Quantification of the carbonaceous matter origin in submicron marine aerosol by13C and 14C isotope analysis

Dual carbon isotope analysis of marine aerosol samples has been performed for the first time demonstrating a potential in organic matter apportionment between three principal sources: marine, terrestrial (non-fossil) and fossil fuel due to unique isotopic signatures. The results presented here, utilising combinations of dual carbon isotope analysis, provides conclusive evidence of a dominant biogenic organic fraction to organic aerosol over biologically active oceans. In particular, the NE Atlantic, which is also subjected to notable anthropogenic influences via pollution transport processes, was found to contain 80 % organic aerosol matter of biogenic origin directly linked to plankton emissions. The remaining carbonaceous aerosol was of terrestrial origin. By contrast, for polluted air advected out from Europe into the NE Atlantic, the source apportionment is 30 % marine biogenic, 40 % fossil fuel, and 30 % continental non-fossil fuel. The dominant marine organic aerosol source in the atmosphere has significant implications for climate change feedback processes. Correspondence to: D. Ceburnis (darius.ceburnis@nuigalway.ie)


Introduction
Aerosol particles in marine air can affect climate by acting as nuclei for cloud condensation (Fors, 2010 #177) (Shaw, 1983;Charlson et al., 1987).It has been long postulated that submicron sizes, where cloud nuclei number concentration is important, predominantly comprises sulphate mass; however, recent studies (O'Dowd et al., 2004) revealed significant enrichments of carbonaceous aerosol during periods of high plankton activity, suggesting a biogenic source from both organically-enriched sea-spray and condensable oceanderived organic vapours.Organic matter has been observed in marine aerosol particles for many decades and has been linked to enrichment of sea-spray by biogenic matter transferred from the sea-surface into the tropospheric boundary layer through bubble-mediated production processes (Blanchard, 1964;Hoffman and Duce, 1977;Middlebrook et al., 1998;Oppo et al., 1999;Russell et al., 2010).The enrichment of organic matter in sea-spray has important implications for marine aerosol haze and cloud layers, ultimately contributing to current and future climate change.Apart from identification of methane-sulphonic acid (MSA) and plankton exudates (lipopolysaccharides) (Facchini et al., 2008a), other postulated biogenic organic species (e.g. Published by Copernicus Publications on behalf of the European Geosciences Union.D. Ceburnis et al.: Quantification of the carbonaceous matter origin in submicron marine aerosol specific carboxylic acids, amines and carbohydrates) (Kawamura and Sakaguchi, 1999;Mochida et al., 2002;Cavalli et al., 2004;Facchini et al., 2008a;Russell et al., 2010) are not necessarily exclusive to marine sources.Indeed, all aforementioned compounds, except MSA and plankton exudates, can be of terrestrial origin as well, either natural, anthropogenic or biomass burning.Marine atmosphere can also be significantly perturbed by terrestrial anthropogenic pollution via long-range transport making Northern Hemisphere more polluted than the Southern due to larger land masses.Chesselet et al. (1981) concluded that more than 80 % of atmospheric particulate organic carbon was of continental origin over remote marine areas in late 1970s.
The first long-term assessment of carbonaceous particulate matter sources on a regional scale has been presented by Gelencser et al. (2007).Using radiocarbon, thermal optical, organic tracer analysis and statistical techniques authors were able to apportion carbonaceous particulate matter between fossil fuel, biomass burning carbon, biological particles and secondary organic aerosol.The approach was further developed in a study of Gilardoni et al. (2011) where more organic tracer species and Quasi-Monte Carlo statistical method were used to apportion between eight sources, and which included seasonal analyses.
Stable carbon isotope analysis has been attempted to apportion marine aerosol organic matter on several occasions (Chesselet et al., 1981;Cachier, 1989;Turekian et al., 2003;Narukawa et al., 2008;Miyazaki et al., 2010).A number of studies have even attempted compound specific stable carbon analysis: Turekian et al. (2003) were able to derive isotope ratio of oxalate attributing it to mostly marine precursors; Li et al. (2010) developed a method for isoprene biomarkers; Fisseha et al. (2006) and Wang and Kawamura (2006) were able to perform stable carbon analysis of organic acids; Kim et al. (2005) and Zhang et al. (2009) developed an isotope analysis of polycyclic aromatic hydrocarbons.However, stable carbon isotope analysis is somewhat inconclusive due to inability to separate continental non-fossil and fossil fuel sources which can only be reliably done by radiocarbon analysis (Szidat et al., 2009).Furthermore, without application of strict sampling criteria to separate clean marine and anthropogenicaly perturbed continental air masses, stable carbon apportionment method has relatively large uncertainty.
The carbon isotope 14 C is produced in the upper atmosphere and enters the biological carbon cycle with the relatively constant initial ratio to 12 C (Currie, 2004;Szidat et al., 2009).On the other hand, 14 C is completely depleted in fossil fuels due to radioactive decay.The 14 C/ 12 C ratio therefore elucidates the contributions of contemporary carbon biomass emissions and fossil fuel emissions.Gustafsson et al. (2009) conclusively determined relative contribution of fossil fuel versus biomass burning in brown clouds over South Asia.Further, the ratio of 13 C/ 12 C elucidates carbon emissions associated with different plants, both terrestrial and oceanic, due to preferential photosynthesis uptake routes of heavier or lighter inorganic carbon isotopes (Smith and Epstein, 1971;Maberly et al., 1992).Thus, the combination of ratios of 12 C, 13 C, and 14 C enables the quantification of different carbon source contributions to carbonaceous samples.Such an approach was used by Raymond (2005) and Avery et al. (2006) in an attempt to apportion carbonaceous material in rainwater, but has not been attempted in aerosol phase.
In this study, dual carbon isotope analysis of fine aerosol particles in marine atmosphere has been conducted for the first time demonstrating its potential in source apportionment between three principal sources: marine, terrestrial (nonfossil) and fossil fuel.

Experimental methods
Marine aerosol samples were collected over the N.E.Atlantic at Mace Head (Jennings et al., 2003;O'Connor et al., 2008) from January to November 2006 and on the R/V Celtic Explorer during June-July 2006 (MAP project).Fine particulate matter samples (D 50 <1.5 µm) were collected on a weekly basis (aiming at two samples per month) using automated sector control system to separate clean marine and polluted air masses.
An automated control system separated clean marine air masses by sampling air within 190-300 deg sector at Mace Head (representing North Atlantic Ocean) when condensation nucleus counts (CN) did not exceed 700 particles cm −3 (measured by TSI 3010 particle counter).Active control of the sampling conditions excluded sampling during occasional short-term spikes of CN 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 BC concentration measured by an Aaethalometer (AE-16, Magee Scientific, single wavelength at 880 nm) did not exceed 50 ng m −3 .Such air masses have been spending the last 48 h (at least) in the marine boundary layer as presented in Fig. 1 for selected samples discussed in Figs. 4 and 5 using the HYSPLIT model (Draxler and Hess, 1997).Taking into account the above sampling conditions clean marine samples were among the cleanest possible to obtain in the anthropogenicaly perturbed Northern Hemisphere.Despite the number of more remote regions like Greenland or Northern Canada, the west coast of Ireland can be as clean as the most remote locations in the Northern Hemisphere when exposed to the established oceanic air masses.
Polluted air masses were sampled during all other conditions but clean and thus represented varying degree of pollution from the European continent.Clearly, "clean marine" and "polluted" definitions are somewhat ambiguous, because clean marine samples can be slightly anthropogenicaly perturbed while polluted samples may contain significant amount of biogenic matter.However, that was exactly the scope of the study trying to quantify those contributions.In total 915 sampling hours of clean marine air and 811 sampling hours of polluted air were sampled during the entire year.Samples were collected on quartz filters (SKC Inc.) pre-fired at 900 • C before sampling using a HiVol Sierra Andersen sampler at 1 m 3 min −1 flow rate, equipped with a cascade impactor unit capable of separating particles at 1.5 µm.Samples were frozen at −20 • C until analysis.D. Ceburnis et al.: Quantification of the carbonaceous matter origin in submicron marine aerosol Twelve clean marine samples were collected throughout different seasons with a typical sample spanning approximately 1 calendar week (but for different number of hours depending on clean sector conditions) to accumulate sufficient mass for analysis.In addition, eight polluted samples were collected during all other non-clean-sector marine conditions.Three of the clean sector samples among those twelve were collected onboard the R/V Celtic Explorer cruising outside continental shelf area off the west coast of Ireland during June-July 2006.Chemical and physical similarities between the samples collected at Mace Head and the R/V Celtic Explorer have been discussed in detail by Rinaldi et al. (2009) concluding that clean marine samples collected at the coastal site (under the strict sampling conditions) were representative of the open ocean environment.
Carbon-13 and carbon-14 isotope analyses were performed to quantify the biogenic marine carbon (i.e.carbon derived from marine plants), continental non-fossil carbon (i.e.carbon derived from terrestrial plant emissions and/or biomass burning emissions), and fossil fuel carbon emissions.The carbon isotopic composition of the carbonaceous aerosol is assumed to be practically stable during transport from source-to-receptor sites, even if they undergo chemical transformations.
Isotopic fractionation has, indeed, been demonstrated by Usdowski and Hoefs (1988) which would compromise the above assumption.The study of Wang and Kawamura (2006) suggested isotopic fractionation based on different isotopic ratios of di-carboxylic acids (DCAs).However, one could obtain different delta 13 C ratios of different DCAs if precursors of various DCAs have different sources, e.g.some produced by secondary processes, some by heterogeneous ageing of primary material.
Physicochemical properties of isotopes arise from quantum mechanical effects with lighter isotopes possessing higher vibration energy levels and, therefore, weaker intermolecular bonds (Hoefs, 2009).Also lighter isotope species have lower vapour pressures facilitating faster phase transitions.Consequently, secondary aerosol formation processes can induce isotopic fractionation due to condensation favouring the lighter isotope although for the large organic molecules the effect would likely be small.However, most of the marine organic aerosol is primary in origin according to the latest research (O'Dowd et al., 2004;Sciare et al., 2009;Rinaldi et al., 2010;Russell et al., 2010).If primary organic matter is preserved and only gaining mass through oxidation (addition of oxygen) then the carbon isotope ratio should remain unchanged.Condensation of lighter organic species would drive isotope ratio to more negative values while evaporation -to more positive.Rinaldi et al. (2010) suggested that there is likely a limited amount of truly secondary organic carbon in marine aerosol with the majority being either primary or processed primary material, hence, little organic mass contributed via condensation processes.Fractionation during evaporation in marine aerosol would only be possible if primary sea spray organics would be losing mass during oxidation processes producing small volatile organic molecules when, for example, breaking unsaturated double bonds.Evaporation of entirely primary organic compounds will not change the isotope ratio and is probably hardly possible due to chemical species present (long chain hydrocarbons, typical of phospholipids (Facchini et al., 2008b)).In summary, current knowledge about the origin and nature of marine organic aerosol matter suggests that isotopic fractionation is expected to be small and within analytical uncertainty of ∼0.2 ‰ in marine organic aerosol.It should be acknowledged that there is a lack of specific targeted studies at isotopic fractionation in the aerosol phase.Last but not least is the absence of an isotopic fractionation signature in polluted air masses (practically stable isotope ratio throughout the year) where secondary processes, condensation and evaporation are established phenomena (refer to discussion of Fig. 4).Overall, while acknowledging the issue of isotopic fractionation it is not clear of how to estimate it quantitatively.
Total carbon (TC) concentrations of all samples were performed by an Analytik Jena Multi N/C2100 elemental analyzer equipped with a solid furnace module.Inside the instrument furnace, the sample was exposed to a constant temperature of 950 • C in 100 % O 2 and the TC was determined as the total evolved CO 2 by a non-dispersive infrared (NDIR) detector.
OC/TC ratios of selected samples for 14 C analysis were determined with a commercial thermo-optical transmission instrument (semi-continuous OC/EC field analyzer RT 3042, Sunset Laboratory Inc Hillsborough, NC, USA) using the EUSAAR-2 protocol (Cavalli et al., 2010).

13 C analyses
13 C isotope analysis was performed using a stable Isotope Ratio Mass Spectrometer (IRMS) calibrated by the primary Pee Dee Belemnite (PDB) standard (Garbaras et al., 2008;Garbaras et al., 2009).The filters were analysed with the elemental analyzer FlashEA 1112 connected to the stable isotope ratio mass spectrometer Thermo Finnigan Delta Plus Advantage.An 1/8th part of a filter was placed into the tin capsule and combusted in the oxidation furnace at the temperature of 1020C in excess of oxygen.Later this gas was transferred into the reduction furnace (650C).The water from the sample in the helium flow was removed by the magnesium perchlorate trap.Then the gas mixture was separated in the column PoraPlot Q (50C).Separated gas was delivered to the mass spectrometer ionization cell through the gas distribution device ConFlow III.Before the analysis of a series of samples the calibration CO 2 gas was delivered to the mass spectrometer until the isotopic ratio uncertainty was better than 0.15 %.
The analysis comprised evaluation of the 13 C to 12 C isotope ratio (R), expressed as δ (delta) values and defined as the standard-normalised difference from the reference standard, and expressed as δ 13 C in parts per mill ( ‰): Three replicates of each sample were analysed. 13C analytical error (±1σ ) was mostly within 1 % (0.5-3.0 %) of the value which is listed in Table 1.Filter blanks were measured and were also estimated from regression analysis.Filter blanks for the measurements were obtained by carrying pre-fired filter to the station and loading it into the sampler, but not exposing it for the duration of the sample.Direct analysis of the filter blank was quite uncertain due to very low total carbon blank of the pre-fired quartz filters.The measured δ 13 C value of the filter blank was −25.4±0.9 ‰ (average value of three blanks).An independent method was used to estimate the filter blank from the regression analysis since the blank δ 13 C value was significantly different from the δ 13 C values of clean marine samples and the filter blank impact would be minimal at high carbon loading.The regression analysis of δ 13 C versus total carbon mass is presented in Fig. 2, which yielded −25.6 ‰ filter blank value with a regression correlation coefficient r = 0.84 (P 0.01).The uncertainty range of the intercept was also clearly outside the range of clean marine samples.Therefore, overall, the two approaches agreed very well.Standard  least squares regression analysis is considered of limited reliability when two variables are of very different magnitude.McArdle (1988) suggested that reduced major axis (RMA) regression should be used when the error rate in x exceeds one-third of the error rate in y as it is in our case.However, RMA produced the same result with a blank value of −25.8±0.4 ‰.It should be noted, that the width of the 95 % confidence limit is also dependent on the absence of very low values (IRMS detection limit is around 200 mV in amplitude of m/z 44) and very high values (TC concentration in the clean sector is generally low).As the blank CO 2 value was very low (about 180 mV) we give preference to the regressed value, but they both were very similar.The field δ 13 C samples were corrected for blank contribution using isotope mixing equation and CO 2 signal in millivolts (mV) assuming a constant contribution from filter field blank: (2) where CO 2(blank) and CO 2(sample) are corresponding raw signals in mV of the measured blank (−25.4‰) and the sample obtained by IRMS.Similarly, δ 13 C marine was calculated to exclude fossil and continental non-fossil contribution using Eq. ( 3) and presented in Table 1 and Fig. 4: where CO 2 (non-marine) was calculated from Fig. 2   to calculate marine δ 13 C ratio assuming a certain amount of carbon being non-marine.The concentration of BC in clean marine air masses is always below 50 ng m −3 (in selectively sampled clean marine air masses as outlined in the Methods section), but some of this carbon must be organic carbon (absorption measurements cannot distinguish between black carbon and other absorbing OC species as found by Kanaya et al. (2008) and Lack et al. (2008)).This calculation was based on the assumption that all black carbon was from nonmarine sources and had a δ 13 C ratio of −28±2 ‰ which is independent from the value of −29 ‰ (fossil fuel) and −26 ‰ (continental non-fossil) obtained by error minimisation approach in chapter 2.3.The range of the reference anthropogenic δ 13 C literature values is rather broad: −29.2 ‰ for black carbon (Rumpel et al., 2006) 1.Some of the corrected marine δ 13 C values are more negative than the reference value of −20 ‰ derived from dual carbon isotope source apportionment which highlights the aforementioned ambiguity and uncertainty in non-marine carbon concentration of 50 ng m −3 .Stable Isotope Ratio Mass Spectrometer (IRMS) raw CO 2 signal in millivolts (mV) was compared against independent total carbon (TC) analysis (above Sect.2).A good agreement between TC and CO 2 signal amplitude (m/z 44 signal in mV) presented in Fig. 3 (r = 0.98, P 0.01) showed that Amplitude CO 2 (m/z 44), mV  IRMS system was well tuned to provide consistent results, even though the IRMS is not designed to provide a quantitative TC concentration.

14 C analyses
In a similar manner to 13 C, 14 C is calculated from radiocarbon analysis to determine the separation of fossil carbon from non-fossil carbon, which was performed by accelerated mass spectrometry (AMS) (Szidat et al., 2006;Szidat et al., 2009).Three cases each of clean marine and polluted sample sub-sets had sufficient total-carbon loadings to be analysed for 14 C due to analytical constraints.In addition, 14 C analysis in polluted samples was also conducted for OC.For clean marine samples, 14 C of OC was estimated from the 14 C measurements of TC by subtraction of the EC component (typically <10 % of TC) under the assumption of a nonfossil contribution of 20 % to EC. OC and TC were combusted for 14 C analysis from the filters in pure oxygen at 340 • C and 650 • C, respectively (Szidat et al., 2004;Szidat et al., 2009).Evolving CO 2 was cryo-trapped and then sealed in glass ampoules. 14C measurement was performed by mixing the CO 2 with He and transferring the mixture it into a cesium sputter gas ion source -constructed for the 200 kV mini-radiocarbon dating system MICADAS (Ruff et al., 2007). 14C values of the filter samples were calculated according to Stuiver and Polach (1977) and corrected for a filter blank using Eq. ( 4).Individual sample measurements are summarised in Table 2.The 14 C results in Table 2 are accompanied by total uncertainties (around 3-5 %) including contributions from the 14 C measurements (mainly from counting statistics) and the correction of the field blank.
where the 13 C isotopic fractionation in the sample, irrespective of environment, is taken into account by normalizing to δ 13 C = −25 ‰ and the standard is corrected for its decay since the reference year 1950.The filter blank for 14 C analysis was 0.19 µg cm −2 obtained by analyzing 35 cm 2 of filter material (mixture of blank 1 and blank 2 filters) gaining approximately 6 µg of carbon, with which we successfully performed the 14 C analysis.Both values (the amount in µg cm −2 and the 14 C result) were attributed with conservatively high uncertainties of 30 % (1 sigma) in order to account for additional blank variability which was not covered by the blank filters.

Isotope mixing equations and source contribution
The source quantification was performed using isotope mixing equations applied to both δ 13 C and 14 C isotope ratios as follows: δ 13 TC = δ 13 TC marine × k 1 (5) 14 TC = 14 TC marine × k 1 (6) where k 1 ,k 2 ,k 3 are contributions of marine, continental (all non-fossil sources, whether natural or man-made) and fossil fuel sources The above equations are solved when the calculated isotopic ratios converge to the measured ratios fixing reference source ratios in all sample solutions, but allowing k coefficients to vary for each sample.The above two equations contain two unknowns for each sample and can be explicitly solved.It should be stressed that Eqs. ( 5) and ( 6) can be solved explicitly only if individual source ratios are fixed (i.e.firmly established).In fact, they are not and that makes the solution slightly uncertain.However, the solution is rather tightly constrained with little degrees of freedom when selecting individual source ratios, primarily because Eq. ( 5) clearly separates the marine source and Eq. ( 6) clearly separates the fossil fuel source due to distinctly different source ratios.One can start with any conceivable set of reference source ratios and then fine tune them by error minimisation approach.For marine, continental and fossil sources, the respective δ 13 C ratios of −20 ‰, −26 ‰, −29 ‰, and 14 C ratios of −50 ‰, 100 ‰, −1000 ‰ were obtained.These ratios were obtained using the error minimisation approach according to Eq. ( 6) until the error was minimised to match the individual analytical error: δ 13 TC calc − δ 13 TC meas δ 13 TC meas 2 + 14 TC calc − 14 TC meas 14 TC meas 2 × 100 where δ 13 TC calc and 14 TC calc are the values obtained by Eqs. ( 5) and ( 6) and δ 13 TC meas and 14 TC meas are corresponding measured values.
In the end there was only one unique set of δ 13 C and 14 C values and corresponding k coefficients to match calculated and measured ratios with the smallest error (individual as well as the total error S).Even more so, that obtained typical source ratios (−20 ‰, −26 ‰, −29 ‰, and −50 ‰, 100 ‰, −1000 ‰) must have been the same for all samples while k coefficients varied between the samples.The above δ 13 C values are consistent with literature values of δ 13 C and 14 C values for marine, continental and fossil fuel sources (Raymond, 2005).The uncertainty of the k coefficients was, indeed, somewhat dependent on the reference source ratio values.By varying reference δ 13 C ratios within ±1 ‰ and 14 C ratios within ±50 ‰, the uncertainty of the k coefficients was typically within 5 %.More of the uncertainty and sensitivity analysis will be discussed later.

Results and discussion
The results of the 13 C isotope analysis are shown in Fig. 4. The clean marine aerosol samples revealed distinctly different δ 13 C isotopic ratios when compared to polluted samples.Marine δ 13 C values were in the range of −20 ‰ to −23.3 ‰, with least-negative values occurring during summer months (periods of high biological activity) while polluted sample values exhibited more-negative values over a much smaller range of −24.7 ‰ to −26.3 ‰ with little seasonality seen.The absence of significant seasonal signal, especially when compared winter and summer values, suggests that isotopic fractionation was of limited importance overall as in polluted air masses secondary particle formation processes, condensation and evaporation (all of which should facilitate isotopic fractionation) are established phenomena (Jimenez et al., 2009).The less-negative values of δ 13 C, corresponding to a higher concentrations of 13   of peak organic aerosol enrichment associated with peak biological activity as has been detailed by Yoon et al. (2007) whose data are presented for the purpose of consistency.This enrichment results from carbon isotopic fractionation occurring during photosynthetic carbon assimilation (Degens et al., 1968) whereby marine plants get enriched in 13 C relative to terrestrial plants due to limited supply of inorganic carbon in water during peak biological activity.The isotopic shift in δ 13 C to less-negative values during summer results from accelerated plankton carbon fixation during bloom conditions (Deuser, 1970).Indeed, many of the more recent studies reported less negative δ 13 C ratios of particulate organic carbon (−26.5 to −20.2 ‰) (Bentaleb et al., 1996;Bauer et al., 2002), protein-like fraction, carbohydrates and dissolved organic carbon in ocean water (−21.5 to −20.5 ‰) (Loh et al., 2004).In clean marine air, larger negative values during winter are most likely due to higher availability of inorganic carbon when biological activity and water temperature are at their lowest, according to mesocosm bloom experiments performed by Benthien et al. (2007).Similarly, the morenegative values during summer were observed in samples during a significant storm onboard the R/V Celtic Explorer which most likely resulted from mixing of organic material from greater depths where the competition for inorganic carbon is lower.Distribution of δ 13 C values suggested a lower estimate of a marine sector δ 13 C value which should be at least −20 ‰ or less negative to accommodate any contribution of terrestrial sources (Table 1).This lower estimate has a significantly less negative value than reported in any other studies (Chesselet et al., 1981;Cachier, 1989;Turekian et al., 2003;Narukawa et al., 2008;Miyazaki et al., 2010).The solution of isotope mixing equations also required the marine δ 13 C value of −20 ‰, demonstrating consistency between 13 C and 14 C measurements.
Using Eq. ( 5) it is possible to separate the marine source contribution to total organic carbon using only δ 13 C data due to distinctly different marine source isotope ratio.It must be noted, however, that such source apportionment would have quite a large error, primarily due to a significant overlap of δ 13 C values of continental non-fossil and fossil fuel sources and no constrain on δ 13 C values of either marine or continental (non-fossil + fossil) sources (Eq.2).Following the above considerations and unconstrained δ 13 C values for marine (−19 to −21 ‰) and continental (−26 to −29 ‰) sources, we get marine source contribution for clean marine samples in the range of 55-96 %.Consequently, even using rather conservative range of δ 13 C values we obtain a large uncertainty in marine source contribution.Therefore, without 14 C it is impossible to separate the contribution of fossil fuel sources and thereby continental non-fossil sources.In a similar way it is not possible to separate the marine sources using 14 C data alone.Combining both isotopes allows for a separation of all three sources in a tightly constrained manner.Using Eqs. ( 5) and ( 6) and a subset of six samples where both isotope ratios were measured, the contribution of the three principal sources was estimated and presented in Fig. 5a.For three of the marine samples, between 70 % and 84 % is marine organic, the remainder being fossil fuel carbon (8-20 %) and continental non-fossil carbon (4-10 %).Table 2 exhibits similar results based on 14 C data (within the uncertainty range) given the absence of continental nonfossil source.This marine contribution even makes up to 76-87 % if considering only the OC fraction (Table 2).A small fraction of continental non-fossil source is likely a remainder of North American emissions, while additional fossil fuel source present in clean marine samples (since it is larger than continental non-fossil source) could be from shipping fossil fuel combustion.
By contrast, non-marine (polluted to a different degree) samples comprised between 21 % and 37 % of continental non-fossil carbon and about 37 % fossil fuel carbon with the remainder being attributed to marine carbon (26 to 42 %).There was a certain degree of variability from sample to sample in both clean marine and polluted sample subsets; however, within the subsets samples were similar to each other.On average, as shown in Fig. 5b, carbon in marine samples comprised 79 % marine biogenic carbon, 14 % fossil fuel carbon and 7 % continental non-fossil carbon while in non-marine air masses 37 % are attributed to fossil fuel, 31 % continental (non-fossil carbon) and 32 % marine biogenic carbon.The presence of a marine source in polluted air masses (and of similar magnitude to the continental non-fossil source) has significant implications in interpreting other experimental results where isotope analysis was not performed.Typically, all organic matter in polluted air masses is attributed to terrestrial (continental non-fossil and fossil fuel) sources.
Regarding the contribution of non-fossil sources in polluted air masses we observe about 60 % of non-fossil carbon which is in agreement with the study of Hodzic et al. (2010).However, that same 60 % is split between marine and continental non-fossil sources, reducing the contribution of the latter.Indeed, in many continental locations, marine source contribution would be negligible, but that is not the case in Ireland which is an island surrounded by biologically active waters in the North Sea, the English Channel and the Irish Sea.
We can hypothesize several reasons for the relatively large contribution of fossil sources in continental air masses (especially during the warm season).Firstly, the United Kingdom (UK) and Ireland have very little forested areas (under 10 % in Ireland and just over 10 % in the United Kingdom compared to an average of 46 % in continental Europe) with even lower percentage of true forest ecosystems.It is well established that forests contribute significantly to SOA formation (Carslaw et al., 2010).Secondly, intensive shipping in the English Channel and the Irish Sea can contribute additional fossil carbon to that originating from traffic.The latter can be supported by the fact that in the same air masses passing over the sea we observe significant contribution of marine sources, presumably from English Channel and the Irish Sea.Continental air masses passing over relatively bare land (mainly grasslands) and intensive shipping areas, can possibly explain the observed pattern.It is worth noting, however, that Heal et al. (2011) also observed large percentage of fossil fuel source over the British Midlands up to and above of 50 %.
The other reason for a relatively large contribution of a fossil fuel source in polluted air masses can be the low solubility of particles containing larger percentage of fossil fuel carbon (assuming larger contribution of elemental carbon and significant degree of external mixture).The particles containing continental non-fossil carbon would be more soluble and conceivably derived from secondary processes, facilitating a more efficient washout from the boundary layer.
This study is a continuation of experimental work on sources and chemical composition of marine aerosols being performed at Mace Head over the last decade or more, but this time using a combination of dual carbon isotope analysis.The concept of clean marine air masses being sampled at Mace Head has been discussed in several papers (Cavalli et al., 2004;Yoon et al., 2007;Rinaldi et al., 2010) showing typical air mass back-trajectories and discussing various elemental tracers.This time in addition to back-trajectories (Fig. 1), demonstrating that clean marine air masses (sampled in this study) had no contact with land for at least 5 days before advecting at Mace Head and spent the last threeto-four days in the boundary layer, we show that our classified marine aerosol also exhibited extremely low radon concentrations ( 222 Rn ∼100-200 mBq m −3 and 220 Rn ∼1-2 mBq m −3 ) (Table 3).The latter fact is a quantitative proof of truly marine air masses with no contact with land for about a week (e.g.Biraud et al., 2000).Considering 222 Rn concentration in polluted (continental) air masses of the order of several thousands of mBq m −3 , values presented for marine samples are generally less then 10 % of the continental ones.Allocating some of the 222 Rn measured in marine air masses to continental shelf emissions (upwind of Mace Head) leaves room for extremely low contribution of North American continental aerosol in clean marine air masses sampled at Mace Head.Our results suggest that the North Atlantic marine boundary layer can be exceptionally clean and largely devoid of anthropogenic material even in the generally polluted Northern Hemisphere most likely due to both dry and wet deposition during long-range transport across the North Atlantic Ocean.Indeed, the situation can be different above the boundary layer where the absence of vigorous mixing and often clouds can preserve significant amounts of the terrestrial carbon.
Previous studies of marine aerosol δ 13 C ratios have indicated marine sources significantly contributing to submicron carbonaceous aerosol: 20 % in the submicron mode and 80 % in the supermicron (Chesselet et al., 1981); 38 % in both the sub-and super-micron modes (Turekian et al., 2003); 45 % in the sub-micron mode (Narukawa et al., 2008); 46-72 % in the sub-micron mode (Miyazaki et al., 2010).All those studies only used δ 13 C ratios and were conducted under conditions of varying degrees of pollution and were not subjected to our strict classification of clean marine air.Our results for polluted marine air are consistent with the aforementioned studies but, our results for clean marine air (up to 84 % of carbon was deemed to be marine in origin) are in contrast to these studies as a result of the unique combinations of both δ 13 C and 14 C ratios and strict classification of clean marine air masses.These results are the first quantitative estimates of the magnitude of marine versus anthropogenic source in marine aerosol and they are very well in line with other measurements performed at the site (Cavalli et al., 2004;O'Dowd et al., 2004;Yoon et al., 2007;Dall'Osto et al., 2010) where a dominant contribution of biogenic organic matter to marine aerosol during periods of high biological activity in oceanic surface waters has been demonstrated.The latest study presented by Ovadnevaite et al. (2011) has revealed an even higher contribution of primary biogenic organic matter to marine aerosol (2-3 µg m −3 ).This contribution is far greater than previously determined by off-line measurement techniques and can even exceed typical terrestrial concentrations (1-3 µg m −3 ) (Dall 'Osto et al., 2010).More specifically, an almost negligible contribution of aromatic compounds in clean marine air masses (Cavalli et al., 2004) compares well with a small percentage of fossil fuel source quantified in this study.Also, significant contribution of MSA in regionally polluted air masses (Dall'Osto et al., 2010) supports a significant contribution of marine source in polluted air masses.

Sensitivity and uncertainty analysis
It was mentioned above that the uncertainty of source specific contributions was, indeed, somewhat dependent on the selected reference source values.However, by varying reference δ 13 C ratios within ±1 ‰ and the 14 C ratio within ±50 ‰, the uncertainty of source specific contributions was typically within 5 %, suggesting that the dual isotope method was rather insensitive to the reference delta values of both isotopes.
The range of marine δ 13 C values found in the literature is from −18 ‰ to −23 ‰ (Williams and Gordon, 1970;Bauer et al., 2002).Quite clearly, a reference marine source δ 13 C value of −23 ‰ can not apply as clean marine samples collected during summer had heavier isotope values without even correcting for anthropogenic influence.Polluted samples could utilise isotope value of −23 ‰, however, we strongly believe that same reference marine source δ 13 C value should apply for all clean and polluted samples.If we used −23 ‰ δ 13 C value in marine samples, we would obtain a discrepancy between measurements and calculations in excess of 1 ‰, which is beyond any conceivable analytical error.It is possible to accommodate a marine source δ 13 C value of −18 ‰, but in this case continental non-fossil and fossil fuel source values would have to be even more negative than the ones used in this study in order to bring closure to the equations.However, a fossil fuel source value of −29 ‰ is already at the lower negative end of literature values as discussed above.The source apportionment was, however, quite insensitive to a small variation of either continental non-fossil or fossil fuel source values, as by definition the 13 C method cannot reliably distinguish between continental non-fossil and fossil fuel sources.After all, Bakwin et al. (1998) presented a range of fossil fuel carbon δ 13 C values from a global network of sites with an average value of −28.3±2.4 ‰, which encompasses both of our values, −26 ‰ and −29 ‰.
A non-modern 14 C value of the reference marine source (−50 ‰) may seem unexpected, but we believe it can be justified.Beaupre and Druffel (2009) indicated for the Pacific Ocean near California that 14 C values for surface ocean waters were stable from 1991-2004.Assuming that this is valid for the Atlantic Ocean, we can take the values from Bauer et al. (2002) without correction so that 14 C for Mace Head should be considered as −170 (±80) ‰.However, these are the values not from the open ocean but from coastal (shallow) waters and from the slope (∼100km from the coast).Taking only the slope values, which are closer to the open ocean conditions, we calculate a value −210 (±60) ‰.This shows that open ocean might even be lower.
14 C (POC) and 14 C (DIC) were 30 ‰ and 60 ‰ respectively for the slope and thus also slightly lower than the biogenic value of the atmosphere, which was 120 ‰ in 1994.
In conclusion, we should assume for our campaign −210 ‰ for DOC (slope waters from Bauer et al. (2002) without any correction for the time between 1994 and 2006), +30 ‰ for DIC (value from Beaupre and Druffel (2009) for 2004 -differences between Pacific and Atlantic Ocean should be negligible for DIC) and 0 ‰ for POC (if DIC decreases by 30 ‰ within the 10 years, we can justify the same for POC).A reference selected value of −50 ‰ would lead to a POC vs. DOC ratio of ∼3:1 following a dominant contribution, in sea spray samples, of water insoluble organic carbon attributed to POC (Facchini et al., 2008b).

Conclusions
Dual carbon isotope analysis method was applied to a subset of six clean marine and polluted aerosol samples demonstrating the ability to discern contribution of three principal sources to atmospheric aerosols: marine, continental nonfossil and fossil fuel.We have conclusively demonstrated a predominant (80 %) marine biogenic source for fine carbonaceous particles in clean marine air over the N.E.Atlantic albeit based on a limited number of samples.This biogenic carbon component resides in particle sizes predominantly contributing to cloud nuclei, pointing to a direct link between plankton and marine cloud-climate interactions.In addition, the anthropogenic contribution to marine organic carbon aerosol is still significant, illustrating the role of hemispheric and long-range transport having notable impacts in nominally pristine marine environments.Marine sources can contribute around 30 % to the amount of total carbon in polluted air masses sampled at Mace Head which has not been quantified so far.The presence of marine source in polluted air masses should not be neglected even though its absolute magnitude may be relatively low.
2.  13 C filter blank estimation by regression analysis.

Fig. 4 .
Fig. 4. Seasonal pattern of δ 13 C in clean marine and polluted samples during 2006 at Mace Head (top).Also shown is the clean marine total carbon concentration (TC) for the 2006 data (middle) and for 2002-2005 (Yoon et al., 2007) (bottom).

Fig. 5 .
Fig. 5. (a) Source contribution to organic matter in marine and polluted air samples in terms of fossil fuel carbon (black), non-fossil fuel continental carbon (dark green), and marine biogenic carbon (blue) sources; (b) same as (a) except for average source contributions over all samples for marine and continental air masses.
* δ 13 C marine was calculated without fossil and continental non-fossil fuel contribution and the uncertainty range is corresponding to the uncertainty of non-marine δ 13 C value (−28±2 ‰).

Table 2 .
Carbon 14 analyses results. . is not determined/measurement not performed * 14 C(OC) was estimated from the 14 C(TC) by subtraction of the EC component under the assumption of a non-fossil contribution of 20 % to EC.
‰ which encompasses all of the above values including the ones of this study (the above value of −28 ‰ and the value of −29 ‰ from Sect.2.3.).It should be noted that none of the 13 C techniques can separate fossil and continental non-fossil sources, therefore, there is an inherent ambiguity about fossil versus continental non-fossil carbon in 13 C analysis.Accounting for the variability of δ 13 C fossil fuel values as −28±2 ‰ a corresponding range of corrected marine δ 13 C values is presented in Table Bakwin et al. (1998))3 C value of −30.9 ‰ (Ulevicius et al., 2010); −27.3 ‰(Huang et al., 2006), −25.6 ‰(Ho et al., 2006), −26 ‰(Widory et al., 2004), −27 ‰(Currie, 2000).However,Bakwin et al. (1998)presented a range of fossil fuel carbon δ 13 C values from a global network of sites with an average value of −28.3±2.4

Table 3 .
Radon concentration (mBq m −3 ) during clean marine sampling periods presented in Fig.4of the paper.