in East Asia, Europe, and southern North America (Figs. 11a, c, and d, respectively), in line with prior studies (e.g., Breider et al., 2014;Yang et al., 2018). Consistent with findings by Schmale et al. (2011), the sulfate-to-organic ratio in the FT was enhanced with increasing influences from East Asia (Fig. 11a). Finally, sources in Europe, southern North America, and northern Canada contribute to the presence of particles containing ammonium (Figs. 11c,d,and b,respectively). Taken together, aerosol particle composition in the summertime Arctic BL and lower FT 405 was driven by a combination of processes occurring along transport pathways and within Arctic regions, respectively. The 21 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. following section presents a detailed analysis of the influences from specific aerosol sources at southern latitudes on aerosol particle composition.  Figure 14. Aerosol composition measured by the ALABAMA and by the HR-ToF-AMS (indicated on the right side) as a function of CO mixing ratios during the southern air mass period. Top panel represents the ALABAMA scaled number concentration (P F · N>320) with right axis referring to EC-containing (black) and left axis referring to nss-nitrate-containing (light blue) particles. Lower panels represent each median (solid line with marker) and interquartile ranges (shaded area) of the HR-ToF-AMS mass concentrations of sulfate (red), nitrate (blue), ammonium (orange), organic matter (light green), and the organic-to-sulfate ratio (dark green). Uncertainty analyses are given in the Supplement Sect. 7.

Vegetation fire and anthropogenic influences on particle composition
Anthropogenic and/or vegetation fire emissions contributed largely to the abundance of EC-and nss-nitrate-containing particles 410 in the summertime Arctic FT, particularly during the southern air mass period. To illustrate this, Fig. 14 shows the aerosol composition as a function of CO mixing ratios during the southern air mass period. Additionally, Fig. 15 provide the aerosol composition as a function of FLEXPART derived air mass residence time over vegetation fires, populated, and industrial areas, respectively. The number concentration of nss-nitrate particles measured by the ALABAMA increased with elevated CO mixing ratios (Fig. 14). In parallel, air mass history shows that nss-nitrate particles were largely abundant when time spent 415 over vegetation fires in northern Canada as well as over populated areas was high (Figs. 15a and b). Consistent with these results, nitrate mass concentrations measured by the HR-ToF-AMS also increased with increasing influence from vegetation fires and population (Figs. 15a and b). Similar to nitrate, enhanced sulfate and ammonium concentrations measured by the HR-ToF-AMS coincided with both high CO mixing ratios and with time spent over vegetation fires (Figs. 14 and 15a). Along with vegetation fires and high CO mixing ratios, the organic-to-sulfate ratio was increasing, indicating the pronounced influence of Top panel represents the ALABAMA scaled number concentration (P F · N>320) with right axis referring to EC-containing (black) and left axis referring to nss-nitrate-containing (light blue) particles. Lower panels represent each median (solid line with marker) and interquartile ranges (shaded area) of the HR-ToF-AMS mass concentrations of sulfate (red), nitrate (blue), ammonium (orange), and organic matter (light green) as well as organic-to-sulfate ratio (dark green). Uncertainty analyses are given in the Supplement Sect. 7. vegetation fire emissions on the presence of organic matter in the summertime Arctic FT.
Contrary to what is expected, measurements by the HR-ToF-AMS during the second period show decreasing concentrations of sulfate, nitrate, ammonium, and organic matter with highest CO mixing ratios (Fig. 14). However, the number fraction of 24 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. nss-nitrate-containing particles (Fig. S22) and the organic-to-sulfate ratio (Fig. 14) were increasing with high CO mixing ratios.
Along with the low-pressure system (see Sect. 3.1.2), the transport of air masses that were characterized by high CO mixing 425 ratios from vegetation fires in northern Canada (Fig. S36) was accompanied with precipitation events, as can be seen in GFS archive data (wetter3, 2020). It is thus conceivable that this transport pathway was linked to intensive particle wash-out events (Garrett et al., 2010(Garrett et al., , 2011Browse et al., 2012).
Enhanced concentrations of nitrate, ammonium, and organic matter as a result of transport from lower latitude sources, mainly from vegetation fires, is consistent with previous observational and modeling studies (Kuhn et al., 2010;Chang et al., 2011;430 Brock et al., 2011;Breider et al., 2014). Interestingly, the ALABAMA analysis shows no significant evidence of influences from vegetation fires on the abundance of EC particles, contrary to earlier findings (e.g., Hirdman et al., 2010;Bourgeois and Bey, 2011;Stohl et al., 2013;Breider et al., 2014;Sobhani et al., 2018). This result might be explained by the following reasons.
It is known from previous field studies that chemically aged BC particles from vegetation fires are thickly coated with organics and inorganics (e.g., Paris et al., 2009;Singh et al., 2010;Kondo et al., 2011;Brock et al., 2011). In addition, a laboratory 435 study by Silva et al. (1999) reported the majority of biomass burning particles to include potassium. Given that potassium is known to produce matrix effects, owing to its low ionization potential (e.g., Silva and Prather, 2000), positive ion signals, other than K + , in biomass burning particle spectra are likely suppressed by the presence of potassium. Further, Moffet and Prather (2009) reported negative ion mass spectra of aged soot to be dominated by nitrate and sulfate signals. Together, it is likely that signals of carbon cluster ions, indicative for the presence of EC/soot, are suppressed by other substances that are abundant 440 in biomass burning particles. This hypothesis is consistent with previous SPMS measurements during field experiments at southern latitudes characterizing biomass burning aerosol (e.g., Pratt et al., 2011;Zauscher et al., 2013;Gunsch et al., 2018).
In detail, mean mass spectra of aged biomass burning particles are generally characterized by the concurrent presence of potassium, sulfate, nitrate, organic acids, and nitrogen-containing organics, confirming findings in our study (see Fig. 7c).
However, rBC mass concentration measured by the SP2 (during three out of five flights) provides evidence that vegetation fires 445 in northern Canada contributed to the presence of rBC in the summertime Arctic FT (see Fig. S15). This finding is consistent with recent studies on the summertime Arctic BC burden (Bourgeois and Bey, 2011;Breider et al., 2014;Zhu et al., 2020).
Besides the influence of vegetation fires, the presence of sulfate, ammonium, nitrate, EC, and organic matter in the summertime Arctic FT can further be linked to anthropogenic emissions from southern latitudes. Air mass history illustrates that elevated concentrations of EC particles occurred when the residence time over populated and/or industrial areas in Europe and southern 450 North America was high (compare Figs. 15b-c and S14 with Figs. 11c and d). Consistent with this analysis, the quantitative analysis of rBC by the SP2 shows qualitatively similar trends for the Arctic air mass period (Fig. S14). Particularly during the southern air mass period, enhanced sulfate, ammonium, nitrate, and organic concentrations coincided with time spent over populated and/or industrial areas (Fig. 15b-c).
The ALABAMA data show that the presence of EC particles in the Arctic can partly be associated with industrial emissions 455 in southern North America (compare Figs. 15c and 11d), likely including flaring emissions (compare Fig. S34 with Fig. 3b and c). Figure S34 illustrates that long range transport from southern North America to Arctic regions might accompanied by emissions from oil production industry in Alberta, Canada (compare with Fig. 3c). This result is consistent with recent modeling 25 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. studies, demonstrating the importance of anthropogenic BC sources in northern Eurasia and North America (e.g., Huang et al., 2010;Bourgeois and Bey, 2011;Sobhani et al., 2018). Particularly, studies by Stohl et al. (2013) and Evangeliou et al. (2018) 460 demonstrated the importance of distant sources in southern North America to affect the abundance of EC-containing particles in the Arctic summer. The authors showed that regions in western Canada (Alberta), operating numerous oil producing facilities, influenced aerosol particle composition in the Arctic summer. However, contributions from sources in South Asia cannot be confirmed by our measurements, which might be related to maximum sampling altitude up to 3.5 km that is less sensitive to Asian pollution sources (e.g., Bourgeois and Bey, 2011;Sharma et al., 2013;Sobhani et al., 2018).

Conclusions
Motivated by the limited knowledge of summertime Arctic aerosol, our study aimed to improve our understanding of sources impacting the vertical structure of aerosol particle chemical composition. Aircraft-based measurements were performed using two complementary aerosol mass spectrometers, the ALABAMA and HR-ToF-AMS, and an SP2 in the summertime Canadian Arctic lower troposphere.

470
Our measurements and air mass trajectory analysis show that the vertical structure of aerosol particle composition in the summertime Arctic lower troposphere was driven by a combination of aerosol processes occurring along northward transport pathways and within Arctic regions. In our prior studies, we could show that particle composition in the isolated stable stratified boundary layer was characterized by the abundance of MSA, sodium, chloride, and organic matter that is partly associated with particulate TMA Köllner et al., 2017). Further, we observed that particulate TMA was externally 475 mixed from sea spray particles and smaller in diameter, suggesting a role of marine-derived secondary organic aerosol formation occurring alongside with primary sea-to-air emissions. Here, the presented results confirm the marine source of TMA by our FLEXPART air mass history analysis. The presence of TMA-containing particles correlated, similar to sea spray particles, with time spent over Arctic open water regions. Overall, this study together with findings from earlier work on the NETCARE 2014 measurements emphasize the importance of marine SOA formation on the vertical structure of particle composition in 480 the summertime Arctic Köllner et al., 2017;Croft et al., 2019).
As new results compared to our prior studies, we discussed particle composition measured during two distinctive synoptic situations in the summertime Arctic lower FT, namely the Arctic and the southern air mass periods. In contrast to the isolated Arctic BL, particle composition aloft was characterized by the presence of nss-nitrate, EC, ammonium, sulfate, and oxidized organics that can partly be attributed to DCA transported long distances from mid-latitude sources. However, particle com-485 position in the FT varied with the synoptic situations, owing to different particle sources and transport pathways. Comparing both periods, the organic, ammonium, and nitrate concentrations measured by HR-ToF-AMS as well as the nitrate fraction by the ALABAMA were higher during the southern air mass period, reflecting the pronounced influence of vegetation fires in northern Canada during this time. Our air mass history analysis showed that elevated concentrations of nitrate, ammonium, rBC, and organic matter occurred when the residence time over vegetation fires in northern Canada was high. Further, those 490 particles were largely present in high CO air masses. Also the organic-to-sulfate ratio coincided with time the air mass spent 26 https://doi.org/10.5194/acp-2020-742 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. over these fires.
Besides the influence of vegetation fires, anthropogenic sources in Europe, North America, and East Asia contributed to the presence of particles containing nitrate, EC, organic matter, ammonium, and sulfate in the summertime Arctic FT. The presence of those particles correlated with time spent over populated and industrial areas in Europe, East Asia, and southern North 495 America. Measurements of enhanced EC concentrations can be traced back to oil production industry, including flaring, in Alberta (Canada), by means of the FLEXPART analysis. Further, particles size of EC and nss-nitrate types were larger compared to TMA-containing particles, suggesting aging processes during long range transport.
Particularly during the first period, our analysis showed that the majority of DCA-containing particles were observed in the Arctic FT and in air masses that spent significant time outside Arctic regions. Long range transport of aerosol and precursors 500 from distant sources thus influenced the abundance of DCA particles in the summertime Arctic FT. These findings add to our understanding of organic acids in summer Arctic aerosol, because previous studies were confined to ground-based and shipborne datasets. Combined with the largely stable stratified nature of the Arctic lower troposphere, previous BL datasets were therefore less sensitive to organic acids from long range transport and related processes.
Together, these results indicate the important role of mid-latitude aerosol sources on summertime Arctic particle composition in 505 the lower FT. Particularly, contributions from biomass burning emissions on summertime Arctic aerosol are likely to increase under scenarios for the future climate (e.g., Randerson et al., 2006;de Groot et al., 2013;Marelle et al., 2018;Evangeliou et al., 2019). To our knowledge, this is the first study providing comprehensive particle chemical composition analysis in the summertime Arctic FT, by using two complementary aerosol mass spectrometers with black carbon measurements from an SP2. However, open questions remain, partly owing to limitations of the current study. Being limited by the spatial and tem-510 poral range of the NETCARE 2014 measurements, it is difficult to asses the representativeness of the findings for the whole Arctic. Future widespread and long-term measurements would help to extend the results to other Arctic regions and seasons.
Lagrangian flight experiments would help to better constrain the relative role of aerosol processing, including wet removal and photochemical aging, on summertime Arctic particle composition. Based on limitations in sampling altitude, future Arctic airborne measurements in the mid to upper troposphere are necessary to further characterize the impact of mid-latitude pollution 515 sources on summertime Arctic aerosol properties. Finally, it is of relevance to further characterize organic compounds in Arctic aerosol along with their sources, formation processes, and impact on clouds.
Data availability. The ALABAMA and FLEXPART data can be accessed by contacting the corresponding author F. Köllner (f.koellner@mpic.de) or the second author J. Schneider (johannes.schneider@mpic.de). Other NETCARE data presented in this publication are publicly available  Special thanks to the whole NETCARE-team for data exchange, discussions and support. The authors thank Alexandre Caseiro (MPIC) for his help and discussions on satellite data and for providing the VIIRS nightfire "Flares Only" product. We thank Oliver Appel (MPIC) for helpful discussions on data analysis of the ALABAMA instrument. Fu, P., Kawamura, K., Usukura, K., and Miura, K.: Dicarboxylic acids, ketocarboxylic acids and glyoxal in the marine aerosols collected during a round-the-world cruise, Mar. Chem., 148, 22 -32, https://doi.org/10.1016/j.marchem.2012.11.002, 2013