Transport of aerosols over the French Riviera – link between ground-based lidar and spaceborne observations

For the first time, a 355 nm backscatter N2-Raman lidar has been deployed on the western part of the French Riviera to investigate the vertical aerosol structure in the troposphere. This lidar system, based at the AERONET site of Toulon–La Garde, performed continuous measurements from 24 June to 17 July 2014, within the framework of the multidisciplinary program Mediterranean Integrated Studies at the Regional and Local Scales (MISTRALS). By coupling these observations with those of the spaceborne instruments Cloud-Aerosol LIdar with Orthogonal Polarization (CALIOP), Spinning Enhanced Visible and InfraRed Imager (SEVIRI), and Moderate Resolution Imaging Spectroradiometers (MODIS), the spatial extents of the aerosol structures are investigated. The origins of the aerosol plumes are determined using back trajectories computed by the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT). This synergy allowed us to highlight plumes of particulate pollutants moving in the low and medium free troposphere (up to ∼ 5 km above the mean sea level) towards the French Riviera. This pollution originates from the Spanish coast, more particularly from Costa Blanca (including Murcia) and Costa Brava–Costa Daurada (including Barcelona). It is mainly due to traffic, but also to petrochemical activities in these two regions. Desert aerosol plumes were also sampled by the lidar. The sources of desert aerosols have been identified as the Grand Erg Occidental and Grand Erg Oriental. During desert dust events, we highlight significant differences in the optical characteristics in terms of the backscatter-to-extinction ratio (BER, inverse of the lidar ratio) between the planetary boundary layer, with 0.024 sr−1 (∼ 42 sr), and the free troposphere, with 0.031 sr−1 (∼ 32 sr). These differences are greatly reduced in the case of pollution aerosol plume transport in the free troposphere (i.e., 0.021 and 0.025 sr−1). Transported pollution aerosols appear to have similar BER to what is emitted locally. Moreover, using the correlation matrix between lidar aerosol extinction profiles as a function of altitude, we find that during transport events in the low free troposphere, aerosols may be transferred into the planetary boundary layer. We also note that the relative humidity, which is generally higher in the planetary boundary layer (> 80 %), is found to have no significant effect on the BER.

Côte d'Azur region" is also the first tourist destination in France after Paris, with 20 million tourists generating over 130 million overnight stays every year, as well as 1.2 million cruise passengers, increasing the traffic pollutants, especially in summer.
Between 1990 and 2005, a marked increasing trend for PM10 ambient concentrations was observed in this area, correlated with an increase in airway diseases (Sicard et al., 2010). The impact of local traffic is predominant in the summer season, but the industries of the Bouche du Rhône area are also large contributors (El Haddad et al., 2013), and the impact of exogenous 5 sources other than Saharan dust is not negligible (Dimitriou and Kassomenos, 2018). In the coastal town of Toulon, centrally located on the French Riviera, Piazzola et al. (2012) have shown during specific events that air masses could be impacted by pollution transported over the Mediterranean. Yet, these source apportionment studies are only based on surface chemical analyses and backtrajectories, which do not take into consideration the complex meteorological environment of the coastline.
In the framework of the multidisciplinary programme Mediterranean Integrated Studies at the Regional and Local Scales 10 (MISTRALS; http://www.mistrals-home.org), in particular for the Chemistry-Aerosol Mediterranean Experiment (ChArMEx, http://charmex.lsce.ipsl.fr) (Mallet et al., 2015), aerosols in the Mediterranean basin have been studied by several authors, whether via their chemical composition (e.g. Cholakian et al., 2018), their optical properties (e.g. Chazette et al., 2016;Granados-Muñoz et al., 2016), their radiative budget (e.g. Nabat et al., 2015;Di Biagio et al., 2016;Sicard et al., 2016), or the identification of their sources (e.g. Chrit et al., 2018). Among these studies, few were conducted in the 15 atmospheric column above the French Mediterranean coast, which may be subject to aerosol loads of very different origins and chemical compositions. These aerosols directly influence the air quality (e.g. Knipping and Dabdub, 2003), as well as the climate balance of the Western Mediterranean Sea (e.g. IPCC, 2014;Nabat et al., 2015), in different ways depending on their nature and the surface albedo. For instance, during the Hydrological Cycle in the Mediterranean Experiment (HyMeX, also part of MISTRAL program), the radiative effect of dust aerosols has been shown to have little impact in the rainfall amounts 20 and location over the Western Mediterranean basin (Flamant et al., 2015).
All of these studies were preceded by early campaigns such as that of the European project Mediteranean Dust Experiment (e.g. Hamonou et al., 1999), or even networked observations such as those of the lidar Earlinet network (e.g. Balis et al., 2000;Pappalardo et al., 2004;Papayannis et al., 2008; and more recently Granados-Muñoz et al., 2016). Coupling in situ measurements and modelling, the vertical structure of the planetary boundary layer under sea-breeze conditions was also 25 invertigated during the ExperimentS to COnstrain Models of atmospheric Pollution and Transport of Emissions (ESCOMPTE, Cros et al., 2004), over the Marseille-Berre area, ~40 km west of the French Riviera.
Little information exists about the description of transboundary transport of aerosols within the free troposphere over the French Riviera. For this reason, a ground-based N2-Raman lidar was installed near this site between 24 June and 16 July, 2014.
The lidar has combined more than 500 hours of continuous operation and has made it possible to carry out a significant study 30 of the aerosol types and origins, in synergy with spaceborne observations and back trajectories modelling. Works performed using lidar measurements over the Balearic Island of Menorca  already highlighted long range transports of aerosols coming from forest fires in North America (see also , from deserts and, to a lesser extent, from pollution sources located on the Costa Brava (Barcelona). During the summer period, we were not able to establish a clear link between the polluted air masses passing over Menorca and those reaching the French Riviera, suggesting another pathway and/or other aerosol sources. Moreover, previous studies carried out on the French Riviera were only based on surface observations (e.g. Piazzola et al., 2012) and have not offered the possibility of clearly identifying the origin of aerosols, which will be shown hereafter.
The experimental strategy is developped in Section 2, where the lidar, the signal processing and the main uncertainty sources 5 is presented. The temporal evolutions of aerosol optical properties and the vertical atmospheric structure over the French Riviera, obeserved from ground-based lidar system, are discussed in Section 3. Section 4 is devoted to the meterological conditions during the field campaign. The long range transport of aerosol plumes highlighted from the lidar measurements is described in Section 5, using the coupling with both spaceborne measurements and back trajectory studies. Section 6 summarizes and concludes. 10

Strategy
The perimeter of the Western Mediterranean is composed of mountains with elevations generally greater than 1000 m above mean sea level (AMSL). This specific morphology facilitates the recirculation of air masses, and therefore aerosols, via the sea breeze / land breeze cycle, the alternation between the katabatic and anabatic winds, and the guiding of the air masses circulation in the valleys and along the sea shore. Sea breeze is an effective means of exchange between the marine or 15 continental boundary layer and the free troposphere. The aerosols trapped in the low and medium free troposphere will then be transported over long distances and may arrive over the Mediterranean coasts where they will reach the surface via free or forced convection processes generated by the mountains. This transport often happens in thin sheet-like plumes whose vertical limits are generally marked by discontinuities of the potential temperature gradient (e.g. Dalaudier et al., 1994;Chazette et al., 2001). The observation of such layers requires a profiler with a high vertical resolution and justified the deployment of a N2-20 Raman lidar on the French Riviera, close to the AERONET (AErosol RObotic NETwork) site of Toulon-La Garde (43.13556° N, 6.00944° E, 50 m elevation), which is representative of peri-urban conditions. The location of the site is given in Figure 1a.
The lidar performed automatic measurements from 24 June to 17 July, 2014. It was remote-controlled from the Paris area.

2.1
The automatic N2-Raman lidar The N2-Raman lidar LAASURS (Lidar Automatic for Atmospheric Surveys using Raman Scattering) is a research instrument composed of three channels for the parallel and perpendicular polarizations with respect to the laser emission, and the inelastic nitrogen vibrational Raman line of the laser induced atmospheric backscattered signal (Chazette et al., 2014;Chazette et al., 2016). The lidar in its confinement is shown in Figure 1b. The emission energy was 16 mJ at the wavelength of 355 nm and 5 fulfilled eye-safety requirements. The overlap function of the lidar is equal to 1 at distances between 150 and 250 m from the emission. The emission comes from an Ultra® Nd:YAG laser manufactured by Quantel, delivering 6 ns width pulses at the repetition rate of 20 Hz. The detection is carried out by photomultiplier tubes and narrowband filters with a bandwidth of 0.2 nm. The signal acquisition is performed using the PXI technology (http://www.ni.com) with a native vertical sampling of 0.75 m. Note that the acquisition uses a pre-trigger for the correction of the sky background. This avoids effects related to the 10 potential variability of the baseline of analog to digital converters.

Inversion scheme
In order to retrieve the optical properties of aerosols transported in the free troposphere, we used a N2-Raman lidar system coupled with the sunphotometer (e.g. Royer et al., 2011;Chazette et al., 2016) of the Toulon-La Garde AERONET (http://aeronet.gsfc.nasa.gov/) site. Indeed, the signal to noise ratio (SNR) of the daytime lidar profiles is insufficient to use 15 the N2-Raman channel as it presents a low emitted energy. Contrariwise, the N2-Raman channel was used during night time for the retrieval of the cumulative aerosol optical thickness (AOT). Combined with the elastic channel, they lead to the retrieval of the aerosol backscatter and extinction coefficients (ABC and AEC), and to their ratio (Chazette et al., 2016). The backscatter to extinction ratio (BER) is equal to the product of the single scattering albedo and of the probability of a photon being backscattered after an interaction between the laser flux and the atmospheric scatterers. It is the inverse of the lidar ratio (LR) 20 often used in the literature. We prefer to consider hereafter the BER, which has a more direct physical meaning. The linear particle depolarization ratio (PDR) is also retrieved as in Chazette et al. (2012). Calculations are performed with a temporal resolution of at least 30 minutes and a vertical resolution of 30 m to improve the SNR in the middle troposphere (between 6 and 7 km AMSL).
Compared to the inversion scheme proposed in Chazette et al. (2016), where an equivalent BER for the entire aerosol column 25 is justified, the frequent presence of two aerosol layers, one in the marine boundary layer and the other in the free troposphere, led us to develop a multi-layer inversion to evaluate independently the BERs of each aerosol layer and verify the relevance of the first approach leading to an equivalent BER. This method is like the previous one, it uses the partial AOT calculated from the N2-Raman channel for each aerosol layers. The transition altitude between the layers is determined manually and the continuity is ensured by a sigmoid function on a thickness of about 1 km between the two layers. Such an approach is possible, 30 especially for the upper layer, if the SNR is larger than 10. This leads us to perform night time profiles with a time average of 5 hours, between 23:00 and 04:00.

Uncertainties
The main uncertainties sources are discussed in Royer et al. (2011). The relative uncertainty on the N2-Raman-derived cumulative AOT is less than 2% for SNR > 10. The uncertainty in the determination of the equivalent BER is in the range of 4-6 10 -3 sr -1 (10-15 sr in terms of LR). Such a value is very dependent on the SNR, which limits the relevant range of the lidar profile, as shown in Table 2 of Dieudonné et al. (2017). The relative uncertainties on the PDR are close to 10% for the AOTs 5 encountered at 355 nm (AOT> 0.2).

Temporal evolution
The temporal evolution of vertical profiles of optical parameters derived from lidar observations are shown in Figures 2 and 3, separated as 2 periods: before and after 7 July. Several days show anomalies of the aerosol load in the free troposphere, 10 whereas others are more common, with a marked signature of the boundary layer cycle. The aerosol layers above 1.5 km AMSL are generally linked to long range transport, especially on 24, 28 June, 1 to 4 and 7 July (high values of AEC in Figure   2b). Over the coastal site, aerosols within the planetary boundary layer (PBL) are mainly from local sources, either sea-spray aerosols generated by breaking waves (e.g. Yoon et al., 2007), or continental component arising from both natural and anthropogenic sources. The relative influence of these aerosol types is modulated by the land/sea breeze cycle (e.g. Piazzola 15 et al., 2012).
The values of PDR are very variable, ranging between ~1% and more than 20%. These highest values are observed in the free troposphere on 24 June and 3-4 July and may be associated with plumes of terrigenous aerosol with non-spherical shapes. In the PBL, we note the existence of vertical streaks which are the signature of thermal up draughts developing during the day.
They can lift local terrigenous particles and even pollens to the PBL top. The PDR is greater, but not very high inside these 20 structures (2-3%). Indeed, these aerosols are most probably mixed with a significant quantity of spherical hygroscopic particles from the sea or from local pollution.
Relatively high AOTs are measured over much of the observation period, with values exceeding 0.2 at 355 nm and peak values greater than 0.5. AOT values below 0.2 correspond to undisturbed periods, i.e. without the presence of aerosol layers in the free troposphere. Hereafter, days are tagged as disturbed with an aerosol charge anomaly in the free troposphere, or as 25 undisturbed in the contrary case. It should be noted that, during the second period of the measurement campaign, there is no significant aerosol load in the free troposphere. During the first period (24 June to 7 July), it is rather the opposite, except from 30 June to 1 July. These two days are grouped together with those of the second period in our analysis. Moreover, these two days show a very strong variation of the BER as it is the case in the beginning of the second period, from 8 to 10 July (0.020 sr -1 during night time and 0.035 sr -1 during daytime). These days are associated with low relative humidity at the ground level: 30 below 50%, as shown in probably not reached (e.g. Randriamiarisoa et al., 2006). The strong variations of BER are certainly attributable to the local breeze regime with a stronger marine contribution during the day and therefore larger aerosols. The daytime value is similar to that found by Flamant et al. (1998), which was ~0.040 sr -1 .

3.2
Variability along the altitude 5 Before proceeding further, we will check whether or not the hypothesis of a constant BER in the whole atmospheric column is justified. An easy way to verify this is to compare the results on the profiles of the cumulative AOT derived from the inversion of the elastic channel and the one from the N2-Raman channel as in Chazette et al. (2017): the selected BER value is close to the real one if the two profiles coincide.
Here, the two-layer method described in Section 2.2 is here applied for night time lidar profile inversion. The profiles of 10 cumulative AOT for the main cases of the campaign are given in Figure 5. We noticed that BER values in the lowermost layers (0.024 sr -1 ) are only significantly different to the ones found in the upper aerosol layers (0.031 sr -1 ) in the case of 24 June. On the same day, the retrieved column-equivalent BER ranges between 0.017 and 0.033 sr -1 , with a mean value at 0.027 sr -1 .
Taking uncertainties into account, these values overlap. We can, however, expect an error on total AOT of about 15 to 20%.
For the other nights, the hypothesis of one column-equivalent BER value is reasonable, although the nature of aerosols can be different between layers. In our case, there is a strong anthropic component associated with the marine aerosols over the 5 measurement site (Piazzola et al., 2012). These aerosols being hydrophilic, both their size and their complex refraction index can change significantly between the PBL and the free troposphere (Randriamiarisoa et al., 2006). Nevertheless, the work of Raut and Chazette (2008b) has shown that a variation in relative humidity does not modify significantly the value of BER for traffic aerosols. Here, the low vertical variability of BER highlighted supports these findings, given that the relative humidity in the PBL was mainly above 80%, whereas it was below 50% in the free troposphere. 10 In addition, the study of correlation matrices between altitudes for i) the whole duration of the campaign, and ii) for the undisturbed period, shows that there are correlated aerosol plumes between the PBL and the free troposphere, which can be due to transfers of aerosols between these two layers. Figure 6 gives a graphical representation of the magnitude of the coefficients in the matrices of both the entire measurement period between 24 June and 17 July, 2014, and the non-perturbated period. For undisturbed cases (Figure 6a), the correlation distance (correlation coefficient > 0.6) does not exceed ~2 km, 15 whereas it largely exceeds 3 km for the whole duration of the campaign. This argues for the conclusion that during transport events in the low free troposphere, aerosols may be transferred into the PBL. Simultaneous transport between the marine boundary layer and the free troposphere is unlikely over the large distance separating the emission and the arrival of the aerosol plumes over the French Riviera. The more relevant hypothesis is that the recirculation of air masses along the sea shore due to relief and a modification of average winds in the PBL between undisturbed and disturbed cases explain these transfers. 20 Furthermore, large correlation distances are found in the middle free troposphere (Figure 6b), where aerosol plumes have been transported over a long distance. For undisturbed situations above 5.5 km AMSL, the patchiness of the data is due to the relatively high noise when aerosol scatters are almost absent.     (Figure 7b). For the undisturbed cases, winds come from North-West in great majority, and there is little marine contribution. We therefore do not observe a marked see/land breeze cycle at the model scale. In this configuration, one can expect the influence of pollution sources to be linked to road 10 traffic, which is intense in the summer season just north of the measurement site, or of possible biomass fires, very frequent in the backcountry at this period. For disturbed cases, the origin of winds near the surface is much more diverse and reflects the breeze cycle with a significant marine contribution from the south sector during the day, and dominant winds from the east or west along the coast. There is always a west/north-west component from the backcountry. In the lower free troposphere, there are also strong differences between disturbed and undisturbed situations. On a larger scale, the circulation of air masses advected over the western Mediterranean mostly depends on the relative positions of the Azores and the Siberian highs. It is strongly modulated by lows travelling east over mid latitudes. Depending 20 on the position of these lows, air masses from the Atlantic (over Gibraltar) and from the Sahara can supply the western Mediterranean coast. Figure 9 gives an illustration of the two different configurations predominantly encountered during the measurement campaign.
In Figure 9a, the meteorological situation of 24 June is presented for level 700 hPa of Era5 reanalyses. It corresponds to elevated values of PDR in the lower free troposphere (Figure 2c). Tropical air masses are channelled by the presence of two 25 highs, one over North Atlantic (Azores high) and the other over the Sahara, as well as a weak low over the Iberian Peninsula and a strong low over Scandinavia. This configuration favours the transport of desert dust aerosols over the Mediterranean Sea (Hamonou et al., 1999). It is repeated on 3-4 July, when strong PDR values are also observed. Concerning the aerosol plumes observed on 1, 2, 4 July (Figure 2b), we found a displacement towards the British Isles and a strengthening of the low near Iceland (Figure 9a), and a weakening of the low near Scandinavia. Tropical air masses are then deviated towards Sardinia, and 30 air masses from the eastern Spanish coast are more often advected above the marine boundary layer, which explains the significant decrease of PDR in the observed aerosol layers.  Figure 8a and discussed hereafter.

Spain's contribution
It is between 28 June and 7 July that aerosol plumes with low depolarization have been observed in the free troposphere (cf. 15 The AOT fields at 550 nm derived from MODIS observations are presented in Figure 11 for the two main dates with aerosol plumes in the free troposphere. The main contributing areas are located around Barcelona and specifically around Murcia, Atmos. Chem. Phys. Discuss., https://doi.org /10.5194/acp-2018-971 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 3 December 2018 c Author(s) 2018. CC BY 4.0 License.
where AOTs above 0.4 are observed. Note that no major forest fire could be identified in the Iberian Peninsula from the MODIS fire product (http://modis-fire.umd.edu/index.php) during the whole measurement period. In Barcelona, road traffic is the foremost cause of pollution, with an urban conglomeration of more than 1.6 million inhabitants (Dall'Osto et al., 2012).
The emissions linked to the automobile industry, petro-chemistry, and shipping activity also contribute in this area. As for Murcia, it is a conglomeration with more than 700,000 inhabitants, in which traffic also has a non-negligible impact, but cannot 5 single-handedly explain the strong signature seen on the MODIS AOT map. It is also a region with intense agriculture, which has been dubbed "Europe's orchard", nonetheless the June period is not suitable for slash-and-burn or muck-spreading.
However, the locality of Escombreras, near Carthagena, in the south-east of Murcia, includes a gigantic seaside oil processing complex, with a refinery and a harbor. This area can therefore be a strong emitter of aerosol precursors: excluding desert dust episodes, MODIS observations often show AOT values above 0.6 in the south of Murcia during summertime. A mix of traffic 10 and industrial emissions can certainly explain the plume observed over Murcia, part of which is transported towards the French Riviera.
There are no CALIOP observations of the Barcelona area. Conversely, there are daytime and night time orbits overflying the region of Murcia, yet none of these orbits are exploitable during the campaign, as they are mostly associated to desert dust episodes, very frequent in the region. Local aerosols therefore cannot be isolated with good accuracy. In order to obtain the 15 CALIOP classification of the aerosols emitted in the Murcia region , we have thus extended our search for a coincident night time orbit until August. The choice of a night time orbit is motivated by the need for higher signal to noise ratio. The only day in summer 2014 when CALIOP overflew Murcia and local aerosols were recorded is 25 August.

Northern Africa contribution
Desert dust aerosol transports are rather frequent during May and June above France, and decrease in frequency over July, while remaining probable. In itself, it is not extraordinary to observe such events during a field campaign with a duration of about one month. The lifting zones located in the Sahara desert are variable, depending on the west-east travel of lows over the Mediterranean basin (Hamonou et al., 1997). Here, we describe the two events sampled by the ground-based lidar, as 5 highlighted by the high PDR values on Figure 2c. This will allow, among others, to evaluate the degree of coherence between the observations of the spaceborne lidar CALIOP and the ground-based lidar in the French Riviera. We have at our disposal, for each of these two dust aerosols events, a single night time CALIPSO orbit passing close to the lidar site. These orbits are traced in Figure 14. The aerosol classification given by version 4.1 of the CALIOP operational algorithm is shown on Figure 15. Note that the plume must have moved between the observations of CALIOP and MODIS 20 since they are 12 to 14 hours apart. However, desert dust aerosols are indeed identified, and the summit altitude of the layers is compatible with the ground-based lidar observations. The CALIOP profiles are inverted with BER values around 0.023 sr -1 (LR = 44 sr) at 532 nm. We have found BERs (LRs) between 0.025 and 0.031 sr -1 (32 and 40 sr) at 355 nm from the ground-based lidar profiles, which remains within the error bars.
Incidentally, Haarig et al. (2017) have not found significant differences between LR values at 355 and 532 nm for desert dust 25 aerosols transported over Barbados, but caution is due since neither the activated sources or the transport duration are similar.
Over the Balearic Islands, Chazette et al. (2016) report values of the BER (LR) between 0.020 and 0.025 sr -1 (40 and 50 sr) for the same type of activated sources. In the synthesis table (Table 1) presented in Dieudonné et al. (2015), we note that BER (LR) values range from 0.017 to 0.029 sr -1 (34 to 58 sr) for pure dust. Thus, there is a wide range of plausible values, which warrants a measurement of BERs (or LRs) as often as possible so as to properly invert lidar profiles. 30 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-971 Manuscript under review for journal Atmos. Chem. Phys.

Other contributions
The French Mediterranean coast is a densely populated area generating trafic and industrial emissions, but also with a frequent occurrence of forest fires (Guieu et al., 2005). A rather local contribution, above 1.5 km AMSL, from the northwest is observed on July 12-13 and 15-17 in Figure 3b, which could be the result of wildfires. Indeed, their occurrence is high during this summer period in the hinterland, whose dry soils are covered mainly by garrigue, populations of cork and holm oaks, 5 interspersed with some coniferous and palm trees. The equivalent BER of this plume is of the order of 0.022 sr -1 and could correspond to a mixture of aerosols of biomass burning and terrigenous as has been observed over the Mediterranean, see Figure 2b by Chazette et al. (2016). Nevertheless, as for the dust aerosols, the likely BER values for biomass burning aerosols are spread over a very wide range and depend on the type of fuel, the nature of soil and the intensity of the fires, all modulated by weather situations. The forest and bush fires are fortunately mastered quickly in this region, which makes it more difficult 10 to detect them via MODIS.

Conclusion
For the first time, a backscatter N2-Raman lidar was implemented on the the French Riviera. Coupled with passive (MODIS and SEVIRI) and active (CALIPSO) spaceborne observations as well as back trajectories modelling, this instrument made it possible to identify pollution aerosol transports in the low/medium free troposphere from the Mediterranean Spanish coast to 15 the French Riviera. Two desert aerosol transport events have also been sampled by the lidar. The likely sources of aerosol plumes trapped in the free troposphere have thus been located. So far, we had not clearly identified the contribution to the AOT of pollution plumes from the eastern Spanish coast. These can represent nearly two-thirds of the total AOT and, given their anthropogenic origins, may have a significant effect on the vertical stability of the atmosphere in the coastal area. The air masses that contain them may also subside and recirculate along the coast, which is lined with mountains. 20 Desert aerosols sampled by the N2-Raman lidar come from the two major sources known in North-West Africa that are Grand Erg Occidental and Grand Erg Oriental. There is no noticeable difference in optical properties retrieved from the lidar measurement between these two aerosol contributions. The backscatter to extinction ratios (lidar ratios) are similar, with values of ~0.027 sr -1 (~37 sr) at 355 nm. They are close to that derived from the CALIOP observations (BER = 0.023 sr -1 or LR = 44 sr at 532 nm). Pollution aerosols encountered in the free troposphere above the measurement site came from the Barcelona 25 region (Costa Brava and Costa Daurada) as well as from the Murcia region (Costa Blanca). This second origin of pollution aerosol plumes was unexpected to this extent. These two sources can be mixed before reaching the western French Riviera, but they are also separately identifiable according to the altitude. Transport altitudes are close to those of desert aerosol plumes, suggesting a similar injection process related to the large temperature difference between the sea surface and the continental air masses. Except for desert aerosols, an important point is the similarity of the BERs in the PBL and in the free troposphere. 30 It is certainly related to the proximity of the types of aerosol production, mainly due to the traffic, but also to a low sensitivity Atmos. Chem. Phys. Discuss., https://doi.org /10.5194/acp-2018-971 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 3 December 2018 c Author(s) 2018. CC BY 4.0 License. of the BER according to the relative humidity, even though the aerosols are hydrophilic, and a likely air mass recirculation in presence of mountain ranges close to the French Riviera coast.
This study relies on a time-limited data set (~1 month, 500 h of lidar measurements), but it raises questions as to the origin of the pollution aerosols that are sampled on the coast by the air quality stations, since the pollution may not be only local and also seems to be imported over the sea at the scale of the larger Western Mediterranean basin. It would be interesting to 5 implement lidar systems more densely and for a longer period in the French Riviera, and further, all along Mediterranean coastlines, which are under a very strong anthropic pressure. The interest of such an approach has been shown elsewhere, where it has been found that few lidar stations were required for an air quality forecasting similar to those constrained by only ground-based in situ measurements (Wang et al., 2013). A first test was also conducted within the framework of ChArMEx using the Earlinet network (Wang et al., 2014), which we may have to complete in the upcoming years to also meet the needs 10 of operational meteorology for forecasting extreme events. maintain the Raman-lidar instrument. The authors would like to thank the AERONET network for sunphotometer products (at https://aeronet.gsfc.nasa.gov/). The authors acknowledge the MODIS Science, Processing and Data Support Teams for producing and providing MODIS data (at https://modis.gsfc.nasa.gov/data/dataprod/), and the Atmospheric Science Data Center (ASDC) at NASA Langley Research Center (LaRC) for the data processing and distribution of CALIPSO products (level 4.10, at https://eosweb.larc.nasa.gov/HORDERBIN/ HTML_Start.cgi). The author would like to thank the entire 20 MSG/SEVIRI team, from ESA, Alcatel Space Industries and Matra Marconi Space. SEVIRI data have been downloaded from the ICARE Data and Services Centre (http://www.icare.univ-lille1.fr/). The NOAA Air Resources Laboratory (ARL) is acknowledged for the provision of the HYSPLIT transport and dispersion model and READY website (http://www.ready.noaa.gov) used in this publication. ECMWF data used in this study have been obtained from the ESPRI/IPSL data server. 25