Marine submicron aerosol gradients , sources and sinks

Aerosol principal sources and sinks over eastern North Atlantic waters were studied through the deployment of an aerosol chemistry gradient sampling system. The chemical gradients of primary and secondary aerosol components – specifically, sea salt (SS), water-insoluble organic matter (WIOM), water-soluble organic matter (WSOM), nitrate, ammonium, oxalate, amines, methanesulfonic acid (MSA) and water-soluble organic nitrogen (WSON) – were examined in great detail. Sea salt fluxes were estimated by the boundary layer box model and ranged from 0.3 to 3.5 ng m−2 s−1 over the wind speed range of 5–12 m s−1 and compared well with the derived fluxes from existing sea salt source parameterisations. The observed seasonal pattern of sea salt gradients was mainly driven by wind stress in addition to the yet unquantified effect of marine OM modifying fractional contributions of SS and OM in sea spray. WIOM gradients were a complex combination of rising and waning biological activity, especially in the flux footprint area, and wind-driven primary sea spray production supporting the coupling of recently developed sea spray and marine OM parameterisations.


Introduction
Marine aerosols contribute significantly to the global radiative budget, and consequently changes in marine aerosol abundance and/or chemical composition have an impact on climate change through both direct and indirect effects.The eastern North Atlantic region is of particular interest due to a combination of storminess, prevailing westerlies bringing marine air masses into continental Europe and biological ac-tivity in surface waters significantly affecting chemical composition of atmospheric particulate matter (O'Dowd et al., 2004).Organic matter (OM) has been observed in marine aerosol particles for many decades and has been linked to fractional contribution of OM transferred from the sea surface into the tropospheric boundary layer through bubblemediated production processes (Blanchard, 1964;Hoffman and Duce, 1977;Middlebrook et al., 1998;Oppo et al., 1999;Russell et al., 2010).There has been significant progress in understanding marine aerosol composition, which has been identified to consist of significant amounts of organic matter (Cavalli et al., 2004;Sciare et al., 2009), both watersoluble and water-insoluble.It has historically progressed from mainly consisting of sea salt (SS) and non-sea-salt sulfate (Charlson et al., 1987;O'Dowd et al., 1997) to complex primary biogenic organic mixtures and states (dissolved, particulate, colloidal or nanogel) (Cavalli et al., 2004;Leck and Bigg, 2005;Russell et al., 2010;Decesari et al., 2011) as well as secondary organic compounds like organic acids (Kawamura and Sakaguchi, 1999;Mochida et al., 2002;Turekian et al., 2003;Rinaldi et al., 2011) and recently discovered biogenic amines (Facchini et al., 2008a;Müller et al., 2009).The findings of Ceburnis et al. (2008) and Facchini et al. (2008b) independently confirmed that water-insoluble organic carbon (WIOC) in marine atmosphere has primary origin while water-soluble organic carbon (WSOC) is mainly secondary or processed primary (Decesari et al., 2011); however, studies of Keene et al. (2007) and Russell et al. (2010) evidenced that even WSOC can largely be of primary origin.
After a significant fraction of marine sea spray particles were found to contain biogenic organic matter compounds (O'Dowd et al., 2004), it became even more important to de-Published by Copernicus Publications on behalf of the European Geosciences Union.D. Ceburnis et al.: Marine submicron aerosol gradients, sources and sinks termine principal sources and sinks of marine organic matter.Tentatively, the source of biogenic marine organic matter has been linked to the ocean surface and driven by a biological activity in surface waters based on a seasonality pattern of organic matter and chlorophyll a (Yoon et al., 2007;Sciare et al., 2009) or regression analysis (O'Dowd et al., 2008;Russell et al., 2010).Furthermore, the first quantitative estimate of submicron aerosol organic matter in an oceanic environment was performed by Ceburnis et al. (2011) using dualcarbon-isotope analysis, showing that over 80 % of organic matter in clean marine air masses is of marine biogenic origin.A pilot study based on the concentration gradient method performed in a marine environment by Ceburnis et al. (2008) revealed that water-soluble organic matter is largely produced by secondary processes while water-insoluble organic matter is of primary origin.
This study is the extension of the study by Ceburnis et al. (2008) through the combination of offline chemical analysis of samples and the measurements of boundary layer height, expansion of the range of chemical species and extension of the timescale to consider seasonal pattern of primary production.Contrary to the above study the flux-gradient method was not attempted due to the uncertainty of the averaged coefficient of turbulent transfer derived from the eddycovariance method, but instead sea salt fluxes were estimated by the boundary layer box model.

Experimental methods
The flux of sea spray aerosols (SSAs) has been studied previously as sea salt mass fluxes or aerosol size and number flux (O'Dowd and De Leeuw, 2007;de Leeuw et al., 2011).Apart from a few studies, the flux experiments have typically focused on super-micron-sized particles.The eddycovariance method for studying submicron particle fluxes was first used by Buzorius et al. (1998) estimating submicron particle fluxes and sinks and has been since applied in a variety of environments: boreal and tropical forest (Buzorius et al., 1998;Ahlm et al., 2009), ocean (Nilsson et al., 2001;Geever et al., 2005;Norris et al., 2008;Brooks et al., 2009), desert (Fratini et al., 2007) and urban areas (Mårtensson et al., 2006;Martin et al., 2009).The eddy-covariance method is typically used to study total particles fluxes.The technique has been modified into the relaxed eddy-covariance method to allow studying size-segregated particle fluxes (Gaman et al., 2004) or the disjunct eddy-covariance method (Held et al., 2007) employing slower response instruments.It should be noted, however, that, while number of sea spray particles is dominated by submicron particles, mass is dominated by super-micron sizes and not a single method is capable of measuring particles around the important boundary of 1 µm.None of the above techniques were suitable for studying chemically resolved fluxes, because chemical analysis typically requires long sampling time (many hours for offline chemical analysis).Most recently, however, the eddycovariance system coupled with a high-resolution aerosol mass spectrometer (AMS) has been used to study chemically resolved fluxes (Nemitz et al., 2008;Farmer et al., 2011), but those were largely limited to areas with a relatively high concentration of species.
The study of chemical gradients in a relatively clean marine atmosphere represents a great challenge due to generally low absolute species concentrations and the lack of appropriate experimental methods.The rationale of choosing the gradient method was based on the fact that persistent fluxes must produce concentration gradients with their sign depending on the source and assuming that recurrent eddies allow sampling for a certain number of hours to meet analytical requirements of chemical species.Additional challenges exist when it comes to reactive species (organic matter) due to chemical transformation during transport to the sampling location or extended sampling durations.A combination of continuous production (or removal) of particles and turbulent eddies of varying magnitude within the boundary layer should establish concentration profiles.The profiles, therefore, are a net result of the competition between upward and downward eddies averaged over time.The persistent surface source will manifest itself in a decreasing concentration away from the source.The absence of the surface source should result in an increasing concentration profile as particles are removed to the surface through deposition processes.
A new set-up to study gradients and corresponding flux estimate was installed at Mace Head Atmospheric Research Station on the west coast of Ireland (Jennings et al., 2003;O'Connor et al., 2008), comprising PM 1 samplers installed at three different heights (3, 10 and 30 m).
Lidar measurements (Jenoptik/Lufft and Vaisala ceilometers) are continuously conducted at Mace Head, and a dedicated algorithm for temporal height tracking (THT) (Haeffelin et al., 2012;Milroy et al., 2012) using the backscatter profiles measured by the lidar was used to identify the surface mixed layers (SMLs) and the decoupled residual layers (DRLs), both important parameters when considering boundary layer filled by primary fluxes.
Chemical fluxes were estimated by a simplified box model according to the method outlined by Ovadnevaite et al. (2012): where C was the concentration measured at 30 m height, H BL was the measured boundary layer height averaged over the sample duration and τ was boundary layer filling time (fixed at 2 days with the uncertainty of ±1 day).
The above method was critically assessed by Lewis and Schwartz (2013), who argued that the method can hardly provide information on wind speed dependence.Indeed, estimated fluxes cannot be strictly related to a particular wind speed because the corresponding boundary layer filling time was not constant and the local wind speed may not have been accurate with regard to the wind speed where primary production occurred.However, the method was only used in this study to roughly estimate the fluxes, without attempting to derive a parameterisation but rather to provide an estimate.
The chlorophyll satellite data (daily, 1 • spatial resolution) were obtained from GlobColour (http://www.globcolour.info).They result from the merging of Medium-Resolution Imaging Spectrometer (MERIS), Moderate Resolution Imaging Spectroradiometer (MODIS) and Sea-viewing Wide Field-of-view Sensor (SeaWiFS) data, using advanced retrieval based on fitting an in-water bio-optical model to the merged set of observed normalised water-leaving radiances.A thorough description of the data treatment can be found in Rinaldi et al. (2013).

Sampling strategy
Meteorological records demonstrate that on average marine westerly air masses account for over 50 % of the time at the station (Cooke et al., 1997;Jennings et al., 2003).The gradient measurement system PM 1 samplers (Sven Leckel Ingenieurbüro GmbH) ran in parallel at a flow rate of 38 L min −1 .Samples were collected in clean marine conditions (wind direction 190 < WD < 300 and condensation particle counter (CPC) concentrations < 700 particles cm −3 ) using an automated sampling system on quartz filters for the analysis of both organic and inorganic components of marine aerosol.The system operated day and night whenever the above clean marine conditions were met.Active control of the sampling conditions excluded sampling during occasional short-term spikes of CPC concentrations either due to coastal nucleation events or occasional local ship traffic.Post-sampling analysis revealed that such air masses did not have contact with land for 4-5 days (as confirmed by air mass back trajectories) and that the black carbon (BC) concentration measured by an aethalometer (AE-16, Magee Scientific, single wavelength at 880 nm) did not exceed 50 ng m −3 .Such air masses have typically spent the last 48 h (at least) in the marine boundary layer as documented by Cavalli et al. (2004) and Ceburnis et al. (2011).The latter study quantitatively demonstrated that in clean marine air masses anthropogenic carbon species typically contributed to 8-20 % of the total carbon mass, which should be applicable to other anthropogenic species due to internally mixed anthropogenic aerosol far from the source.It is important to note that clean marine samples collected at Mace Head are representative of the open-ocean environment considering chemical and physical similarities between open-ocean and coastal (Mace Head) samples (Rinaldi et al., 2009).The marine air criteria used at Mace Head were demonstrated to be sufficient at ensuring that anthropogenic and coastal effects are minimised to guarantee a dominant, if not at times overwhelming, natural marine aerosol signal as detailed in the study of O'Dowd et al. (2014).

Offline chemical analysis and concentration gradients
Fifteen PM 1 gradient samples were collected during a 13month period in clean marine conditions as listed in Table 1.
Normalised averaged concentration profiles of all measured chemical species were obtained as follows: for each aerosol component, only samples for which concentrations above the detection limit were observed at all three sampling altitudes were used in data analysis.Normalisation was done by dividing the concentration at every height by the sum concentration of three levels, thus giving the same weight to every profile for averaging purposes.After normalisation, the profile of each mass category was averaged, resulting in statistically meaningful variances around the mean value, and presented as an average and its standard deviation.The normalised averaged concentration profiles allowed classification and categorisation of the profiles, but the normalised data were not used for calculating gradients and fluxes.The main features were similar to the ones documented by Ceburnis et al. ( 2008): decreasing concentration with height, or negative gradient, was common of species produced at the surface by primary processes, while increasing concentration with height, or positive gradient, was common of species produced by secondary processes in the atmosphere aloft or within the marine boundary layer.
Concentration gradients of various chemical species were obtained by a linear fit of the concentration profile (except WSOM).A detailed discussion of potential influence of local sources (surf zone) to the gradient can be found in Ceburnis et al. (2008) and is reconsidered in the "Results and discussion" section.

Errors and uncertainties
The uncertainty of estimated boundary layer fluxes was dependent on the cumulative uncertainty of the measured species concentration, boundary layer height and boundary layer filling time: where the assigned boundary layer filling time uncertainty (50 %) was dominating the combined uncertainty due to relatively small concentration uncertainty (5 %) and boundary layer height uncertainty (10 %).
The relative uncertainty of the organic matter fractional contribution to sea spray (OM ss = WIOM/(WIOM + SS)), where the variable WIOM appeared in both nominator and denominator and WIOM represented total sea spray OM, resulted in a more complicated equation of the combined propagated uncertainty of the OM fractional contribution: where the ratio in front of the square root is the fractional contribution of sea salt in sea spray, resulting in the frac-tional uncertainty of the OM fractional contribution dependent on the sea salt fractional contribution and, therefore, always smaller than the additively combined fractional uncertainty of sea salt and WIOM measurement.
The uncertainty of the fitted functional relationships obtained from the discretely measured values was presented with the 95 % confidence bands, which was conceptually different from the fractional uncertainties of individual values.The confidence bands also helped to define the best-fitted function (e.g.linear or power law) as unrealistic fits had very low or no confidence at all.Typically, the confidence bands become narrower as the number of points increases and/or their scatter decreases.The presentation of the confidence bands provided the physical meaning of the points residing outside the confidence bands.An individual point which is outside the confidence bands suggests a higher order of the relationship or an unaccounted-for freak error.Several such cases will be discussed accordingly.

Results and discussions
The measurements at three different heights allow resolving the vertical concentration profiles of different chemical species and the magnitude of the sources and sinks shaping the profiles.Most of them were non-linear but well interpretable since concentration and flux footprints were studied in detail in the previous pilot study of Ceburnis et al. (2008).It is important to note that the footprint of the measured absolute concentration was many tens to hundreds of kilometres offshore while the footprint of the concentration gradient was within about 10 km from the measurement location, i.e. coastal waters (Ceburnis et al., 2008).The surf-zone emissions may have had certain influence on the concentrations of sea salt or sea spray at the lowest level of 3 m, particularly for low wind speeds, practically disappearing at higher winds (O'Dowd et al., 2014), but had little or no impact on secondary organic aerosol.The different distances of the flux footprint arise from emissions contributing to the concentration at different heights.The flux footprint of the 90 % concentration difference between 3 and 10 m is 0.2-1.2km, while the footprint of the 90 % of the difference between 10 and 30 m extends to 5 km (Fig. 1, Ceburnis et al., 2008).The remaining 10 % of the contribution extends well beyond 5 km, with perhaps 10 km distance being a safe approximation.A condensation potential could have also contributed to the concentration differences of certain species as the time required for the air parcel to cover 10 km distance is about 15 min, which is more than sufficient to achieve gas-aerosol equilibrium (e.g.Meng and Seinfeld, 1996;O'Dowd et al., 2000).

Primary components
The concentration profile of sea salt (top left in Fig. 1) was unambiguously surface-sourced or primary; i.e. concentration was decreasing vertically.Some of the individual profiles were sharper than others, but all were primary with only three exceptions where the profiles were distorted at lower heights possibly partly due to measurement errors and partly due to boundary layer dynamics and changes in sea state during the sampling period (ascending and descending wind regimes).However, as was stated earlier, surf-zone emissions could have had influenced the concentration value at the lowest level of 3 m.
Interestingly, similar "negative gradient" concentration profiles were obtained for nitrate and oxalate.However, those profiles were slightly but repeatedly (systematically) distorted; i.e. the concentrations of oxalate and nitrate significantly diverged from the sea salt one at the lowest sampling height of 3 m while following the sea salt profile above 10 m.It is well established that nitrate is produced by secondary processes and mainly manifesting itself through condensed nitric acid on pre-existing sea salt particles in the absence of anthropogenic ammonium nitrate.Sea salt particles at the lowest level were the freshest, having the closest flux footprint and, consequently, adsorbed the least amount of condensable nitric or oxalic acid compared to higher levels.Similarly to nitrate, oxalic acid could have been condensing on pre-existing sea salt particles as well despite more diverse chemical pathways of oxalic acid (some of the oxalate could also be produced by oxidation of organic matter inside sea spray particles (Rinaldi et al., 2011) and, therefore, manifesting itself as a "primary" species.The concentration profile of oxalic acid was similar to that of nitrate and could indicate that a significant amount of oxalate is produced in the atmosphere aloft, subsequently condensing onto primary sea spray particles due to its acidic nature.
The WIOM concentration profiles were split between three main categories: production (five profiles), removal (six profiles) and mixed profiles (four profiles) (bottom right of Fig. 1).Given that fractional contribution of OM in primary sea spray is related to the enrichment of organic matter at the ocean surface, this range of behaviour can be interpreted in terms of the location of biologically active region relative to the flux footprint.The wind speed has been reported to have an effect on fractional contribution of OM, but quan-titative effect is unclear and will be discussed in more detail in Sect.3.4.The biologically active water patches within the flux footprint (∼ 10 km from the measurement location) were responsible whether WIOM was produced or removed from the surface layer, or a combination of both processes occurred.Therefore, a mixed profile pointed at the production at a longer distance from the coast and the removal close to the measurement location.Thus the removal profile was pointing both at the deposition within the flux footprint area and/or the absence of biological activity in surface waters within the flux footprint area.The WIOM production by the secondary processes cannot be completely excluded either, but we have no evidence of that.It is worth noting that the production profiles were observed in early spring (March until early May), when biological activity is high at the coast, and during late summer (late July-August), when biological activity has a second maximum identified by the chlorophyll proxy (Yoon et al., 2007).In contrast, the removal profile was observed during late spring and early summer, when biological activity retreats away from the coast into the open ocean.Despite a general pattern of the evolution of biological activity presented by Yoon et al. (2007) it should be stressed that biological activity is very patchy all over the ocean, including coastal areas, and the phytoplankton blooms are generally governed by the availability of nutrients, which themselves are supplied by ocean currents and upwelling and become unpredictable on a day-to-week timescale.

Secondary components
The inorganic secondary species (nssSO 4 and NH 4 ) are presented in the top right of Fig. 1, along with an aerosol neutralisation profile considering only ammonium and sulfate which will be discussed later.The ammonium profile was clearly secondary, as expected, due to ammonia being the principal gaseous neutralising agent in the marine boundary layer.It should be noted that the concentration profile of nssSO 4 was pretty constant and did not follow that of the ammonium profile as could be expected considering that sulfuric acid is the main acidic species in the marine boundary layer, typically neutralised by ammonium.NssSO 4 was calculated as the difference between two relatively large numbers (total measured SO 4 minus sea-salt SO 4 as inferred from a conservative tracer such as Na ion).As sea salt concentration changed quite dramatically with height, especially in moderate to high wind speed during winter, some ambiguity must be acknowledged before interpreting the nssSO 4 profile.In fact, if the winter sulfate profiles had been excluded from the average, that would have improved the average profile.In any event nssSO 4 concentrations at three different heights were not significantly different, preventing any conclusions with respect to apparently secondary nssSO 4 .The uncertainty in non-sea-salt sulfate determination can be the reason for the difference with respect to the profile of ammonium.Looking at the profiles, it can be observed that marine aerosol sampled www.atmos-chem-phys.net/16/12425/2016/Atmos.Chem.Phys., 16, 12425-12439, 2016 at Mace Head is more neutralised at 30 m than closer to the sea level (Fig. 1 top right and Fig. 2), even though neutralisation with respect to sulfuric acid is never complete, due to scarcity of ammonia in the marine boundary layer.Figure 2 shows calculated ammonium (considering neutralisation by sulfate only as nitrate was more likely to be neutralised by sodium (causing chloride depletion) due to scarcity of ammonia in the marine boundary layer) versus measured ammonium, revealing significant but consistent differences in neutralisation pattern at three different heights.The neutralisation profile can be driven by the gaseous ammonia vertical profile, which we have no hint about, or can be an indication of the importance of in-cloud processes of sulfate neutralisation considering also that measurements at the lowest level were somewhat perturbed due to surf-zone fluxes.In fact, if the neutralisation of acidic sulfates occurred prevalently in clouds, after scavenging of gaseous ammonia into acidic droplets, this process would occur more likely at the top of the marine boundary layer, where cloud layers form, justifying the observed neutralisation profile.
The secondary organic species (MSA, WSOM and WSON) are presented in the bottom left of Fig. 1.The MSA exhibited a "mixed profile" with steep increase of concentration between 3 and 10 m, typical of secondary products and decreasing profile between 10 and 30 m, likely due to condensation of MSA on sea salt particles (Hopkins et al., 2008) that causes an apparently primary profile.A clear secondary profile was observed for WSOM also, reaffirming the conclusion of Ceburnis et al. (2008) on the secondary origin of WSOM.The water-soluble organic nitrogen concentration pattern is presented in the bottom left of Fig. 1.WSON presents a mixed profile; therefore, it is not possible to attribute it to primary or secondary formation processes unambiguously.WSON concentration in aerosol samples is generally difficult to quantify as it is calculated as the difference between the TN and the WSIN -both numbers of similar magnitude.As a result, only 7 complete profiles could be derived out of 15 samples and should, therefore, be considered cautiously (8 profiles were discarded as incomplete, i.e. missing determined concentration at one or two levels).Along with WSON, aliphatic amines were analysed follow- ing Facchini et al. (2008a).WSON, DMA and DEA are minor constituents of marine aerosol, together typically accounting for 10 % of secondary organic aerosol (Facchini et al., 2008a).While the magnitude of their absolute concentrations may be misleading -amines can be important species facilitating new particle production in the marine atmosphere (Dall'Osto et al., 2012) -quantification of their concentration by offline chemical analysis is always challenging.Mostly concentrations of DMA and DEA at the lowest height were below detection limit, and therefore no profile can be provided for these species with confidence.However, the fact that detectable concentrations were always observed at 30 m strongly suggests a secondary origin for DMA and DEA.
The well-established aerosol chemical compounds such as nitrate, oxalate and MSA as well as less-well-established WSON were all studied for the first time using the fluxgradient method.The concentration profiles of the above compounds have not demonstrated that the species were secondary, despite well-established knowledge of their secondary formation in the atmosphere aloft (boundary layer, clouds or free troposphere) (Seinfeld and Pandis, 2006;Facchini et al., 2008a;Rinaldi et al., 2011).Figure 3 is presented to elucidate an apparent primary profile of nitrate and oxalate which is due to the aforementioned species condensing or reacting with sea spray particles.MSA by contrast has the weakest if any relationship with sea salt.Figure 3 (top left) presents the relationship between nitrate and sea salt mass, which appears to be linear with the exception of two-three outliers.The outliers likely appeared due to the presence of trace amounts of ammonium nitrate.Ammonium nitrate is generally considered as an anthropogenic species and can be present in trace amounts due to pollution background.The trace amount was really small, 20-30 ng m −3 of nitrate, only reaffirming the cleanness of the marine atmosphere studied at Mace Head.Despite a strong similarity in concentration pattern of nitrate and primary sea salt it is inconceivable that a significant amount of primary nitrate can be produced (nitrate is a tracer nutrient in sea water); therefore, it must be derived by condensation of nitric acid on pre-existing sea salt.
The relationship of sea salt and oxalate (top right plot of Fig. 3) was slightly different from nitrate and somewhat similar to MSA.While oxalate can indeed condense on preexisting sea salt particles, its chemical pathways of secondary production are different and more diverse than those of nitrate as was detailed by Rinaldi et al. (2011).Oxalate can also be present in sea spray particles via oxidation of organic matter in sea spray and, therefore, dependent on the biological activity of the ocean.As opposed to nitrate, the oxalate was not enhanced in the presence of copious amounts of sea salt particles, suggesting that oxalic acid is not an everpresent species in the boundary layer which would readily condense on sea salt.The same was true for MSA, which showed even less of a relationship with the sea salt mass (bottom left of Fig. 3).MSA production is photochemically driven and time-limited considering the gradient footprint of 0.2-10 km in the coastal zone.WSON is a relatively less studied class of chemical compounds, of which amines are the best known compounds (Facchini et al., 2008a).The observed concentrations of DEA, DMA and WSON were very similar to the ones documented by Facchini et al. (2008a) in clean marine air masses.Both WSON and the sum of DMA and DEA exhibited a relationship with WSOC (bottom right of Fig. 3); however, only WSOC and WSON were correlated at a significant level (r = 0.58, P < 0.01).Note that the sum of amines is presented in absolute concentration while that of WSON is presented as a mass of nitrogen.The comparison between the WSON and the sum of amines suggested that the amines were likely the dominant species of WSON, albeit difficult to determine due to the detection limit as noted above.The WSOC-WSON-DEA-DMA relationship is presented in Fig. 3 (bottom right) for exploratory purposes as these inter-relationships have not been examined or discussed in the context of marine aerosol.et al. (2011) suggested that fractional contribution of organic matter in sea spray particles depends not only on the biological activity in oceanic surface waters but also on the wind speed at the point of emission.The data of this study were examined according to the approach of Gantt et al. (2011).Figure 4 presents the inter-relationship between the fractional organic matter contribution to sea spray (OM ss = WIOM/(WIOM + SS)); wind speed using the data set of this study, which were not part of the data set used by Gantt et al. (2011); and the chlorophyll a concentration in the open-ocean area upwind from Mace Head as examined in www.atmos-chem-phys.net/16/12425/2016/ Figure 3. Plots of sea salt and secondary species which resembled primary production concentration pattern: SS vs. NO 3 (top left), SS vs. oxalate (top right), SS vs. MSA (bottom left) and WSOC vs. WSON (also plotted as the sum of dimethylamine and diethylamine) (bottom right).Note that WSOC and WSON concentration are presented as micrograms (µg) of carbon and nitrogen mass, respectively, while all other species are reported in absolute species concentrations.Rinaldi et al. (2013).The open-ocean region was 10 × 10 • , or roughly 1000 × 1000 km, upwind from Mace Head.Only WIOM was taken into account in calculating the fractional contribution of OM in sea spray.Notwithstanding the fact that a fraction of measured WSOM was associated with sea spray and formed by processing primary WIOM, quantitative assessment is beyond current knowledge.Both relationships were statistically significant (P 0.01) and explained 58 % of the variance (top plots), suggesting an overlap.The obtained relationships agree well with the relationship reported by Rinaldi et al. (2013) based on an extended data set (reaching 70 % OM fractional contribution at 1.0 µg m −3 ).Further, when the former relationship is coloured by the chlorophyll a concentration in the oceanic region upfront of the measurement location at Mace Head, no apparent pattern can be discerned (bottom plot) apart from general mutual relationship.It can be concluded that while the OM ss dependence on wind speed is significant it may actually be weaker than the OM ss -chlorophyll a relationship due to inter-dependence of wind speed and chlorophyll a -wind speed is higher in winter when chlorophyll a concentration is at its lowest and vice versa -thereby contributing to the excessive variance of OM ss and wind speed.Note that the seasonal relationship between wind speed and chlorophyll is simply a coincidence.For example, the effect of wind speed could be elucidated if very low OM fractional contributions were often observed during summer or high fractional contributions during winter, which was never the case with very few exceptions (only two).However, it is hardly a coincidence that the two points (top right plot in Fig. 4) with rather similar chlorophyll a concentration (∼ 0.4 µg m −3 ) residing outside the 95 % confidence bands are the ones characterised with the lowest and the highest wind speed, reaffirming that the effect of wind speed is real, albeit difficult to separate from the OM ss -chlorophyll a relationship.In conclusion, wind stress is the driver of primary sea spray production, but biological productivity is modifying sea spray chemical composition.Last but not least, it is important to note that the chlorophyll a concentration is only useful as a proxy of biological activity which can affect a fraction of primary organic matter in sea spray in different ways depending on the trophic level interactions (O'Dowd et al., 2015).

Flux estimates and implications for existing sea spray source functions
The SML height obtained from lidar measurements varied in the range of 846-1102 m among the eight periods for which overlapping lidar measurements were available.An occasional formation of nocturnal boundary layer was ignored here due to the nature and resolution of the gradient samples.
It is important to note that the boundary layer filling time constant τ is a feature of a particular low-pressure system(s) arriving at the point of observation in a connecting flow.However, the sample deployment time (7 days) and the actual number of sampled hours within a particular sector prevented estimating the sample-dependent constant τ , which was set at 2 days with the uncertainty of ±1 day.SML measurements were available for 10 out of 15 gradient samples.The sea salt fluxes estimated using Eq.(1) (and using sea salt concentration at 30 m height) ranged from 0.3 to 3.5 ng m −2 s −1 over the wind speed range of 5-12 m s −1 .Concurrently, estimated WIOM fluxes ranged from 0.3 to 2.2 ng m −2 s −1 over the same wind speed range and were inversely correlated with SS fluxes; i.e. the highest SS flux estimates were accompanied by the lowest WIOM fluxes, conforming to fractional OM considerations in Sect.3.2.Given the uncertainty of the estimated sea salt fluxes over the wind speed range it was necessary to compare it with the available sea spray source functions.Equally important was to cover a wide range of methods used to derive fluxes.Figure 5 presents the source functions for which submicron sea salt mass could have been calculated and includes the following: Callaghan (2013), Clarke et al. (2006), Fuentes et al. (2010), Gong-Monahan (Gong, 2003), Martensson et al. (2003), Ovadnevaite et al. (2012Ovadnevaite et al. ( , 2014b) ) and the boundary layer box model estimates of this study.Clarke et al. (2006), Fuentes et al. (2010) and Mårtensson et al. (2003) parameterisations were derived in either laboratory conditions or in situ surf-zone breaking waves and coupled with Monahan and Muircheartaigh (1980) whitecap parameterisation to yield the flux-wind speed relationship.All of the above  2008) paper.One of the primary findings of the Callaghan (2013) work was the importance of choosing the correct whitecap timescale for the discrete whitecap method in particular.Finally, Ovadnevaite et al. (2012Ovadnevaite et al. ( , 2014b) ) and the gradient method of this study used ambient measurement data (real-time AMS sea salt measurements, SMPS measurements and PM 1 gradient measurements, respectively) but were completely independent of each other and different in terms of the utilised methods.It should be noted that, despite the fact that the latter methods estimated net fluxes as opposed to production fluxes measured in the laboratory experiments, deposition fluxes are typically small, in the order of 2-4 % in submicron particle range (Hoppel et al., 2002).The presented parameterisations fall into two regimes as seen in Fig. 4: Clarke et al. (2006), Fuentes et al. (2010), Gong (2003) and Mårtensson et al. (2003) parameterisations exhibit a significantly higher wind speed dependency compared to the more recent parameterisations by Callaghan (2013) or Ovadnevaite et al. (2012Ovadnevaite et al. ( , 2014b)).The split into regimes is even more apparent on a linear flux scale.It must be noted that up until now the majority of global-or regional-scale models used one of the former four parameterisations (Gong, 2003;Mårtensson et al., 2003;Clarke et al., 2006;Fuentes et al., 2010), typically resulting in the over-estimated mass concentrations (e.g.Textor et al., 2006;de Leeuw et al., 2011).It should also be noted that the applicability of the Clarke et al. (2006), Fuentes et al. (2010) and Mårtensson et al. (2003) parameterisations becomes more questionable for higher wind speeds as the divergence between the more recent parameterisations becomes progressively greater and the slope of the dependency curve becomes unrealistically steep.
Figure 5 reiterates the conclusion made by Ovadnevaite et al. ( 2012) that the improvements were needed in both the whitecap parameterisation, now addressed by Callaghan (2013), and the more realistic differential aerosol productivity term recently advanced by Ovadnevaite et al. (2014b).It is reasonable to suggest that the laboratory experiments or the in situ surf-zone breaking waves were most likely unable to realistically replicate air entrainment by the open-ocean breaking waves and consequently formed bubble plumes, resulting in unrealistic whitecap coverage and/or size distributions.The most recently developed parameterisation by Ovadnevaite et al. (2014b) advanced even further by introducing a Reynolds number instead of a commonly used wind speed, thereby removing the uncertainty related to the sea wave state (during rising or waning winds) and implicitly containing sea surface water temperature and salinity, which have both been implicated in altering aerosol production (Mårtensson et al., 2003;Jaeglé et al., 2011;Zábori et al., 2012).
It can be argued that the new whitecap parameterisation of Callaghan et al. (2013) coupled with Clarke et al. (2006), Fuentes et al. (2010) and Mårtensson et al. (2003) parameterisations would bring all of them closer to the more recent parameterisations; however, it is important to make few distinctive comments.While the Gong-Monahan parameterisation has decreased the sea salt mass flux when cou-pled with Callaghan (2013) whitecap parameterisation instead of the original Monahan (Monahan et al., 1982) whitecap parameterisation, the size-resolved flux remains unrealistic due to the arbitrary adjusted submicron size distribution below 0.2 µm (Gong, 2003).A single mode centred at around 100 nm fails to reproduce submicron size distributions observed in ambient air in the stormy maritime boundary layer (Ovadnevaite et al., 2014b).Similarly, Clarke et al. (2006), Fuentes et al. (2010) and Mårtensson et al. (2003) parameterisations, even when coupled with Callaghan (2013) whitecap parameterisation, would still predict concentrations far in excess of observed concentrations.The size-resolved fluxes are crucial in predicting direct and indirect climate effects and have to be benchmarked against the ambient rather than the laboratory measurements unless both reasonably agree.

Seasonality of observed concentrations, gradients and estimated fluxes
The sampling strategy aimed at capturing two samples per month provided that clean marine conditions were prevailing and each sample lasted on average 50 % of the time during the calendar week.In reality, 15 samples were collected covering a full year (April 2008-May 2009) as listed in Table 1.The observed seasonal cycle may not have been typical but allowed examining gradients and corresponding flux estimates associated with varying oceanic conditions throughout the calendar year.The observed chemical species concentrations have been typical of those documented at Mace Head by Yoon et al. (2007) and Ovadnevaite et al. (2014a).Sea salt concentrations and respective estimated fluxes by the box model were generally largest in winter (0.2-0.85 µg m −3 and 1.2-3.5 ng m −2 s −1 , respectively) and smallest in summer (0.08-0.55 µg m −3 and 0.3-1.1 ng m −2 s −1 , respectively) which was mainly due to the wind pattern over the eastern North Atlantic (Jennings et al., 2003;O'Dowd et al., 2014).However, occurrence of a deep low-pressure system in e.g.September 2008 with corresponding high winds resulted in high sea salt concentrations and large fluxes despite the seasonal pattern suggesting otherwise.On the other hand, it has been suggested that sea salt can be replaced in primary sea spray by primary marine OM (Oppo et al., 1999;Facchini et al., 2008b), in which case sea salt fluxes estimated from observed concentrations would become smaller.Vaishya et al. (2012) showed that aerosol scattering dependence on the wind is different between contrasting seasons, suggesting the effect of primary marine OM on sea spray production.However, considering the uncertainties of the estimated flux by the box model, a quantitative proof of the aforementioned effect was not possible.
The WIOM concentrations and gradients revealed a much more complex pattern.The absolute concentrations were lower in winter (0.06-0.19 µg m −3 ) and higher in summer (0.1-0.44 µg m −3 ), following the pattern of oceanic biologi-cal activity recently reaffirmed by Ovadnevaite et al. (2014a).The seasonal variation of WIOM gradients and estimated fluxes, however, was different as the gradients depended on biological activity in the flux footprint region (0.2-10 km from the coast), while the fluxes depended both on the biological activity and wind-speed-dependent sea spray production in the flux footprint area.The three distinct profiles of WIOM gradients presented in Fig. 3 clustered in characteristic periods.The removal gradient prevailed in late spring and early summer, when biological activity was waning close to the coast.Yoon et al. (2007) demonstrated that biological activity revealed by chlorophyll proxy typically started at the coast early in the season and then gradually moved offshore and northward, thereby affecting the WIOM gradients and corresponding fluxes.The production gradient manifested itself during late summer and early spring, reaffirming conclusions made by Yoon et al. (2007) about the presence of two or more phytoplankton bloom peaks during the biologically active season.The mixed WIOM profile prevailed during autumn, when biological activity was waning over the eastern North Atlantic, but at the same time shifting closer to the coast.The spatial resolution of satellite chlorophyll data and the large errors associated with coastal interfaces in particular (Darecki and Stramski, 2004;Gregg and Casey, 2007) prevented exploring the relationship between coastal biological activity and WIOM gradients and WIOM fractional contribution to sea spray.The previous chapter demonstrated that open-ocean biological activity revealed by the chlorophyll proxy upwind from Mace Head correlated well with the WIOM fractional contribution to sea spray, validating the seasonal pattern of WIOM gradients and fluxes.Therefore, despite WIOM gradient profiles being found to be dependent on biological activity in the flux footprint area (0.2-10 km), that did not invalidate a relationship between WIOM and chlorophyll in the open ocean over the eastern North Atlantic.Recently, Long et al. (2014) demonstrated a diurnal signal in primary marine OM production, suggesting that sunlight-mediated biogenic surfactants may have a previously overlooked role.The time resolution of the gradient samples (weekly) and randomness of clean sector sampling during day and night prevented exploring the effect in this study.However, the results of this study do not contradict the above study either as the primary marine OM production would be enhanced in summer compared to other seasons following radiation pattern.

Conclusions
Marine aerosol sources, sinks and estimated sea salt fluxes were studied over the entire year by the gradient method.The chemical gradients of primary species, such as sea salt, and more generally sea spray were found to show strong production flux.The fractional contribution of organic matter in submicron aerosol depended linearly on chlorophyll a concentration, conforming to previously published relationships.The study of certain secondary species (nitrate, oxalate, MSA, WSON) was performed for the first time, revealing not only their mainly secondary origin but also interactions with primary sea spray.The seasonal pattern of concentrations, gradients and estimated fluxes by the box model highlighted complex interactions between biological activity, especially in the flux footprint area and wind-driven sea spray production.The estimated sea salt mass fluxes by the box model compared well with sea salt source functions which used carefully selected ambient measurement data.The critical evaluation of the range of available flux-wind speed parameterisations highlighted significant advances in the development of the sea spray source functions for the benefit of global climate models.

Data availability
The data are available from corresponding author by request due to their uniqueness and further use in future papers.
The Supplement related to this article is available online at doi:10.5194/acp-16-12425-2016-supplement.

Figure 1 .
Figure 1.The gradient profiles of chemical species studied: species resembling primary production (top left); inorganic species resembling secondary production (top right); organic secondary species (bottom left); and water-insoluble organic matter split into production, removal and mixed profiles (bottom right).

Figure 2 .
Figure 2. A scatter plot of sulfate neutralisation by ammonium with respect to sampling height.

Figure 4 .
Figure 4. Effect of wind speed and chlorophyll a concentration on the fractional contribution of organic matter (OM ss ): OM ss vs. WS (top left), OM ss vs. chlorophyll a (top right) and OM ss vs. WS coloured by chlorophyll a (bottom).Individual uncertainties of the flux and wind speed are marked with caps, while the grey area denotes 95 % confidence bands of the fitted parameterisation.

Figure 5 .
Figure 5.A comparison of the most-often-used and recently developed sea spray and wind speed parameterisations in log scale (left) and linear scale (right).The estimated boundary layer fluxes by the box model (green circles) were not fitted against the wind speed due to the uncertainty related to the boundary layer filling time constant constituting the bulk of the total uncertainty.

Table 1 .
Gradient sample weekly collection timescale and the number of hours each sample was sampled during a 13-month period in2008-2009.

Table 2 .
Absolute concentration ranges of measured chemical species.