The HR-ToF-AMS (referred to as AMS in the following text) can provide the size-resolved chemical composition of non-refractory submicron particles (Canagaratna et al., 2007; DeCarlo et al., 2006). Its size cut is within the range of near unity transmission efficiency of the main container inlet during R/V Polarstern cruises (Fig. S1). To minimize the inlet loss between the AMS and MPSS, the AMS was located next to the MPSS, and the inlet of AMS was connected directly in front of the MPSS inlet. The two instruments shared a Nafion dryer to maintain the RH of the sampling flow below 40 %. The AMS was operated alternatively between V- and W-modes at a time resolution of 2 min. Regular calibration was performed according to reported methods (Jimenez et al., 2003). Collection efficiency (CE) of 0.7 was applied based on intercomparisons between the (1) AMS and MPSS and (2) AMS and offline measurements, as already described in Huang et al. (2017).

The default components measured by the AMS include organics, sulfate, nitrate, ammonium and chloride. In particular, the marine biogenic tracer MSA was quantified by the AMS based on standard calibrations and validated by collocated offline measurements (Huang et al., 2017). To reduce the signal noise, AMS data at 20 min resolution are calculated and used in the following analysis except as noted. Table 2 provides the detection limits (DLs) of detected AMS species at 20 min resolution (for calculation details, see Huang et al., 2017). In addition, a total uncertainty of approximately 30 % is estimated for AMS measurements, including 10 % for the inlet system, 20 % for the ionization efficiency calibration and 20 % for the collection efficiency (Poulain et al., 2014).

The AMS data measured during R/V Polarstern cruises were analyzed using the software Squirrel v1.54 for the unit resolution and Pika v1.13 for the high resolution, both downloaded from the ToF-AMS webpage (http://cires1.colorado.edu/jimenez-group/ToFAMSResources/ToFSoftware/index.html, last access: 15 December 2018). The software was based on Igor Pro (version 6.22A, WaveMetrics Inc., Portland, OR, USA).

A PM1 high-volume DIGITEL filter sampler was deployed to sample aerosol particles at a 24 h time resolution (midnight to midnight, UTC) during R/V Polarstern CR1, CR2 and CR3, operating at a flow rate of 500 L min-1. The daily aerosol masses were collected on quartz fiber filters (150 mm, Munktell and Filtrak, MK 360, Bärenstein, Germany), which were pretreated at 105 ∘C for 24 h before being measured. All filter samples were refrigerated at -20 ∘C until analysis. The total aerosol particle mass was determined by the weight of the clean and particle-loaded filter, and the particle-loaded filter was separated into several aliquots for different analyses. Inorganic ions were measured after aqueous filter extraction (25 % of the filter in 20 mL, filtered with a 0.45 µm syringe) with ion chromatography (IC; ICS3000, Dionex, Sunnyvale, CA, USA). Organic carbon (OC) and elemental carbon (EC) were analyzed by a thermo-optical method with a Sunset Laboratory Dual-Optical Carbonaceous Analyzer (Sunset Laboratory Inc., Tigard, OR, USA) using the EUSAAR 2 protocol (van Pinxteren et al., 2017). In total, 86 ambient samples from CR1 (25 samples), CR2 (30 samples) and CR3 (31 samples) were analyzed, while 45 effective samples were considered in this study due to the exclusion of filters contaminated by ship exhausts and/or seawater.

Ship contamination exclusion. During all four cruises, the ship's own exhausts (including emissions from the engine and kitchen) from the main chimney occasionally affected the measurements, especially when the wind came from behind the ship. The specific contamination periods are identified based on extremely high concentrations of combustion tracers (e.g., eBC and organics) and meteorological parameters (e.g., wind direction), as illustrated in Figs. S2 and S3. Accordingly, the ship contamination range of the relative wind direction (RWD) was determined to be between 135 and 250∘. With this criterion, 85.2 % of ambient AMS data (2 min resolution) remained as “clean” data during all four cruises. In the following text, all the data analyses are performed on the clean data.

Intercomparisons between AMS and PM1 offline. Prior to investigating the particle chemical composition during the R/V Polarstern measurements, intercomparisons between AMS and parallel measurements were performed to assure the AMS data quality (Figs. S4 and S5). The mass concentration of the individual species from the AMS was compared to that from the offline measurements (Fig. S4). Organic matter (OM) provided by AMS measurement is correlated with OC from filter measurements (R2=0.53). The slope of 1.75, i.e., the OM / OC ratio, is similar to those found in offshore areas, e.g., 1.8 in the Gulf of Mexico (Russell et al., 2009). An excellent linear correlation was found for sulfate (slope =1.16, R2=0.94), and the correlation for ammonium was moderate (slope =0.80, R2=0.63). No correlation was found for nitrate and chloride. A few values of nitrate from the AMS are higher than those from offline measurements, probably due to the evaporative loss of nitrate. The contribution of organic nitrates remains unknown but might not be negligible. Chloride is underestimated by the AMS, because this instrument partly detects sea salt, which is the main source of the chloride measured during R/V Polarstern cruises. Considering that nitrate and chloride account for only a tiny fraction of the measured particle mass concentration, and the sea salt is estimated specifically, the AMS can still properly detect a major part of the MBL aerosols in our measurements.

Mass closure with MPSS. The total particle mass concentration from the AMS is calculated as the sum of the mass concentrations of default AMS species (i.e., organics, sulfate, nitrate and ammonium) completed by estimated sea salt and eBC. This concentration is compared to the calculated particle mass concentration derived from the particle number size distribution (Fig. S5). Taking sea salt into consideration can slightly improve the correlation between the two different techniques (the slope varies from 0.81 to 0.85; R2 varies from 0.72 to 0.77; Fig. S5). The intercomparisons also demonstrated that a constant CE of 0.7 for AMS measurements could achieve better agreement between AMS and external measurements, whereas the composition-dependent collection efficiency introduced by Middlebrook et al. (2012) does not seem appropriate for R/V Polarstern measurements because of the high acidity and low ammonium nitrate content of the MBL aerosol particles (Huang et al., 2017).

Sea salt estimation. The important primary marine product, sea salt, was estimated using the method from Ovadnevaite et al. (2012), who successfully quantified sea salt aerosol concentration with HR-ToF-AMS in Mace Head, Ireland. The reported method is based on the typical sea salt ion NaCl+ (m/z 57.95) and a scaling factor of 51 derived from laboratory calibration using artificial seawater (Ovadnevaite et al., 2012). In R/V Polarstern measurements, NaCl+ fragment was well identified by the high-resolution mode. Nevertheless, the linear correlation between NaCl+ from the AMS and sea salt derived from PM1 filter samples showed a mild relationship with R2 of 0.38 and yielded a scaling factor of 62 (±6). To be consistent with the literature (Ovadnevaite et al., 2012, 2014; Schmale et al., 2013), the scaling factor of 51 from the reference is applied to the sea salt surrogate (NaCl+) to estimate the sea salt mass concentration in this study. With this factor, the AMS-estimated sea salt concentrations are quite close to those measured by the offline method, showing a slope of 1.01, and the main scattering points are located within the uncertainty range of the Mace Head measurements (Fig. S6). Additionally, the p value of the regression of AMS-derived sea salt with offline results is 0.009 (Spearman's correlation test), indicating that sea salt concentrations from the AMS and offline methods are correlated significantly. The uncertainties of this estimation method in R/V Polarstern measurements probably link to the different sampling (inlet) locations as well as the discrepant analysis methods between the AMS and offline measurements.

An overview of the chemical composition of Atlantic aerosol particles during four cruises is given in Fig. 2, including AMS default species (organics, sulfate, nitrate and ammonium), specially quantified marine products (MSA and sea salt) as well as eBC. Because the ship tracks on latitude are nearly monotonic to the cruise time, particle chemical composition as a function of latitude shows both temporal variation and latitudinal distribution. Air mass back trajectories indicate that the air masses captured during R/V Polarstern cruises either came from the ocean or were influenced by the continents (Europe, Africa or South America), as marked in Fig. 2. Additionally, the statistics of seasonal variation of PM1 chemical composition over the North (>5∘, i.e., 5∘ N) and South (<-5∘, i.e., 5∘ S) Atlantic Ocean are provided in Table 3. Because there is no clear seasonal difference in the chemical composition of measured PM1 in the area near the equator, the average, median, standard deviation and percentage in total were calculated for the region from 5∘ S to 5∘ N in latitude (defined as tropical in this study), regardless of the season, and are also shown in Table 3.

Over all measurements, the total measured submicron particle mass concentration varied over a large range, from 0.22 to 14.15 µg m-3, with a median of 1.83 µg m-3. The reported concentrations sit between the clean marine case (from 0.27 to 1.05 µg m-3) measured at the coastal station Mace Head, Ireland (Ovadnevaite et al., 2014), and the case of mixed marine and continental air masses (from 3.69 to 4.17 µg m-3) over the North Atlantic (Dall'Osto et al., 2010). This difference suggests that the detected aerosols during the R/V Polarstern cruises were from mixed sources, as evidenced by the air mass origins. The average chemical composition of the measured particles indicated that sulfate was the major contributor to the total particle mass concentration (50±13 %), followed by organics (21±9 %), sea salt (12±11 %), ammonium (9±4 %), eBC (5±4 %) and nitrate (3±2 %). The mass concentrations of total particles and individual species showed large variations associated with time and were obviously not distributed normally. Hence, the median, rather than the average, of the mass concentration is used below.

During all four cruises, sulfate dominated the particle mass most of time, showing the highest median mass concentration (0.78 µg m-3) of all six species. Sulfate could contribute up to 85 % of the total particle mass concentration. Seasonal discrepancies were found for sulfate mass concentration (a higher level in spring than in autumn), especially over the South Atlantic (Table 3). A similar seasonal variation of the marine biogenic tracer MSA was also observed (Huang et al., 2017), suggesting that biogenic sources (i.e., phytoplankton) contributed significantly to sulfate (Charlson et al., 1987; Hoffmann et al., 2016). In the tropical Atlantic, sulfate appeared at a slightly lower value than that in spring but was still much higher than that in autumn. Nonetheless, anthropogenic sources of sulfate cannot be ruled out and can even be important, considering that long-range transport has been found to be a major source of sulfate (more important than DMS) over the tropical Pacific (Simpson et al., 2014).

Organics (here including MSA) are the second most abundant species (median: 0.26 µg m-3). Organics did not show pronounced variation associated with seasons but did vary with location. The mass concentration of organics was elevated in air masses under continental influence. During CR2 and CR4, organics dominated the total measured particle mass concentration (up to 59 %) in the region between 0 and 15∘ N. This resulted in substantial organic aerosols over the tropical Atlantic (median: 0.34 µg m-3). Moreover, the non-marine species eBC tended to increase together with organics, especially near the equator, although taking up only a tiny portion of the total particle mass in the whole measurements, which indicates that anthropogenic emissions may be a significant contributor to the organic aerosols over this region of the Atlantic Ocean.

The marine primary product, sea salt, was found to play a minor role (6 %–17 % in different regions) in the measured submicron particles during R/V Polarstern cruises. However, the mass concentration of sea salt could reach 1.63 µg m-3 while taking up to 66 % of the total particle mass loading associated with elevated wind speed. This finding is comparable to the importance of sea salt reported over the northeastern Atlantic (66 %–84 %; Ovadnevaite et al., 2014) and in the sub-Antarctic islands (47 %; Schmale et al., 2013) for high wind speed (>10 m s-1).

The ammonium concentrations did not exhibit a seasonal difference between spring and autumn. The highest median value was found over the tropical Atlantic, followed by the North Atlantic, while the lowest median value was in the Southern Hemisphere. Both continental emissions via long-range transport and marine organisms could be the origin of the ammonium or its precursor ammonia in the MBL (Adams et al., 1999). Nitrate showed low mass concentration (0.04 µg m-3), which is not surprising, because it both has no marine sources and would mainly occur in supermicron particles – nitrate would react with sulfuric acid and escape from submicron particles in the form of gaseous nitric acid, which will relocate to larger particles, e.g., sea salt particles (Saltzman, 2009).

The MOOA factor contributes an average of 16 % to the total OA mass, with a median mass concentration of 0.04 µg m-3. This factor is well correlated with the marine secondary organic aerosol (SOA) tracer MSA (R2=0.83; Fig. 3) and is consequently linked to the oxidation of DMS emitted by phytoplankton (Charlson et al., 1987; Gondwe et al., 2003). One characteristic of the MOOA factor is a high contribution from CxSy+ ions (7 %), which mainly include the MSA identified ions CH3SO2+ (m/z 78.985), CH2SO2+ (m/z 77.978), CH4SO3+ (m/z 95.988), CHS+ (m/z 44.980) and CH2S+ (m/z 45.988; e.g., Huang et al., 2017). This leads to a high S / C ratio (0.030), which is 10 to 30 times higher than that of other factors (Fig. 3). Note that the S / C ratios derived from the PMF analysis tool have to be used with caution because of calculation uncertainties (Aiken et al., 2007), but they can still provide an indication of the significance of sulfur when calculated with the same tool among the factors from the same data set. Even though organic sulfur species play a remarkable role, oxygenated organic fragments are still the major species, accounting for 52 % of MOOA mass loading (CxHyO+ 30 %, CxHyOz+ 22 %; Fig. 4a). These fragment families are followed by hydrocarbon fragments, making up 30 % of the MOOA mass loading. CH3+ is the most abundant fragment of the CxHy+ family, contributing 43 % to the total CxHy+ group mass loading in the MOOA. As a result, MOOA shows the highest H / C ratio (1.73) of all five OA factors, similar to those reported previously: 1.57 (marine OA) in Paris during summer (Crippa et al., 2013) and 1.8 (MOOA) on Bird Island in the sub-Antarctic region (Schmale et al., 2013). This suggests that a high level of hydrocarbon ions (in particular CH3+), together with CxSy+, could be an important characteristic for marine-sourced SOA. The MOOA includes almost all organic fragments observed in MSA, as well as some additional oxygenated fragments, which indicates that the MOOA factor consists not only of MSA but also of other organic components either emitted from the same source or produced via similar processes as MSA. This finding is also proved by the relationship between the mass concentration of MOOA and MSA (slope =2.19, R2=0.83). Significant oxygen-containing ions, such as CO2+ (m/z 43.990) and COOH+ (m/z 44.998), could be related to carboxylic acids, which are important in the composition of secondary marine aerosol (Decesari et al., 2011; Fu et al., 2011). The possible precursors of these carboxylic acids could be isoprene and monoterpenes emitted by marine phytoplankton (Bonsang et al., 1992; Shaw et al., 2010; Yassaa et al., 2008), which can be easily oxidized to form highly oxygenated products such as 2-methylglyceric acid, 2-methyltetrol and pinic acids (Claeys et al., 2010; Fu et al., 2011). It is well known that isoprene and monoterpene oxidation also leads to the formation of organosulfate compounds (Claeys et al., 2010; Fu et al., 2011; Iinuma et al., 2007; Surratt et al., 2007, 2008), which can contribute to the CxSy+ fragments observed in the MOOA factor.

The secondary origin of MOOA was supported by its diurnal cycle (Fig. 4), showing a small but clear elevation in the afternoon, and reaching the maximum (0.05 µg m-3) at 16:00 UTC, almost when global radiation started declining. To focus on the atmospheric behavior of MOOA and exclude the influence of other chemical compositions, an “MOOA-dominating period” was selected for a case study (approximately 57 h from 19:40 UTC, 18 November 2012 to 04:20 UTC, 21 November 2012). As shown in Fig. 6, the MOOA played an important role in the total OA during the selected period, constituting an average of 78 % (up to 100 %) of the total OA mass concentration. The MOOA had strong correlation with MSA with R2 of 0.81, almost the same as the overall coefficient (R2=0.83). Note that the latter one is slightly higher mainly due to greater variation of both MOOA and MSA concentrations during all four cruises. This covariation also resulted in a quite stable MSA / MOOA ratio of 52%±9% during the MOOA-dominating period. NO3, NH4 and eBC showed low mass concentrations (median: 0.04, 0.08 and 0.05 µg m-3, respectively) close to their DLs, indicating a negligible impact from anthropogenic emissions during this period. The diurnal pattern for this specific period (Fig. 6b), with a minimum of 0.11 µg m-3 (MOOA mass concentration) at 07:00 UTC and a maximum of 0.25 µg m-3 at 16:00 UTC, was more noticeable than the average case (Fig. 4b). Similar diurnal cycles are observed for MSA and sulfate, suggesting that MOOA, MSA and sulfate are formed via the same secondary pathway (Charlson et al., 1987; Gondwe et al., 2003; von Glasow and Crutzen, 2004). Model studies found that the DMS is mainly (84 % globally) removed via photooxidation by OH radicals (Kloster et al., 2006) or oxidized by O3 in aqueous-phase reactions to yield a significant amount of MSA (Hoffmann et al., 2016). These findings suggest that DMS oxidation is controlled by photochemical processes and that its products should show a daytime maximum associated with solar radiation (as well as global radiation, as shown in Fig. 6b).

The enhancement of MOOA mass concentration was observed to be independent of the air mass categories, and most MOOA peaks were associated with marine air masses (Fig. 3). The average MOOA mass concentration with marine air masses was 0.05±0.05 µg m-3 with a maximum of 0.31 µg m-3, both higher than those for MOOA from air masses with continental influence, i.e., 0.04±0.03 µg m-3 with a maximum of 0.19 µg m-3. This finding also supports that MOOA originate mainly from the ocean. Related to the DMS oxidation, the MOOA factor shows prominent seasonality and higher contributions to the total OA mass concentration in spring than in autumn; moreover, it shows higher contributions over the South Atlantic than the North Atlantic in spring (Fig. 5). In particular, the MOOA could be the exclusive contributor to the OA at approximately 16∘ S in spring, linked to the high biological activity fueled by the Benguela upwelling system, which is a northward-flowing ocean current along the west coast of southern Africa from Cape Point (Nelson and Hutchings, 1983), bringing up the nutrients from deep cold waters.

Because the MOOA component is successfully traced by MSA, the relationship between the MSA and MOOA should be applicable to estimate DMS-related SOA over the Atlantic Ocean. The correlations between MSA and MOOA in spring, autumn and the tropical season (with unclear seasonal variation, 5∘ N to 5∘ S latitude) are shown in Fig. 7. The scattering points are fitted using linear orthogonal distance regression (ODR). Overall, the correlations between MSA and MOOA in three cases are robust (R2=0.85, 0.53, 0.88), consistent (slope =0.57, 0.56, 0.56), and independent of regions and seasons (for spring and autumn). We therefore infer that the relation between MSA and its concomitant (DMS-related) SOA is roughly stable over the Atlantic Ocean and suggest estimating MOOA mass concentration as the product of the MSA concentration multiplied by a factor of 1.79, which may be useful for a better estimation of marine DMS-related SOA both in field measurements and in models.

The MNOA component showed a unique time series and poor correlation with the other four factors (all R2 were below 0.13). The mass concentration of MNOA varied from below the DL to 0.47 µg m-3 (median 0.03 µg m-3) and contributed 16 % of the OA. The MNOA is characterized by a remarkable contribution from organonitrogen fragments (17 %, Fig. 4), mainly related to CxHyN+, such as C2H6N+ (accounting for 5.3 % of total m/z 44 intensity), C2H7N+ (2.6 % of m/z 45) and CH4N+ (2.7 % of m/z 30), which are at least 1 order of magnitude higher (intensity fraction in the located m/z) than in other factors. Nevertheless, this factor is still dominated by oxygenated fragments including CxHyO+ (30 %) and CxHyOz+ (28 %, Fig. 4), and the diurnal variation of MNOA shows a broad afternoon peak with a maximum at 16:00 UTC (Fig. 4), similar to that of the amine-related secondary factor in New York City showing a diurnal pattern with maximum at 12:00 EST (Eastern Standard Time; Sun et al., 2011). Both of these findings may indicate that secondary formation could be one of the possible pathways for MNOA generation.

N-containing OA (NOA) factors from PMF analysis have been found in many studies and can be related to various origins highly dependent on local sources, such as gentoo penguin hatching activities (Schmale et al., 2013) and local primary (industrial) emissions (Aiken et al., 2009). During the R/V Polarstern campaign, the MNOA is correlated with neither eBC (R2=0.17) nor NO3 (R2=0.06), excluding the possibility of combustion and anthropogenic (continental) sources. Meanwhile, high similarity (R2=0.70) of the mass spectral profile is found between the R/V Polarstern MNOA and the NOA in Sun et al. (2011), who attributed that factor in New York City to marine and local industry emissions and stressed the possibility of gas-to-particle conversion via the reactions of acidic gases and the gaseous amines. It is worth noting that the characteristic CxHyN+ fragments are actually different in these two studies: C3H8N+ (m/z 58) and C2H4N+ (m/z 42) dominated the CxHyN+ group in the New York study, while C2H6N+ (m/z 44), C2H7N+ (m/z 45) and CH4N+ (m/z 30) played a more important role in R/V Polarstern measurements. These C1–C3 CxHyN+ fragments may originate from low-molecular-weight aliphatic amines such as methylamine (CH5N), dimethylamine (C2H7N) and trimethylamine (C3H9N), which have been regarded as the main species in organonitrogen fragments from marine sources (Gibb et al., 1999; Müller et al., 2009) as well as important biogenic SOA precursors over the ocean (Dall'Osto et al., 2012; Facchini et al., 2008; Müller et al., 2009). In addition, a large number of CxHyN+ fragments results in the highest N / C atom ratio (0.124) among all five OA factors, and this N / C ratio is close to the one used in a global 3-D chemistry-transport model by Kanakidou et al. (2012) to trace the organonitrogen from ocean sources (0.15), while it is different from the ratio for biomass burning (BB) and anthropogenic sources (0.3). This may also support the idea that the MNOA originates from the ocean, likely transformed from phytoplankton-emitted gaseous amines.

The MNOA shows no correlation with marine secondary product tracer MSA (R2=0.01). This absence of correlation has already been reported in several marine measurements of particulate amines with biogenic origins (Facchini et al., 2008; Miyazaki et al., 2011; Müller et al., 2009), although acid–base reactions were considered a key process of gas-to-particle conversion for amines over the ocean (Facchini et al., 2008; Sun et al., 2011). Given that gaseous amines can enter the particle phase in different ways (Ge et al., 2011), other pathways, including oxidation, nucleation and condensation, should also be considered.

During each cruise, the MNOA had a low mass concentration with no clear seasonality. However, its high significance can be found when the total OA mass concentration was extremely low, generally below 0.5 µg m-3 (Fig. 5), e.g., 20∘ S to 20∘ N in CR1 and CR3. It is interesting that during CR3, the MNOA contributed to nearly half of the total OA between the equator and 20∘ S in association with very clean air masses from the open ocean, which may confirm its marine origin. We also noticed that the MNOA had a similar variation trend during all four cruises, slowly increasing and then decreasing broadly between 20∘ S and 20∘ N. This variation was generally coincident with the change in the water temperature (Fig. S10). The covariation might be explained by the biogenic sources of the MNOA – temperature could positively affect metabolism rate, including N excretion of marine microorganisms (Ikeda, 2014).

The only true primary factor identified, MHOA, is related to marine primary emissions. It is characterized by a high contribution of hydrocarbon ions (64 % of total factor mass concentration, Fig. 4), which is usually considered a feature of primary emissions. One-fourth of the MHOA mass was identified as oxygenated organic compounds (CxHyO+, CxHyOz+). This finding is not surprising, because carboxylic acids can also be supplied via the primary pathway and have been identified as a distinct type of primary marine organic matter (Hawkins and Russell, 2010). A small but pronounced portion (5 %) of CxHyN+ fragments was observed in total MHOA mass loadings, consistent with the findings in previous measurements that amines also existed in primary marine OA (Frossard et al., 2014; Quinn et al., 2015).

The time series of MHOA mass concentration was spiky and varied from < DL to 0.67 µg m-3, with a median of 0.04 µg m-3. It contributed 19 % to the total OA mass concentration, comparable to two marine secondary factors (MOOA 16 %, MNOA 16 %). No clear diurnal pattern was found for the MHOA, indicating that the contribution of photooxidation could possibly be ignored for this factor.

Considering that the ship contamination periods have been eliminated and the MHOA was not correlated with eBC (R2=0.04), fresh ship emissions were unlikely to have been a contributor to this primary factor. The primary emissions from the ocean were considered a main source of the MHOA because of the similarity (R2=0.61) in the mass spectral profile between this factor and primary marine OA during high biological activity at Mace Head, Ireland (Ovadnevaite et al., 2011). For the entire measurement period, the MHOA was not correlated with sea salt (R2=0.01) or the wind speed (R2=0.01), even though a similar variation trend (but no strong linear correlation, R2<0.4) was found in several periods, such as the southern parts (from 33 to 0∘ S) of CR2 and CR4 (Fig. S11). This trend does not conflict with the speculation that MHOA is related to marine primary emissions, because the mass fraction of organics in the sea spray aerosol was found to be size-dependent, increasing with decreasing particle size (Gantt et al., 2011; Quinn et al., 2015). The enrichment factor of organic compounds, i.e., the ratio between the organic carbon in sea spray aerosols and that in seawater, is also largely influenced by particle size (Quinn et al., 2015). In addition, the transfer of organic matter from seawater to the particles is chemoselective and more complicated than it is for inorganic sea salt (Schmitt-Kopplin et al., 2012). During all four cruises, the MHOA factor did not show any location dependence but played a more prominent role in regions with extremely low OA mass concentrations, e.g., the southern part of CR1 and the whole period of CR3 (Fig. 5). This finding could support the hypothesis that the MHOA originates from the ocean.

Although measured over the ocean, 19 % of the organic aerosol mass was contributed by continental emissions from human activities, presented by the Anth-OOA factor with a similar median (0.04 µg m-3) but the highest maximum (2.70 µg m-3) compared to the other OA factors. The Anth-OOA factor was identified by the anthropogenic tracer NO3 regarding time series (R2=0.52; Fig. 3) and resemblance of mass spectral profiles to several reported continental OOA factors (Table 4). The Anth-OOA in this study was in excellent agreement (R2=0.98) with a factor of continental organics observed over the central Arctic Ocean (Chang et al., 2011). It was also similar (R2=0.67) to the average OOA factor based on nine urban measurements (Ng et al., 2011). As shown in Fig. 4, the Anth-OOA factor was dominated by CxHyO+ fragments (contributing 40 % to the total factor mass), followed by CxHyOz+ (26 %). Significant contributions were observed from CO2+ (m/z 44, 21 % of total Anth-OOA mass loading) and C2H3O+ (m/z 43, 6 %), comparable to values in previous studies in urban and rural areas (e.g., Ng et al., 2011; Poulain et al., 2011a). Another carboxylic acid fragment, CO+ (m/z 28, 21 % of total Anth-OOA mass loading), was estimated according to the CO2+ ion; thus CO+ is not specifically discussed in the present work. There was no clear diurnal pattern for the Anth-OOA factor (Fig. 4). The Anth-OOA mass concentration dropped in the early morning from 05:00 UTC, reached a minimum at ∼ 10:00 UTC, then stayed stable until 00:00 UTC. This change may be explained by the rising mixing layer in the morning, which dilutes the particle concentration, and/or increasing temperature after sunrise, and drives volatile species from particles into the gas phase. Hence, the Anth-OOA is unlikely to be contributed to by local photochemical formation but rather by continental outflow from anthropogenic activities via long-range transport.

Significant elevation of Anth-OOA mass concentrations (up to 2.70 µg m-3) was associated mainly with continental air masses (Fig. 3). The median value of the Anth-OOA mass concentration in marine air masses was 0.03 µg m-3 and higher (0.08 µg m-3) during periods influenced by continental air masses. Additionally, the contribution of Anth-OOA to the total OA significantly increased, even when close to Africa and Europe (Fig. 8). This confirms that the continental outflow is the main source of the Anth-OOA factor.

Of all five factors, the Comb-OOA is the only one correlated with eBC (R2=0.69; Fig. 3), pointing to combustion sources. It was the most abundant component of the measured OA on average, contributing 30 % of the total OA mass. The median Comb-OOA mass concentration was 0.07 µg m-3 (< DL to 1.38 µg m-3). Similar to the Anth-OOA, most Comb-OOA peaks appeared together with continental air masses (Fig. 3), indicating a non-marine source. One significant characteristic of the Comb-OOA factor was its highly oxygenated level. Oxygenated fragments including CxHyO+ and CxHyOz+ accounted for 73 % of the total Comb-OOA mass concentration (Fig. 4). This factor contained remarkably high CO2+ (the ratio of CO2+ to total Comb-OOA =f44=32 %, indicating a possible origin from organic acids. As a result, the Comb-OOA exhibited the highest O / C ratio (O / C = 1.35) and lowest H / C ratio (H / C =0.94) of all OA factors. The Comb-OOA was also contributed by pronounced N-containing fragments: CxHyN+ (4 %) and CxHyOzNw+ (2 %). Despite the small contribution to the Comb-OOA mass concentration, CxHyOzNw+ fragments were primarily distributed in this factor (46 % of total CxHyOzNw+ intensity).

The Comb-OOA factor was attributed to aged aerosol particles from combustion emissions, mainly because of its highly oxygenated level and covariance with combustion tracer eBC. Additionally, this factor had a similar mass spectrum to the OA below clouds measured by a flight impacted by ship emissions (R2=0.73; Coggon et al., 2012) and the photooxidation products of diluted diesel generator exhaust (R2=0.46; Sage et al., 2008), rather than any fresh combustion emissions, e.g., the fresh ship emissions factor (R2=0.00; Chang et al., 2011), the average BB organic aerosols (BBOA; R2=0.09; Ng et al., 2011) and the average HOA factor (R2=0.01; Ng et al., 2011). This finding is reasonable, because the influence of direct ship emissions from the R/V Polarstern has been removed. Sage et al. (2008) found that SOA could be quickly formed from fresh diesel exhausts, and the aged products were increasingly oxidized with time, resembling the profile of atmospheric aged OA within a few hours of oxidation. Given that, most of the time, the R/V Polarstern traveled along major shipping routes (Fig. S12), a possible source of the Comb-OOA could be the transported ship exhausts from other ships.

The BB emissions via long-range transport are likely to be the other important source of the Comb-OOA, indicated by a good correlation with CxHyOzNw+ fragments (from AMS, R2=0.83) and potassium (K+, from offline measurements, R2=0.61), as shown in Fig. 8. The CxHyOzNw+ ions have been related to the photooxidation product of m-cresol, a typical wood burning emission (Iinuma et al., 2010; Poulain et al., 2011a), and potassium is commonly regarded as an unreactive tracer of BB emissions. Nevertheless, some frequently used BB tracers, C2H4O2+ and C3H5O2+ (Capes et al., 2008; Cubison et al., 2011), were extremely low in our Comb-OOA factor (the ratio of C2H4O2+ to total Comb-OOA =f60=0.06 %, the ratio of C3H5O2+ to total Comb-OOA =f73=0 %), even though the f60 increased to 0.18 % during the high-Comb-OOA period (the factor mass concentration and mass fraction in total OA reached 1.02 µg m-3 and 47 %). The absence of these BB tracers may be caused by decay in the aging process during long-range transport. Fresh BB emissions can be quickly photochemically aged (in several hours), resulting in the formation of new OA with a higher degree of oxygenation and a significant decrease in BB tracers as well as saturated hydrocarbon compounds. This change has been observed in both laboratory (Bertrand et al., 2018; Grieshop et al., 2009) and field measurements (Capes et al., 2008; DeCarlo et al., 2010).

The significance of the Comb-OOA, i.e., the aged particles from combustion emissions, are dependent on regions and seasons (Fig. 5). More than 50 % of the total OA was contributed by aged particles from combustion emissions, especially during the two November cruises (CR2 and CR4) between 15∘ N and 15∘ S, when the ship was near western and central Africa. It is well known that Africa is the single biggest continental source of BB emissions with strong seasonality (Roberts et al., 2009). In western and central Africa, more intense fires occurred in November than in April and May according to the fire maps (Fig. S13), consistent with higher mass concentrations of Comb-OOA in CR2 and CR4 than during the other two cruises. In addition, the mass concentration of the Comb-OOA increased when the ship was near the English Channel, where marine traffic is very busy (Fig. S12).