Biogenic and Anthropogenic sources of Arctic Aerosols

There are limited measurements of the chemical composition, abundance, and sources of black carbon (BC) containing particles in the high Arctic. To address this, we report 93 days of Soot Particle 20 Aerosol Mass Spectrometer (SP-AMS) data collected in the high Arctic. The period spans from February 20th until May 23rd 2015 at Villum Research Station (VRS) in Northern Greenland (81°36’ N). Particulate sulfate (SO42-) accounted for 66% of the non-refractory PM1, which amounted to 2.3 μg m-3 as an average value observed during the campaign. The second most abundant species was organic matter (24%), averaging 0.55 μg m-3. Both organic aerosol (OA) and PM1, estimated from the sum of all collected 25 species, showed a marked decrease throughout May in accordance with Arctic haze leveling off. The refractory black carbon (rBC) concentration averaged 0.1 μg m-3 over the entire campaign. Positive Matrix Factorization (PMF) of the OA mass spectra yielded three factors: (1) a Hydrocarbonlike Organic Aerosol (HOA) factor, which was dominated by primary aerosols and accounted for 12% of OA mass; (2) an Arctic haze Organic Aerosol (AOA) factor, which accounted for 64% of the OA and 30 dominated until mid-April while being nearly absent from the end of May; and (3) a more oxygenated Marine Organic Aerosol (MOA) factor, which accounted for 22% of OA. AOA correlated significantly with SO42-, suggesting the main part of that factor being secondary OA. The MOA emerged late at the end of March, where it increased with solar radiation and reduced sea ice extent, and dominated OA for the rest of the campaign until the end of May. Important differences are observed among the factors, 35 including the highest O/C ratio (0.95) and S/C ratio (0.011) for MOA – the marine related factor. Our data supports current understanding of the Arctic summer aerosols, driven mainly by secondary aerosol formation, but with an important contribution from marine emissions. In view of a changing Arctic climate with changing sea-ice extent, biogenic processes, and corresponding source strengths, highly Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2019-130 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 25 March 2019 c © Author(s) 2019. CC BY 4.0 License.

time-resolved data are urgently needed in order to elucidate the components dominating aerosol 40 concentrations.

Introduction
Climate change driven by anthropogenic emission of greenhouse gases seriously impacts the Arctic.
Areas such as the Arctic have experienced average temperature increases of twice the global mean during the last 100 years (IPCC, 2013). Warming has led to destabilization of permafrost (AMAP, 2017) and a 45 longer melting season resulting in a critical decrease in the sea-ice extent (Stroeve et al., 2007). The latter changes the Earth's albedo and results in positive sea-ice and snow-albedo feedbacks causing further warming (Lenton, 2012). In addition to long-lived greenhouse gases such as CO2, atmospheric aerosols also have an impact on the radiation balance of the Earth. Aerosols affect the radiative balance in various ways. They can absorb and scatter solar radiation, causing either warming or cooling of the atmosphere, 50 respectively. Aerosols can also impact the properties of clouds, for example affecting cloud reflectivity, by serving as cloud and ice-condensation nuclei (Twomey, 1977).
It is well established that the aerosol concentration in the Arctic atmosphere is seasonally varying resulting in higher loadings during winter and spring, compared to summer and fall, often referred to as "the Arctic haze" (Heidam et al., 2004;Tunved et al., 2013;Heidam et al., 1999;Quinn et al., 2007;55 Barrie et al., 1981;Heidam, 1984). This is explained by a greater accessibility to the lower troposphere in the Arctic from anthropogenic source regions outside the Arctic due to an expansion of the polar dome (AMAP, 2011) in winter and spring. In addition, during the Arctic winter strong temperature inversions create stable stratification where aerosol removal processes are strongly reduced prolonging their atmospheric lifetime (Stohl, 2006;Sodemann et al., 2011;AMAP, 2011). The air masses inside the 60 wintertime dome are extremely dry, limiting aerosol wet deposition, while low turbulence exchange caused by the stratification and slow vertical exchange reduces the dry deposition of aerosols (Sodemann et al., 2011;Stohl, 2006;Abbatt et al., 2018). The Arctic haze is observed during spring and is visible as a distinct pollution layer (Heidam et al., 1999;Law and Stohl, 2007;Stohl, 2006;Heidam et al., 2004).
BC is the most important aerosol at absorbing solar radiation in the atmosphere. Of particular concern 80 for the Arctic, when BC is deposited on snow and ice-covered surfaces it changes the albedo, leading to increased absorption of solar radiation and direct heating of the surface (Bond et al., 2013). Consequently, melting accelerates giving BC an important role especially in an Arctic context (Bond et al., 2013;Quinn et al., 2008;AMAP, 2011). Long-range transport of BC to the Arctic is very effective in mid-winter, when removal processes are slowest. Transport reaches a minimum in March -April and wet deposition 85 becomes the most important removal process in the later spring (Abbatt et al., 2018). Still, natural emissions from vegetation fires can be considerable in spring and early summer (Mahmood et al., 2016).
Overall, this leads to a general seasonal cycle with the highest concentrations of BC observed between January and April and the lowest concentrations throughout the summer, but with periodic spikes in concentration throughout the summer (Sharma et al., 2006). OA is also an important component of Arctic 90 aerosol and is composed of many different molecules derived from either primary emissions or from secondary production. Consequently, there are often many distinct sources of OA. OA can typically contribute up to one third of PM1 in the Arctic though few studies have characterized this component in detail (Barrett et al., 2015;Brock et al., 2011;Frossard et al., 2011;Kawamura et al., 2010;Quinn et al., 2002;Shaw et al., 2010). Total OC is relatively constant or decreasing with time in late winter. However, 95 during spring it increases suggesting that there is photochemical production of OA .
There is a need for more detailed measurements of OA composition in the Arctic to better understand the key sources and how these vary with time . Due to aerosols' climatic importance it is crucial to expand the knowledge regarding their chemical and physical properties in the Arctic to reduce the current uncertainty (IPCC, 2013) with respect to the overall effect of aerosols on Earth's 100 energy budget.
It is crucial to understand natural sources in addition to anthropogenic sources of Arctic aerosols. Marine and coastal marine locations constitue a large part of Arctic, and marine aerosols is a source of inorganic and organic aerosols. Production of primary marine aerosols is known to correlate with wind speed and possibly also other mechanisms . Primary marine organic aerosols in Arctic regions 105 are believed to consist of water soluble or surface active organic compounds present in the surface water, or water insoluble microgels Leck and Bigg, 2005;Orellana et al., 2011). Sea salt aerosols play an important role for the climate in spring and autumn (Abbatt et al., 2018). Methane sulfonic acid (MSA), an oxidation product of dimethyl sulfide (DMS) is abundant in spring and summer (Abbatt et al., 2018) and is a key indicator of secondary marine aerosols. Increases in MSA levels has 110 been associated with marginal sea ice moving North (Laing et al., 2013;Quinn et al., 2009;Sharma et al., 2012). In fact, DMS emissions in the Arctic have increased by 30% per decade the last two decades due to both increased temperatures and decreased ice cover (Abbatt et al., 2018). A relationship between MSA and the frequency of new particle formation has also been recently demonstrated based on longterm observations (Dall'Osto et al., 2017). This suggest that DMS is important for summertime particle 115 formation. Another important natural source of Arctic aerosols is ammonia, which is believed to originate Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2019-130 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 25 March 2019 c Author(s) 2019. CC BY 4.0 License. from migrating sea bird colonies (Croft et al., 2016). Modeling studies have been shown to better capture particle burst and growth when an ammonia source from sea birds were included (Croft et al., 2018;Croft et al., 2016). Additionally, ammonia can also be transported from boreal wildfires from lower latitudes.
Many previous Arctic studies have been based on off-line analysis and filter measurements of ambient 120 aerosols with a relatively low time resolution of hours up to a week (Heidam et al., 1999;Heidam et al., 2004;Skov et al., 2006;Quinn et al., 2007;Massling et al., 2015;Leaitch et al., 2018;Sharma et al., 2012;Quinn et al., 2009). Beside the low time resolution, two disadvantages of these types of measurements can be evaporate loss or adsorption of semi-volatile compounds (Lee et al., 2013;Dillner et al., 2009). Highly time-resolved in-situ measurements can reduce these artifacts while also enabling 125 the possibility to observe the variations and trends of different chemical species on a much shorter timescale. In this way, it is possible to look into the processes behind the observed levels. In the last decade, Aerosol Mass Spectrometry (AMS) (Canagaratna et al., 2007;DeCarlo et al., 2006;Jimenez et al., 2003;Drewnick et al., 2005;Jayne et al., 2000) has been widely used as an on-line method for quantitative analysis of chemical composition of atmospheric particles. With the addition of a laser vaporizer (Onasch 130 et al., 2012), its application has been extended to include refractory aerosol components, including refractory black carbon (rBC).
In this study, the time dependent concentrations of sub-micrometer particle composition including OA, SO4 2-, NO3 -, NH4 + , chloride (Cl) and rBC are reported at the high Arctic site VRS in Northern Greenland.
The measurements were conducted by application of a soot particle aerosol mass spectrometer (SP-AMS) 135 and auxiliary measurements during the Arctic spring 2015, when concentrations are expected to peak.
This study presents three months of data using an SP-AMS in the high Arctic. The objectives are to gain better insight into the processes influencing the chemical composition of high Arctic aerosols and to allocate potential sources and source regions. The latter was investigated through positive matrix factorization (PMF) of the organic aerosol mass spectra from the SP-AMS. 140

Sampling site
The atmospheric measurements were carried out at VRS located at the Danish military station, Station Nord in North Greenland ( Figure S1, 81 0 36'N, 16 0 40'W, 24 m above mean sea level). VRS is situated in a region with a dry and cold climate where the annual precipitation is 188 mm and the annual mean 145 temperature is -21 °C. The dominating wind direction is southwestern with an average wind speed of 4 m s -1 as apparent from Figure S1 (Rasch et al., 2016;Nguyen et al., 2013). The SP-AMS data were sampled in an atmospheric observatory containing two laboratories whereas data from a multi-angle absorption photometer (MAAP) and a filter pack sampler was collected in a smaller co-located hut (Flygers hut) -both equipped with particle and gas inlets. The two measurement sites are located 2.5 km 150 southeast of the military station and are only 300 meters apart. Given the close proximity of the two laboratories and the lack of hyper-local sources, we expect both to sample largely the same air mass. A high-volume sampler (HVS) provided filter samples for off-line analysis. The HVS was located at the outskirts of the military station, hence 2.5 km from the main sampling site. All particulate measurements in the Atmospheric Observatory were conducted by drawing air through a slightly heated (absolute 5 °C) 155 particle inlet custom-built by TROPOS. Sampling took place during a campaign within CRAICC (Cryosphere-Atmosphere Interactions in a Changing Arctic Climate) and extended over a three months period from 20 February until 23 May 2015.

The soot-particle aerosol mass spectrometer
An SP-AMS (Aerodyne Research Inc.) was deployed at VRS for measuring mass concentration and 160 chemical composition of submicrometer aerosols with a time resolution of two minutes. The SP-AMS is described in detail elsewhere (Onasch et al., 2012). In brief, the instrument samples aerosols into a vacuum chamber through an aerodynamic particle lens, which creates a narrow particle beam. In the vacuum chamber, the aerosols accelerate to a velocity depending on their vacuum aerodynamic diameter enabling analysis of the aerosol size distribution. Subsequently, the aerosols undergo vaporization, 165 ionization with 70 eV electron impact, and detection with time-of-flight mass spectrometry. The vaporization in the SP-AMS can occur in two ways: (1) impaction on a tungsten surface at a temperature of 600 °C, or (2) intersection with the beam of a continuous-wave 1064 nm intracavity Nd:YAG laser.
The laser extends the application of the AMS to include refractory particulate matter (R-PM) since it enables vaporization of strongly infrared light absorbing particles, such as refractory BC (Onasch et al., 170 2012). In this study, high-resolution (HR) mass concentrations of SO4 2-, NO3 -, NH4 + , organics, Cl and rBC are obtained from the SP-AMS.
The SP-AMS was operated in two minutes laser off and two minutes laser on in V-mode and alternated between the mass spectrum mode and the particle time-of-flight (pToF) to obtain submicrometer particles (PM1) and particle size distribution, respectively. Non-refractory species are reported for time periods 175 where the laser was off. The flow rate was inspected with a Gilian Gilibrator (Sensidyne) and pToF size calibration with ammonium nitrate particles was performed at the beginning and at the end of the field study. During the first part of the campaign, ionization efficiency (IE) calibrations with ammonium nitrate particles were conducted on a weekly basis and during the last part every second week. To establish the detection limit and to enable adjustments of the fragmentation tables a high-efficiency 180 particulate air (HEPA) filter was applied on a daily basis for a period of 30 to 60 minutes with a time resolution of 2 minutes. The lower detection limit of the different species was determined as three times the standard deviation of the mass concentration during the HEPA filter periods ( Table 1). The data were analyzed with the standard AMS Igor Pro-based (version 6.35 Wavemetrics, Inc) software tools SQUIRREL (version 1.57G) and PIKA (version 1.16H), available at http://cires1.colorado.edu/jimenez-185 group/ToFAMSResources/ToFSoftware/index.html. The analysis followed the principles described in study is most likely primarily a sum of organic Cl and NH4Cl due to the acidic environment at VRS.
However, the partitioning of chloride between different specices has not been investigated further, since it is not within the scope of this study. A RIE for rBC of 0.46 was found from calibrations with Regal Black (a commercial carbon black). The appropriateness of this RIE for ambient Arctic rBC is discussed 195 further below (Section 2.4). Calibrations with Regal Black and ammonium nitrate were done with the same frequency. Fragment ions from organic species can overlap with some of the marker ions for rBC.
To minimize the organic contribution to the nominal rBC signal (especially at C1 + an organic contribution was evident), C3 + was used to quantify rBC. Thus, the C3 + signal was scaled with a factor of 1/0.55 to match the fraction in the Regal Black mass spectra (Martinsson et al., 2015). The applied collection 200 efficiency (CE) for non-refractory PM and rBC will be discussed in more detail in a subsequent section.

Auxiliary equipment
The aerosol light absorption was measured using a MAAP (Model 5012 Thermo Scientific) operated at a flow rate of 1 m 3 hour -1 with an inlet without a size cut-off. Aerosols were sampled on a filter in which the light absorption at 670 nm was measured by a photometer. Detailed information about the instrument 205 can be found in Petzold and Schonlinner (2004) and previous MAAP measurements from VRS are published in Massling et al. (2015). The BC concentration is determined from the relationship between the aerosol light absorption coefficient and a specific aerosol absorption coefficient (Petzold and Schonlinner, 2004). The specific absorption coefficient describes BCs ability to absorb solar radiation at a specific wavelength, which depends on the age of the aerosol (Petzold et al., 1997;Sharma et al., 2002) 210 and is often determined based on correlations with thermal-optical measurements of elemental carbon (EC) (Sharma et al., 2004). In this study, the MAAP's default value of 6.6 m 2 /g has been applied based on Massling et al. (2015). Uncertainty in the conversion factor likely impacts the reported absolute concentrations, but not the temporal variability. In addition, a scanning mobility particle sizer (SMPS) measured the particle number size distribution, which was used for validating the SP-AMS results. 215 Description of validation can be found in Supporting Information.

Comparison between instruments
A collection efficiency (CE) adjustment is normally applied to AMS data, which accounts for particle loss in the instrument caused by the inlet and the aerodynamic lens, beam divergence, and particle bounce effects (Canagaratna et al., 2007;Onasch et al., 2012). In this study, the parameterization developed by 220 Middlebrook et al. (2012) has been used where a time dependent CE is determined based on the aerosols chemical composition. Previous studies have shown an increasing CE with particle acidity, the content of nitrate, and relative humidity (Quinn et al., 2006;Jayne et al., 2000;Matthew et al., 2008). The time dependent CE varied with the majority of values between 0.8 and 1 ( Figure S2). In this study, the high CE was due to acidic aerosols. This is also evident from Figure S3.a showing that the theoretical predicted 225 NH4 + concentration necessary for neutralizing the mass concentration of inorganic anions is much larger than the actual NH4 + concentration measured by the SP-AMS (slope = 0.15). The acidity is explained by the high amount of sulfuric acid. Applying the RIE for rBC of 0.46 determined from Regal Black calibrations, a good correlation between rBC and BCMAAP is found ( Figure S3.b). While there is a strong linear relationship between the two (R 2 230 = 0.83), the BCMAAP was about three times larger than the SP-AMS rBC (slope = 0.33 ± 0.02). This indicates that the actual RIE for rBC was lower than the value of 0.46 determined during laboratory calibrations. A lower RIE can be explained by different particle size and a more complex morphology of the Arctic soot compared to the Regal Black used for calibration. An effective RIE is determined for rBC by forcing the SP-AMS measurements to match the MAAP measurements. For rBC an effective RIE of 235 0.15 (= 0.33 * 0.46) is hence applied in this study.
Comparison of the total PM1 mass concentration (sum of OA, SO4 2-, NH4 + , NO3 -, Cl and rBC) with the calculated total volume from the SMPS assuming spherical particles was carried out to validate the SP-AMS results. The SMPS was operated to characterize particles having mobility diameters between 9 and 870 nm. This corresponds to a larger size range than sampled by the SP-AMS, which has 100 % 240 transmission efficiency within aerodynamic diameters between 70 and 600 nm, and adjustment from aerodynamic diameter to mobility diameter further brings the SP-AMS into the SMPS range (DeCarlo et al., 2006;Allan et al., 2003). However, previous studies (Nguyen et al., 2016;Lange et al., 2018) have shown that the dominant particle size range at VRS during winter and spring months is within detection range of the SP-AMS. Thus, the number of particles from the SMPS exceeding the size span measured 245 by the SP-AMS should be relatively small and thereby not influence the results, since particles in the lower end of the size distribution do not significantly contribute to volume. There was a generally reasonable temporal correspondence between the two measurements. Although there were some periods where they differed notably it were within the expected range given the accuracy of the two instruments.
A more detailed discussion about the comparison between the two instruments is presented in Supporting 250 Information ( Figure S5).

Positive Matrix Factorization
PMF analysis (Paatero, 1997;Paatero and Tapper, 1994;Lanz et al., 2007;Ulbrich et al., 2009)  values from 12 to 100. The detailed procedure is described elsewhere (Ulbrich et al., 2009;Zhang et al., 2011). The input HR mass spectra and error matrix with the appropriate ion fragments were generated in PIKA, where the error matrix was calculated as the sum of the quadrature of the electronic noise and 260 Poisson counting for each ion (Allan et al., 2003). Isotopes were removed from both the data and error matrix since they would give additional weight to the parent ion in the PMF analysis.
As described in Ulbrich et al. (2009) "weak" ions with a signal-to-noise ratio (SNR) between 0.2 and 2 were down-weighted by a factor of 2 whereas "bad" ions with a SNR below 0.2 were removed from the data and error matrix. The PMF was executed in exploration mode with a range of factors (between 1 265 and 5). The robustness of the solutions were tested by setting different random starting points (SEED: 0 to 10, steps = 1) (Zhang et al., 2011). The detailed procedures for choosing the best solution were based on Zhang et al. (2011). A solution with three factors was identified after evaluating Q/Qexp and residuals, interpreting the mass spectra and investigating the temporal correlation between the factor time series and potential tracer species (Ulbrich et al., 2009;Zhang et al., 2011). FPEAK and seed values were 270 changed to test the stability of the three-factor solution and based on the diagnostic plots a three-factor solution was selected with a FPEAK and seed value of zero ( Figure S7). A 4-factor solution was scientifically not meaningful with respect to chemistry and returned an O/C ratio >> 1 for one of the factors. Hence we do not observe a fourth "continental" factor, which has been previously observed during the ASCOS cruise track in the summer/autumn season around Svalbard (Chang et al., 2011). If 275 present, the continental factor is most likely of negligible abundance for which reason the PMF-analysis cannot differentiate it from other Oxygenated Organic Aerosol (OOA). Detailed information regarding the factor combination can be found in Supporting Information.  Figure S6. Figure 1c shows the time dependent mass fraction of the different species. The total measured PM1 concentration during the field study is relatively high, averaging 2.3 µg m -3 . It should 285 be emphasized that this average does not consider particulate water, NaCl, and elements such as K, Ca, Si, Al and Fe. These elements may additionally contribute 0.1 -0.2 µg/m 3 to PM1 Heidam et al., 2004). The measurement period covers the Arctic late winter and spring where high aerosol loadings are expected due to the favorable conditions for long-range transport of aerosols from midlatitudes and slow particle removal rates. With regard to PM1 concentration we hence observe the typical 290 Arctic haze phenomenon. Generally, the area around VRS is dominated by winds from southwest , which is also evident during this campaign ( Figure S1). As expected no diurnal pattern is observed for any of the chemical species indicating that the aerosols are regional and likely predominately from long-range transport.
During the entire campaign, SO4 2is the dominant species that on average makes up almost 70% of the 295 PM1 mass concentration measured by the SP-AMS (average 1.5 µg m -3 , Figure 1.b-c). This is in accordance with previous findings for SO4 2at VRS based on measurements with lower time-resolution Fenger et al., 2013;Heidam et al., 2004). Atmospheric SO4 2is mainly formed as secondary inorganic aerosols and only a minor fraction is from primary emissions (Massling et al., 2015).
Secondary SO4 2is dominated by atmospheric oxidation of sulfur dioxide (SO2) and to a minor extent 300 DMS (as the long-range transport is occurring over sea ice), and is dependent on the oxidative capacity of the atmosphere e.g. the concentration of hydroxyl radicals (OH). Secondary long-range transported to VRS) of SO2 leads to higher concentration of SO4 2from March, where solar radiation is sufficient with peak radiation exceeding 100 W/m 2 (Figure 3). This is consistent with results reported from other 305 Arctic sites (Quinn et al., 2007;Gong et al., 2010;Heidam et al., 2004;Skov et al., 2017). Previous studies suggest that the main source of SO2 and SO4 2at VRS is long-range transport of anthropogenic emissions mainly originating from Siberia (Heidam et al., 2004;Nguyen et al., 2013). In winter and early spring, direct emissions of sea-salt sulfate and photo-oxidation of oceanic emissions of DMS were expected to play a minor role since the ocean surrounding VRS is frozen at that time of the year (Heidam 310 et al., 2004). From the beginning of April, the sea ice extent of the Northern Hemisphere is markedly reduced, and at the same time solar radiation increases ( Figure 3). In this period, we observe MSA as an ion in the SP-AMS at m/z 78.9854. MSA is formed by atmospheric oxidation of DMS, which results from bacterial breakdown of dimethylsulfoniopropionate produced by marine phytoplankton and microalgae (Carpenter et al., 2012). DMS emerges steadily and peaks in the end of April (see Section 315 3.2). Oxidation of DMS may involve the hydroxyl radical, ozone, and halogen radicals such as Cl and BrO (Barnes et al., 2006;Hoffmann et al., 2016).
In this study, the OA fraction is the second largest contributor to PM1 with an average concentration of 0.6 µg m -3 . Weekly averages showed a clear decrease from mid-April relative to the spring season concentrations (Figure 1). The OA time dependent concentration shows relatively large peaks during 320 shorter time periods, which in some cases can be attributed to a change in wind direction from Southwesterly to Northerly winds (around 10˚, Figure S1). While these wind directions were registered on a few occasions they potentially provided local pollution from the military station located three kilometers away from the measurement site. These peaks have not been discarded and the impacts of local pollution will be discussed further in Section 3.

325
Particulate NH4 + is found in much lower concentrations compared to OA and SO4 2with an average concentration of 0.09 µg m -3 . For the campaign, a significant correlation is found between SO4 2and NH4 + . However, it is known that SO4 2and NH4 + do not originate from the same sources. SO2, a key precursor to SO4 2-, originates from combustion of fossil fuel and is oxidized to SO4 2in the atmosphere.
In contrast, ammonia (NH3) which is the precursor of NH4 + , derives largely from long-range transport 330 from farms and more locally from sea bird colonies (Croft et al., 2016). The strong correlation between SO4 2and NH4 + (R 2 = 0.70) suggests that the acidity of the particles is reasonably constant with time. This is furthermore in agreement with the general assumption that NH4 + is bound irreversibly to SO4 2-(e.g. Seinfeld and Pandis, 1998), in this case as ammonium bisulfate. Particle bound NH4 + has a much longer lifetime than NH3 and therefore it is transported as NH4 + even to the high Arctic. 335 The average concentration of NO3and Cl are 0.03 and 0.02 µg m -3 , respectively, which is close to the detection limits. These concentration levels are lower compared to what has previously been observed at VRS (Fenger et al., 2013;Heidam et al., 2004). However, the SP-AMS does not measure refractory chlorine species, such as NaCl. Moreover, Fenger et al. (2013) found that the overall size distribution of chloride and NO3differed from SO4 2-, with Cl and NO3mainly found in supermicrometer particles (> 340 1 µm) not detectable by SP-AMS. These particles were suggested to originate from local/regional sources (frost flowers). Only during certain periods with specific wind directions NO3and Cl were found in accumulation mode particles, which were ascribed to long-range transported particles (Fenger et al., 2013).
The average rBC concentration of 0.1 µg m -3 is above the lower detection limit (0.01 µg m -3 ) and the 345 highest rBC loadings are found in the first month of the campaign (February). As with OA, some of the spikes in the rBC time series are related to a change in wind direction and likely the result of local pollution from the military station. All data are included here and missing time periods of rBC (during April and May) are due to technical problems with the SP-AMS laser. BC is primarily emitted from both anthropogenic and natural combustion sources (Bond et al., 2013). Upon emission, aerosols containing 350 BC grow by condensation and coagulation into the accumulation mode. Particles in the accumulation mode have the longest lifetime with respect to dry deposition and thus particles can be transported over longer distances during the Arctic haze period when precipitation is scarce (Bond et al., 2013;AMAP, 2011;Massling et al., 2015). These accumulation mode BC-containing particles may serve as cloud seeds in the late spring, when precipitation begins to be important in the Arctic (Garrett et al., 2011). Further, 355 condensational growth of the BC-containing particles may increase the absorption by these particles (Cappa et al., 2012;Liu et al., 2015). Previous studies have found a correlation between BC and SO4 2at different Arctic stations (Massling et al., 2015;Eckhardt et al., 2015;Hirdman et al., 2010). These studies suggest that the two species are internally mixed and possibly undergo similar transport patterns.
Furthermore, comparable correlation slopes were found for the different Arctic locations, which suggests 360 that source regions of BC and SO4 2are similar for the entire Arctic. An even more recent study suggests that only a minor part of ambient aerosols contained rBC inclusions (Kodros et al., 2018). We also find a significant correlation between the two species (students t-test, level of significance 99.995), consistent with previous studies. However, we also find that the R 2 value is relatively low (0.18). The reason for this is that there are periods with particularly high rBC concentrations, likely originating from local 365 emission sources (e.g. the military base), which will be investigated further in the following section.
Additionally, in April and May SO4 2from DMS oxidation will make up a larger fraction of total SO4 2-, and thereby reduce the ratio between rBC and SO4 2-, which is also evident from Figure S4.

Source Apportionment
The PMF analysis was conducted for the HR OA mass spectra with one to five PMF factors and a three-370 factor solution was chosen (more details can be found in Supporting Information). Figure 2 shows the mass spectral profiles of the three different factors for the entire campaign period. Figure 3 illustrates time series for the factors and Table 2 shows the correlation of each factor with tracer species, respectively. Figure  The HOA factor is characterized by hydrocarbon fragments especially at m/z 41, 43, 55, 57, 67, 69 and 380 71 (C3H5 + , C3H7 + , C4H7 + , C4H9 + , C5H7 + , C5H9 + , C5H11 + , respectively) from chemically reduced organic emissions. The O/C ratio of 0.11, high signal at m/z 57 and the absence of CO2 + is a characteristic of primary combustion sources of fossil origin, which is similar to other HOA factors found in previous studies (Zhang et al., 2005;Aiken et al., 2009) and at other Arctic locations (Frossard et al., 2011). The very small contribution of the CO2 + ion at m/z = 44 and the very small abundances of typical biomass 385 burning OA (BBOA) marker ions at m/z 60 (C2H4O2 + ) and m/z 73 (C3H5O2 + ) in the HOA factor spectrum suggests that the HOA factor is not mixed with BBOA. This finding is consistent with previous results that indicate BBOA levels are typically very low, based on measurements of levoglucosan in the Arctic, (Zangrando et al., 2013). The time series of HOA and rBC showed a moderate correlation (R 2 = 0.35), which is consistent with the HOA factor being of primary origin. The relatively low R 2 value (Table 2)  390 can be partly explained by rBC being internally mixed with SO4 2and transported with the AOA factor.
The HOA time series is generally higher in concentration at the beginning of the measurement period ( Figure 4). The time series of HOA reveals a number of shorter periods with high mass loading, which could be caused by local pollution from the military station 2 km north of the measurement site due to a change in wind direction, or exhaust plumes from snow scooters and heavy-duty vehicles occasionally 395 clearing the road nearby the measurement station for snow (see windrose, Figure S1). It is not trivial to distinguish local events and in this case the possible local contamination was investigated by comparing high HOA peaks (> 0.45 µg m -3 ) with size distribution measurements from the SMPS (Lange et al., 2018). Periods which were attributed to local contamination accounted for less than 1% of OA concentration. Therefore, essentially the entire HOA concentration is assigned to long-range 400 transportation, possibly sources with different ratios of HOA and rBC which would explain the moderate correlation between HOA and rBC.
The AOA is the most abundant factor from the beginning of the campaign through mid-April and accounts for 64% of OA mass for the entire field study (Figure 2b). The CO2 + ion contributes notably to the mass spectrum and the O:C is 0.63, indicating that this factor is likely secondary in origin. The 405 dominating OA during the Arctic haze period is thus SOA, which results from long-range transport into the region during winter/spring. AOA is abundant during February to mid-April though lower concentrations are observed around middle of March. At the end of April and onwards the factor essentially disappears, which is in agreement with increasing wet deposition in the spring and a contracting polar dome impairing long-range transport into North Greenland (Abbatt et al., 2018). 410 Generally, an OOA factor mainly consists of SOA but can also include oxygenated organic species from primary emissions (Zhang et al., 2005). In this case the AOA factor correlates significantly (level 99.995) with SO4 2-, which is mainly formed by atmospheric oxidation of SO2 suggesting the main part of the factor being SOA. The correlation is especially good until mid-April after which SO4 2beigns to correlate with MOA. The O/C ratio of 0.63 also indicates a less oxidized and fresher SOA factor, or an SOA 415 formed from generally larger precursor VOCs, similar to what has been found in previous studies (O/C between 0.52 -0.64, (Aiken et al., 2008)). The AOA mass spectrum also included mass spectral peaks at m/z 60.021 (C2H4O2 + ) and 73.029 (C3H5O2). These fragments are often taken as being indicative of anhydrous sugar such as levoglucosan, and thereby suggest that biomass burning makes some contribution to Arctic OA. However, in this study biomass burning cannot be verified, since the 420 abundance of C2H4O2 + did not exceed the expected contribution from SOA (Aiken et al., 2008;Aiken et al., 2009;Cubison et al., 2011;Lee et al., 2010;Saarnio et al., 2013). Biomass burning is generally assumed to play a significant role in the context of the composition of the Arctic aerosol (Stohl et al., 2013) where recent publication using isotopes of carbon reports biomass burning or biofuel use to account for up to 57% of EC at the Arctic station Zeppelin at Svalbard during high pollution events in 425 winter (Winiger et al., 2015). However, levoglucosan is prone to atmospheric oxidation by hydroxide radicals (OH) (Hennigan et al., 2010;Hoffmann et al., 2010), which could degrade the markers during transport to North Greenland. This can explain the low abundance of levoglucosan markers measured in this study.
The MOA factor has a mass spectrum dominated by m/z 28 and 44 (CO + and CO2 + ), that is a more 430 oxygenated OA factor due to the presence of e.g. organic acids and acid derived species, such as esters (Duplissy et al., 2011). A high O/C-ratio of 0.95 reveals that the factor is highly oxidized and photochemically aged. The MOA spectrum resembles a marine organic plume previously published from Mace Head, Ireland containing both primary and secondary organic aerosols of marine origin (Ovadnevaite et al., 2011). This spectrum and MOA in this study are different from the marine organic 435 aerosol factor published during the ASCOS expedition in the Central Arctic Ocean (Chang et al., 2011), which shows a closer resemblance with the mass spectrum of pure MSA. In the MSA spectrum, m/z 15, 48, 64 and 79 are dominating peaks, which was also observed in the marine factor from the ASCOS expedition. The distinct peak at m/z 78.9854 is specific for MSA (Huang et al., 2017), and reveals that MOA has a secondary biogenic source (Becagli et al., 2013). The resemblance of MOA from this study 440 with the mass spectrum from Mace Head indicates, that MOA is not solely a secondary marine source, but is most likely also composed by primary marine organic aerosols e.g. from sea spray (Ovadnevaite et al., 2011;Fu et al., 2015). few minutes, only, above the mountains at the horizon. Polar daytime initiates photochemistry and hence the production of OH radicals (Seinfeld and Pandis, 2006) and reactive halogen radicals (Hoffmann et al., 2016;Barnes et al., 2006). From mid-April, the sun is above the horizon all day until the beginning of September. Still solar radiation varies over the day and hence the OH production. In contrast, the concentration of OH during build up of Arctic haze is correspondingly low with ozone being the major 450 oxidant during the dark winter. In Figure 3, the daily averaged solar radiation (W m -2 ) and sea ice extent (km 2 ) on the Northern Hemisphere are shown together with the time series of MOA. While MOA is less abundant during February and March, this factor greatly increases in April, when radiation exceeds approximately 100 W m -2 . In April, the highest OA concentrations is observed where AOA accounts for around 70% of OA (Figure 4). In May, MOA becomes the dominating OA while AOA nearly disappears. 455 At the same time we observe the lowest concentration of OA consisting of 75% MOA (Figure 4). This is significantly higher than observed at Alert (Narukawa et al., 2008). Until the beginning of April, the sea ice extent is constant at around 14.5 million km 2 on the Northern Hemisphere (Figure 3). Hereafter, about a month after the onset of polar daytime, the sea ice surface area starts to decline. After 6 weeks starting from a constant sea-ice extent in mid-May, it is reduced by 2 million km 2 corresponding to a 460 14% loss of ice-covered surface area. Consequently, more open waters allow for higher DMS emissions and atmospheric oxidation of DMS to MSA involving OH. Also open leads and marginal ice zones provide primary marine aerosols . Indeed, previous findings suggest that biogenic productivity in open oceans and sea ice zones and the emission of DMS are responsible for increased new particle formation, as sea ice pack extent retreats (Dall'Osto et al., 2017). Quinn and co-workers 465 reported increased concentrations of MSA at Barrow from 2000 to 2009 associated with the northward migration of the marginal ice zone (Quinn et al., 2009;Sharma et al., 2012;Laing et al., 2013). Of the four northernmost year-round manned observatories at Alert, Mount Zeppelin, VRS and Barrow, the highest MSA concentrations are measured at Mount Zeppelin, likely due to its proximity to open waters around Svalbard, which are a significant source of DMS from May to August (e.g. Lana et al. (2011)). 470 This contrasts with the ice situation around VRS, which is ice covered most of the year.
Considering the stronger oxidizing environment starting in April, we expect MOA to be abundant until autumn, and possibly co-exist with an emerging continental factor as reported during the ASCOS cruise track in late summer/autumn (Chang et al., 2011). MOA constitutes only 22% of OA on average during our measurement period. However, the radiative impact may be greater than the other OA types because 475 it emerged after polar sunrise and persisted during polar daytime for which reason they are optically active 24 hours a day although solar radiation has a diurnal cycle. Moreover, MOA is by far the most abundant OA from end of April and onwards. The observed transition between AOA and MOA is in agreement with Narukawa et al. (2008), who observed a transition between fossil fuel influenced OA to marine OA. MOA is not only secondary but may contain oxidation products of DMS and other VOCs 480 from oceanic origin, and primary components including colloidal gels (Croft et al., 2018;Leck and Bigg, 2005;Orellana et al., 2011). In line with our findings, modelling at several sites in the Canadian Arctic suggested that marine OA may account for more than half of the summertime OA (Croft et al., 2018).
Biogenic marine aerosols can scatter solar radiation, which will result in a negative radiative forcing.
Biogenic marine aerosols can also coat soot particles, which may be transported from wild fires (AMAP, 485 2015), which could impact the CCN activity and absorption by the soot particles, with the latter potentially enhancing the warming influence of the particles (Lange et al., 2018). These findings encourage further studies of optical properties and chemical composition and physico-chemical parameters as CCN ability or hygroscopicity of aerosols prevailing during polar daytime.

Conclusion 490
In the transition from polar night to polar day we concluded SO4 2to be the most abundant species in submicrometer aerosols averaging 1.5 µg m -3 during February to May 2015 and decreasing throughout the campaign period. This is in accordance with previous findings from VRS and Svalbard (Udisti et al., 2016) where SO4 2has been apportioned to be 75% anthropogenic, while natural contributions from crustal, sea salt and biogenic sources contributed minorly by 3%, 12% and 12%. While not previously 495 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2019-130 Manuscript under review for journal Atmos. Chem. Phys. OA burden decreases substantially. In contrast, the MOA is nearly non-existant until early spring but is then by far the dominating OA from the end of April and onwards. The fact that MOA emerges at a time where long-range transport is impaired by increased deposition and a contracting polar dome indicates that the sources to this factor are more Arctic regional in nature. This demonstrates the importance of biogenic sources in the Arctic, especially in the spring. In view of changing biogenic processes and 515 corresponding source strengths of aerosol precursors in a changing Arctic climate with changing sea-ice extent, additional high time resolution measurements are urgently needed in order to elucidate the organic components dominating aerosol summer mass and number concentrations.

Supporting information
Supporting information describes site information, supplementary instruments, collection efficiency, 520 validation of SP-AMS data, and key diagnostics for the PMF solution.

Author contribution
Ingeborg E. Nielsen and Jacob K. Nøjgaard carried out the field measurements, and Ingeborg did the analysis of the SP-AMS data. Jacob and Ingeborg carried out the PMF analysis and took lead in writing