The seawater samples were taken from a fishing boat, launching from Baía das Gatas, São Vicente. A glass plate with a sampling area of 2500 cm2 was vertically immersed in the seawater and then slowly drawn upwards to take the SML. The surface films adhered to the surface of the glass plate and were removed with Teflon wipers directly into a bottle. This glass plate approach is described in detail by Cunliffe (2014). The ULW was sampled at a depth of 1 m into a plastic bottle fitted on a telescopic rod. To avoid influences from the SML, the bottles were opened underwater at the intended sampling depth. All seawater samples were stored in plastic bottles at -20 ∘C until the time of analysis. All materials for the seawater sampling were pre-cleaned with a 10 % HCl solution and high-purity water.

Size-segregated aerosol particles were sampled using five-stage Berner-type impactors (Hauke, Gmunden, Austria) at the top of a 30 m sampling tower at the CVAO since this location best represents the conditions above the ocean in accordance with previous studies. The internal boundary layer (IBL), which can form when air passes a surface with changing roughness (i.e., the transfer from open water to island), is mainly beneath 30 m (Niedermeier et al., 2014). Moreover, aerosol particles were sampled on the top of MV (744 m a.s.l.). To avoid the condensation of atmospheric water in the aerosol particle sampling substrate, a conditioning unit consisting of a 3 m long tube was installed between the impactor inlet and the sampling unit. By heating the sampled air, the high relative humidity of the ambient air before collecting the aerosol particles was set to 75 %–80 %. The temperature difference between the ambient air at the impactor inlet and the sampled air after the conditioning unit was below 9 K (van Pinxteren et al., 2020). The Berner impactors were operated with a flow rate of 75 L min-1 for 24 h, and pre-baked aluminum foils (350 ∘C for 2 h) were used as substrate material. The five stage Berner impactor includes stage 1 (B1): 0.05–0.14 µm; stage 2 (B2): 0.14–0.42 µm; stage 3 (B3): 0.42–1.2 µm; stage 4 (B4): 1.2–3.5 µm and stage 5 (B5): 3.5–10 µm. When it comes to the segregated aerosol particle samples, our study differentiates between the ones of submicron size (B1, B2, B3), the ones of supermicron size (B4, B5) and the ones of PM10 (B1–5). After the sampling, the aluminum foils were stored in aluminum boxes at -20 ∘C until the time of analysis. It needs to be pointed out that the Berner impactors ran continuously; thus, the impactor on MV sampled aerosol particles also during cloud events. However, due to the preconditioning unit, the cloud droplets were efficiently removed before the aerosol particles were collected on the aluminum foils.

At the MV station, an acrylic glass Caltech Active Strand Cloud water Collector version 2 (CASCC2), according to Demoz et al. (1996), was used to sample cloud water. During a “cloud event”, the bottles were changed every 2–3 h, whereas on the other days, the sampling time was, e.g., overnight (every 12 h). For each sampling, the Teflon rods used were pre-cleaned with a 10 % HCl solution followed by high-purity water. The liquid water content (LWC) of the cloud was measured continuously by a particle volume monitor (PVM-100, Gerber Scientific, USA). Finally, the collected cloud water was sampled in pre-cleaned plastic bottles and stored at -20 ∘C until the time of analysis.

For the dissolved organic carbon and total dissolved nitrogen (DOC/TDN) content and the analysis of inorganic ions, the seawater samples were first filtered (0.45 µm syringe filter) and then quantified with a TOC-VCPH analyzer (Shimadzu, Japan) or ion chromatography (ICS3000, Dionex, Sunnyvale, CA, USA), as described in van Pinxteren et al. (2017). As the seawater samples must first undergo a desalination step for the FAA analysis, 32 mL (SML samples) or 48 mL (ULW samples) were desalinated using Dionex™ OnGuard™ II Ag/H cartridges (Thermo Fisher Scientific™, Waltham, Massachusetts, USA). The volume of the desalinated samples was reduced to several microliters using a vacuum concentrator at T=30 ∘C (miVac sample Duo, GeneVac Ltd., Ipswich, United Kingdom) with a recovery rate of >86 %. The 0.2 µm syringe filters (Acrodisc GHP, 25 mm; Pall Corporation, New York, USA) enabled the filtering of the enriched samples; then, a derivatization was performed with the AccQ-Tag™ precolumn derivatization method (Waters, Eschborn, Germany). Besides, the FAA analysis includes the determination of glycine (Gly), L-alanine (Ala), L-serine (Ser), L-glutamic acid (Glu), L-threonine (Thr), L-proline (Pro), L-tyrosine (Tyr), L-valine (Val), L-phenylalanine (Phe), L-aspartic acid (Asp), L-isoleucine (Ile), L-leucine (Leu), L-methionine (Met), L-glutamine (Gln) and γ-aminobutyric acid (GABA) (purity ≥ 99 %; Sigma-Aldrich, St. Louis, Missouri, USA). Ultra-high performance liquid chromatography with electrospray ionization and Orbitrap mass spectrometry (UHPLC/ESI-Orbitrap-MS) performed the analytical measurements of the derivatized FAAs. The UHPLC system (Vanquish Horizon UHPLC system, Thermo Fisher Scientific™, Waltham, Massachusetts, USA) was coupled with an ESI-Orbitrap mass spectrometer (Q Exactive™ plus, Thermo Fisher Scientific™, Waltham, Massachusetts, USA). The samples were subsequently separated through an ACQUITY UPLC^® HSS T3 column (Waters, Eschborn, Germany) with dimensions of 1.8 µm (particle size), 2.1 mm (inner diameter) and 100 mm (length) at a constant temperature of 30 ∘C and a detection in positive mode. The eluent composition consisted of (A) 0.2 vol % acetic acid in high-purity water (Millipore Elix 3 and Element A10, Merck Millipore, Darmstadt, Germany) and (B) acetonitrile (Optima^® LC/MS Grade, Fisher Scientific, Hampton, New Hampshire, USA). The flow rate of the eluent was 0.3 mL min-1. The eluent gradient program was 5 % B for 1 min, from 5 % B to 100 % B in 16 min, 100 % B for 2 min constant and from 100 % B to 5 % B in 0.1 min, and 5 % B was then kept constant for 3.9 min. This analytical procedure can be used for amines, too, as described in van Pinxteren et al. (2019). The FAA concentrations were determined via external calibration. Since no chiral column was used in the UHPLC separation, we cannot differentiate between L- and D-amino acids in our ambient samples. Each seawater sample was measured as a duplicate with a relative standard deviation <10 % and under consideration of the blank samples for seawater. They consist of high-purity water, which was filled in pre-cleaned plastic bottles and handled the same as the seawater samples. The limit of quantification (LOQ) of the individual FAAs in seawater samples is in good agreement with the FAA analysis of seawater samples (e.g., Kuznetsova et al., 2004) and are listed in Table S1. The LOQs were mostly below 10 nmol L-1; however, GABA and Met exhibited LOQs of 24.2 and 16.8 nmol L-1, respectively (due to high blank values). A quantification of some FAAs in seawater, mainly in the ULW with its generally lower FAA concentrations compared to the SML, is therefore partly limited.

For analyzing the size-segregated aerosol particle samples, the substrate material of each stage was extracted in 3 mL of high-purity water (Millipore Elix 3 and Element A10, Merck Millipore, Darmstadt, Germany). The aqueous particle extracts were divided into aliquots for the analysis of water-soluble organic carbon and total dissolved nitrogen (WSOC/TDN), inorganic ions, and amino acids. The aliquots for WSOC/TDN were first filtered with a 0.45 µm syringe filter and then determined by a TOC-VCPH analyzer (Shimadzu, Japan), as described in van Pinxteren et al. (2012). For the analysis of inorganic ions, the aliquots (250 µL) were filtered (0.45 µm syringe filter) and investigated using ion chromatography (ICS3000, Dionex, Sunnyvale, CA, USA), as outlined in Müller et al. (2010). The aliquot (1.5 mL) of the aqueous particle extracts for FAA analysis was reduced to several microliters with a vacuum concentrator at T=30 ∘C (miVac sample Duo, GeneVac Ltd., Ipswich, United Kingdom), filtered using 0.2 µm syringe filters, and derivatized and analyzed using the UHPLC/ESI-Orbitrap-MS method, as explained in Sect. 2.2.1 for seawater samples. FAA concentrations were calculated via external calibration; each sample was measured in duplicate with a relative standard deviation <10 % and under consideration of field blanks. For generating field blanks, pre-baked aluminum foils without active sampling were cut and prepared the same as field samples, including extraction and measurements for WSOC/TDN, inorganic ions and amino acid analysis. All values presented here for aerosol particle samples are field blank corrected. The LOQs of the individual FAAs in aerosol particle samples are listed in Table S1 and are in good agreement with the sensitivity of other analytical methods for FAAs in aerosol particles (e.g., Matsumoto and Uematsu, 2005). Although a variance in LOQs between the individual FAAs is apparent, FAAs with relatively high LOQs (39.5 pg m-3) such as Ala, GABA, and Asp in submicron and supermicron aerosol particles could be quantified (as discussed in Sect. 3.2 and 3.4).

The analysis of mineral dust tracers on nucleopore foils sampled with the Berner impactor was performed with the total reflection X-ray fluorescence S2 PICOFOX (Bruker AXS, Berlin, Germany) equipped with a Mo-X-ray source on polished quartz substrates, as can be seen in Fomba et al. (2013). The particulate mass (PM) of the aerosol particle samples was determined by weighing the filter before and after sampling (van Pinxteren et al., 2015). Back trajectory analyses provided information regarding the origins of air masses. Back trajectories of 7 d were calculated on an hourly basis within the sampling intervals using the NOAA HYSPLIT model (HYbrid Single-Particle Lagrangian Integrated Trajectory; http://www.arl.noaa.gov/ready/hysplit4.html (last access: 21 December 2020), 26 November 2016) in the ensemble mode at an arrival height of 500 ± 200 m (van Pinxteren et al., 2010); van Pinxteren et al. (2020) provide more details. The calculated backward trajectories are representative of both aerosol particle sampling stations (CVAO and MV).

The cloud water samples were processed the same way as seawater samples for the analysis of DOC/TDN and inorganic ions (Sect. 2.2.1). For the amino acid analysis, the volume of cloud water samples (44 mL) was reduced to several milliliters using a vacuum concentrator at T=30 ∘C (miVac sample Duo, GeneVac Ltd., Ipswich, United Kingdom). After the filtration with 0.2 µm syringe filters (Acrodisc-GHP; 25 mm, Pall Corporation, New York, USA), an aliquot of the prepared cloud water was derivatized based on the AccQ-Tag™ precolumn derivatization method (Waters, Eschborn, Germany). The analytical measurements of the derivatized FAAs were performed with UHPLC/ESI-Orbitrap-MS (Sect. 2.2.1). The cloud water samples were measured as duplicates with a relative standard deviation <10 %. Via external calibration, the amino acid concentrations under the consideration of the cloud water blanks were calculated. The blank samples of cloud water were generated by rinsing the pre-cleaned Teflon rods with high-purity water after their installation in the cloud water sampler. Then, the blank samples were handled the same way as the field cloud water samples, including the derivatization and analytical separation, as described in Sect. 2.2.1. Overall, the LOQs of the individual FAAs in cloud water samples are in good agreement with the reported sensitivity of the FAA analysis in cloud water (Bianco et al., 2016) and listed in Table S1. Since the LOQs of the FAAs in cloud water are below 0.3 ng m-3 and often below 0.06 ng m-3, a limitation of the FAA composition in cloud water due to the LOQs is rather unlikely despite the variance of FAA concentrations (11.2–489.9 ng m-3) in cloud water (Sect. 3.3). To calculate the atmospheric concentration of FAAs in cloud water, the measured concentrations were multiplied by the measured liquid water content (LWC) of the clouds, as Fomba et al. (2015) applied previously.

The enrichment factor in the SML (EFSML) was calculated by dividing the concentration of the analyte in the SML with the concentration of the analyte in the ULW using the following Eq. (1): 1EFSML=c(analyte)SMLc(analyte)ULW. Accordingly, both an enrichment in the SML with EFSML>1 and a depletion in the SML with EFSML<1 are indicated.

The FAA concentration in the ULW was assumed to be based on the concentration (LOQ/2) of individual amino acids for seawater samples from the same campaign day when individual FAAs could be quantified in the SML samples but not in the corresponding ULW ones due to FAA values below the LOQs (listed in Table S1). For the calculation of this estimated EFSML, specially marked in the following, the concentration of 25.2 nmol L-1 was used for c(analyte)ULW in Eq. (1).

To calculate the enrichment factor of the individual analytes in different matrices (M), the concentration of the analyte in matrix 1 (M1) relative to the sodium (Na+) concentration in M1 was divided by the analyte concentration in matrix 2 (M2) relative to the Na+ concentration in M2 using Eq. (2): 2EFM1=c(analyte)M1/cNa+M1c(analyte)M2/cNa+M2. The aerosol enrichment factor (EFaer) was calculated in each of the five Berner stages (Bx with x=1–5 as M1) using the respective analyte or Na+ concentration in relation to the SML or the ULW as M2. For this purpose, the aerosol particle concentrations, typically sampled at a 24 h interval, were combined with SML and ULW concentrations which had been collected during the aerosol particle sampling period. The analyte concentration in each size class of size-segregated aerosol particle samples (B1–5) was combined with the analyte concentration in SML and ULW. The calculation of the EFaer was limited to the availability of data in both matrices – size-segregated aerosol particles and SML and/or ULW samples. The EFaer could only be calculated if both the analyte concentration and the sodium concentration could be quantified in the size-segregated aerosol particles and the corresponding SML and/or ULW samples. To calculate the enrichment factor in cloud water (EFCW), the concentration of the analyte and of Na+ in the cloud water was considered as M1 and those of the SML or the ULW as M2. The determination of EFaer was possible for n=3 samples on the basis of both SML and ULW. The EFCW could only be determined for n=1 sample though, based on the SML and/or ULW measurements. Section 3.4 discusses both the EFaer and the EFCW in more detail.

Free amino acids (FAAs) were measured in the seawater as a source region of FAAs in primary marine aerosol particles. Figure 1b shows the measured ∑FAA concentration in the SML and the ULW samples, together with their enrichment factor EFSML (Eq. 1).

∑FAA included all investigated amino acids (listed in Sect. 2.2.1) except for Met, Gln and GABA. As discussed in Sect. 2.2.1, GABA and Met have the highest LOQs of the analytical method used here, which may be one reason why these two analytes could not be quantified in the seawater samples (ULW and SML).

Looking at the percentage composition within the ULW (10.1 % hydrophilic, 57.0 % neutral, 32.8 % hydrophobic) and the SML (10.6 % hydrophilic, 61.7 % neutral, 27.7 % hydrophobic), the values are similar to each other. However, the concentration of ∑FAA varied between 0.01 and 1.10 µmol L-1 in the ULW and between 0.13 and 3.64 µmol L-1 in the SML. Interestingly, in the second half of the campaign, the ∑FAA concentration was higher than in the first part. Previous studies in different oceanic areas (Kuznetsova and Lee, 2002; Kuznetsova et al., 2004; Reinthaler et al., 2008; van Pinxteren et al., 2012; Engel and Galgani, 2016) have already reported a generally strong variability in ∑FAA concentrations, especially in the SML. Reinthaler et al. (2008) concluded that the SML in the open ocean is a highly variable environment with high concentrations of dissolved FAAs and their high enrichment in the SML is without clear diel variations in their concentrations. Nevertheless, the variance of the ∑FAA concentrations in the SML and ULW observed here could neither be explained by the variance of DOC or TDN values nor by wind speed and chl a concentrations (see Fig. 1, Tables S2, S5) since no correlation between these parameters and the concentration or enrichment of FAAs was found. This is consistent with other publications which observed that the amino acid concentration in seawater is not related to environmental parameters such as wind, humidity and light (Kuznetsova et al., 2004; van Pinxteren et al., 2012). The results of the individual FAA concentrations in seawater (ULW, SML) and their EFSML, listed in Table S3, show clear differences between the individual amino acids and the amino acid classes. The most highly enriched amino acids in the SML are the neutral ones with values of up to 203 compared to the hydrophilic (EFSML: 2–98) and hydrophobic (EFSML: 1–96) amino acids. This may be related to the fact that Ser, Thr and Gly, as part of the neutral amino acids, are main components of cell wall proteins (Hecky et al., 1973). The direct release of FAAs through cell lysis and the associated destruction of the cell wall can thus explain the increased enrichment of neutral amino acids in the SML. Our study confirmed that the SML is often non-uniformly enriched with FAAs as outlined from previous observations (Kuznetsova and Lee, 2002; Reinthaler et al., 2008; van Pinxteren et al., 2012; Engel and Galgani, 2016). Different factors, such as the transport of FAAs from the ULW to the SML, the in situ production by an extracellular hydrolysis of CAA or a direct release of FAAs by cell lysis, probably cause the observed enrichment of FAAs in the SML. Kuznetsova and Lee (2002) showed that the rapid extracellular hydrolysis of CAAs in the SML was not the cause of the non-uniform enrichment in the SML. Moreover, they suggested that the intracellular pools of organisms rich in dissolved free and combined amino acids (DFAAs and DCAAs) compared to seawater can be leached out by stressed microorganisms, resulting in the release of DFAAs which in turn influences the pools of both DFAAs and DCAAs in seawater. Based on previous studies, the transportation and releasing mechanisms seem most likely to be the reasons for the observed enrichment of FAAs. However, further experiments are required to finally elucidate the most important drivers causing the enrichment.

Altogether, it can be concluded that there is some variability in the FAA concentration in the SML and in the ULW with a clear trend of its strong enrichment in the SML. The fact that the FAA concentrations were in accordance with the ones measured at the same location in November 2013 (0.64 µmol L-1; Table S4) supports the suggestion that the FAA concentrations reported here can be considered representative of the NATR region as part of the North Atlantic Ocean. These concentrations are generally similar when comparing them to FAA concentrations in other marine regions (Kuznetsova and Lee, 2002; Reinthaler et al., 2008). Reinthaler et al. (2008) considered concentrations of dissolved FAAs of 0.02–0.13 µmol L-1 (ULW) and of 0.43–11.58 µmol L-1 (SML) in the subtropical Atlantic Ocean, as well as values of 0.07–0.60 µmol L-1 (ULW) and of 0.77–3.76 µmol L-1 (SML) in the western Mediterranean Sea. Consequently, the FAA concentrations in the NATR region, with its very low surface chlorophyll and a greater annual than seasonal variability are on the same order of magnitude compared to other marine regions (i.e., subtropical Atlantic and western Mediterranean Sea; Reinthaler et al., 2008).

DOC and TDN concentrations and their enrichment in the SML (EFSML) are listed in Table S5. The contribution of ∑FAA to DOC or TDN in seawater has been calculated (taking into account the carbon and nitrogen content of the amino acids; Table S6) and is also listed in Table S5. The carbon content of ∑FAA contributed to the DOC with values between 0.1 % and 7.6 % in the seawater samples and a median of 2.4 % (n=17), differing between 2.8 % (n=11) in the SML and 1.8 % (n=6) in the ULW samples. Looking at the nitrogen content from ∑FAA to TDN in the seawater samples, 0.1 %–42.4 % of the TDN consisted of ∑FAA with a median of 8.3 % (n=18). In the SML, ∑FAA contributed on average 11.9 % (n=11) to TDN, whereas they contributed in the ULW 3.2 % (n=7). The observed daily variations within the contribution of ∑FAA to DOC/TDN, result from the daily variations in ∑FAA concentrations in seawater (Fig. 1) and of DOC/TDN (Table S5). In the SML of the Atlantic Ocean and the western Mediterranean Sea, the DFAAs contributed ∼ 12 % of the DOC and ∼ 30 % of the dissolved organic nitrogen (DON) (Reinthaler et al., 2008). Our results regarding the contribution to DOC were of the same order of magnitude but slightly lower than those of Reinthaler et al. (2008).

To obtain a first indication of the particle origin that might help to explain the differences in the particle composition concerning amino acids, the particles were associated with the origin of the air masses and with marine and dust tracers. Overall, the CVAO station experienced northeasterly trade winds during this campaign, which are typical for this season within this region (Fomba et al., 2014; van Pinxteren et al., 2020). According to physical and chemical specifications such as the air mass origins, particulate MSA concentrations and MSA/sulfate ratios, as well as particulate mass concentrations of dust tracers, aerosol particles predominantly of marine origin with low to medium dust influences were observed. The dust and marine tracers of the aerosol particles considered here are discussed in more detail in the Supplement (Table S8 and in “aerosol particles: dust and marine tracers”). Further information on the classification of the air masses and distinct concentrations of dust tracers are given in the overview paper of this campaign (van Pinxteren et al., 2020).

The lower panel of Fig. 2b shows the atmospheric concentration of FAAs in each Berner stage at the CVAO, whereas the upper panel represents the concentration in the submicron, the supermicron and PM aerosol particle size ranges. In the submicron aerosol particles, the concentration of ∑FAA was between 1.3 ng m-3 (1 October 2017) and 6.3 ng m-3 (7 October 2017). While the concentration ∑FAA varied between 0.2 ng m-3 (6 October 2017) and 1.4 ng m-3 (22 September 2017) in the supermicron size range, the highest atmospheric concentrations of ∑FAA were found in the submicron aerosol particles (mean of 3.2 ng m-3) compared to the supermicron ones (mean of 0.6 ng m-3). Daily variations in the ∑FAA content of the investigated size-segregated aerosol particle samples were observed; the ∑FAA tended to increase slightly throughout the campaign. OM parameterization studies showed that wind speed and chl a concentrations were the most important parameters for the regulation of the OM production in sea spray aerosol particles (Gantt et al., 2011; Rinaldi et al., 2013; van Pinxteren et al., 2017). Correlations between the ∑FAA concentrations of the size-segregated aerosol particles (considered as submicron, supermicron and PM10) and the wind speed were not observed for data reported here (Fig. 2, Table S2). However, the available wind speed and wind direction data represented an average value of 24 h. Therefore, short, pronounced changes in the wind speed that might have affected the amino acid transfer would not have been visible in the averaged wind speed value. The major source of bubbles is whitecaps or breaking waves that occur when the wind speed exceeds 3–4 m s-1 (Blanchard, 1975), which was continuously reported during the campaign. Hence, the high wind speeds, together with the constantly observed breaking waves, indicated that the wind intensity in this region might be consistently sufficient to transfer the amino acids from the ocean into the atmosphere. No significant correlation could be observed between the ∑FAA concentration of size-segregated aerosol particle samples (submicron, supermicron and PM10) and the chl a concentration in seawater. Nevertheless, the increasing chl a concentration throughout the campaign (Fig. 2, Table S2) could be a reason for the slight increase in the concentrations of ∑FAA in seawater and in submicron aerosol particles, indicating a possible connection between ocean and atmosphere, e.g., the transfer of amino acids from the ocean into the atmosphere.

Overall, the concentrations reported here agree well with other FAA studies on marine aerosol particles. Matsumoto and Uematsu (2005) found averaged total concentrations of dissolved FAAs of 4.5 ng m-3 in aerosol particles (average of <2.5 and >2.5 µm) in the western North Pacific Ocean. Moreover, Wedyan and Preston (2008) observed an average concentration of dissolved FAA of 2.5 ng m-3 on total suspended particles (TSPs) during a transect ship cruise in the Atlantic Ocean. For Antarctic aerosol particles, the observed mean total FAA concentration in size-segregated aerosol particle samples (<0.49–10 µm) at the Mario Zucchelli Station was 4.6 ng m-3 (Barbaro et al., 2015). Hence, regarding the sum of FAAs, a striking similarity was found between FAA concentrations in different parts of the ocean that probably underlay different influences (e.g., pristine region in the Southern Ocean, continental-influenced aerosol particles in the North Pacific Ocean).

∑FAA included all investigated amino acids (listed in Sect. 2.2.1) except for Met and Gln, analytes which were not detected in the size-segregated aerosol particle samples. The most abundant FAA was Gly, which was consistently found in submicron and supermicron aerosol particles, followed by Ala and Ser. However, towards the end of the campaign (4–7 October 2017), a high contribution of the hydrophilic FAAs GABA and Asp was detected (shown in the upper panel of Fig. 2b), which caused the slight increase in the total FAA concentration.

The higher complexity of the FAA composition in the submicron aerosol particles could only be determined because the analytical method applied here was able to quantify the individual molecular FAA species. Such differentiation would not be possible with methods that determine the proteins as a sum parameter (e.g., the often applied Bradford method).

The high abundance of Gly in the aerosol particles is in good agreement with the Gly occurrence in other marine studies. Gly seems to be the dominant FAA, independent of whether the particles were sampled in the Arctic (Scalabrin et al., 2012), Antarctic (Barbaro et al., 2015) or in the North Pacific (Matsumoto and Uematsu, 2005) or whether they were attributed to a local marine source (Wedyan and Preston, 2008; Barbaro et al., 2015) or were rather under continental or long-range influences (Matsumoto and Uematsu, 2005). Other abundant FAAs reported at the different locations are arginine (not analyzed here), Ser and Ala; the latter ones are also dominant FAAs found in the study here reported. Consequently, the usage of the major FAAs as chemical biomarkers seems to be restricted to some extent due to their lack of source specificity. The high abundance of GABA found in the second half of the campaign has neither been partly regarded (i.e., included as a standard compound) in some marine studies nor yet been reported in ambient marine aerosol particles but seems to be special for this location. However, the reasons for the high concentration of hydrophilic FAAs within these respective sampling days remain unclear since no change in the environmental parameters determined (e.g., wind speed, wind direction, chl a concentration; Fig. 2a) was observed. In addition, we considered further FAA physicochemical parameters, such as the octanol–water partition coefficient (KOW), the topological polar surface area (TPSA), which describes the surface activity, and the density (Table S9), to describe the concentration changes. However, no statistically relevant correlations between the FAA concentration or composition and physicochemical parameters were found here either. Our observations could not clarify possible additional (i.e., non-marine) sources leading to the higher concentrations and complexity in the FAA composition. The dynamics behind the varying FAA concentrations and compositions at this location seem to be complex.

Following this hydropathy classification, the submicron aerosol particles consisted on average of 5 % hydrophobic, 15 % hydrophilic and 80 % neutral amino acids, while the supermicron aerosol particles contained on average only 7 % hydrophobic and 93 % neutral amino acids (Table S7). During the campaign, an increase in the contribution of hydrophilic amino acids was observed with a maximum of 55 % on 7 October 2017. Barbaro et al. (2015) reported that hydrophilic components were predominant (60 %) in locally produced marine Antarctic aerosol particles, whereas hydrophobic compounds were the rather dominant aerosol particles collected at the continental station (23 % and 27 %). The relatively high content of hydrophilic FAAs during certain periods of the campaign points at least to some influence of local oceanic sources.

In consideration of the carbon or nitrogen content of the amino acids (Table S6), the contribution of ∑FAA to WSOC and water-soluble organic nitrogen (WSON) in the size-segregated aerosol particles was calculated (Table S10). In the submicron size range, ∑FAA contributed up to 5.3 % (average 1.1 %) to WSOC, while in the supermicron range, ∑FAA only contributed up to 0.04 % to WSOC. Looking at ∑FAA's total contribution to WSOC (PM10), 0.7 % of WSOC consists of ∑FAA, which is in good agreement with the value of the study by Mandalakis et al. (2011). Considering the nitrogen content of the amino acids, ∑FAA contributed to the estimated WSON (WSON = 25 % of measured TDN concentrations according to Lesworth et al., 2010) an average of 0.4 % in the submicron and of 0.05 % in the supermicron size ranges. The observed daily variations in the contribution of ∑FAA to WSOC/WSON were derived from the daily variations in the atmospheric concentrations of ∑FAA (Fig. 2) and of WSOC/ WSON (Table S10). In summary, ∑FAA contributed up to 5.3 % to WSOC and 1.8 % to WSON when it comes to the submicron aerosol particles (7 October 2017) and up to 0.15 % to WSOC and 0.1 % to WSON for the supermicron aerosol particles. These percentages were on the same order of magnitude as for other organic compound groups, e.g., amines. It was shown by van Pinxteren et al. (2019) that amines contributed on average 5 % to the submicron WSOC content in marine aerosol particles. The percentage of ∑FAA to WSOC (up to 5.3 %) in the submicron aerosol particles especially demonstrated that FAA comprised a substantial fraction of submicron WSOC in marine aerosol particles.

From the MV samples, FAAs and additional parameters such as PM, WSOC, sodium and MSA were investigated. The results are listed in Table S11. The submicron aerosol particles at MV had an averaged ∑FAA concentration of 1.5 ng m-3 (0.8–1.9 ng m-3) which was about 3 times lower compared to the ∑FAA concentration at the CVAO. The ∑FAA concentration in the supermicron aerosol particles at MV (1.2 ng m-3; 0.2–2.9 ng m-3) was similar to the respective concentration at the CVAO. Additional online measurements of particle size number distributions (PSNDs) at the CVAO and MV, described in Gong et al. (2020), were in good agreement with one another during cloud-free times. This indicated that, for cloud-free conditions, the aerosol particles measured at ground level (30 m) within the IBL, which is mainly below 30 m (Niedermeier et al., 2014), represented the aerosol particles at cloud level. Thus, the aerosol particles within the marine boundary layer (MBL) were well mixed, and Mt. Verde was most of the time within the MBL (van Pinxteren et al., 2020). However, as described above, the Berner measurements were (continuously) taken during cloud-free and during cloudy times. The concentration and composition of the aerosol particles can therefore be affected by the clouds that formed and disappeared consistently during the sampling period of the aerosol particles on Mt. Verde (for further details on the frequency of the cloud events see Gong et al., 2020, and van Pinxteren et al., 2020). There was also no rain during the entire campaign. Furthermore, aging processes may occur during the upwind of the aerosol particles from the CVAO to the MV station, which takes about 4 h considering an average vertical wind of 5 cm s-1 (van Pinxteren et al., 2020). The particles at MV exhibited lower particle masses, as well as lower concentrations of the aerosol particle constituents. The decrease in concentrations of ∑FAA, PM, sodium, MSA and WSOC was reduced by a factor of 3 to 4 with respect to the submicron aerosol particles. However, no uniform depletion ratio between their concentrations at the CVAO and MV was found for the supermicron aerosol particles (Table S11). While the PM of the supermicron particles was reduced by a factor of 4 at MV (similar to the submicron aerosol particles), sodium and WSOC were depleted more strongly (factor of 11–12) compared to their respective concentrations at the CVAO. This suggests that the submicron particles were rather uniformly affected and depleted, likely by cloud processes, while the supermicron particles were influenced by clouds, and potentially other sources, in a non-uniform way. Nevertheless, the abundance of the marine tracers (sodium, MSA), together with the presence of FAAs in the aerosol particles (which mainly had a similar composition compared to the oceanic and ground-based particulate FAAs), indicated an oceanic contribution to the aerosol particles at cloud level.

The hydrophilic amino acids (Asp, Glu, GABA) comprised a significant fraction in the ULW and the SML, as well as in the (submicron) aerosol particles and in cloud water (Fig. 4a–d). They were not detected in the supermicron aerosol particles. A conspicuous finding is the high concentration of GABA, which is present exclusively in the submicron aerosol particles (B1 and B2: 0.05–0.42 µm) at the CVAO. Despite the relatively high LOQ of GABA in seawater (Table S1), a major abundance of GABA in seawater would be detectable. GABA is a metabolic product of the microbiological decarboxylation of Glu, which has been detected in all marine compartments. Active microbial enzymes on nascent sea spray aerosol have recently been reported by Malfatti et al. (2019). The abundance of GABA in the submicron aerosol particles suggests that either GABA could have been produced by the microbiological decarboxylation of Glu by present (marine) microorganisms in the aerosol particles or that GABA was transferred from the seawater to the atmosphere. However, GABA could not be found in seawater (ULW and SML), and this is not related to the sensitivity of the analytical method. Hence, a very enhanced oceanic transfer of GABA would be needed to explain this finding. Such an enhanced transfer was, however, not observed for the other hydrophilic amino acids (Glu and Asp), and their percentage composition was not strongly different regarding seawater and submicron aerosol particles at the CVAO. Unless the oceanic transfer of GABA is very different compared to other hydrophilic amino acids, this pathway does not explain the high abundance of GABA in the submicron aerosol particles at the CVAO.

Together with the facts that GABA is a known indicator for the microbiological decomposition of OM (Dauwe et al., 1999; Engel et al., 2018) and microorganisms are known to be present in marine aerosol particles even in the submicron size range (Rastelli et al., 2017), the formation of GABA in the aerosol particles might be related to an in situ formation. Interestingly, GABA was not detected in cloud water samples, although bacteria were found during the campaign in cloud water (van Pinxteren et al., 2020) whose presence has been reported in the literature (Jardine, 2009; Vaïtilingom et al., 2013; Jiaxian et al., 2019). It remains speculative whether GABA was degraded in cloud water despite its rather long lifetime (remote cloud case: 28.8 h; Table S13) or whether it was not produced by the bacteria in cloud water. Asp has been detected in all marine compartments and showed high cloud water concentrations. Correlations between Asp with diatoms and zooplankton have been reported for the marine environment (Hammer and Kattner, 1986). Hence, the occurrence of Asp in the marine environment can be attributed to a biogenic origin, while the high concentrations of Asp in cloud water (Figs. 3 and 4d) might be related to an oceanic source. In the Antarctic, Barbaro et al. (2015) attributed the hydrophilic amino acid fraction mainly to locally produced aerosol particles. According to the biogenic sources of the hydrophilic acids and their characteristics observed here, a local marine source for Asp and Glu, together with the biogenic formation of GABA in the aerosol particles in the Cabo Verde islands, could be prevalent.

Neutral amino acids were generally the amino acid group with the highest concentration in all investigated marine compartments, accounting for more than 50 % of the FAA total (Fig. 4a–d). Ser and Gly were the dominant representatives of this group. It is remarkable that especially the aerosol particles in the larger size range (e.g., supermicron aerosol particles: B4, B5) at both sampling stations are less complex in amino acid composition and almost exclusively dominated by Gly, followed by Ser and Ala (Fig. 4b, c). Gly is discussed in the literature as a photochemical degradation product of other existing amino acids, and this comparatively more stable amino acid (Gly) thus becomes a major component of the FAA composition (Barbaro et al., 2015). Compared to other amino acids, Gly and Ser have a very low atmospheric reactivity (McGregor and Anastasio, 2001) and therefore a higher mean lifetime τ (Gly: 0.48 h; Ser: 0.24 h; remote aerosol case; Table S13). Due to its atmospheric stability, Gly is proposed as an indicator for long-range transport (Barbaro et al., 2015, and references therein) and has a very low atmospheric reactivity (McGregor and Anastasio, 2001). However, our results clearly show that Gly and Ser are also present in seawater to a high extent, likely resulting from the siliceous exoskeleton of diatom cell walls (e.g., Hecky et al., 1973). Hence, besides long-range transport, a transfer from the ocean via bubble bursting might be an additional likely source of the stable, long-lived FAAs in the atmosphere. The neutral amino acid Pro has been reported to be of biogenic origin in the marine environment and was detected in seawater (Fig. 4a), in submicron aerosol particles at the CVAO (Fig. 4b) and in cloud water (Fig. 4d). Fischer et al. (2004) demonstrated that Pro can be used to identify the presence of algal spores in aerosol particles and might thus be used as a tracer for an oceanic source. The presence of Pro in all marine compartments suggests a transfer from the ocean into the atmosphere up to cloud level. This is supported by the comparatively low atmospheric reactivity of Pro (remote aerosol case: 0.24 h; Table S13). Finally, the hydrophobic FAAs Ile, Leu and Thr were found in all marine compartments in low concentrations. They are classified as relatively reactive amino acids and their abundance has been attributed to local or semi-local sources consequently (e.g., Mashayekhy Rad et al., 2019). Their low but constant abundance in all marine matrices again indicates a bubble-bursting transfer.

Aromatic FAAs like Phe and Tyr were present in seawater but not in the aerosol particles nor in cloud water samples. It could be assumed that either these aromatic FAAs were not transferred from the ocean into the atmosphere, they reacted already after their transfer due to chemical transformation reactions, or they were not detected because of their low atmospheric concentration. The mean lifetimes τ of Phe (0.014 h) and Tyr (0.007 h) (Table S13) showed that both FAAs had a comparatively high atmospheric reactivity (τ<1 min) in remote aerosol case conditions. Hence, a rapid chemical reaction of these compounds is most likely. Moreover, previous studies reported low atmospheric concentrations of Tyr and Phe in aerosol particles. Barbaro et al. (2011) found Phe (0.5 ng m-3) and Tyr (0.3 ng m-3) with a contribution of <1 % to ∑FAA (∑FAA: 42.5 ng m-3) on TSP samples in the urban background (Venice, Italy). In our study at the CVAO, the mean value of ∑FAA in PM10 aerosol particles was 3.8 ng m-3 (Sect. 3.2). Assuming that Phe and Tyr were contributing a very small fraction to ∑FAA, as reported in Barbaro et al. (2011), their concentrations would be below the detection limit and would thus probably not be detected. It can be concluded that either the aromatic FAAs could not be quantified in aerosol particles due to the sensitivity of the analytical method used here or they could react very quickly in the atmosphere and would therefore not be detected.

A high similarity regarding the FAA species within the different marine compartments could be observed, although some differences could also be identified (e.g., GABA). Together with the high concentration of ocean-derived compounds (Na+, MSA) in the aerosol particles and cloud water, this indicates a coupling between the FAAs in the ocean and the atmosphere. A quantitative metric for comparing compounds in the ocean and in the atmosphere is the EFaer (Eq. 2). The concept is mainly applied to closed systems (e.g., Quinn et al., 2015; Rastelli et al., 2017) because FAA formation or degradation pathways in the aerosol particles including biological or photochemical atmospheric reactions, and possible transport from sources other than marine, are excluded in this parameter. However, for comparison purposes, it might be useful to calculate the EFaer also from open systems, as done, e.g., by Russell et al. (2010) or van Pinxteren et al. (2017). The averaged EFaer of ∑FAA in the individual Berner stages of the case study at the CVAO based on SML and ULW concentrations are shown in Fig. 5.

The EFaer of ∑FAA, based on SML, were in the supermicron size range (1×101, B5, 1×101, B4) and several orders of magnitude smaller than in the submicron range (4×102, B3, 2×104, B2, 5×103, B1). Furthermore, the calculated EFaer, based on ULW, was up to 1 order of magnitude higher in the aerosol particles than the EFaer, based on the SML. This is due to the different FAA concentrations in seawater (Sect. 3.1) as the sodium values were very similar at 12.45 g L-1 in the ULW and 12.53 g L-1 in the SML. It was shown by van Pinxteren et al. (2017) that the EFaer of the WSOC in the submicron marine ambient aerosol particles at Cabo Verde ranged between 103 and 105. The averaged EFaer of the WSOC during our campaign was between 2×103 and 1×104 in the submicron range, between 3×102 and 4×102 in the supermicron range (Table S14), and in good agreement with van Pinxteren et al. (2017). Comparing the EFaer of ∑FAA (1×101–2×104) with the EFaer of WSOC (1×101–2×104) in the submicron range, both EFaer values are on the same order of magnitude. Moreover, similar percentages of ∑FAA were observed for the DOC in the SML (up to 7.6 %) (Sect. 3.1) and for the WSOC in submicron aerosol particles (up to 5.3 %) (Sect. 3.2).

Previous studies have shown that OM ejected into the atmosphere during bubble bursting results in the formation of sea spray aerosol particles containing OM similar to SML (Russell et al., 2010; Cunliffe et al., 2013, and references therein). Especially the film droplets have been reported to be enriched in OM and are suggested to transfer OM from the SML into submicron aerosol particles (Wilson et al., 2015). The supermicron aerosol particles tend to form from the larger jet droplets and thus represent the ULW composition (Blanchard, 1975; Wilson et al., 2015). We cannot derive mechanistic transfer characterizations from the ambient measurements performed here. Nevertheless, the constant FAA enrichment in the SML, together with the strong FAA enrichment in the submicron aerosol particles, strongly suggests that film droplets form the submicron particles. However, Wang et al. (2017) showed that jet drops (which transfer OM from the ULW) also have the potential to contribute significantly to the formation of submicron sea spray aerosol particles; so jet droplets can also contribute to FAA formation.

Applying the concept of the enrichment factor to cloud water and calculating the EFCW (Eq. 2), EFCW(∑FAA)=4×103 (based on SML) and 1×104 (based on ULW) could be determined. As mentioned in Sect. 3.3, several atmospheric processing (aging), oceanic transfer and biogenic-driven processes might contribute to this high enrichment and need to be addressed in future studies. The high FAA concentrations and enrichments might have implications for OM processing through clouds and are worth further studying.