Continental pollution in the western Mediterranean basin : vertical profiles of aerosol and trace gases measured over the sea during TRAQA 2012 and SAFMED 2013

In this study we present airborne observations of aerosol and trace gases obtained over the sea in the western Mediterranean basin during the TRAQA (TRansport and Air QuAlity) and SAFMED (Secondary Aerosol Formation in the MEDiterranean) campaigns in summer 2012 and 2013. A total of 23 vertical profiles were measured up to 5000 m above sea level over an extended area (40–45 N and 2W–12 E) including the Gulf of Genoa, southern France, the Gulf of Lion, and the Spanish coast. During TRAQA and SAFMED the study area experienced a wide range of meteorological conditions which favoured pollution export from different sources located around the basin. Also, several events of dust outflows were measured during the campaigns. Observations from the present study show that continental pollution largely affects the western Mediterranean both close to coastal regions and in the open sea as far as ∼ 250 km from the coastline. The measured aerosol scattering coefficient varies between ∼ 20 and 120 Mm, while carbon monoxide (CO) and ozone (O3) mixing ratios are in the range of 60–165 and 30–85 ppbv, respectively. Pollution reaches 3000–4000 m in altitude and presents a very complex and highly stratified structure characterized by fresh and aged layers both in the boundary layer and in the free troposphere. Within pollution plumes the measured particle concentration in the Aitken (0.004–0.1 μm) and accumulation (0.1–1.0 μm) modes is between ∼ 30 and 5000–6000 scm (standard cm), which is comparable to the aerosol concentration measured in continental areas under pollution conditions. Additionally, our measurements indicate the presence of highly concentrated Aitken layers (10 000–15 000 scm) observed both close to the surface and in the free troposphere, possibly linked to the influence of new particle formation (NPF) episodes over the basin.

to 5000 m a.s.l. over an extended area (40)(41)(42)(43)(44)(45) • N latitude and 2 • W-12 • E longitude) including the Gulf of Genoa, Southern France, the Gulf of Lion, and the Spanish coast. TRAQA and SAFMED successfully measured a wide range of meteorological conditions which favoured the pollution export from different sources located around the basin. Also, several events of dust outflows were measured during the campaigns.
Observations from the present study indicate that continental pollution largely affects the Western Mediterranean both close to coastal regions and in the open sea as far as ∼250 km from the coastline. Aerosol layers not specifically linked with Saharan dust outflows are distributed ubiquitously which indicates quite elevated levels of background pollution throughout the Western Basin. The measured aerosol scatter- 15 ing coefficient varies between ∼20 and 120 M m −1 , while carbon monoxide (CO) and ozone (O 3 ) mixing ratios are in the range of 60-170 and 30-85 ppbv, respectively. Pollution reaches 3000-4000 m in altitude and presents a very complex and highly stratified structure characterized by fresh and aged layers both in the boundary layer and in the free troposphere. Within pollution plumes the measured particle concentration 20 in the Aitken (0.004-0.1 µm) and accumulation (0.1-1.0 µm) modes is between ∼ 100 and 5000-6000 s cm −3 (standard cm −3 ), which is comparable to the aerosol concentration measured in continental urban areas. Additionally, our measurements indicate the presence of highly concentrated Aitken layers (10 000-15 000 s cm −3 ) observed both close to the surface and in the free troposphere, possibly linked to the influence of new

Introduction
Atmospheric aerosols play an important role on climate through their participation to several chemical, dynamical, and radiative processes. At present, still large uncertainties persist in the estimation of the aerosol direct and indirect effects mainly due to the difficulty of fully characterizing their spatial and vertical distribution and properties 5 (Boucher et al., 2013). The Mediterranean region is a complex area where atmospheric aerosols of different origins and types may be found (Pace et al., 2006;Kallos et al., 2007;Gkikas et al., 2012). High levels of anthropogenic aerosol particles and pollutants are measured in the Mediterranean (Lelieveld et al., 2002), which is also indicated as one of the main 10 hot spots for air quality issues (Monks et al., 2009).
The North-Western part of the Mediterranean Basin, due to its proximity to highly polluted industrialized areas (such as the Po Valley in northern Italy and the Fos/Berre in southern France) and large coastal cities (Barcelona, Genoa, Marseilles, Nice, or Valencia), is frequently affected by continental outflows and severe pollution episodes 15 (Mallet et al., 2005;Pérez et al., 2008;Pey et al., 2010). The strength of these episodes is particularly intense during summer when stable meteorological conditions and the high level of insolation promote photochemical reactions and the build-up of ozone and other pollutants (e.g. Millán et al., 2000).
A number of studies have investigated the dynamics of pollution export over the Soriano et al., 2001;Mallet et al., 2005). In particular, the interaction between synoptic circulation and local dynamics, such as orography and sea breezes, has been shown to strongly impact the vertical distribution, layering, and aging of particles along coastal regions (e.g. Millan et al., 1997;Gangoiti et al., 2001;Pérez et al., 2004;Velchev et al., 2011). 10 The capability of reproducing this complexity by air quality models represents a real challenge (Jimenez et al., 2006;Jiménez-Guerrero et al., 2008), and experimental observations gives a fundamental support to test the performances of the model outputs over the Western Mediterranean environment.
The large set of observations conducted in the last decades has permitted to acquire 15 a detailed characterisation of pollution aerosols in the surroundings of the Western Basin. However, at the present time we miss an extensive representation of the mean aerosol load, distribution, and properties in the whole region, in particular over the remote sea. In addition, there is a significant lack of observations over some key areas, as for example the Gulf of Genoa, directly under the influence of the outflow from the 20 highly polluted Po Valley (Velchev et al., 2011). In this study we present measurements of aerosols and trace gas vertical profiles acquired during 24 scientific flights performed with the ATR-42 French research aircraft during the TRAQA (TRansport and Air QuAlity) and SAFMED (Secondary Aerosol Formation in the MEDiterranean) campaigns in summers 2012 and 2013 25 in the framework of the Chemistry-Aerosol Mediterranean Experiment (CHARMEX, https://charmex.lsce.ipsl.fr/). The TRAQA and SAFMED flights explored an extended region of the Western Mediterranean Basin between 40-45 • N latitude and 2 • W-12 • E longitude including the Gulf of Genoa, Southern France, the Gulf of Lion, and the Span- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ish coasts. Measurements were performed over the sea at various distances from the coastline with lidar and in situ instruments. During TRAQA and SAFMED the Western Basin was interested by diverse synoptic conditions which led to the occurrence of different pollution export regimes (Mistral/Tramontane events, outflow from the Po Valley and the Iberian Peninsula) and allowed sampling atmospheric aerosols of different 5 origin and types.
The main objective of the present work is to provide extensive observations of the vertical distribution of aerosols and trace gases related to the export of anthropogenic pollution at the regional scale of the Western Mediterranean Basin. The detailed knowledge of the vertical structure of the atmosphere is very important to understand the 10 impact of continental pollution over the basin.
The paper is organized as follows: in Sects. 2, 3, and 4 we describe the flight trajectories and strategy during TRAQA and SAFMED, the in situ measurements carried out on board the ATR-42 aircraft, and the meteorological conditions observed during the campaigns. In Sect. 5 we present the results. The aerosols and trace gases ver- 15 tical profiles are shown in Sects. 5.1 and 5.2. Section 5.3 is dedicated to analyse the variability of the pollution plume composition and atmospheric structure also in link with the different outflow conditions. Airborne measurements in presence of layers with high concentrations of fine particles are discussed in Sect. 5.4. The main conclusions are reported in Sect. 6. 20 2 Overview over flights Figure 1 shows the trajectories of the flights performed during the TRAQA (20 June-13 July 2012) and the SAFMED (24 July-1 August 2013) campaigns. Research flights were performed with the SAFIRE (Service des Avions Français Instruments pour la Recherche en Environnement, http://www.safire.fr/) tropospheric aircraft ATR-42. The 25 aircraft has a maximum endurance of 4 h. The flight altitude ranges between a minimum of ∼ 60 m over the sea, to a maximum of ∼ 5000 m above sea level (a.s.l.  Table 1. 10 The general flight strategy consisted in plane flights with lidar observations and vertical ascents/descents to sound the vertical atmospheric column (from ∼ 60-100 m to 3000-5000 m a.s.l.) and identify main meteorological and aerosol features, followed by straight levelled runs (SLRs) within the detected aerosol layers. In this study we focus on vertical profiles data. A total of 23 profiles were acquired in 20-30 min each by 15 performing a spiral trajectory ∼ 10-20 km wide. Fig. 1 also identifies the geographical position of each sounding. As shown in Fig. 1 the profiles were performed at different distances from the coastline, from a minimum of ∼ 5-10 km for V31 and V32 to more than ∼ 250 km for V20 and V25, and covered almost all the different sectors of the Western Basin.

Measurements and methods
The basic equipment of the ATR-42 aircraft includes sensors for the measurements of meteorological parameters (pressure, temperature, relative humidity, wind components), radiative fluxes (down-and up-welling shortwave and longwave radiation), and carbon monoxide (CO) and ozone (O 3 ) mixing ratios. Introduction ATR-42 (∼ 93 m s −1 ), samples air at a volumetric flow of ∼ 350 L min −1 . The 50 % passing efficiency of the inlet was tested to be 12 µm diameter. Various sampling lines depart from AVIRAD to connect to different instruments mounted inside the aircraft cabin: (i) a 3-wavelength nephelometer (TSI Inc., model 3563) for the measurement of the dry particle volume total scattering (σ s ) and hemispherical backscattering (σ bs ) coeffi-5 cients at 450, 550, and 700 nm, (ii) a 7-wavelengths aethalometer (Magee Sci., model AE31) for the measurement of the particle absorption coefficient (σ a ) at 370, 470, 520, 590, 660, 880, and 950 nm, (iii) an optical particle counter (OPC) (GRIMM Inc., model 1.129) for the measurement of the particle number concentration over 32 size classes between 0.3 and 32 µm in diameter, (iv) a Condensation Particle Counter (CPC, TSI Inc., model 3775) for the measurement of the total particle number concentration in the diameter range 0.004-1.0 µm; and (v) 3 lines for aerosol sampling on filter membranes and a 4-stage cascade impactor (Dekati Inc.) to measure the bulk and sizesegregated particle composition. In addition, the ATR-42 was equipped with a Passive Cavity Aerosol Spectrometer Probe (PCASP, model 100X) OPC for the measurement 15 of the aerosol number concentration over 31 size classes between 0.1-3.0 µm. The PCASP was installed outside the cabin on the left side of the aircraft fuselage.
In this study we consider measurements of the (i) aerosol scattering coefficient from the nephelometer, (ii) particle concentration from the CPC and PCASP instruments (GRIMM data are not considered since they are available only below ∼ 350 m dur-20 ing TRAQA), (iii) CO and O 3 trace gases from the MOZART analyser, and (iv) meteorological parameters from the ATR-42 sensors. A more detailed description of the nephelometer, CPC, PCASP, and MOZART measurements and their data analysis is provided in the following sections.
The present analysis is based only on measurements obtained in cloud free condi-

Aerosol scattering coefficient
A three-wavelength integrating nephelometer has been used to measure the dry particle volume total scattering (σ s ) coefficient at 450, 550, and 700 nm. The sampling flow rate was 30 L min −1 . Data were acquired at 6 s resolution during TRAQA and 1 s resolution during SAFMED. The instrument was calibrated prior the campaign with 5 free-particle air and CO 2 as gases of low and high known scattering coefficient. Nephelometer measurements have been corrected for angular truncation and Lambertian non-idealities by applying the formulae by Anderson and Ogren (1998). The measurement uncertainty on σ s is calculated taking into account for the photon counting, gas calibration, and angular corrections uncertainties (Anderson et al., 1996;Anderson and 10 Ogreen, 1998). The total uncertainty on σ s is estimated to be lower than 10 % at the three wavelengths. The nephelometer measured the scattering coefficient in dry air conditions. This is due to the heating of the airflow while entering the aircraft cabin and the temperature in the cavity of the instrument. The relative humidity measured during the flights in- 15 side the nephelometer was < 25 % in more than ninety percent of cases, with values up to ∼ 40 % occasionally observed at very low altitudes (< 200 m) over the sea surface. A possible underestimation of the scattering coefficient may thus occur in case of hygroscopic aerosols, especially under high relative humidity conditions in the atmosphere.

20
The particle scattering Ångström exponent (α s ) has been calculated from spectral nephelometer measurements with a power-law fit of the measured scattering coefficients vs. wavelength.

Aerosol particle number concentration
The vertical profiles of the total particle number concentration in the Aitken (dN Aitken , PCASP measured at a sample flow of 1.5 and 0.06 L min −1 , respectively, and with a time resolution of 1 s for the PCASP and 5 and 1 s for the CPC during TRAQA and SAFMED, respectively. The PCASP was factory calibrated with monodisperse polystyrene sphere latex (PSL) whose complex refractive index at the instrument operating wavelengths 5 (632.8 nm) is 1.59-0i. The measured sphere-equivalent optical diameter has been converted to a sphere-equivalent geometrical diameter (D g ) by taking into account the complex refractive index of the sampled aerosol (Liu and Daum, 2000). Given that in the very large majority of cases the aerosol sampling during TRAQA and SAFMED was associated to the export of pollution plumes, only pollution aerosols have been 10 considered for OPC correction. Note that these data are not optimized for dust or marine aerosol observations. A large interval of values (n ∼ 1.50-1.72, k ∼ 0.001-0.1 for UV-visible wavelengths) are reported in the literature for the real and the imaginary parts of the refractive index for anthropogenic aerosols over Europe (e.g., Ebert et al., 2002Ebert et al., , 2004Müller et al., 2002;Mallet et al., 2003Mallet et al., , 2011Chazette et al., 2005;Raut 15 and Chazette, 2008). For our calculations at 632.8 nm we have fixed the imaginary part of the refractive index to 0.01, thus representing a mean absorbing aerosol, and then we have varied the real part between its minimum (1.50) and maximum (1.72) reported value. D g is then set at the mean ± one SD of the values obtained for the different values of n. We assume in these calculations that the refractive index does not 20 vary with height. After refractive index correction the D g range for the PCASP becomes 0.10-4.47 µm, with an uncertainty between 1 and 25 %. The smallest and the largest size bins of the OPC, for which the minimum and maximum edges respectively are not defined, have been excluded from the datasets, thus reducing the PCASP D g range to 0.11-4.17 µm. 25 Once corrected for the refractive index, PCASP data have been combined with those from the CPC to calculate dN Aitken , dN Acc , and dN Coarse . Values for dN Acc and dN Coarse are obtained by integrating the PCASP number concentrations in the 0.1-1.0 and 1.0-4.17 µm ranges, while dN Aitken is estimated as the difference between CPC concentra- tion and the integral of PCASP data between 0.1 and 3.0 µm. The comparison between the PCASP and the GRIMM below 350 m altitude indicates that the former underestimates by about 50 % the aerosol concentration in the range 0.4-1.0 µm (the accuracy of the GRIMM has been verified by optical closure study against simultaneous aircraft nephelometer measurements). This is estimated to induce a ∼ 20 % underestimation of 5 the dN Acc calculated here. Conversely, the PCASP underestimation in the 0.4-1.0 µm range has almost a negligible impact on dN Aitken . CPC measurements, and so dN Aitken data, were not available during SAFMED flights V49, V50, and part of V51. 10 Carbon monoxide (CO) and ozone (O 3 ) mixing ratios were measured by the MOZART instrument described in detail by Nedélec et al. (2003). CO is a long-lived tracer for air masses influenced by combustion processes, whereas O 3 in the troposphere is a photochemical product of the oxidation of CO and volatile organic compounds (VOCs) in the presence of nitrogen oxides (NO x ). CO and O 3 are measured at a resolution of 30 15 and 4 s, respectively. The nominal uncertainty is ±5 % for CO and ±2 % for O 3 (Nedélec et al., 2003). However, a recent airborne intercomparison in May 2014 in the framework of the French ChemCalInt project and the TGOE European Joint Research Activity has suggested a greater uncertainty (up to 30 %) on CO measurement by MOZART onboard the ATR-42 (A. Borbon, personal communication, 2015). Trace gas analysis will 20 focus mostly on the vertical distribution of the ∆O 3 /∆CO ratio rather than absolute concentrations (see Sect. 5.3) and the uncertainty on CO should not affect data interpretation.

STP conversion
In order to compare measurements obtained at different altitudes the data presented 25 here are reported at standard temperature and pressure (STP) using T = 293. 15  Hence, the scattering coefficient is scaled to STP conditions and the particle concentrations are given as particles per standard cm −3 (s cm −3 ). For a generic parameter x measured at the temperature T and pressure p, the conversion at STP is done with the formula: (1) CO and O 3 do not need to be corrected for STP since the mixing ratio does not depend on temperature and pressure.

Meteorological parameters
The vertical profiles of the pressure (p), the temperature (T ), the relative humidity (RH) 10 and the wind components towards the east and the north (U, V ) have been used to analyse the atmospheric structure during flights. Starting from the measured parameters the potential temperature (θ) has been also calculated as θ = T p 0 /p 0.286 with p 0 = 1013.2 mbar. For each profile the height of the marine aerosol boundary layer (MABL) and planetary boundary layer (BL) has been estimated visually by looking at 15 the vertical gradients of T , θ, and RH. Meteorological parameters have been also used to calculate the vertical profiles of the gradient Richardson number (Ri ): (2) In Eq. (1) g is the gravitational acceleration and z is the height. The Ri number is the ratio between the buoyancy force and the wind shear and it is used to indicate dynamic 20 stability and the formation of clear air turbulence. Turbulence can develop when Ri is below the critical threshold Ri crit = 0.25, while it is inhibited for Ri > 1 (e.g., Wallace and Hobbs, 2006

Tracking the air mass back-trajectories
The Lagrangian trajectory model FLEXTRA (FLEXible TRAjectories, Stohl et al., 1995) has been used in selected cases to track the origin of sampled air masses. Five days 5 three-dimensional back-trajectories have been calculated using the ECMWF (European Centre for Medium-Range Weather Forecast) operational analysis with a 0.5 • by 0.5 • horizontal resolution and up to 30 vertical model levels below 4000 m. The model specific humidity and potential vorticity is also interpolated along the trajectory path.
4 Meteorological conditions, aerosol load, and pollution export regimes 10 In order to characterize the general aerosol conditions encountered over the Western Mediterranean Basin during the TRAQA and the SAFMED campaigns we have plotted the time-series of the aerosol optical depth (τ) at 440 nm and the 440-870 nm Ångström exponent (α) measured at the three AERONET stations of Barcelona, Frioul, and Ersa located around the basin. Data are shown in Fig. 2 and correspond to the pe- 15 riod of the campaign of measurements plus 10 days before and after. Table 1 reports the date, location, and main meteorological and export conditions encountered during TRAQA and SAFMED flights. The TRAQA campaign in 2012 was characterized by very variable meteorological conditions which prevented the accumulation of high levels of pollutants over the basin.

20
The aerosol optical depth was below 0.2 before the beginning of the campaign over the whole basin and increased to ∼ 0.3-0.5 (with 1 < α < 2) in the periods 23-26 June and 3-13 July. Isolated peaks of τ were measured in correspondence of two Saharan dust intrusion events which occurred on the 17-23 June (τ ∼ 0.6) and 29 June 2012 (τ ∼ 1.4). Different wind regimes occurred during TRAQA and favoured the continental Introduction  (Grell et al., 2005) at 925 mbar are shown in Fig. 3 for 26 June and 3 July 2012. Main observed meteorological/export conditions can be summarized as follows: (i) on 26-27 June north/north-easterly winds blew across northern Italy determining an air mass outflow towards the Gulf of Genoa (measurements 5 on flights V18-V19-V21), (ii) on the same days a strong Mistral-Tramontane episode favoured the dispersion of pollutants towards the central part of the Western Basin. Measurements during the event were performed during flight V20, (iii) on 3-4 July the wind regime was dominated by weasterly/south-weasterly winds mostly blowing at the surface across the Iberian Peninsula and southwestern France. This condition allowed 10 measuring the export of pollution from the Spanish coasts, in particular close to the area of Barcelona (flights V24-V25-V26, see Fig. 3). Additionally, flight V31 sounded the atmospheric structure close to the Spanish coasts reaching the southern urban area of Valencia, (iv) moderate Mistral episodes occurred on the 6-7 and 11 July 2012. In those cases the Mistral wind combined with a persistent westerly flow thus yield- 15 ing pollution export towards the central and central-eastern part of the Western Basin, as measured during flights V27-V28-V30-V32, (v) finally, Saharan dust aerosols were sampled during flights V16 and V20 (episode of the 17-23 June) and flights V22 and V23 (episode of the 29 June). During SAFMED the meteorological conditions were more stable and two distinct 20 phases were observed: (i) a stable anticyclone affected the whole Western Mediterranean area during the first half of July until the 26th, thus possibly favouring a more pronounced accumulation of pollution in this part of the basin. Relatively high values of both τ (∼ 0.2-0.8) and α (∼ 1-2.5) were measured at the three sites of Barcelona, Frioul, and Ersa in this period, (ii) a cyclonic system moving from the Atlantic region 25 towards Europe then affected the Western Basin on 28-29 July 2013. Very clean conditions (τ < 0.1-0.2) were measured afterwards over the entire region until the end of the SAFMED campaign. Winds were mostly westerly/south-westerly in the first period of the campaign (24-29 July 2013, flights V46, V47, V48, V49, V50), which means that ACPD 15,2015 Continental pollution in the Western Mediterranean Basin the sampled air flow came mostly from the sea. Then, from 30 July to 1 August 2013 a north-easterly flow affected the SAFMED investigated area thus promoting the export of pollution from Northern Italy towards the Gulf of Genoa (flights V51, V52). A strong Mistral event (29 July-1 August) and two Saharan dust outbreaks (27-28 July and 1 August) affected the Western Basin, however not influencing vertical profile observa-5 tions during SAFMED. In order to identify the distribution of observations during TRAQA and SAFMED as a function of the aerosol type we have plotted in Figure 4 the distribution of the measured scattering coefficient σ s at 450, 550, and 700 nm as a function of the calculated scattering Ångström exponent α s for all vertical profiles. The plot shows a similar scat-10 tering intensity between cases dominated by coarse particles (α s < 0.5-1.0), such as desert dust, and those dominated by fine particles (α s > 1.0-1.5), such as pollution aerosols. For both dust and pollution σ s peaks at about 100-120 M m −1 . The frequency of occurrence of α s shows that pollution plumes represent the large majority of the cases observed, with more than 70 % of measurements with α s > 1.0. Figure 5 shows the box and whisker plots of the aerosol scattering coefficient σ s at 450, 550, and 700 nm, particle number concentration in the accumulation (dN Acc ) and coarse (dN Coarse ) diameter ranges, and CO and O 3 measured in the boundary layer (BL) and in the free troposphere (FT) within pollution plumes for all the different vertical 20 soundings analysed in this study. This plot summarizes the range of values observed during TRAQA and SAFMED. Even within the single BL and FT the different parameters show a large variability that will be explored in the following paragraphs.  Figure 6 shows the vertical profiles of σ s , dN Acc , and dN Coarse during TRAQA and SAFMED flights. The date, time and coordinates of each profile, as well as the heights of the top of the marine and planetary boundary layer (MABL and BL) estimated from meteorological data are also indicated in the plot.

5
For the different vertical soundings the particle concentrations dN Acc and dN Coarse vary in the range ∼ 100-3000 s cm −3 and ∼ 5-4000 s cm −3 , respectively, for plumes with σ s between 10 and 120 M m −1 . The profile of the aerosol scattering coefficient is mostly correlated to dN Acc , and this also reflects the pronounced spectral variability of the scattering coefficient, typical of pollution/anthropogenic particles. dN Coarse also 10 contributes to the scattering signal in some cases especially at high altitudes (see V16, V20, V21, V22, and V23 above ∼ 2000 m). These observations are associated to the dust intrusion episodes which occurred in the Western Mediterranean Basin during TRAQA, which however will not be analysed in detail here. Aerosol layers affected by dust have been labelled with a "D" in Fig. 6. 15 Maxima of the scattering coefficient have been measured for TRAQA flights V21 and V23 (∼ 120 M m −1 for pollution in the BL and ∼ 100 M m −1 in the dust layer), whereas flights V46-V48-V49, during the first and more polluted phase of SAFMED, are the richest in dN Acc (1500-3000 s cm −3 over the whole column). Minima of σ s and dN Acc are obtained for flight V51 at the beginning of the second SAFMED phase.

20
Pollution plumes observed in the different flights extend from the boundary layer to the free troposphere up to 3000-4000 m altitude. The vertical structure of the aerosol scattering coefficient/particle concentration is linked to the variability of the atmospheric thermodynamic structure and is generally characterized by a first layer confined in the MABL (< 400 m, profiles V16, V20, V22, V25, V48, V51), followed by one or more  . The highest values of both the scattering coefficient and dN Acc for pollution are found within the MABL or BL in most cases, while a local minimum of σ s and dN Acc is generally identified at the top of the BL. The scattering coefficient and the particle concentration measured in the FT are comparable with the values observed in the BL, and in few cases even larger (V25, V26, V30). Only in one case (profile V31) σ s and 5 dN Acc decrease monotonically with height. The aerosol vertical distribution, both in the BL and in the FT, often presents a strongly stratified structure characterized by the presence of several thin sub-layers within one main identified aerosol plume, as it can be seen in particular in the dN Acc profiles (V20, V21, V22, V25, V46, V49). The particle concentration in the Aitken mode (0.004-0.1 µm; dN Aitken , not shown in Fig. 6) is generally below 5000-6000 s cm −3 at all altitudes up to 4000 m within pollution plumes. dN Aitken is correlated with dN Acc in most of the observed cases, which indicates the common source of particles in these two size ranges. Few layers exceeding ∼ 10 000-15 000 s cm −3 are observed occasionally both in the BL and in the FT. These will be discussed in more detail in Sect. 5.4.

15
The dN Acc and dN Aitken measurements within the BL and in the FT over the sea are comparable with the values measured close to the surface at urban continental sites under pollution conditions (Petzold et al., 2002;Mallet et al., 2005;Weigner et al., 2006;Junkermann, 2009;Hamburger et al., 2012). This suggests that the export towards the basin favours the redistribution of the pollution plumes along the vertical. Because of 20 mixing in the BL, measured concentrations within the BL can be as high as those observed close to the surface at urban continental site under pollution conditions. Values of dN as high as in the BL are observed in the FT because of transport in specific conditions, as discussed below.
The observations of aerosol profiles obtained during TRAQA and SAFMED are rep- 25 resentative of the complex transport regimes which characterizes the export towards the Western Basin and that is mostly determined by the interaction between regional meteorology and local dynamics (e.g., Gangoiti et al., 2001). A first example is associated to the measurements in the area of Barcelona. As discussed in Pérez et al. (2004)  the presence of mountains up to ∼ 500-3000 m altitude a few kilometres inland favours, during summertime, the recirculation of pollutants along the coasts of Spain. In these cases, the aerosols emitted at the surface in coastal areas are transported inland and uplifted by sea breezes and mountain winds then the plumes are re-injected at different altitudes and distances from the coast. During the TRAQA flights V24, V25, and  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ficient R 2 ∼ 0.5-0.8) within measured pollution plumes, and also correlated with σ s and N Acc both in the BL and in the FT, which indicates photochemically active plumes. CO is generally higher in the BL, and shows absolute maxima in the lowest levels (V20, V21, V24, V28, V46), then it decreases in the FT. Ozone presents a more complicated vertical structure due to the different photochemical and dynamical processes which 5 control its formation and distribution. At first, local peaks of O 3 correlated with CO are observed in correspondence of pollution plumes both in the BL and in the FT. An absolute maximum of O 3 is sometimes found near the top of the BL (V24, V25, V30) possibly due to aged air masses trapped in the boundary layer. Isolated peaks of O 3 (∼ 75-80 ppbv) not correlated with aerosols and CO are also measured in few cases 10 above 3000-3500 m (V21, V25, V27, V28, V52). The analysis of back-trajectories indicates that these high-altitude ozone layers are associated to the descent of air masses travelling at about 7-8 km, which thus may suggest a downward transport from the upper troposphere or the tropopause region due to a stratosphere-troposphere exchange (Ancellet and Ravetta, 2005). Finally, absolute minima of O 3 (∼ 15-30 ppbv) 15 are measured within the dust layers during flights V20 and V21, maybe related to the dust/ozone heterogeneous reactions which leads to O 3 destruction, as documented in several studies (Bonasoni et al., 2004;Haywood et al., 2011).

∆O 3 /∆CO and dN Aitken /dN Acc ratios and variability of pollution plume composition 20
Using the O 3 , CO, dN Aitken and dN Acc measurements we have estimated: -The O 3 -CO enhancement ratio (∆O 3 /∆CO), i.e. the ratio of the ozone to carbon monoxide variations compared to their baseline values. The ∆O 3 /∆CO enhancement ratio is frequently used to estimate the efficiency of O 3 formation and its export (Parrish et al., 1993;Zhang et al., 2006). From our observations (Fig. 7 -The Aitken to accumulation number ratio (dN Aitken /dN Acc ), which defines the relative importance of particles in the Aitken and accumulation modes. dN Aitken is generally associated to gas-to-particle conversion and nucleation events and is higher in fresh plumes, while it decreases with the increasing of the plume lifetime due to coagulation or condensation of water-soluble chemical species on the 5 particle surface (Kulmala et al., 2004).
The combination of ∆O 3 /∆CO and dN Aitken /dN Acc has been used to retrieve additional information on the atmospheric vertical structure, layering, and particle aging. Within detected pollution plumes the ∆CO and ∆O 3 reach up to 100-120 ppbv and 45-55 ppbv, respectively, with a corresponding ∆O 3 /∆CO ratio which varies in the 10 range ∼ 0.10-2.0 for all cases. These values are comparable with the range of observations available in the literature for fresh and moderately aged pollution plumes in the BL and in the lower FT (Chin et al., 1994;Parrish et al., 1998;Zhang et al., 2006;Cristofanelli et al., 2013). dN Aitken /dN Acc is between about 1 and 20 in most of pollution cases, which indicates the presence of both fresh layers rich in Aitken parti- 15 cles and aged plumes poor in Aitken particles. Extremely high values of dN Aitken /dN Acc (∼ 50-100) are measured in few cases in layers with very low dN Acc concentrations.
The large variability in ∆O 3 /∆CO and dN Aitken /dN Acc indicates a strong heterogeneity in terms of composition and lifetime for the different observed plumes. This heterogeneity reflects the complexity in terms of sources, production processes, and transport

V19: export from northern Italy/Po Valley
The profile shown for flight V19 (Fig. 8a) is characterized by the presence of two different aerosol structures: the first one below 800 m, characterized by low dN Aitken and relatively high O 3 concentrations (dN Aitken /dN Acc ∼ 1-5, ∆O 3 /∆CO ∼ 0.4-1.5), possibly associated to moderately aged pollution, and the second one between 800 and 5 2600 m, very rich in fine particles (dN Aitken /dN Acc ∼ 5-15) so possibly associated to fresh emissions. The export of fresh pollution at high altitudes from northern Italy as observed in V19 may be associated to the peculiar orography of this region and the uplift of continental air masses. This is confirmed by the analysis of the back-trajectories ( Fig. 9)  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 5.3.2 V20: export during a Mistral/Tramontane event V20 provides an example of export during a Mistral/Tramontane event. As shown in Fig. 8b, winds from the northwest direction are measured at all altitudes during flight V20. The aerosol profile in the BL is characterized in the first 400 m by the presence of a layer richer in CO (100 ppbv close to the surface) and dN Aitken (dN Aitken /dN Acc > 20) 5 possibly linked to fresh pollution, followed by the alternation of several layers characterized by a variable dN Aitken (1000-6000 s cm −3 ) and lower CO (∼ 70 ppbv). For all these layers the O 3 is very low (∼ 30-40 ppbv) and the ∆O 3 /∆CO ratio is < 0.6-0.8. At higher altitudes, between 1400 and 2000 m, we observe a layer enriched in O 3 (∆O 3 /∆CO ∼ 1-2) in correspondence of an almost aerosol-free region. This enriched 10 ozone layer might be possibly associated to a downward transport from higher tropospheric layers, as also suggested by the back-trajectories (Fig. 9), as well as to the mixing with ozone rich layers along the air mass trajectory. Larger particles, from longrange transport of Saharan dust at latitudes below 30 • N, are measured between 2000 and 3000 m, with a minimum of O 3 (∼ 15-20 ppbv) registered within the layer. Several 15 other flights were performed during Mistral/Tramontane episodes (V27, V28, V30, V32) and show, similarly to V20, the presence of several layers both in the BL and the FT.

V24: export from the Barcelona area
Measurements during V24 may be taken as representative of local recirculation (Pérez et al., 2004). In the V24 profile in Fig. 8c we may recognize up to 5 differ- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 1.5). A third, more aged, sublayer (dN Aitken /dN Acc ∼ 2-5, ∆O 3 /∆CO ∼ 0.8-1.0) is observed within the BL between 600 and 1000 m. The FT is characterized by the presence of moderately aged plumes from ∼ 1000 to 2800 m (dN Aitken /dN Acc ∼ 2-10, ∆O 3 /∆CO ∼ 0.2-0.8), and a very aged plume at 2800-3800 m almost deprived in Aitken particles and richer in O 3 (dN Aitken /dN Acc < 1, ∆O 3 /∆CO ∼ 0.6-1.5). A marked 5 local minimum is observed at the top of the BL for σ s , dN Acc , dN Aitken , CO, and O 3 , suggesting the presence of air masses with different origin between the BL and the FT. This is also confirmed by the analysis of the back-trajectories (Fig. 9) which indicates a low level air masses coming from the Spanish coasts in the BL, and air masses travelling at higher altitudes in the FT. In particular, the layer at 2800-3800 m is possibly 10 associated to an intercontinental transport from Northern America, as shown in the trajectory ending at 3500 m. A similar structure characterized by the alternation of fresher and more aged plumes in the BL and FT is also observed in V25 for which aerosol layers are detected up to 4000 m altitude. The detailed analysis of these three events evidences the complexity of the atmo-15 spheric structure over the Western Mediterranean Basin in link with the different dynamical processes involved.

Layers with enhanced Aitken mode particle numbers
Isolated layers with dN Aitken ∼ 10 000-15 000 s cm −3 have been observed in few cases both in the BL and in the FT. The vertical profiles of dN Aitken only for some selected 20 cases are shown in Fig. 10. For about half of the observed events (V16 at ∼ 200-400 m, V21 at ∼ 400-800 m, V28 at∼ 250 m, and V31 at∼ 1000-3000 m; only V28 and V31 shown in Fig. 10) the dN Aitken layer presents a good correlation with dN Acc , CO, and O 3 which suggests that the layer has been transported from a region directly emitting in this size range. 25 The most remarkable example is V31 (Fig. 10a) that the air mass comes from the western-southwestern direction above 1000 m, so the dN Aitken layer can be directly related to pollution export from the urban region of Valencia.
In all the other cases the high dN Aitken layer appears generally not related to simultaneous dN Acc and O 3 increase. Two of these cases (V16 at ∼ 800-1000 m and V28 at 5 ∼ 100 m) occur in the BL.
For the V28 layer (Fig. 10b) the dN Aitken is correlated with CO which might indicate the influence of local emissions close to the surface level. CO values are relatively high (140-160 ppbv) within the layer. It has been often assumed that new particle formation events (NPF) only occur in almost clean environments (e.g., O'Dowd et al., 2010;Sel-10 legri et al., 2010), and that they are suppressed under polluted conditions. In a recent study, Brines et al. (2014) show the occurrence of NPF events also in urban areas with high level of pollution in the Mediterranean region. So, we explore the possibility of NPF in our observations. Given the size ranges of the CPC and PCASP, however, we cannot discriminate within dN Aitken the particle concentration in the sole 4-20 nm range, 15 i.e. the size range involved in nucleation. So it is not possible to directly associate the V28 observations to NPF. In order to obtain a qualitative indication of the possible occurrence of NPF, we have looked at the air mass dynamics within the layer. Several studies suggest, in fact, that NPF might be favoured by turbulence and air mass mixing (e.g., Nilsson et al., 2001;Wehner et al., 2010). We have thus looked at the gradient 20 Richardson number (Ri ) which gives information on the atmospheric dynamical stability. Vertical profiles of Ri are also shown in Fig. 10. For V28 the vertical profile of Ri indicates that below 200 m the Ri number is consistently below zero, which suggests well established turbulent conditions possibly favouring NPF in this layer.
In other two cases (V19, Fig. 10c, and V26, Fig. 10d are favorable for turbulence within the layer, and this may indicate also in this case the possible role of NPF. Finally, a case of high dN Aitken concentration has been also observed in correspondence of dust particles between ∼ 3000 and 4000 m (V23b, Fig. 10e). This layer can be possibly linked to the photochemically-induced nucleation which may occur in pres-5 ence of dust and SO 2 as hypothesised in a recent study by Dupart et al. (2012) and observed by Nie et al. (2014).

Conclusions
The data presented in this paper gives an overview of the distribution of aerosols and trace gases within the tropospheric column up to 5000 m above the Western Mediter-10 ranean Basin.
These data add to the very few available measurements of aerosol and trace gases vertical profiles over the sea surface in the Central (e.g., Junkermann, 2001;Meloni et al., 2003;Di Iorio et al., 2003;Pace et al., 2014) and Eastern (e.g., Formenti et al., 2002;Dulac and Chazette, 2003) parts of the basin thus contributing to improve the 15 description of the atmospheric composition and structure over the whole Mediterranean area.
Observations from the present study indicate that continental pollution strongly affects the composition and structure of the Western Mediterranean Basin both close to coastal regions and in the open sea. Pollution layers extend up to 250 km far from 20 the coasts and reach up to 3000-4000 m altitude, presenting a complex and highly stratified structure. The measured particle concentration is comparable with the values reported for continental Europe (Petzold et al., 2002;Junkermann, 2009;Hamburger et al., 2012). In addition, the geographical distribution of aerosols and trace gases observed in this study appears quite homogeneous within the investigated area, suggesting a relatively similar contribution from the various sources located around the north-western basin. Pollution plumes with different compositions, origins, and lifetimes are observed in link with the different observed dynamical export conditions and meteorological regimes. The aerosol and trace gas observations during TRAQA and SAFMED are consistent with the results of former campaigns and with the interpretation of observed or well known air-masses dynamics and meteorological phenomena that can occur in 5 the Western Basin (Flamant and Pelon, 1996;Millan et al., 1997;Gangoiti et al., 2001;Pérez et al., 2004;Mallet et al., 2005).

ACPD
The large heterogeneity in aerosol compositions, origins, and lifetimes as documented in this study can reflect in a large heterogeneity of aerosol optical properties, with consequences for their direct radiative effect in this part of the basin. This aspect 10 will be investigated in a companion paper analysing the TRAQA and SAFMED in situ measurements of the aerosol absorption and scattering properties and their variability.