Samples were taken during the MarParCloud (“Marine biological production, organic aerosol particles and marine clouds: a process chain”) campaign that took place from 13 September to 13 October 2017 at the Cabo Verde archipelago island Sao Vicente located in the Eastern Tropical North Atlantic (ETNA). A detailed overview of the campaign, background, goals, and first results are available in van Pinxteren et al. (2020). Measurements were performed at the Cape Verde Atmospheric Observatory (CVAO) as described in more detail elsewhere (Triesch et al., 2021a, b; van Pinxteren et al., 2020). The CVAO is located directly at the shoreline at the northeastern tip of the São Vicente island at 10 m  above seal level (a.s.l.) (Carpenter et al., 2010; Fomba et al., 2014). Owing to the trade winds, this site is free from local island pollution and provides reference conditions for studies of ocean–atmosphere interactions as there is a constant north-westerly wind from the open ocean towards the observatory. However, it also lies within the Saharan dust outflow corridor, and mainly in the winter months (January and February), dust outbreaks frequently occur.

Total suspended aerosol particles (TSPs) for TEP analysis and PM10 sampling for analysis of further aerosol constituents (inorganic ions, INPs, dust) was performed on top of a 30-m sampling tower of the CVAO. Tower measurements there mainly represent the conditions above the ocean because 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 below 30 m (Niedermeier et al., 2014). During the MarParCloud campaign, the marine boundary layer (MBL) was well mixed as indicated by an almost uniform particle number size distribution within the MBL (Gong et al., 2020b; van Pinxteren et al., 2020). Information on the meteorological conditions during the sampling period is given in Table S1 in the Supplement.

The TSPs were sampled with a filter sampler consisting of a filter holder equipped with a 0.2-µm pore-sized polycarbonate (PC) filter mounted to a pump. The PC filters had been cleaned with 10 % HCl and rinsed with ultrapure water (resistivity = 18.2 MΩ cm) before application. Sampling usually took place for 24 h and the flow of the pump was between 5 and 10 L min-1 and frequently measured with a flowmeter. Total air volumes between 10 and 15 m3 were sampled. In seawater TEP analysis, filtration is usually performed at a gentle pressure of 0.2 bar (Engel, 2009) which corresponds to a max flow rate of 21 or 38 L min-1. The flow rate of aerosol sampling was max. 10 L min-1 and therefore TEP losses during aerosol particle sampling were not expected.

PM10 particles were sampled with a high-volume sampler (Digitel, Riemer, Germany) equipped with preheated (105 ∘C for 24 h) 150-mm quartz fibre filters (Munktell, MK 360) at a flow rate of 700 L min-1, described in detail elsewhere (van Pinxteren et al., 2020). The sampling times for TSPs as well as PM10 were usually set to 24 h.

Cloud water was sampled on Mt. Verde, which is the highest point of the São Vicente island (744 m), situated in the northeast of the island (16∘52.11′ N, 24∘56.02′ W) and northwest to the CVAO (van Pinxteren et al., 2020). Again, Mt. Verde experiences direct trade winds from the ocean with no significant influence of anthropogenic activities from the island (Carpenter et al., 2010). Bulk cloud water was collected using a compact Caltech Active Strand Cloudwater Collectors (CASCC2) equipped with acid-cleaned Teflon®strands (508 µm in diameter). Cloud droplets were caught on the strands and gravitationally channelled into an Nalgene bottle. The 50 % lower size cut for the CASCC2 is approximately 3.5 µm in diameter. Much of the liquid water content (LWC) in clouds is contained of drops between 10 and 30 µm in diameter and the CASCC2 is predicted to collect drops in this size range with an efficiency greater than 80 % (Demoz et al., 1996).

Three cloud water samples collected on 20 September 2017, 28 September 2017, and 4 October 2017 were analysed for the TEP number concentrations. They were filtered (150–200 mL) through 0.2-µm pore-sized filters for TEP analysis using the same filter type and conditions as applied for the aerosol particle staining. All equipment that was in contact with the cloud water samples (Teflon^® strands, sampling bottles, filters) had been cleaned with 10 % HCl and rinsed with ultrapure water (resistivity = 18.2 MΩ cm) before each application, as recommended in Engel (2009).

To investigate a direct oceanic transfer of TEPs via bubble bursting, TSPs were sampled from a plunging waterfall tank experiment that is described in detail in the MarParCloud overview paper (van Pinxteren et al., 2020, Supplement section). The tank was designed to study the bubble-driven transfer of organic matter from the bulk water into the aerosol phase. It consists of a 1400 L basin with a 500 L aerosol chamber on top. The bubble-driven transport of organic matter was induced using a skimmer on a plunging waterfall. A stainless steel inlet was inserted into the headspace of the tank and connected with three filter holders for offline aerosol particle sampling without size segregation (TSPs). The filter system for TEP analysis was equipped with a 0.2-µm pore-sized, acid-cleaned polycarbonate (PC) filter mounted to a pump. Sampling usually took place for ∼24 h, the flow of the pump was between 5 and 10 L min-1 and frequently measured with a flowmeter. Total volumes between 9 and 10 m3 were sampled. The sampling procedure was therefore identical to the ambient TEP filter sampling. Another filter holder was equipped with a preheated 47-mm quartz fibre filter (Munktell, MK 360) for sodium analysis. The stainless steel inlet was additionally connected to a TROPOS-type Scanning Mobility Particle Sizer (Wiedensohler et al., 2012) for online aerosol measurements. This method of aerosol generation resulted in an efficient generation of nascent sea-spray aerosol particles with an aerosol particle size distribution centred around 100 nm (van Pinxteren et al., 2020).

The filters obtained from ambient and tank-generated TSP aerosol particle sampling and cloud water filtrations were stained with 3 mL of an Alcian blue stock solution (0.02 g Alcian blue in 100 mL of acetic acid solution, pH 2.5) for 5 s yielding an insoluble non-ionic pigment and afterward rinsed with milliQ water. The dye Alcian blue consists of a macromolecule with a central copper phthalocyanine ring linked to four isothiouronium groups via thioether bonds (Passow and Alldredge, 1995). The isothiouronium groups are strong bases and account for the cationic nature. The exact staining mechanism is not resolved but it is believed that the cationic isothiouronium groups bond via electrostatic linkages (ionic bonds) with the polyanionic molecules of the TEP molecule; hence, the carboxylic and sulfonic side groups are stained. Alcian Blue can also react with carbohydrate-conjuncted proteins at proteoglycans, but not with nucleic acids and neutral biopolymers (Villacorte et al., 2015). After staining, the filters were kept at -20 ∘C and transported to the laboratories of TROPOS.

For microscopic analysis, the protocol following Engel (2009) was applied. In short, abundance, area, and size-frequency distribution of TEP were determined using a light microscope (Zeiss Axio Scope A.1) connected to a camera (ColorView III). Filters were screened at 200× magnification. About 10 pictures were taken randomly from each filter in two perpendicular cross-sections (five pictures each cross-section; dimension 2576×1932 pixels, 8-bit colour depth) and microscopic pictures of TEPs in cloud water are shown in Fig. 1. Images were then semi-automatically analysed using ImageJ (Version 1.44). A minimum threshold value of 16 µm2 was set for particle size during particle analysis to remove the detection of non-aggregate material by the programme. This resulted in a minimum particle size of 4.5 µm (assuming spherical particles).

Blank filters were taken for aerosol sampling (inserting filters in the aerosol sampler without probing them) and cloud water (filtering reagent water over a pre-cleaned filter), stained and treated the same way as the microscopic analysis. Blank number concentrations were on average 6 % of the cloud water results and between 5 % and 20 % for aerosol results and the blank values were subtracted from the samples.

The analysis of inorganic ions from PM10 samples was performed with ion chromatography and conductivity detection. Aqueous extracts of the aerosol samples were made by ca. 25 % of the PM10 filter in 1.5 mL ultra-pure water (resistivity = 18.2 MΩ cm) for 1 h. After the filtration (0.45-µm syringe filter) of the extracts sodium (Na+), calcium (Ca2+), magnesium (Mg2+), were analysed by using ion chromatography (Dionex ICS-6000, Thermo Scientific). The cations were separated in an isocratic mode (eluent: 36 mM methanesulfonic acid) on a Dionex IonPac CS16-4 µm column (2×250 mm) that was combined with a Dionex IonPac CG16-4 µm guard column (2×50 mm). The detection limits for the determined ions were between 5 and 20 µg L-1 (Zeppenfeld et al., 2021).

Non-sea-salt calcium was calculated from the ion ratio of Ca2+/Na+ in seawater of 0.038 (Turekian, 1968). Dust concentrations were estimated from the aerosol particle mass concentrations as the residual mass after the subtraction of all analytical concentrations from the PM10 mass, as described elsewhere (Fomba et al., 2014). Trace metal content was determined using a total reflection X-ray fluorescence S2 PICOFOX (Bruker AXS, Berlin, Germany) spectrometer equipped with a Molybdenum X-ray source (Fomba et al., 2013). The cloud LWC was measured using a particle volume monitor (PVM-100, Gerber Scientific, USA), which was mounted at the same height as the cloud water samplers.

Ice nucleating particle number concentrations (NINP) were measured using two droplet freezing techniques (LINA: Leipzig Ice Nucleation Array and INDA: Ice Nucleation Droplet Array) in different marine compartments. The uncertainties of NINP are given by the 5 % to 95 % confidence interval and the results are presented in Gong et al. (2020a).

All the samples of this study are summarised in Table 1. In addition to samples from the MarParCloud campaign, surface seawater samples obtained from the ETNA (Engel et al., 2020) were considered.

To determine enrichment or depletion of TEPs in the atmosphere (i.e. on the aerosol particles and in the cloud water) in relation to the TEP concentration in the ocean water, the concept of the aerosol enrichment factor can be applied. To this end, the concentration of the compound of interest in each compartment is related to the respective sodium mass concentration, as sodium is regarded as a conservative sea salt tracer transferred to the atmosphere in the process of bubble bursting (Sander et al., 2003). This concept is usually applied to calculate the enrichment of a compound in the aerosol particles (EFaer.) in relation to seawater (Quinn et al., 2015), but was recently extended to calculate the enrichment of organic compounds in cloud water (EFcloud) in relation to seawater (Triesch et al., 2021a). Therefore, the enrichment factor is defined as EFatm. (atmosphere enrichments factor) in Eq. (1). 1EFatm.=c(TEP)atm/cNa+massatmc(TEP)seawater/cNa+massseawater. For Eq. (1), TEP number concentrations were converted to TEP volume concentrations. To this end, for atmospheric and for oceanic samples, particle number concentrations of TEPs were extracted from the size distribution spectra and volume concentrations were calculated (assuming spherical particles). More detail on the conversion can be found in the Supplement (Tables S2–S5).

Within the 3-week sampling period, TEPs varied within one order of magnitude between 7×102 and 3×104 #TEP m-3 in the aerosol particles and between 4×106 and 9×106 #TEP L-1 in the cloud water (analysed diameter size range: ∼4.5 to ∼30 µm) as shown in Fig. 2. The cloud water concentrations were converted to atmospheric concentrations using the measured LWC of the cloud water (0.39 g m-3) and resulted in concentrations of 2–4×103 #TEP m-3 (Table S4). Comparing the #TEP concentrations in cloud water with those in the ambient aerosol particles suggested that about 20 % of the ambient TEP particles might be activated to cloud droplets when a cloud forms.

In addition, TEPs were measured in four aerosol particle samples from the plunging waterfall tank and the concentrations varied between 4×102 and 3×103 #TEP m-3 (Table S3). Although the TEP concentrations in ambient aerosol particles and cloud water were not significantly different (ANOVA, one-way, p=0.054 at a 0.05 level), the tank-generated TEP concentrations were significantly lower than the ambient aerosol TEP concentrations (ANOVA, one-way, p=0.004 at a 0.05 level). The TEP number concentrations measured in the different atmospheric compartments, the ambient aerosol particles, the tank-generated aerosol particles, and the cloud water are summarised in Fig. 3a and the individual values are presented in the Tables S2–S4.

In addition to the total number concentrations, TEP number size distribution was derived from all ambient aerosol particle samples and are shown in Fig. 4a–d in both, linear and logarithmic form. In addition, the TEP number size distribution of one cloud water sample is presented in Fig. 4e and f. All samples exhibited very similar trends in their size distribution, with higher number concentrations for smaller sizes.

From the observed size distributions, it can be assumed that the number concentrations will continue to increase towards smaller sizes. A comparison of TEP number concentrations in the ambient aerosol particles or cloud water with literature values is challenging owing to the availability of very few studies and different sample types and size ranges regarded in different studies. However, the trend in the TEP number size distributions observed here is consistent with studies by Kuznetsova et al. (2005) showing increased TEP concentrations in simulated sea spray regarding particle sizes from 50 µm to 10 µm in diameter. In addition, TEP mass concentrations showed a similar trend, with higher concentrations towards smaller particle sizes (size range 0.1–1 µm, Aller et al., 2017), that was, however not as pronounced as for the TEP number concentrations observed here.

Regarding polymer gels in general, a strong increase with decreasing sizes was observed in cloud water in the high Arctic (north of 80∘ N) in late summer using a very sensitive microscopic technique with epifluorescence (Orellana et al., 2011). 2×109 micrometre-sized polymer gels per mL-1 and 2–6×1011 nanometre-sized polymer gels per mL-1 were observed and the majority of the particles were smaller than 100 nm (Orellana et al., 2011). The measurements from Orellana et al. (2011) regarded a much smaller particle diameter range (down to the nanometre scale) than the present work and are therefore not directly comparable. However, from the logarithmic TEP number concentration vs. diameter relationship (Fig. 4) we calculated TEP number concentrations for smaller particle ranges (sub-micrometre size range). TEP number concentrations between 4.2×104 #TEP m-3 (low “TEP5” case, equation from Fig. 4b) and 1.6×106 #TEP m-3 (high “TEP10” case, equation from Fig. 4d) are calculated for 1-µm large particles. The high but varying concentrations for the two cases underlines the need for more measurements in the submicron range to derive robust numbers. Similarly, a concentration of 3.0×108 #TEP L-1 for 1-µm large particles in cloud water were calculated and 2.1×1010 #TEP L-1 for 200-nm large particles might exist in the submicron-size range (following the equation from Fig. 4f).

These calculations show that the number of gel-like particles in the high Arctic was still several orders of magnitude higher than TEP particles in the tropical Atlantic, e.g. 1010 #TEP L-1 (200-nm particles) in tropical cloud water observed here vs. 1011 #polymer gels per mL-1 (=1014 #polymer gels per L-1) from Orellana et al. (2011). If the TEPs in the tropical atmosphere comprise only a small subgroup of the total polymer gel number, or if the total amount of gel-like particles is generally higher in polar regions remains to be investigated.

From a recent study of TEP number concentrations in different oceanic regions, TEP number concentrations in surface waters (10-m depth) of the East Tropical North Atlantic (ETNA) were obtained (Engel et al., 2020). ETNA is the region that geographically includes the Cabo Verde islands. The oceanic TEP number concentrations are shown in Fig. 5. The TEPs in the ocean showed a similar size distribution to the TEPs in the atmosphere (i.e. aerosol particles and cloud water, Fig. 4) with increasing TEP number concentrations towards smaller particle sizes (Table S5 and more details in Engel et al., 2020).

A detailed comparison of #TEPs in the ocean and in the atmosphere regarding the identical size bins showed that the #TEP distribution among the different size bins was much more balanced for seawater than for aerosol particles. In aerosol particles, on average 52 % of the #TEPs were located in the smallest size bin analysed (4.5–7 µm) and show a sharp decrease towards the second size bin (which contained 24 % of the TEPs) (Fig. 6). For the seawater TEPs, however, around 35 % of the #TEPs were found in the first size bin and the relative contribution decreased uniformly towards the larger size bins (Fig. 6). This distribution is also visible in the correlation curves of Figs. 4b, d, f, and 5b. The correlation curves for the aerosol particles (and cloud water) have a steeper slope than the curve obtained for seawater TEPs. This could imply that (i) the transfer of TEPs from the ocean to the atmosphere is most efficient for small size ranges, (ii) larger TEPs are converted to smaller TEPs in the atmosphere (e.g. break down), and/or (iii) atmospheric in-situ formation mechanism of TEPs preferably occur in smaller particle size ranges. These considerations will be further evaluated in Sect. 3.3.

To compare seawater and atmospheric TEP concentrations in terms of enrichment or depletion, the atmospheric enrichment factor EFatm. (Eq. 1) was calculated. However, the TEP number concentrations in the ocean surface water were obtained from an additional measurement campaign, taking place in the biologically productive Mauritanian Upwelling region in the year 2012, hence, at another time and season (Table 1). Compared with other oceanic regions, the TEP values from the Mauritanian Upwelling region were at the higher end (Engel et al., 2020). The region around the CVAO is rather oligotrophic and chlorophyll-a values during the MarParCloud campaign were relatively low with 0.1 up to 0.6 µg L-1 (van Pinxteren et al., 2020). As TEP production is often connected to phytoplankton activity, the TEP concentration at the CVAO might be lower than that of more productive regions (Robinson et al., 2019a). A previous study showed that the total TEP number concentrations (covering TEP sizes between 1 and 200 µm) at the Cabo Verde islands (south of São Vicente at 16∘44.4′ N, 25∘09.4′ W) were by a factor of 2 lower than the data reported here, in detail 0.9×107 L-1 (Engel et al., 2015) vs. 2×107 L-1 (Table S5). Lower TEP concentrations would result in higher EFatm. (Eq. 1) regarding the ambient as well as the tank measurements as the same type of seawater was used for the calculations. Hence, the EFatm. reported here represent lower limits.

In order to compare the same TEP diameters in all compartments, the size range between 5 µm (lower limit for atmospheric measurements) and 10 µm (typical upper limit for ambient aerosol particles) was regarded and converted from number to volume concentration (more details in Tables S2–S4 and Fig. S1 in the Supplement). For ocean water, TEP number concentrations of 3.5×103 #TEP mL-1 (= 3.5×106 #TEP L-1) and a TEP volume concentration of 4.6×105 µm3 TEP mL-1 (= 3.5×108 µm3 TEP L-1) were obtained. The respective values for the TEP volume concentration of ambient and tank-generated aerosol particles, as well as for the cloud water are listed in Tables S2–S4 and illustrated in Fig. 3b. As mentioned above, the factors given here are subject to some uncertainties and represent lower limits. An error discussion is introduced in the Supplement as an appendix to Table S2. It is clearly visible that the EFaer.ambient are significantly higher than the EFaer.tank (ANOVA, one-way, p=0.0017 at a 0.05 level) with average values of 9×103 and 50, respectively. The average EFcloud was 1×103. This means that the enrichment of TEPs derived from the plunging waterfall tank, representing the bubble-bursting transfer, is about two orders of magnitude lower than the enrichment of TEPs in the ambient aerosol particles.

It should be noted that the lower enrichment in the tank resulted from the lower TEP number concentrations in the generated aerosol particles, as the sodium concentrations in the tank aerosol were even higher than in the ambient particles (Table S3). This suggests that, although an artificial tank study cannot represent the ambient environment, the generation of sea-spray aerosol was in progress; however, TEP transfer seemed not to be pronounced.

In the following, the enrichment factors obtained here will be discussed in more detail considering studies available from the literature.

Atmospheric enrichment of ocean-derived OM has often been reported (e.g. Facchini et al., 2008; Keene et al., 2007; O'Dowd et al., 2004; Schmitt-Kopplin et al., 2012; Triesch et al., 2021a, b; van Pinxteren et al., 2017). Submicron particles are usually strongly enriched with OM with aerosol enrichment factors EFaer. of 103 up to 105 (Quinn et al., 2015 and references therein). The enrichment in supermicron aerosol particles is, however, significantly lower. Laboratory studies showed enrichment of OM in the order of 102 (Hoffman and Duce, 1976; Keene et al., 2007; Quinn et al., 2015). From the MarParCloud campaign, enrichment factors of free amino acids were between 10 and 30 in ambient supermicron particles (Triesch et al., 2021a). Kuznetsova et al. (2005) reported TEP enrichments in freshly produced sea spray with EFaer.=44±22 based on TEP number concentration. Consequently, the  EFaer.tank (50±35) reported here are in line with published enrichment factors for OM in general and TEPs specifically. However, the EFaer.ambient (9×103) were orders of magnitude higher than reported enrichment factors for supermicron aerosol particles. Enrichment factors of OM in cloud water are scarce; we recently reported an enrichment of 103–104 for free amino acids in cloud water from the MarParCloud campaign (Triesch et al., 2021a) that was higher than the EFcloud observed here.

The concept of the aerosol enrichment factor originates from controlled tank experiments where a direct transfer of compounds from the ocean via sea-spray aerosol formation occurs. Obviously, this does not automatically correspond to the ambient environment as mixing processes, ageing, and further transformation reactions are not accounted for. However, the EFaer.ambient, which is much bigger than the EFaer.tank, and the comparison of EFcloud towards former literature data, clearly show the presence of significantly more TEPs in ambient aerosol and cloud water than in oceanic seawater, which will be discussed in detail in the following section.

Different kinds of ice-nucleating macromolecules have been found in a certain range of biological species and consist of a variety of chemical structures including proteins, polysaccharides (Pummer et al., 2015), and lipids (DeMott et al., 2018). TEPs, consisting of polysaccharidic chains bridged with divalent cations, may therefore possess good properties to act as INPs; however, such a link has not yet been shown in field experiments.

During the MarParCloud campaign INP number concentration (NINP) was measured in different marine compartments and the results are presented in Gong et al. (2020a). By combining INP concentration in the seawater, aerosol particles, and cloud water, it was found that NINP in the atmosphere was at least four orders of magnitude higher than what would be expected if all airborne INPs were to originate from sea spray. The measurements indicated that other sources besides the ocean, such as mineral dust or other long-ranged transported particles, contributed to the local INP concentration. However, some indications for contributions of biological particles to the INP population were obtained (details in Gong et al., 2020a). Nevertheless, the sources of INPs could not be revealed in detail.

In the present study, quantitative INP data (presented in Gong et al., 2020a) and TEP data measured from the same campaign were compared. To this end, INP concentrations achieved from PM10 quartz-fibre filters taken at the CVAO during the same period as the TSP filters were compared with the TEP measurements. In addition, cloud water INP and TEP data obtained from the same samples were combined.

Transparent exopolymer particle number concentrations were on average between 103 and 104 m-3 in the ambient aerosol particles, whereas INP number concentrations at -15 ∘C were between 10 and 102 m-3 (Gong et al., 2020a). It is interesting to note that the TEP concentrations in the ambient aerosol particles were about two orders of magnitude higher than the INP concentrations. Similar findings were obtained for the cloud water comparisons; TEP concentrations (∼106 L-1) were on average two orders of magnitude higher than INP number concentrations at -15 ∘C in cloud water (∼104 L-1) (Gong et al., 2020a).

The correlation between INP (active at -15 ∘C) and TEP concentrations was weak, with R2=0.3 (Fig. 7f), showing that a direct link between INP and the entire TEP number concentrations was not very pronounced. It needs to be underlined that TEP concentrations below a particle size of 4.5 µm are not included here and according to the size distribution, the TEP concentrations are increasing towards smaller sizes. Most of the TEPs reported here were in the supermicron size range between ∼4.5 and 14 µm (Fig. 4). However, the biologically active NINP at Cabo Verde was mainly present in the supermicron mode (>1 µm) (Gong et al., 2020a); hence, a comparison with the TEP particle concentrations above 5 µm seems justified. Nevertheless, future studies should concentrate on the exact same size ranges for TEPs and INPs.

The INP functionalities of biomolecules are not straightforward and whether a macromolecule acts as an INP is dependent on many factors, such as its size, the proper position of functional groups, and their allocation (Pummer et al., 2015). Typically, not the entire surface of an INP but rather specific areas (active sites) participate in ice nucleation. This means that despite TEPs likely providing INP properties, only a fraction of TEPs, if any, may be able to act as INPs. This hypothesis is supported by the findings that marine gels exhibit hydrophobic and hydrophilic surface-active segments, strongly suggesting a dichotomous, non-uniform behaviour of polymer gels (Leck et al., 2013; Orellana et al., 2011; Ovadnevaite et al., 2011). As mentioned in the sections “Abiotic formation” and “Biotic formation”, TEPs are often attached to, or colonised with bacteria. Bacteria themselves have been shown to provide excellent INP functionalities (Pandey et al., 2016) and TEPs may act as a carrying medium for INPs, such as bacteria. Bacteria concentrations were higher than TEP concentrations and also higher than INP concentrations. However, only a fraction of all bacteria (0.5 %–25 %) is associated with TEPs and, vice versa, not all TEPs are colonised by bacteria (Passow, 2002b). There is an indication that especially in oligotrophic waters, as the Cabo Verde islands are, the fraction of bacteria attached to TEPs is comparably low (Schuster and Herndl, 1995). Hence, the concentration range of bacteria-colonised TEPs in relation to INPs is worth further consideration. This may help to unravel whether a functional relationship between bacteria-colonised TEPs and INPs exists and whether a certain number of TEPs contain fragments in the biological INP population that, beyond dust, play a role in the Cabo Verde atmosphere.