Air–sea exchange and gas–particle partitioning of polycyclic aromatic hydrocarbons in the Mediterranean

Abstract. Polycyclic aromatic hydrocarbon (PAH) concentration in air of the central and eastern Mediterranean in summer 2010 was 1.45 (0.30–3.25) ng m−3 (sum of 25 PAHs), with 8 (1–17)% in the particulate phase, almost exclusively associated with particles Capsule: In late summer the seawater surface in the Mediterranean has turned into a temporary secondary source of PAH, obviously related to biomass burning in the region.


S1.2 Diffusive air-sea exchange flux calculation
A sensitivity analysis was done to explore the influence of the variabilities of air and seawater temperatures and wind speed (expressed as their standard deviations) during individual sample duration on the air-sea exchange flux (Table S2). eventually also emitted to air from coal and crop residues combustion (Bi et al., 2008;Shen et al., 2012), and eventually emitted to seawater influenced by pulp or paper mill effluents (Leppänen and Oikari, 1999) or by diagenesis (Alexander et al., 1995) in the region.
However, these sources of RET to air and seawater are neglected as expected to contribute insignificantly and to show less inter-annual variability. Moreover, advection of RET into the model domain e.g., from fires in the western Mediterranean is neglected for simplicity.
Temperature and wind speed data were taken from the Iraklion meteorological station (35°20'N / 25°11'E, 39 m above sea level), located close to the centre of the model domain.
Wind speed data were extrapolated to 10 m above sea level assuming neutral conditions all the time (Stull, 1988). Input data are listed in Table S3. Only wind speeds of on-shore winds were considered representative, while periods (hourly data) of off-shore winds observed at Iraklion were rejected, leading to gaps in the time series of predicted F aw . No experimental data for RET lifetime in seawater exist. Degradation rate in seawater is uncertain. It was derived from a model estimated halflife against hydrocarbon biodegradability in freshwater (56 days; BioHCwin; USEPA, 2009), which could be much longer for seawater. A factor of 10 is often applied to estimate lifetime in seawater from data in freshwater (EU, 1996).
Gaseous air and seawater concentrations and the air-sea exchange flux, F aw , are output.
Two scenarios 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 (Table S3).
UEPS considers lower estimates for the atmospheric and seawater mixing layers, the degradation rate in seawater (k OC ) and the export (settling) velocity in seawater (v exp ) and an upper estimate of the of fire-related PM 2.5 emission flux.   (Table S4) are consistent with those based on high-volume sampling (Table 1a), this is not the case for RET (higher concentration in particulate phase size fraction corresponding to <0.25 µm than as total atmospheric concentration from high-volume sampling). This is unexplained and may be related to loss of RET from the QFF.  summer) no trend, in particular no reversal of air-sea exchange is indicated by these two data sets, 3 years apart.

S2.2.2 RET
Under UEPS (Fig. S3), for 6 out of 12 observed (i.e., fugacity ratio-derived) F aw (all > 0) agreement within one order of magnitude is found (underpredicting), the wrong sign (F aw < 0) is predicted for 2 such cases (31.8.2010 and 2.9.2010) and no prediction was possible for 4 such cases. Note that because of a high frequency of nocturnal off-shore winds at the coastal station from where wind speed data were adopted (land breeze at Iraklion, Crete), data gaps in the simulated time series of F aw occur more often during night-time than during day time (visible in Fig. S3). Because of the diurnal variation of temperature and wind speed these data gaps are often corresponding with maxima rather than minima of predicted F aw .
Underprediction could be due to neglected emissions to air and seawater in the region other than fire related (no or little seasonality) or neglect of advection into the model region (similar seasonality as captured emissions). Therefore, also the amplitude of the high frequency (daily) fluctuations could be underestimated. On the other hand, F aw derived from observed concentrations C a and C w is uncertain, too. The biggest contribution is expected to be caused by sampling air and water not simultaneously (but combining short seawater sampling intervals with 10-20 h air sampling periods, often starting or ending when seawater samples were collected).
A sensitivity analysis (section S1.2) was performed to quantify the uncertainty of the calculated flux, F aw , accounting for the variabilities of wind speed, air and seawater temperatures during sampling periods (Table S2). F aw is found most sensitive to wind speed, changes on average for all the samples about 160% when adding or subtracting one SD of wind speed (hourly data) from the mean. The flux is much less sensitive to variation of the air and seawater temperatures, leading to changes of approximately 2 and 9%, respectively, when adding or subtracting one SD from the mean. While the sensitivity of F aw to wind speed would be even higher when based on higher time-resolved data, hourly data appear appropriate considering mixing times of surface waters. This sensitivity to input uncertainties may explain part of the underestimate, but not up to one order of magnitude.