Within the seawater analysis, we measured the dissolved amino acids (DAAs) and dissolved carbohydrates (DCHO) in the DOC fraction, as DOC represents by far the largest pool of organic material in the ocean (Riebesell et al., 2011). DOC is the fraction of OC that passes through a GF/F filter of 0.7 µm pore size (Wurl, 2009); however, filter pore sizes between 0.2 and 1.2 µm are often applied e.g. due to requirements of the analytical system (e.g. Romankevich, 1984; Zäncker et al., 2017). The lipid measurements (from the same samples) were taken from Triesch et al. (2021b), are included in the DOC fraction as well, being filtered via a 0.7 µm pore-sized filter (GF/F, Whatmann) and extracted in an organic solvent, and refer to dissolved lipids (DLs).

For the aerosol particles (PM1), we focused on the water-soluble organic (WSOC) fraction of the amino acids (AAaer.) and carbohydrates (CHOaer.). The lipids, however, were extracted with an organic solvent and filtered (analogous to seawater), meaning they comprise the organic-soluble organic fraction of the aerosol particles (Lipidsaer.).

For the analysis of the DCHO and CHOaer. as well as the DAAs and AAaer., molecular-resolved techniques were applied. DCHO in filtered (0.2 µm Millex syringe filters) and desalinated SML and bulk water samples and CHOaer. in the filtered PM1 extracts were quantified using high-performance anion-exchange chromatography coupled with pulsed amperometric detection (HPAEC-PAD) (Zeppenfeld et al., 2020, 2021). After an acid hydrolysis (0.8 M HCl, 100 ∘C, 20 h), the monosaccharides fucose (Fuc), rhamnose (Rha), arabinose (Ara), galactose (Gal), glucose (Glc), xylose (Xyl), mannose (Man), galactosamine (GalN), glucosamine (GlcN), muramic acid (MurAc), galacturonic acid (GalAc), and glucuronic acid (GlcAc) were measured.

For the DAA analysis, seawater samples (25.5 mL) were desalinated and concentrated to a few millilitres as described in Triesch et al. (2021a). For the AAaer. from PM1 particles, an aqueous extract of the amino acids was prepared by shaking a piece of the filter in 2 mL of water. After a filtration step (filter pore size: 0.2 µm), 25 µL of ascorbic acid (20 mg mL-1, purity 99 %, Sigma-Aldrich, St. Louis, Missouri, USA) was added to a 200 µL aliquot of the desalted seawater/aqueous filter extract to avoid the oxidation of the obtained amino acids as suggested in Mandalakis et al. (2010). Following the addition of 250 µL of HCl (Supra-quality, ROTIPURAN^® Supra 35 %, Carl Roth, Karlsruhe, Germany), the hydrolysis was performed at 110 ∘C for 20 h. After cooling to room temperature, the hydrolysed filtrate was evaporated, resolved in 500 µL milliQ-water (Millipore Elix 3 and Element A10, Merck Millipore, Darmstadt, Germany), filtered, derivatized using the AccQ-Tag™ pre-column derivatization method (Waters, Eschborn, Germany), and measured by ultra-high-performance liquid chromatography with electrospray ionization and Orbitrap mass spectrometry (UHPLC/ESI-Orbitrap-MS), as described in Triesch et al. (2021a). The analytes include the amino acids glycine (Gly), alanine (Ala), serine (Ser), glutamic acid (Glu), threonine (Thr), proline (Pro), tyrosine (Tyr), valine (Val), phenylalanine (Phe), aspartic acid (Asp), isoleucine (Ile), leucine (Leu), methionine (Met), glutamine (Gln), and γ-aminobutyric acid (GABA) (purity ≥99 %, Sigma-Aldrich, St. Louis, Missouri, USA). The DL and Lipidsaer. measurements were taken from Triesch et al. (2021b), where the analysis was done with a semi-molecular technique. For a better understanding of the data, a short description is given in the following: the dissolved fraction of seawater samples (obtained after filtration through pre-combusted 0.7 µm GF/F filters) and aerosol filters were extracted with dichloromethane according to specific protocols after adding 2-hexadecanone as an internal standard, as described in more detail in Triesch et al. (2021b). The extract was analysed with thin-layer chromatography (TLC). Lipid classes were separated on Chromarods SIII and calibrated with an external calibration with a mixture of standard lipids by a chromatograph flame ionization detector (FID) Iatroscan MKVI (Iatron, Japan). The separation scheme included elution steps in the solvent systems with increasing polarity. The lipid classes included hydrocarbons (HC), fatty acid methyl esters (ME), free fatty acids (FFA), alcohols (ALC), 1,3-diacylglycerols (1,3 DG), 1,2-diacylglycerols (1,2 DG), monoacylglycerols (MG), wax esters (WE), triacylglycerols (TG), pigments (PIG), phospholipids (PP) including phosphatidylglycerols (PG), phosphatidylethanolamine (PE), phosphatidylcholines (PC), and glycolipids (GL), which cover sulfoquinovosyl diacylglycerols (SQDG), monogalactosyl diacylglycerols (MGDG), digalactosyl diacylglycerols (DGDG), and sterols (ST). It needs to be underlined that, as no single lipid compound but rather lipid groups (based on varying polarity in the TLC system) were measured, the lipid results can be classified as analysis on a semi-molecular level.

OC on the aerosol particles (PM1 samples) was measured by means of a thermal-optical method using the Sunset Laboratory Dual-Optical Carbonaceous Analyzer (Sunset Laboratory Inc., USA) from a filter piece with an area of 1.5 cm2. The EUSAAR 2 temperature protocol was utilized, and a charring correction was applied (Cavalli et al., 2010). The correction value for pyrolytic carbon was determined based on measurements of a sample transmission using a 678 nm laser. Samples were thermally desorbed from the filter medium under an inert He atmosphere followed by oxidizing an O2/He atmosphere while applying carefully controlled heating ramps. A flame ionization detector was used to quantify methane following a catalytic methanation of CO2.

Sodium was measured from filtered (0.45 µm syringe filter) aqueous extracts of the PM1 samples using ion chromatography (more details in Zeppenfeld et al., 2021; van Pinxteren et al., 2022). The details on sampling and the sample treatment are summarized in Table S1.

Atmospheric concentrations were calculated from the measured analyte concentrations on the filter or in the filter extract. The masses of the analytes on the filter were extrapolated to the total filter and related to the collected air volume that had flown through the filter.

Detection limits were 0.5–2.5 µg L-1 (DAAs) and 0.1–0.2 pg m-3 (AAaer.) (Triesch et al., 2021a, c) and 0.13–0.7 µg L-1 (DCHO) and 0.5–5 pg m-3 (CHOaer.) (Zeppenfeld et al., 2021). Regarding the lipids, the method was carefully optimized for seawater analysis (Gašparović et al., 2015, 2017) and adopted for aerosol particle analytics as described in Triesch et al. (2021b). The detection limits were determined as the analyte concentrations corresponding to a signal-to-noise (S/N) ratio of 3 ranging from 0.06 to 0.33 µg as reported previously in Penezić et al. (2022).

For the seawater analysis, field blanks were taken by filling high-purity water in pre-cleaned plastic bottles and handled in the same way as the seawater samples. The seawater blanks were in general below 10 % and, for the lipids, always below 15 % (Triesch et al., 2021b). All data are blank-corrected by subtracting the field blank values from the samples.

Field blanks for aerosol particles were prepared using pre-baked quartz fibre filters without an active sampling and treated according to the same procedure as the field samples. The concentrations of the target analytes were calculated by external calibration. Each sample was measured twice with a relative standard deviation of typically <10 %, and field blanks, which were for most compounds negligible and for the lipid classes always below 20 % of the real aerosol particle sample, were subtracted. All presented values are corrected for the field blank.

The SML enrichment of DOC components is generally attributed to diffusion, turbulent mixing, scavenging, and transport of surface-active matter from rising gas bubbles in the water column (Liss and Duce, 1997). Within the groups investigated here, the DLs are the most hydrophobic compounds and are generally classified as highly surface-active compounds (Burrows et al., 2014). Although the surface-activity parameters (e.g. octanol–water partition coefficient, density, topological polar surface area) of the individual lipids differ among each other (values in Triesch et al., 2021b), the lipids are overall more non-polar and surface-active compared to the carbohydrates and amino acids (values in Triesch et al., 2021a). Nevertheless, the enrichment of the DLs in the SML was significantly lower compared to the carbohydrates and amino acids. This was obvious from a comparison of the averaged values and from single samples obtained on the same dates (3 and 7 October 2017, values in Tables S2 and S3 and Triesch et al., 2021b).

One explanation for the finding lies in the sampled SML thickness. With the glass plate technique, an SML thickness of about 100 µm has been sampled and reported for the CVAO area (van Pinxteren et al., 2017). Hence, the 100 µm-thick SML might be very well-mixed with regards to the soluble amino acids and carbohydrates, but the surface-active compounds, such as lipids, are potentially located at the very top and form a thin (nanometre-thick) monolayer. In the literature, the SML is described either as a series of sub-layers of wet and dry surfactants (Hardy, 1982) or as a gelatinous matrix (Sieburth, 1983). Independent of the model, it can be expected that a gradient along the surface likely forms with surfactants at the very top of the layer. The formation of a lipid-rich nanolayer at the very top agrees with surface-sensitive spectroscopy measurements that are able to tackle the uppermost layer and finds strong indications of a nanolayer dominated by soluble surfactants (Lass and Friedrichs, 2011) and hydrophobic low-molecular-weight lipids (Frka et al., 2012). The nanolayer, however, cannot be probed with currently applied SML sampling methods. Therefore, the measured SML concentrations may represent a very diluted (likely highly lipid-enriched) layer. Consequently, the SML structure is even more complex, which needs to be considered, particularly when discussing lipid enrichments in the SML. Here, a combination of bulk measurements with dedicated surface probing appears highly desirable.

Regarding the DAAs in detail, it is interesting to note that some compounds are exclusively present in the SML, as mentioned above. They belong to hydrophilic (Glu), hydrophobic (Iso), and neutral (Tyr) fractions of amino acids, underlining that their occurrence in the SML might not be related (solely) to their physicochemical properties. Besides an air-bubble-driven transfer to the surface, enrichment in the SML can be supported by co-adsorption mechanisms. Less surface-active compounds (e.g. amino acids and carbohydrates) can be attached due to ionic interactions/coulomb interactions to the head groups of the air-bubble-attached surfactants (e.g. lipids) that mediate their enrichment in the SML (Burrows et al., 2016; Hasenecz et al., 2019; Link et al., 2019; Schill et al., 2018). Co-adsorption can provide an explanation for the high occurrence of non-surface-active, very soluble compounds such as carbohydrates. A recent laboratory study showed different mechanisms for the co-adsorption of polysaccharides that form a second calcium-bridge sub-layer underneath the monolayer, whereas monosaccharides intercalate and induce reorganization within the nanolayer (Vazques de Vasquez et al., 2022). However, in the current study, only a small SML enrichment of the DCHO was observed and, hence, no indication of a strong co-adsorption.

Further explanations for the accumulation of dissolved compounds require in situ formation or degradation by SML-specific reactions that might be triggered by distinct environmental conditions in the SML. Biotic pathways and abiotic SML-specific (photo)chemical reactions may strongly impact OC cycling at the sea surface (Liss and Duce, 1997). The high abundance of the amino acid Glu in the SML observed here was also reported in the FAA fraction by Triesch et al. (2021a) and can likely be explained by in situ formation. In incubated cloud water, as another marine compartment, Glu has been shown to be produced via biotic and abiotic mechanisms, e.g. via the oxidation of proline (Jaber et al., 2021, and references therein), and such processes may be relevant in the ocean surface as well. Regarding biotic processes, it is well-known that microorganisms have complex and highly interconnected enzymatic networks and can biodegrade or biosynthesize organic compounds (Kyoto Encyclopedia of Genes and Genomes – KEEG – pathway). Kuznetsova and Lee (2002) suggested that stressed microorganisms, rich in dissolved combined amino acids, may be leached and release them, which in turn affects the pools of both these compounds in seawater. Although such formation mechanisms generally happen in the upper ocean, there are indications of SML-specific processes. Along a transect from upwelling regions toward oligotrophic gyres, it was found that while in the bulk water a clear trend toward degradation of amino acids was observed, the production and degradation patterns of amino acids in the SML were much more complex (Reinthaler et al., 2008). This is indicative of the role of the SML in the production of labile DOC driven by coupled microbial and photochemical processes. Similarly, Kuznetsova and Lee (2001) observed that peptide turnover was always faster in the SML than in sub-surface waters, likely due to the greater concentrations of DOC in the SML. The authors concluded that the accumulation of organic and inorganic compounds in the SML leads to a more nutritious medium for microbial growth and consequently enzymatic hydrolytic activity compared to the bulk water. Connecting this to the results presented here, this might suggest that changes induced by abiotic and biotic processing need to be considered when regarding the SML composition. Although such reactions likely also affect lipids and carbohydrates, they seem to be most pronounced for amino acids.

A further mechanism contributing to the high and variable SML enrichment of the DAAs at the current location might be microbial nitrogen fixation at the sea–air interface. Measurements showed that cyanobacteria are a very pronounced phytoplankton group in this region (Franklin et al., 2009; Hepach et al., 2014; Zindler et al., 2012), which was dominant during the MarParCloud campaign (van Pinxteren et al., 2020). Cyanobacteria are able to take up nitrogen from the atmosphere (Zehr, 2011). Earlier studies showed that cyanobacteria-fixed nitrogen is incorporated into amino acids (specifically glutamine; Carpenter et al., 1992). The calculated net amino acid release from cyanobacterial colonies (Trichodesmium thiebautii) revealed that nitrogen fixation and the biogeochemical turnover of ambient amino acids are an important source of recently fixed (“new”) nitrogen within the oceanic surface water (Capone et al., 1994). These considerations are, however, highly speculative and demand further studies to investigate whether nitrogen fixation and biosynthesis via cyanobacteria, which is often observed in sub-tropical and tropical oceans (Montoya et al., 2007), might establish a considerable route for amino acid formation and enrichment in the SML from the atmospheric side.

Although SML enrichment factors for amino acids, carbohydrates, and lipids have been reported in the available literature, they have not previously been shown in such detailed analysis for samples collected from the same site as shown here. From this study it can be concluded that the amino acids are strongly enriched in the SML compared to carbohydrates and lipids, even under the same environmental conditions. In a recent study, we showed a strong enhancement of other nitrogen-containing species (aliphatic amines) in the SML at this location, while the amine concentration in the bulk water was often not detectable (van Pinxteren et al., 2019). This suggests that the pronounced SML enrichment specifically exists for nitrogen-containing organic species. In addition, the absence of a relation of the SML enrichment to physical compound parameters (e.g. hydrophobicity) suggests that enrichment processes based on physicochemical properties (e.g. surface activity) alone do not drive SML enrichment. Rather, an SML in situ formation mechanism impacts the abundance of amino acids and likely nitrogen-containing organic species in general.

The finding that the Lipidsaer. were much more concentrated than the AAaer. and the CHOaer. resulted in a very different pattern compared to the similar seawater concentrations (Fig. 1 versus Fig. 3). To quantitatively compare the seawater and the aerosol concentrations, the EFaer. were calculated (values in Table 1).

For the amino acids, the EFaer. were between 9.2×102 (related to the SML) and 2.1×103 (related to the bulk water) and on average 1.5×103. For the carbohydrates, the EFaer. were between 1.3×103 (related to the SML) and 1.4×103 (related to the bulk water) and on average 1.3×103 and therefore similar to the EFaer. of the amino acids. For the lipids, however, the EFaer. were 2 orders of magnitude higher (EFaer.=1.4×105, related to the SML; EFaer.=1.7×105, related to the bulk water; EFaer.=1.6×105 on average, Table 1).

The overall high enrichment of OC in the aerosol particles is explained by complex, as-yet resolved interactions at the ocean surface where organic matter is enriched relative to sodium during the formation of film and jet droplets. Burrows et al. (2014) applied a conceptual model (“slab” model) where all organics partition to the surface of a “slab” of oceanic water or to both the outer and inner surfaces of a bubble film. The organic enrichment is therefore significantly higher for the thinner bubble films (bubble film thicknesses: 0.01 to 1 µm) than for the thicker SML (typically sampled SML thicknesses: 20 to 400 µm). This mechanism can explain EFaer. of OC in sub-micron aerosol particles of 102 to 103 compared to the SML (Burrows et al., 2014). However, EFaer. from ambient and laboratory-controlled observations show that, for some compounds, even higher EFaer. are obtained. In a controlled tank study, Rastelli et al. (2017) found strong enrichments for lipids (up to 1.4×105) as well as for proteins (up to 1.2×105) and carbohydrates (up to 1.0×105, Table 1). A recent controlled bubble-bursting laboratory study showed that amino acid enrichments can be up to 107 in sub-micron SSA between 0.029 and 0.060 µm (Triesch et al., 2021c). Similarly, Schmitt-Kopplin et al. (2012) showed that surface-active biomolecules are preferentially transferred from surface water into the atmosphere via bubble bursting. The ambient enrichment factor of the lipids (105) shown here and in Triesch et al. (2021b) agreed well with laboratory-derived ones (Rastelli et al., 2017), indicating that the transfer mechanisms simulated in laboratory experiments agree with observations performed here in the field. Hence, the high surface activity and/or lipophilic character of the lipid classes might explain their strong (chemo-selective) transfer to the aerosol particles. Even though the lipid composition on the aerosol particles varied slightly from the seawater concentration (Triesch et al., 2021b), their transfer is likely driven by their physicochemical properties (high surface activity and/or the lipophilic character). For the amino acids and carbohydrates, however, more complex mechanisms may determine their transfer to the atmosphere. Rastelli et al. (2017) suggested that diverse biological processes on the ocean drive the properties of proteins and carbohydrates on the ocean surface and in the atmosphere. Moreover, these compounds are known to be involved in marine gel-like particle formation, such as transparent polymer particles (TEPs) and Coomassie stained particles (CSPs), observed in the ocean and more recently in the atmosphere (Aller et al., 2017; Kuznetsova et al., 2005; van Pinxteren et al., 2022), adding more complexity to the system. Hence, not only a sea-to-air transfer but also atmospheric in situ formation and degradation might determine the concentration of the OC and notably of the amino acid and carbohydrates. This suggests that atmospheric processing plays an important role besides the physically driven bubble-bursting sea–air transfer of OC.

OC concentrations in marine aerosol particles during this campaign varied between 0.13 and 0.31 µg m-3 with an average value of 0.20 µg m-3 (values in Table S7). This agreed well with previous OC measurements from the CVAO that were on average 0.27 µg m-3 OC (van Pinxteren et al., 2017). To date, only a small percentage of OC on marine aerosol particles has been characterized on a molecular level, and organic biomarkers often comprise only a few percent of the OC (Chen et al., 2021). Fu et al. (2011) measured more than 140 different single organic species in marine aerosol from different oceanic areas; however, the identified species comprised less than 5.7 % of the OC. Taking together the OC components described here (Lipidsaer., AAaer., CHOaer.), the contribution of the identified components to the OC was calculated. Furthermore, the OC contribution of recently identified components from previous campaigns within the Cabo Verde region, in detail, aliphatic amines, methane-sulfonic acid (MSA), oxalic acid, and carbonyls (van Pinxteren et al., 2015), was included. The OC contribution of the single compounds and compound groups are shown in Fig. 5 (values in Table S8). Altogether, about 48 % of the average OC could be explained by the identified components. Regarding the maximum (0.31 µg m-3) and minimum (0.13 µg m-3) OC concentrations within the campaign, the OC contributions of the respective compounds are between 31 % (lower limit) and 74 % (upper limit). The major identified OC fractions (related to the average OC) were the Lipidsaer., with 43 %. They were followed by the aliphatic amines (4 %) that are in good agreement with a recent CVAO study, where they contributed on average 5 % to the (water-soluble) OC (van Pinxteren et al., 2019). MSA (0.9 %) and oxalic acid (0.3 %) were minor OC contributors. Similarly, the CHOaer. and the AAaer. made up minor percentages with 0.3 % and 0.4 % respectively. Regarding the lipids, it needs to be considered that the analysis performed here was not based on the detection of individual analytes but on an organic solvent extraction of the particle constituents and extract separation by solvents with different polarities applied in the TLC. The analytical method has been optimized for seawater analysis. Within atmospheric processing, additional organic compounds can form, which might contain a hydrophobic part and which are potentially included in the lipid analysis performed here. However, the large similarity of the lipid groups within the seawater and the aerosol particles as well as the agreeing concentrations of the single lipid groups (FFA, ALC) with measurements from other marine stations with molecular techniques (gas chromatography–mass spectrometry – GC-MS) suggests that the same compound classes were present in the particles. Future analysis of the lipid fraction with mass spectrometric techniques will help to better resolve this issue.

About 50 % of the aerosol OC remained uncharacterized on a molecular and semi-molecular level. The non-identified OC part may contain larger macro-molecules that might be composed of particulate or non-soluble forms in water (carbohydrates and proteins) or organic solvents (lipids) that were removed in the performed analysis during the sample preparation step. In addition, the unknown part might include component groups that belong to the soluble carbohydrates or amino acids but that are either too stable or too labile for the sample preparation procedure (e.g. within the hydrolysis step). Moreover, other complex molecules that cannot be captured with the methods applied here likely add to the unknown fraction including optical active parts summarized as chromophore dissolved organic matter (CDOM), humic-like substances (HULIS), brown carbon, and water-soluble pigments. As mentioned above, NMR analysis showed that SSA contains a large fraction of lipopolysaccharides comprising complex, macro-molecular groups of sugars, esters, carbonyls, acids, and lipids (Facchini et al., 2008). However, these components have not yet been analysed in aerosol particles using chromatographic techniques. The uncharacterized part may also contain particulate OC compounds, such as larger aggregates of marine gels or gel-like particles like TEPs. High TEP number concentrations in aerosol particles were recently identified in Cabo Verde (van Pinxteren et al., 2022), and high mass concentrations of TEPs (e.g. 1.2 µg m-3 for PM1) were found in the western North Atlantic atmosphere (Aller et al., 2017).

Kieber et al. (2016) suggested that the major OC component in sub-micron sea spray particles is recalcitrant, and recently Beaupre et al. (2019) proposed that 19 % to 40 % of the OC associated with freshly produced marine aerosol particles was refractory dissolved OC. This percentage agrees with the non-identified OC part from the present study. However, we cannot identify or classify the remaining OC fraction or attribute it with certainty to the recalcitrant OC. Further studies of the OC, in tropical and other areas of the world, are needed to continue resolving OC and related transfer and formation processes. Nevertheless, the potentially recalcitrant OC fraction in the oligotrophic region does not seem to constitute the majority of OC as reported by Kieber et al. (2016).