Interactive comment on “ Polycyclic aromatic hydrocarbons in atmospheric aerosols and air – sea exchange in the Mediterranean ”

Introduction Conclusions References


Introduction
The marine atmospheric environment is a receptor for polycyclic aromatic hydrocarbons (PAHs) which are advected from combustion sources on land (power plants, biomass burning, road transport).Marine sources may be significant near transport routes (ship exhaust).Long-range transport from urban and industrial sources on land are the predominant sources of PAHs in the Mediterranean (Masclet et al., 1988;Tsapakis et. al, 2003 and2006;Tsapakis and Stephanou, 2005a).A number of PAHs are semivolatile (vapour pressures at 298 K in the range 10 −6 -10 −2 Pa) and, hence partition between the gas and particulate phases of the atmospheric aerosol, influenced by temperature, particulate phase chemical composition and particle size (Keyte et al., 2013).Upon deposition to surface water PAHs partition between the aqueous and particulate (colloidal and sinking) phases and may bioaccumulate in marine food chains (Lipiatou and Saliot, 1991;Dachs et al., 1997;Tsapakis et. al, 2003;Berrojalbiz et al., 2011).They were also found enriched in the sea-surface microlayer relative to Introduction

Conclusions References
Tables Figures

Back Close
Full subsurface water (Lim et al., 2007;Guitart et al., 2010).Semivolatile PAHs may be subject to re-volatilisation from the sea surface (reversal of air-sea exchange), similar to chlorinated semivolatile organics (Bidleman and McConnell, 1995), in case high concentrations in surface water would build up.This had been predicted by multicompartmental modelling for 2-4 ring PAHs for polluted coastal waters and also the open ocean (Greenfield and Davis, 2005;Lammel et al., 2009a) and was indeed observed in coastal waters off the northeastern United States (Lohmann et al., 2011).Field studies in the open sea found net-deposition to prevail whenever determined (e.g.Tsapakis et al., 2006;Balasubramanian and He, 2010;Guitart et al., 2010;Castro-Jiménez et al., 2012;Mai, 2012).However, some 3-4 ring parent PAHs, among them fluorene (FLU), fluoranthene (FLT) and pyrene (PYR), were reported to be close to phase equilibrium in the Mediterranean, Black and North seas (Castro-Jiménez et al., 2012;Mai, 2012), and net volatilisation of FLT and PYR was observed in the open southeastern Mediterranean Sea in spring 2007 (Castro-Jiménez et al., 2012).
The aim of the measurements on board RV Urania was to investigate levels of organic pollutants in summer in Mediterranean air and gain insights about sources and phase partitioning in the aerosol of these substances.This study is on the cycling of PAHs in the marine atmosphere.

Sampling
Samples were taken during the RV Urania cruise, 27 August-12 September 2010 (see Supplement (SM), Fig. S1).PAHs were collected in the gaseous and particulate phases using high volume samplers (Digitel) equipped with one glass fibre filter (Whatman) and one polyurethane foam (PUF) plug (Gumotex Břeclav, density 0.030 g cm PAH sampling on GFF and in PUF can be subject to losses related to oxidation of sorbed PAH by ozone (Tsapakis and Stephanou, 2003).This artifact is species-specific and the more pronounced the higher the ozone concentration and the longer the sampling time.Among the PAHs addressed benzo(a)pyrene and pyrene have been identified as particularly vulnerable to oxidation.Based on such sampling artefact quantification studies (Tsapakis and Stephanou, 2003;Galarneau et al., 2006) and ozone levels (Table 1a) and sampling times (Table S1) we expect that total PAHs are underestimated by up to 50 % in the gas-phase and by up to 25 % in the particulate phase.
With the aim to characterize the potential influences of ship-bourne emissions on the samples, passive air samplers with PUF disks were exposed at 5 different locations on board during 16 days.The PAH levels indicated by these samplers did not show significant differences, hence the ship-based contamination is negligible.

Organic trace substances analyses and quality assurance
For organic analysis all samples were extracted with dichloromethane in an automatic extractor (Büchi B-811).Surrogate recovery standards (D8-naphthalene, D10phenanthrene, D12-perylene) were spiked on each PUF and GFF prior to extraction.
The volume was reduced after extraction under a gentle nitrogen stream at ambient temperature, and fractionation achieved on a silica gel column.Introduction

Conclusions References
Tables Figures
Each one GFF and PUF field blank, b, were taken for high-volume sampling under as far as possible identical manipulations on site.No QFF field blank was taken for impactor sampling, but instead the mean of values of the QFF substrates of the 2 uppermost impactor stages (in total 5) was taken.The respective b value was subtracted from sample values.The limit of quantification needs to take the accuracy of the blank level into account.In lack of a measure for the variation of the field blank, the relative The instrument limit of quantification (LOQ), which is based on the lowest concentration of calibration standards used, was 0.5 ng, corresponding to 0.5-2.5 pg m −3 for high-volume samples, ≈ 8 pg m −3 for impactor samples, 6-10 pg m −3 for semivolatile PAHs determined in passive air samples and up to 200 pg m −3 for non volatile PAHs in passive air samples.
Water samples (2-2.5 L) were extracted immediately after their collection using solid phase extraction on C 18 Empore discs using a vacuum manifold device.Disks were stored closed in glass vials in a freezer and transported to the processing laboratory, PAHs were eluted from disks using 40 mL of dichloromethane.The above listed PAHs were analysed the same way as the the air samples.LOQ was 0.1 ng L −1 .

Other trace constituents and meteorological parameters
Ozone was measured with an absorption method (Teledyne-API model 400A UV) on the top deck (10 m above sea surface).Meteorological parameters (air temperature, humidity, wind direction and velocity) and oceanographic parameters were determined aboard.Introduction

Conclusions References
Tables Figures

Back Close
Full

Models of gas-particle partitioning
The data set is used to test gas-particle partitioning models for semivolatile organics in terms of the organics' mass size distribution and size dependent particulate matter (PM) composition.The models assume different processes to determine gas-particle partitioning, i.e. an adsorption model (Junge-Pankow; Pankow, 1987), and two absorption models (i.e.K OA models; Finizio et al., 1997;Harner and Bidleman, 1998).
Absorption is into particulate OM.We refrain from testing dual adsorption and absorption models (e.g.Lohmann and Lammel, 2004) as no soot data or PM chemical composition data are available.Particulate mass fraction, θ, and partitioning coefficient, K p , are defined by the concentrations in the 2 phases: with c p and c g in units of ng m −3 , c p representing the whole particle size spectrum.
Different models describe different processes to quantify differences in ad-and absorption between compounds.The Junge-Pankow model uses the vapour pressure of the subcooled liquid p 0 L , θ = c J S/(p 0 L + c J S), (data taken from Lei et al., 2002), c J should be approximately 171 Pa cm for PAHs (Pankow, 1987).The aerosol particle surface, S, was not measured and a typical value for maritime aerosols is adopted instead (4.32×10 −7 cm −1 ; Jaenicke, 1988).Harner and Bidleman, 1998, use

Air-sea diffusive mass exchange calculations
State of phase equilibrium is addressed by fugacity calculation, based on the Whitman two-film model (Liss and Slater, 1974;Bidleman and McConnell, 1995).The fugacity Introduction

Conclusions References
Tables Figures

Back Close
Full ratio (FR) is calculated as: with gas-phase concentration C a (ng m 3 ), dissolved aqueous concentration C w (ng m 3 ), universal gas constant R (Pa m 3 mol −1 K −1 ), water temperature and salinity corrected Henry's law constant H Tw, salt (Pa m 3 mol −1 ), and air temperature T a (K).Values 0.3 < FR < 3.0 are conservatively considered to not safely differ from phase equilibrium, as propagating from the uncertainty of the Henry's law constant, H Tw, salt , and measured concentrations (e.g., Bruhn et al., 2003;Castro-Jiménez et al., 2012;Zhong et al., 2012).This conservative uncertainty margin is also adopted here, while FR > 3.0 indicates net deposition and FR < 0.3 net volatilisation.The diffusive air-seawater gas exchange flux (F aw , ng m −2 day −1 ) is calculated according to the Whitman two-film model (Bidleman and McConnell, 1995;Schwarzenbach et al., 2003): with air-water gas exchange mass transfer coefficient k ol (m h −1 ), accounting for resistances to mass transfer in both water (k w , m h −1 ) and air (k a , m h −1 ) (see Bidleman and McConnell, 1995;Zhong et al., 2012; and references therein).

Non-steady state 2-box model
The air-sea mass exchange flux of RET is simulated by a non-steady state zerodimensional model of intercompartmental mass exchange (Lammel, 2004).This 2-box model predicts concentrations by integration of two coupled ordinary differential equations that solve the mass balances for the two compartments, namely the atmospheric marine boundary layer (MBL) and seawater surface mixed layer.Processes considered in air are dry (particle) deposition, removal from air by reaction with the hydroxyl radical, and air-sea mass exchange flux (dry gaseous deposition), while in seawater Introduction

Conclusions References
Tables Figures

Back Close
Full export (settling) velocity, deposition flux from air, air-sea mass exchange flux (volatilisation), and degradation (as 1st order process) are considered.All input parameters are listed in the SM, Table S2.
Atmospheric depositions related to emissions from open fires are assumed to provide the only source for seawater RET.These are available as daily time series for the East Mediterranean domain (28-45 • N, 8-30 • E) through the fire-related PM 2.5 emissions as provided by the Global Fire Assimilation System (GFASv1.0;Kaiser et al., 2012) in combination with an emission factor (207 mg RET in PM 2.5 (kg fuel burnt) −1 ; Schmidl et al., 2008).The fire emissions are averaged over the domain and assumed to disperse within the MBL only.This is justified due to the assumed underestimation of the fire related emissions and ignorance of other (emission) sources.The 2-box model is run for the years 2005-2010, for the east Mediterranean domain (28-45 • N, 8-30 • E) with a 1 h time resolution.Fluxes in the range F em = (0.30 ± 1.46) ng m −2 h −1 (positive defined upward) are simulated (using the initially estimated parameter set, Table S2).
GFAS uses global satellite observations of fire radiative power to estimate daily dry matter combustion rates and fire emission fluxes.The GFAS system partly corrects for observational gaps (e.g.due to cloud cover) and detects fires in all biomes, except for very small fires (lower detection limit of around 100-1000 m 2 effective fire area).

Analysis of long-range advection of air
So-called retroplumes are generated using operational weather prediction model data and a Lagrangian particle dispersion model, FLEXPART (Stohl et al., 1998(Stohl et al., , 2005)).Hereby, 50 000 particles per hour were released and followed backwards in time for 5 days.The model output is a 3-D distribution of residence time.Introduction

Conclusions References
Tables Figures

Back Close
Full 3 Results and discussion

PAH concentrations in air and seawater
The mean total (i.e., sum of gaseous and particulate) Σ25 PAHs concentration is 1.45 ng m −3 (time-weighted), and ranged from 0.30-3.25 ng m −3 (high-volume samples).The spatial variability of PAH levels in the Mediterranean is large, determined by long-range advection (Tsapakis and Stephanou, 2005a;Tsapakis et al., 2006).
The levels found in this study in the southeastern Mediterranean are for most substances lower than found earlier (Table 2).The variability of advection or advection in combination with different routes taken by a vessel being influenced differently by coastal or ship emission plumes may explain this difference.Diagnostic ratios (BAA/(BAA + CHR), FLT/(FLT + PYR); Dvorská et al., 2011) in some of the samples (No. 2, 4, 7, 8, and 15) reflect the influence of traffic and industrial sources.We investigated the potential source distribution of individual samples collected along the cruise (Sect.2.6) and found that indeed maxima of PAH concentrations corresponded with air masses having resided over large urban areas, and, vice versa, low concentrations corresponded with no or benign hit of such areas (illustrated in Fig. S5).This perception is supported by the ozone data i.e., 53 (47-65) ppbv during influence from urban areas but 37 (33-62) ppbv otherwise.It had been pointed out that the source distribution around the Mediterranean may cause a west-east gradient, leading to higher concentrations found in the Ionian Sea and Sicily region (ISS) than in the southeastern Mediterranean (SEM; Berrojalbiz et al., 2011).This gradient is somewhat reflected in our results, as levels in the ISS exceeded levels in the SEM (Table 2a).
The PAH concentrations in seawater near Crete (samples No.

Gas-particle partitioning
Only a small mass fraction of the total, θ = 0.08, is found in the particulate phase, confirming earlier findings from remote sites in the region (Tsapakis and Stephanou, 2005a;Tsapakis et al., 2006; Table 3c).The particulate mass fraction, θ, of four semivolatile PAHs varied considerably along track (see SM Fig. S2).θ is thought to be strongly influenced by temperature and doubling per 13 K cooling was found in a Mediterranean environment (Lammel et al., 2010b) apart from PM composition.An exploration of the temperature dependence is not possible, because of only little temperature variation during the cruise (T mean_sample = 21 − 27 • C, Fig. S2).For similar temperatures higher θ values had been observed at sites on the region influenced by urban and industrial sources (Mandalakis et al., 2002;Tsapakis and Stephanou, 2005b;Akyüz and Cabuk, 2010), which is probably related to the influence of higher organic and soot PM mass fractions.Gas-particle partitioning models (Table 3) underpredict θ, except the Finizio et al., 1997, model for one substance, TPH.θ predicted by the Junge-Pankow (JP) model comes closest.Due to the lack of OC/EC data an extended examination is not possible.
In size-segregated samples PAH was almost exclusively quantified in particle size fractions < 0.25 µm a.e.d.(< LOQ in the other stages; Table 1a).No similar measurements from the open sea or even clean coastal sites exist to compare with.However, the observation here is in line with observations at continental sites in central and southern Europe: there, mass median diameters of PAHs were found to be in the accumulation range, mostly 0.5-1.4µm (Schnelle et al., 1995;Kiss et al., 1998;Lammel et al., 2010b and2010c), but also a second, coarse mode was found (up to 2.4 µm; Chrysikou et al., 2009).

Fugacity ratio and air-sea exchange flux
Fugacity ratios (Fig. 1a) and vertical fluxes (Fig. 1b) could be quantified for FLT, PYR and RET.The uncertainty window of FR = f a /f w = 0.3 − 3.0 is based on the uncertainty Introduction

Conclusions References
Tables Figures

Back Close
Full FLT and PYR are indicated to be close to phase equilibrium, with most of the FR values within the uncertainty range, one sample (No. 1) indicating deposition of FLT and one or two (No. 9 and 13) indicating volatilisation of FLT and PYR, respectively.In comparison with earlier observations of FLT and PYR air-sea exchange in the SEM in 2001/02 and 2007(Tsapakis et al., 2006;;Castro-Jiménez et al., 2012) and considering spatial and temporal variabilities no trend, in particular no reversal of air-sea exchange is indicated.This comparison is detailed in the SM, S2.2.1.
RET, however, is found net-volatilisational throughout most of the cruise (Fig. 1).
Among the highest fluxes (> 50 ng m 3 d −1 ) are some samples with very low FR, < 0.03.Fugacity of RET from water is supported by its Henry's law coefficient (11 Pa m 3 mol −1 at 298 K) which is higher than for CHR (0.53 Pa m 3 mol −1 ) and FLT (2.0 Pa m 3 mol −1 ).
RET is commonly considered as biomarker for coniferous wood combustion (Ramdahl, 1983).A decrease in wildfires could explain the suspected RET volatilisation.Integrated over the domain and the year 2010, fires released 7.2 PJ fire radiative energy, which translates into around 22.2 Gg of PM 2.5 emitted (Fig. 2).Compared to the PM 2.5 emissions of the years 2003 to 2012, the year 2010 had the lowest emissions, equivalent to 46 % of the 2003-2012 mean, and only 18 % of the peak emissions of the year 2007 (Fig. 2d).As typical for the East Mediterranean region, the fire season in 2010 started by the end of June and ended by early October.The Urania cruise measurements took place between 27.8.and 12.9., i.e. towards the end of the main burning season (Fig. 2c).During the first half of the Urania cruise, widespread fire activity was observable in the entire domain, with most intense fires occurring in Southern Italy, Sicilia and along the East coast of the Adriatic and the Ionian Sea (notably in Albania and Greece) (Fig. 2a).
The hypothesis that seasonal depositional input of RET into the surface waters during the fire season (summer) triggers reversal of diffusive air-sea exchange, at least in the year 2010, are tested by box model (Sects.2.5 and S1.3) runs.Two scenar-Introduction

Conclusions References
Tables Figures

Back Close
Full ios are considered, an "Initially Estimated Parameter Set" (IEPS) representing mean values for environmental parameters, and an "Upper Estimate Parameter Set" (UEPS) which represents realistic environmental conditions favouring seawater pollution (SM, Table S3).Simulated diffusive air-sea exchange flux, F aw , during 2005-2010 initialised by the UEPS is shown in Fig. 3a and by the IEPS in the SM, Fig. S3, and during the observations (cruise of RV Urania, 27 August-9 September 2010) initialised by the UEPS in Fig. 3b.
The model confirms the hypothesis that seasonal depositional input of RET into the surface waters during the fire season (July-September, typically in the range sured C a and C w , seems to be dominated by biomass burning in the region in the previous fire season.F aw is predicted highly fluctuating, also during the observational period (Fig. 3b).Even under UEPS the model is underpredicting F aw (Fig. 3b).The sensitivity to input uncertainties (SM S1.2) may explain part of the underestimate, but not up to one order of magnitude.

Conclusions
PAH pollution of the atmospheric Mediterranean environment was below previous observations at the beginning of the decade (2001/02;Tsapakis and Stephanou, 2005a;Tsapakis et al., 2006), also considering possible losses during sampling.This might reflect emission reductions.
Three gas-particle partitioning models were tested and found to underpredict the particulate mass fraction in most of the samples (four PAHs i.e., BAA, TPH, CHR and Introduction

Conclusions References
Tables Figures

Back Close
Full BBF).Although input parameters were incomplete these results confirm the earlier insight that additional processes on the molecular level need to be included, beyond adsorption (Junge-Pankow model) and absorption in OM (K oa models), namely both adsorption and absorption (Lohmann and Lammel, 2004) or even a complete description of molecular interactions between sorbate and PM matrix (Goss and Schwarzenbach, 2001).
Simulations with a non-steady state 2-box model confirm the hypothesis that seasonal depositional input of RET from biomass burning into the surface waters during summer is followed by a period of flux reversal.The volatilisation flux is smaller in magnitude than the net-deposition flux during the previous months, but correspondingly, i.e. higher after intense fire seasons.Future negative emission trends or interannual variability of regional sources may trigger the sea to become a secondary PAH source through reversal of diffusive air-sea exchange.Because of non-diffusive emission from the sea surface, such as aerosol suspension from sea spray and bubble bursting (Woolf, 1997;Qureshi et al., 2009;Albert et al., 2012), the true volatilisation may have exceeded the diffusive flux significantly.

Conclusions References
Tables Figures

Back Close
Full Discussion Paper | Discussion Paper | Discussion Paper | −3 , 50 mm diameter, cleaned by extraction in acetone and dichloromethane, 8 h each, placed in a glass cartridge) in series.Particle size was classified in the particulate phase us-Discussion Paper | Discussion Paper | Discussion Paper | ing high-volume filter sampling (F = 68 m 3 h −1 , model HVS110, Baghirra, Prague) and low-volume impactor sampling (F = 0.54 m 3 h −1 , Sioutas 5-stage cascade, PM 10 inlet, cutoffs 2.5, 1.0.0.5, 0.25 µm of aerodynamic particle size and back-up filter, impaction on quartz fibre filters (QFF), SKC Inc., Eighty Four, USA, sampler Baghirra PM 10−35 ).In total 15 high-volume filter samples, exposed 8-36 h (230-1060 m 3 of air), and 3 low-volume impactor samples, exposed 5 d, were collected.Water sampling was performed using the stainless steel ROSETTE active sampling device equipped with 24 Niskin bottles (volume of 10 L) deployed in water at 1.5 m depth for surface water sampling.
Discussion Paper | Discussion Paper | Discussion Paper | standard deviation (SD) of field blanks from earlier field campaigns, (σ c /b c ), on a highmountain site (high-volume sampling summer 2007, n = 5; Lammel et al., 2009b) and in the Mediterranean (impactor sampling summer 2008, n = 6; Lammel et al., 2010a) was used (σ = (σ c /b c )×b).The identical samplers, sampling and analysis protocols for Discussion Paper | Discussion Paper | Discussion Paper |all analytes had been applied.Values below the sum of the field blank value (from this campaign) and 3 relative SDs of the field blank values (from the previous campaigns) were considered < LOQ(LOQ = b + 3σ).NAP and ACY were excluded from the data set, because of the lack of blank values.The field blank values of most other analytes were below instrument LOQ in high-volume PUF and GFF samples.However, higher field LOQs, up to (6-25) pg m −3 (according to sampled volume of air) resulted forANT,  PYR and RET, pg m −3 for ACE, FLN, PHE and FLT in PUF.Field LOQs of PAHs in impactor QFF samples were below instrumental LOQ for most substances, but in the range (8-15) pg m −3 for ACE, ANT, and FLT, ≈ 55 pg m −3 for FLN, and 120-140 pg m −3 for NAP and PHE.
Discussion Paper | Discussion Paper | Discussion Paper | the log K OA and f OM : log K p = log K OA + log f OM − 11.91; and Finizio et al., 1997, uses only the K OA as predictor (data taken from Ma et al., 2010): log K p = 0.79× log K OA − 10.01.The range of the fraction of OM used here is based on Putaud et al., 2004 (16 % lower limit) and Spindler et al., 2012 (45 % upper limit).
Discussion Paper | Discussion Paper | Discussion Paper | followed by a period of prevailing flux reversal, typically F aw = 10 −2 − 10 0 ng m −2 d −1 , which in the years 2008-2010 started in October and lasted until the onset of the fire season, but eventually started later in the years 2005-2007 (at least under IEPS).The volatilisation flux is predicted smaller in magnitude than the net-deposition flux during the fire season, but correspondingly, i.e. higher after intense fire seasons.The high RET volatilisation flux, indicated by mea- Discussion Paper | Discussion Paper | Discussion Paper |