Seasonal variation and origins of volatile organic compounds observed during two years at a western Mediterranean remote background site (Ersa, Cape Corsica)

An original time series of about 300 atmospheric measurements of a wide range of volatile organic compounds (VOCs) has been obtained at a remote Mediterranean station on the northern tip of Corsica Island (Ersa, France) over 25 15 months from June 2012 to June 2014. This study presents the seasonal variabilities of 35 selected VOCs, and their various associated sources. The VOC abundance was largely dominated by oxygenated VOCs (OVOCs) along with primary anthropogenic VOCs having a long lifetime in the atmosphere. VOC temporal variations were then examined. Primarily of local origin, biogenic VOCs exhibited notable seasonal and interannual variations, related to temperature and solar radiation. Anthropogenic compounds showed increased concentrations in winter (JFM months) followed by a decrease in spring/summer 20 (AMJ/JAS months), and higher winter concentration levels in 2013 than in 2014 by up to 0.3 μg m in the cases of propane, acetylene and benzene. OVOC concentrations were generally high in summertime, mainly due to secondary anthropogenic/biogenic and primary biogenic sources, whereas their lower concentrations during fall and winter were potentially more influenced by primary/secondary anthropogenic sources. Moreover, an apportionment factorial analysis was applied to a database comprising a selection of 14 individual or grouped VOCs by means of the positive matrix factorization 25 (PMF) technique. A PMF 5-factor solution was taken on. It includes an anthropogenic factor (which contributed 39% to the total concentration of the VOCs selected in the PMF analysis) connected to the regional background pollution, three other anthropogenic factors (namely short-lived anthropogenic sources, evaporative sources, and long-lived combustion sources; which together accounted for 57%) originating from either nearby or more distant emission areas (such as Italy and south of France), and a local biogenic source (4%). Variations in these main sources impacting VOC concentrations observed at the 30 Ersa station were also investigated at seasonal and interannual scales. In spring and summer, VOC concentrations observed at Ersa were the lowest in the 2-yr period, despite higher biogenic source contributions. During these seasons, anthropogenic sources advected to Ersa were largely influenced by chemical transformations and vertical dispersion phenomena and were

vertical dispersion phenomena and were mainly of regional origins. During fall and winter, anthropogenic sources showed higher accumulated contributions when European air masses were advected to Ersa and could be associated with potential emission areas located in Italy and possibly more distant ones in central Europe. Higher VOC winter concentrations in 2013 than in 2014 could be related to anthropogenic source contribution variations in anthropogenic sources probably governed by their emission strength of the main anthropogenic sources identified in this study together with external parameters, i.e. weaker 5 dispersion phenomena and the pollutant depletion. High frequency observations collected during several intensive field campaigns conducted at Ersa during the three summers 2012-2014 confirmed findings drawn from bi-weekly samples of the 2-yr period in terms of summer concentration levels and source apportionment. However, they also suggested that higher sampling frequency and temporal resolution, in particular to observe VOC concentration variations during the daily cycle, would have beenare needed necessary to confirm the deconvolution of the different anthropogenic sources identified by 10 following the PMF approach. Finally, comparisons of the 25 months of Ersa observations at Ersa with VOC measurements conducted at 17 other European monitoring stations highlighted the representativeness of the Ersa station for monitoring seasonal variations in VOC regional pollution impacting continental Europe. Nevertheless, winter VOC winter concentration levels can significantly vary between sites, pointing out spatial variations in anthropogenic source contributions. As a result, Ersa concentration variations in winter were more representative of VOC regional pollution impacting central Europe. 15 Moreover, iInterannual and spatial variations in VOC winter concentration levels variations in winter were also significantly impacted by synoptic phenomena influencing meteorological conditions observed in continental Europe, suggesting that short observation periods may reflect the variability of the identified parameters under the specific meteorological conditions of the study period.

Introduction 20
The main trace pollutants in the atmosphere encompass a multitude of volatile organic compounds (VOCs), with lifetimes varying from minutes to months (e.g., Atkinson, 2000). Their distribution is principally owing to (i) multiple natural and anthropogenic sources, which release VOCs directly to the atmosphere. At a global scale, natural emissions are quantitatively larger than anthropogenic ones (Guenther et al., 2000) and the largest natural source is considered to be the vegetation (Finlayson-Pitts & Pitts, 2000;Guenther et al., 2000Guenther et al., , 2006. In urban areas, numerous anthropogenic sources can abundantly 25 emit various VOCs (Friedrich and Obermeier, 1999). Once in the atmosphere, VOC temporal and spatial variabilities are notably influenced by (ii) mixing processes along with (iii) removal processes or chemical transformations (Atkinson, 2000;Atkinson and Arey, 2003). Accordingly, with a view to thoroughlyextensively characterize VOC sources, it is meaningful to examine their chemical composition, in addition to identifyingand identify the factors controlling their variations at different time scales. 30 VOC regional distributions are eminently changingchange considerably as a result of various confounding factors, namely the emission strength of numerous potential sources, diverse atmospheric lifetimes and removal mechanisms, transport process and fluctuating fluctuations in meteorological conditions. Therefore, these elements underline the necessity to carry out long-term VOC measurements. There are gGrowing efforts now are currently being made to conduct European background measurements over several seasons (e.g., Seco et al., 2011), one year (such as Helmig et al., 2008;Legreid et al., 2008) and even several years (Solberg et al., 1996, 2001and Tørseth et al., 2012 at several European sites; Hakola et al., 2006 andHellén et al., 2015 in Scandinavia;Dollard et al., 2007;Grant et al., 2011 andMalley et al., 2015 in United Kingdom;Borbon et al., 5 2004;Sauvage et al., 2009 andWaked et al., 2016 in France;Plass-Dülmer et al., 2002 in Germany;Navazo et al., 2008 in Iberian Peninsula;Lo Vullo et al., 2016 in Italy). These research studies principally explored the effectiveness of emission regulations efficiency and, links between tropospheric ozone production and changes in VOC concentration levels. They alsos, and assessed seasonal variations and regional distributions in of VOC concentrations. Nonetheless, investigations on of principal factors governing temporal and spatial variations in VOC concentration levels in the European background 10 atmosphere remain scarce. However, the consideration of the influence of (i) source emission strength variations (built upon a factorial analysise.g., Lanz et al., 2009 andLo Vullo et al., 2016), (ii) long-range transport of pollution (e.g., by the examination of air mass trajectories combined with measured concentrations measured at a study site; Sauvage et al., 2009) and (iii) fluctuations in meteorological conditions (which are prone to disperse the pollutants on a regional or long-range scale over long distances through by convective and advective transport) can supply relevant information to deal more in depth with 15 the evaluation of seasonal variations and regional distribution of VOC concentrations in the European background atmosphere.
Particulate and gaseous pollutants detrimentally affect the Mediterranean atmosphere. Accordingly, they are prone to increase aerosol and/or ozone concentration levels in the Mediterranean, regularly higher compared to most regions of continental Europe, and primarily during summer (Doche et al., 2014;Nabat et al., 2013;Safieddine et al., 2014). The Mediterranean region is known to be a noteworthy climate change "hot spot", which is expected to go through severe warming 20 and drying in the 21 st century (Giorgi, 2006;Kopf, 2010;Lelieveld et al., 2014). As a consequence, this can have serious consequences on the release of VOCs from biogenic and anthropogenic sources along with their fate in the atmosphere, with uncertain predicted impacts (Colette et al., 2012(Colette et al., , 2013Jaidan et al., 2018). Actually, the examination of air composition, concentration levels and trends in the Mediterranean region persist to be challenging, primarily on account ofdue to the lack of extensive in-situ observations. Given this context, as part of the multidisciplinary regional research program MISTRALS 25 (Mediterranean Integrated Studies at Regional and Local Scales; http://mistrals-home.org/, last access: 11/10/2020), the project ChArMEx (the Chemistry-Aerosol Mediterranean Experiment, http://charmex.lsce.ipsl.fr, last access: 11/10/2020; Dulac, 2014) aims at assessing the current and future state of the atmospheric environment in the Mediterranean along with examining its repercussions on the regional climate, air quality and marine biogeochemistry. In Within the framework of ChArMEx, several observation periods were conducted at the Ersa station, a remote site considered to be representative of the northwestern 30 Mediterranean basin, in order to explain variations in VOC concentrations affecting the western Mediterranean atmosphere. Michoud et al. (2017) characterized the variations in VOC concentrations observed at Ersa in summer 2013 (from 15 July to 5 August 2013) by identifying and examining their sources.
In this article, we have presented and discussed the factors controlling seasonal and interannual variations of a selection of VOCs observed at the Ersa station over more than two years (from early June 2012 to late June 2014). To this end, this study describes (i) the concentration levels of the targeted VOCs, (ii) their temporal variations at seasonal and interannual scales, (iii) the identification and characteristics of their main sources by statistical modelling, (iv) the evaluation of their source contributions on seasonal bases, together with (v) the representativeness of the Ersa station in terms of seasonal 5 variations in VOC concentrations impacting continental Europe.

Study site
Located in the northwestern part of the Mediterranean Sea, Corsica Island is a French territory situated 11 km northerly from Sardinian coasts, 90 km easterly from Tuscany (Italy) and 170 km southerly from the French Riviera (France). Being the 4th 10 largest Mediterranean island, its land corresponds to an area of 8681 km 2 encompassed by around 1000 km of coastline (Encyclopaedia Britannica, 2018). Corsica contrasts to other Mediterranean islands due to the importance of its forest cover (about a fifth of the island).
Within the framework of the ChArMEx project, an enhanced observation period has been set-up at a ground-based station in the north of Corsica (Ersa; 42.969°N, 9.380°E) over 25 months, from early June 2012 to late June 2014., with tThe 15 aim of was to provideing a high quality controlled climatically relevant gas/aerosol database following the recommendations and criteria of international atmospheric chemistry networks, i. e., the Aerosol, Clouds and Trace gases Research Infrastructure (ACTRIS -https://www.actris.eu/; last access: 11/10/2020), the European Monitoring and Evaluation Program (EMEPhttp://www.emep.int/; last access: 11/10/2020 - Tørseth et al., 2012), and the Global Atmosphere Watch of the World Meteorological Organization (WMO-GAW -http://www.wmo.int/pages/prog/arep/gaw/gaw_home_en.html; last access: 20 11/10/2020). This The Ersa remote site is part of the Corsican Observatory for Research and Studies on Climate and Atmosphere-ocean environment (CORSiCA -https://corsica.obs-mip.fr/, last access: 11/10/2020; Lambert et al., 2011) and is located on the highest point of a ridge equipped with windmills (see the orographic description of the surroundings in Cholakian et al., 2018), at an altitude of 533 m above sea level (a.s.l.). Given its position on the north of the 40-km long Cape Corsican peninsula ( Fig. 1), the Mediterranean Sea is clearly visible from the sampling site on west, north, and east sides (2.5-6 km 25 from the sea; see also the figure presented in Michoud et al., 2017). The station was initially set up in order to monitor and examine pollutions advected to the receptor siteErsa by air masses transported over the Mediterranean and originating from the Marseille-Fos-Berre region (France; Cachier et al., 2005), the Rhone Valley (France), and the Po Valley (Italy; Royer et al., 2010), namely largely industrialized regions. The Ersa station is about 30 km north of Bastia (Fig. 1), the second largest Corsican city (44121 inhabitants; census 2012) and the main harbour. An international airport (Bastia-Poretta) is located 16 km 30 further south of Bastia city centre. More than two millions of passengers transited in Corsica per Bastia during the tourist season (May-September) in 2013, (ORT Corse, 2013; http://www.corse.developpementdurable.gouv.fr/IMG/pdf/Ete2013.pdf, last access: 11/10/2020). However, as the Cape Corsican peninsula benefits in the south from a mountain range (peaking between 1000 and 1500 m a.s.l.) acting as a natural barrier, the sampling site is therefore not affected by transported pollutions originating from Bastia. Only small rural villages and a small local fishing harbour (Centuri) are found within 5 km of the measurement site. Additionally, the Ersa station is accessible by a dead end road serving only the 5 windmill site, surrounded by vegetation made up of Mediterranean maquis, a shrubland biome characteristically consisting of densely growing evergreen shrubs, and also roamed by a herd of goats from a nearby farm. Some forests (78 % of holm oaks, with some cork oaks and chestnuts) are also located nearby, thus ensuring that local anthropogenic pollution does not contaminate in-situ observations. As a result, the Ersa station can be characterized as a remote background Mediterranean site.

VOC measurements
During a study period of two years, non-methane hydrocarbons (NMHCs) and OVOCs (carbonyl compounds) were measured routinely employing complementary off-line methods. Four-hour-integrated (09:00-13:00 or 12:00-16:00 UTC) ambient air samples were collected bi-weekly (every Monday and Thursday) into steel canisters and on sorbent cartridges. The inlets were roughly 1.5 m above the roof of a container mainly housing trace gas analysers. Table 1 describes VOC measurements set up  15 throughout the observation period and Fig. S1 specifies their collection periods.
As generally realized in the EMEP network, 24 C2-C9 NMHCs were collected into Silcosteel canisters of a volume of 6 L, conforming to the TO-14 technique, which is considered adequate for the measurement of many non-polar VOCs (US-EPA, 1997). 152 air samples were realized with a homemade device (PRECOV) for sampling air at a steady flow rate regulated to 24 mL min -1 by canisters previously placed under vacuum. NMHC analysis was performed by a gas chromatograph coupled 20 with a flame ionization detector (GC-FID) within three weeks following the sampling. Separation was performed by a system of dual capillary columns supplied with a switching device: the first column was a CP Sil5CB (50 m x 0.25 mm x 1 µm), suitable for the elution of VOCs from six to nine carbon atoms and the other one was a Plot Al2O3/Na2SO4 (50 m x 0.32 mm x 5 µm), in order to effectively elute VOCs from two to five carbon atoms. Four main steps constituted the quality assurance/quality control program: (i) the implementation of standard operating procedures, (ii) canister cleaning and 25 certification (blank levels < 0.02 ppb), (iii) regular intercomparison exercises and (iv) sampling tests carried out in field conditions and concomitant with in-situ measurements (Sauvage et al., 2009).
About 150 air samples were gathered using sorbent cartridges (63 air samples on multi-sorbent cartridges and 89 additional ones on 2,4-dinitrophenylhydrazine -DNPH -cartridges), by means of an automatic clean room sampling system (ACROSS, TERA Environment, Crolles, France). C1-C16 VOCs were collected via a 0.635 cm diameter 3-m long PFA line. 30 They weare then trapped into one of the two cartridge types: a multi-sorbent one consisted of carbopack C (200 mg) and carbopack B (200 mg; marketed under the name of carbotrap 202 by Perkin-Elmer, Wellesley, Massachusetts, USA), orand a Sep-Pak DNPH-Silica one (proposed by Waters Corporation, Milford, Massachusetts, USA). These off-line techniques are further characterized in Detournay et al. (2011) and their satisfying use in-situ has already been discussed by Detournay et al. (2013) and Ait-Helal et al. (2014). Succinctly here, the sampling of 44 C5-C16 NMHCs, comprising alkanes, alkenes, aromatic compounds and six monoterpenes, as well as six C6-C11 n-aldehydes, was conducted at a flow rate fixed at 200 mL min -1 and using the multi-sorbent cartridges. These latter were preliminary prepared by means of a RTA oven (French acronym for 5 "régénérateur d'adsorbant thermique"manufactured by TERA Environment, Crolles, France) in order to condition them during 24 h with purified air at heated to 250 °C and at flow rate regulated at 10 mL min -1 . In parallelConcomitantly, 15 additional C1-C8 carbonyl compounds were collected at a flow rate fixed at 1.5 L min -1 using the DNPH cartridges. During the field campaign, several ozone scrubbers have beenwere successively inserted in the sampling lines in order to limit any eventual ozonolysis of the measured VOCs: a MnO2 ozone scrubber was retained for the multi-sorbent cartridges while KI 10 ozone scrubber was placed upstream of the DNPH cartridges. Moreover, stainless-steel particle filters of 2 µm diameter porosity (Swagelok) were installed in order to prevent particle sampling. Then, VOC samples were transferred to the laboratory to be analysed within 6 weeks of their collection using a GC-FID (for the multi-adsorbent cartridges) or by high-performance liquid chromatograph connected to an ultraviolet detector (HPLC-UV; for the DNPH cartridges).
The reproducibility of each analytical instrument has beenwas frequently checked, firstly by analysing a standard, 15 and examining results by plotting them on a control chart realized for each compound. The VOC detection limit was determined as 3 times the standard deviation of the blank variation. Obtained dDetection limits in this study were all below 0.05 µg m -3 for the steel canisters and the DNPH cartridges, and of 0.01 µg m -3 for the multi-sorbent cartridges. The uncertainties for each species were evaluated respecting the ACTRIS-2 guidelines for the uncertainty evaluation (Reimann et al., 2018), considering precision, detection limit and systematic errors in the measurements. Evaluated rRelative uncertainties assessed in this study 20 ranged from 7% to 43% concerning for the steel canisters, between 7% and 73% for the multi-sorbent cartridges and from 6% to 41% concerning for the DNPH cartridges. Finally, the VOC dataset was validated following the ACTRIS protocol (Reimann et al., 2018).
Among the 71 different VOCs monitored at Ersa during the observation period, 35 VOCs were finally selected in this study following the methodology described in Sect. S1 of the Supplement. 25

Ancillary measurements
Other trace gases (CO and O3) and meteorological parameters were ancillary monitored at the Ersa site during the observation period. CO was measured from 22 November 2012 to 16 December 2013 by a cavity ring-down spectroscopy analyser (G2401; Picarro, Santa Clara, California, USA) at a time resolution of 5 min. O3 was measured from 31 May 2012 to 26 December 2013 by means ofusing an ultraviolet absorption analyser (TEI 49i manufactured by Thermo Environmental Instruments Inc.,  30 Waltham, Massachusetts, USA) at a time resolution of 5 min. Meteorological parameters (temperature, pressure, relative humidity, wind speed, wind direction and totaldirect and diffuse -solar radiation) were measured every minute from 8 June 2012 to 14 August 2012, and every 5 min from 15 August 2012 to 11 July 2014, with a weather station (CR1000 manufactured by Campbell Scientific Europe, Antony, France) placed at approximately 1.5 m above an adjacent container roof to the one which housed trace gas instruments. Note that ancillary trace gas and meteorological parameter results presented in this study are 4-hour averages concurrent to periods when the VOC sampling periods were realized (see Fig. S1).

Identification and contribution of major sources of VOCs
In order to characterize NMHC concentrations measured at Ersa with steel canisters (the reasons for this VOC selection are 5 presented in Sect. S2 of the Supplement), we apportioned them within their sources in this study using the positive matrix factorization approach (PMF; Paatero, 1997;Paatero and Tapper, 1994) applied to our concentration dataset. The PMF mathematical theory has already been presented in Debevec et al. (2017) and is therefore reminded in Sect. S2 of the Supplement. We used the PMF version 5.0, an enhanced tool developed by the Environmental Protection Agency (EPA) and including the multilinear engine program version 2 (ME-2; Paatero, 1999), and followed the guidance on the its use of PMF 10 (Norris et al., 2014). Using NMHC inputs composed of 152 atmospheric data points of 14 variables (13 single primary HCNMs NMHCs and another one resulting of from the grouping of C8 aromatic compounds) and following the methodology presented in Sect S2, a five-factor PMF solution has been selected in this study.

Classification of air mass origins 15
In order to identify and classify air-mass origins, we analysed back trajectories calculated by the on-line version of the HYSPLIT Lagrangian model (the Hybrid Single Particle Lagrangian Integrated Trajectory Model developed by the National Oceanic and Atmospheric Administration -NOAA -Air Resources Laboratory; Draxler and Hess, 1998;Stein et al., 2015) using Ersa as the receptor site (arrival altitude at Ersa: 600 m a.s.l.) were analysed. For each 4h-atmospheric data point of the field campaign used for the factorial analysis, five back-trajectories of 48 h were computed using GDAS one-degree resolution 20 meteorological data, in order to follow the same methodology as Michoud et al. (2017). The first back trajectory of a set corresponds to the hour when the air sampling was initiated (i.e. 09:00 or 12:00 UTCsee Table 1) and the 4 other ones were calculated every following hour. The time step between each point along the back-trajectories was fixed at 1 hour.
Then, the computed back trajectories were visually classified. Having several back trajectories per sample allows allowed us to check if whether air masses transported toat the station over 4 hours were globally of the same origin. As a result, 25 sSamples associated with air masses showing contrasted trajectories (e. g., due to a transitory state between two different origins) were classified as of mixed origins (9% of the air masses) and discarded from this study (9% of the air masses).
Remaining air masses were then manually classified into five trajectory clusters (marine, Corsica-Sardinia, Europe, France and Spain - Fig. 2 and Table 2) in function ofdepending on their pathway when they reached the Ersa station, their residence time over each potential source region and the length of their trajectories. Additionally, air masses of each cluster were sub-30 divided in function ofdepending on their distance travelled during their 48-h course in order to highlight potential more distant sources from local ones. This sub-division is also given in Table 2 in order to pinpoint differences in transport times.

Identification of potential emission areas
Since the initial origin of an air mass cannot be unquestionably assessed using wind measured at a receptor site, PMF source type contributions from the PMF were coupled with back-trajectories in order to investigate potential emission regions 5 contributing to long-distance pollution transport to the Ersa site. To achieve this, the concentration field (CF) statistical method established by Seibert et al (1994) was chosen in the present study. The CF principle has already been presented in Debevec et al. (2017) and is therefore only reminded in Sect. S3 of the Supplement.
For each VOC observation, 723-hday back-trajectories together with meteorological parameters of interest (i. e., precipitation), were retrieved from the GDAS meteorological fields with a PC-based version of the HYSPLIT lagrangian 10 Lagrangian model (version 4.4 revised in February April 20186), following the same methodology as that used for the 48-h back-trajectories previously presented. The arrival time of trajectories at the Ersa station corresponds to the hour when half of the sampling was done carried out (i.e. 11:00 or 14:00 UTCsee Table 1). Note that longer back-trajectories were considered for CF analyses than those for air mass origin classification (Sect. 2.4.1), in order to be in the same conditions as Michoud et al. (2017) and hence to have comparative results between the two Ersa VOC studies. 15 CF analyses applied to VOC source contributions were carried out by means of the ZeFir tool (version 3.750; . Back-trajectories have beenwere shortened (i.e. the ZeFir tool considered shorter back-trajectories than 72 hours) when a precipitation higher than 0.1 mm was encountered along the trajectory (Bressi et al., 2014). As also done by Michoud et al. (2017), back-trajectories have been also shortened when air mass altitudes gone beyondexceeded 1500 m a.s.l. in order to discard biases related to the significant dilution impacting air masses reaching the free troposphere. A better statistical 20 significance of the CF results is commonly considered for grid cells with a higher number of crossing trajectory points. As a result, some studies applied an empirical weighing function so as to limit the possible influence of high elevated concentrations, which may be observed during occasional episodes with uncommon trajectories (e.g., Bressi et al., 2014;Waked et al., 2014Waked et al., , 2018, on cells having a low number of trajectory points and hence could influence cells having a low number of trajectory points. We preliminary tried to apply this weighing function in this study. Exploratory tests revealed that 25 CF results with the empirical weighing function only highlighted local contributions, given the number of air masses considered inof this study. The farther a cell is was from the Ersa station, the lower was its corresponding value (number of points of the total number of back-trajectories contained in the ij th grid cell, Sect. S3 of the Supplement), and the more the weighing function tended toward downweighting the low valueresults related to this cell. Therefore, CF results discussed in this study were realized without weighing and these limitations should be taken into account when examining CF analyses, 30 which are hence considered as indicative information.
Finally, the spatial coverage of grid cells is was set from (9° W; 32° N) to (27° E; 54° N), with a grid resolution of 0.3° x 0.3°. Allocated contributions were smoothed following a factor (corresponding to the strength of a Gaussian filter) set to 5 to take into account the uncertainties in the back-trajectory path (Charron et al., 2000).

Meteorological conditions 5
Seasonal variations in pollutant ambient concentrations are commonly recognized to be significantly governed by meteorological parameters (namely temperature, total solar radiation, relative humidity and wind speed). Their mMonthly variations in meteorological parameters are depicted in Fig. 3. As the field measurement period covered two years (i.e. from in air flux orientation originally due to the sudden stratospheric warming of the stratospheric polar vortex (Coy and Pawson, 2015)the stratospheric polar vortex underwent a sudden stratospheric warming (SSW; Coy and Pawson, 2015) in early January 2013, having repercussions on the tropospheric polar vortex, which collapsed several times towards Europe. As a result, air flux orientation was modified from north to east, bringing cold air, and hence causing a particularly rigorous European winter 25 2013. On the other hand, most of the western European countries experienced a mild winter in 2014 characterized by its lack of cold outbreaks and nights, and caused by an anomalous atmospheric circulation (Rasmijn et al., 2016;Van Oldenborgh et al., 2015;Watson et al., 2016). Globally, rRelative humidity followed opposite seasonal variations to temperature and solar radiation. In June 2012, 5 aAir in June was dryer compared toin 2012 than in June 2013 and 2014 (mean relative humidity of 57 ±15%, 77 ±16% and 67 ±33% infor June 2012, 2013 and 2014, respectively). The wind speed did not show a clear seasonal variation over the two years studied. Slightly higher wWind speeds were slightly higher noticed in April and May, which could induce higher dispersion of air pollutants and favour advect air pollutantstheir advection to the Ersa station byfrom more the most distant sources to the Ersa station. 10

Air mass origins
Occurrences of air mass origins which have influenced Ersa throughout the observation period are indicated in Table 2. The Ersa station was predominantly under the influence ofinfluenced by continental air masses coming from Europe (corresponding to cluster 3, 31%), France (cluster 4, 26%), Corsica-Sardinia (cluster 2, 14%) and Spain (cluster 5, 5%) and to a lesser extent by air masses predominantly of predominant marine origin (cluster 1, 15%). Each of these five clusters is mostly associated 15 with a particular trajectory sector (e.g., south for air masses originating from Corsica and/or Sardinia, see  (Table 2), as observed by Michoud et al. (2017). Continental air masses spent less time over the sea than marine ones.
Nonetheless, tTransit times of continental air masses over the sea differed depending on in function of how they are were 20 categorized. Air masses originating from Corsica-Sardinia, France and Europe have spent 0-8 hours (median values - Table   2) above the sea before reaching the Ersa station, while the air masses originating from Spain have spent about 36 hours. These contrasting transit times may denote both distinctive atmospheric processing times for the air masses and different oceanic source influences on VOC concentrations observed at the Ersa station.
In particular, European and French air masses showed lower transit times over the sea (median values of 6 h and 25 8 hours, respectively; Table 2) when their trajectories weare categorized as long; compared to short ones (23 h and 19 hours, respectively). These findings are based on the fact that an air mass trajectory classified as short has a closer distance between two of its succeeding successive trajectory points compared to another one classified as long. Due to the Ersa location in the Mediterranean Sea, the air masses having trajectories characterized categorized as long have spent more time tolonger periods above the sea before reaching the Ersa site. Note that European and French air masses were more frequently characterized by 30 long trajectories (accounting for 20% of the air masses observed at Ersa during the studied period, for each) than short ones (11 and 6%, respectively). On the other handMoreover, marine air masses having short and long trajectories have both shown showed long transit times (median values of 40-48 hours - Table 2). and Corsican-Sardinian air masses were only concerned characterized by long trajectories.

VOC mixing ratios
Statistical results on of concentrations of the 35 VOCs selected in this study (see Sect. S1 in the Supplement) are summarized in Table 3. Their average concentration levels as a function of the measurement sampling times (09:00-13:00 or 12:00-16:00) 5 are also indicated in Table S1. These VOCs were organized into three principal categories: biogenic, anthropogenic, and oxygenated VOCs (5, 16 and 14 targeted species, respectively; Table 3). Isoprene and four monoterpenes were classified into BVOCs, while the other primary hydrocarbons (alkanes, alkenes, alkynes and aromatic compounds) were included into anthropogenic NMHCs, since their emissions are especially in connection with human activities. OVOCs have been presented separately, as these compounds come from both biogenic and anthropogenic (primary and secondary) sources. OVOCs were 10 the most abundant, accounting for 65% of the total concentration of the 35 compounds VOCs selected in this study. They were mainly composed of acetone (contribution of 51% to the OVOC cumulated cumulative concentration). Anthropogenic NMHCs also contributed significantly (26%) to the total VOC concentration of the 35 measured VOCs and principally consisted of ethane and propane (which represented 34 and 17% of the anthropogenic NMHC cumulative concentrationmass, respectively) as well as n-butane (7%). The high contribution of species with which generally have the longest lifetimes in the atmosphere 15 (see Sect. 3.45) is consistent with the remote location of the Ersa site and is in agreement with Michoud et al. (2017). BVOCs only contributed little to the total VOC concentration on annual average (49% on annual average, 13% in summer), reaching 13% in summer. They were mainly composed of isoprene and α-pinene (contribution of 44 and 32% to the BVOC mass, respectively). These compounds are among the major BVOCs in terms of emission intensity for the Mediterranean vegetation (Owen et al., 2001) and accounted for half of isoprenoid concentrations recorded during the intensive field campaign conducted 20 at Ersa in summer 2013 at Ersa (Debevec et al., 2018;Kalogridis, 2014). On the contrary, a larger α-terpinene contribution was noticed during the summer 2013 intensive field campaign than the 2-yr observation period. Note that speciated monoterpenes were measured differently during the summer 2013 field campaign, by means of an automatic analyser (see Sect. S4 in the Supplement).

VOC variability 25
Monthly and interannual variations of primary (anthropogenic and biogenic) NMHCs along with OVOCs observed at Ersa selected in this study (Sect. S1) are discussed in this section. Seasonal VOC concentration levels are indicated in Table 4. In addition,Note that the results of comparison between the VOC monitoring measurements investigated in this study with and concurrent field campaign measurements performed during the summers 2012-2014 (is presented in Sect. S4 of the Supplement), in order to check supported the representativeness of the 2-yr observation period with regard to its summer 30 concentration levels.

Biogenic VOCs
Concentration variations of three selected BVOCs, isoprene, α-pinene and camphene, were analysed at different timescales (monthly/interannual variations; Fig. 4). These BVOCs exhibited high concentrations from June to August, consistently with temperature and solar radiation variations (see Sect. 3.1). Indeed, throughout the summer 2013 SOPobservation period, Michoud et al., (2017) and Kalogridis (2014) observed that emissions of isoprene and the sum of monoterpenes were merely 5 mainly governed by temperature and solar radiation, considering supported both by the diurnal variations in their concentrations (Geron et al., 2000a;2000b;Guenther et al., 2000) and their correlations with environmental parameters.
High concentrations of camphene and α-terpinene were also noticed in August 2013 (0.2 ±0.1 and 0.3 ±0.3 µg m -3 , respectively; Additionally, significant concentrations of α-pinene were noticed from September to November (Fig. 4), while isoprene concentrations were close to the detection limit and temperature and solar radiation were decreasing. However, solar radiation decreased much quicker than temperature in fall (Fig. 3), which could suggest additional temperature-dependant emissions (Laothawornkitkul et al., 2009), contrarily to those prevailing in summer, have influenced α-pinene fall concentrations. Moreover, the lower ozone fall concentrations than in summer (O3 concentration variations are depicted in Fig.   S2 of the Supplement) also pointed out a weaker degradation of α-pinene in fall. Note that the interpretation of interannual variations in BVOC measurements is based on a limited number of sampling days during the study period and different 5 collection times (Table 1 and Sect. 2.2.1). It should then therefore be considered cautiously, given variable day-to-day and strong diurnal BVOC concentration variations which were observed during the summer 2013 observation period (Kalogridis, 2014).

Anthropogenic VOCs
Variations of a selection of NMHCs, are depicted in Fig. 5. These compounds illustrateing contrasted reactivity, ( according 10 tobased on their atmospheric lifetimes (estimated according to considering their photochemical reaction ratesd with OH radicals defined in Atkinson, (1990) and Atkinson and Arey, (2003)), are depicted in Fig. 5. Despite lifetimes in the atmosphere ranging from a few hours to some days, all selected NMHCs were characterized by similar seasonal variations, with an increasing winter trend followed by a decrease in spring/summer ( Fig. 5 and Table 4), at the exception of n-hexane, propene and C8 aromatics (which were the most reactive species of the NMHCs selected in this study and which havehad the lowest 15 concentrations; -Tables 3 and 4). NMHC concentrations were higher in winter than in summer, up to 5 times higher in the case of acetylene (Table 4). Note that ethane concentration levels were still relatively important during summer (mean concentration of 1.0 ±0.2 µg m -3 ) while other NMHCs showed concentrations below 0.4 µg m -3 .
Moreover, NMHCs exhibited different concentration levels during the two studied winter periods (Fig. 5). Mean winter NMHC concentrations were higher in 2013 than in 2014 by up to 0.3 µg m -3 in the cases of propane, acetylene and 20 benzene (relative differences of 15%, 42% and 42%, respectively). These latter compounds and ethane had the longest lifetimes among those selected in this study. However, ethane concentrations recorded at Ersa did not show any interannual variation over the two years studiedstudy period (Fig. 5).

Oxygenated VOCs
Variations of selected OVOCs, illustrating contrasted reactivity (following the according to their atmospheric lifetimes 25 considering their photochemical reaction rates with OH radicals defined in Atkinson, 1990 andArey, 2003same methodology as that applied to NMHCs; Sect. 3.4.2) were are depicted presented in Fig. 6. Formaldehyde, acetaldehyde, glyoxal, methylglyoxal and C6-C11 aldehydes have relatively short lifetimes into the atmosphere (photochemical reaction rate with OH radicals of 9-30 10 -12 cm 3 molecule -1 s -1 ) and hence they can result from relatively close sources. On the other hand, acetone and methyl ethyl ketone (MEK) have the longest atmospheric lifetime (0.17-1.22 10 -12 cm 3 molecule -1 s -1 ) of among 30 the OVOCs selected in this study, and hence they can also result from distant sources and/or be formed within emissionenriched air masses before they reachadvected to the Ersa station.
Firstly, formaldehyde, methylglyoxal and n-hexanal have shown similar seasonal variations (Fig. 6), with high summer and spring concentrations (Table 4), suggesting an important contribution of primary/secondary biogenic sources to their concentrations. Fu et al. (2008) found that the largest global sources for methylglyoxal were isoprene and to a lesser extent acetone, this latter source can contribute to its background concentrations. Besides photochemical production, n-hexanal and formaldehyde can be notably emitted by many plant species (Guenther et al., 2000;Kesselmeier and Staudt, 1998;Wild 5 et al, 2003). Acetaldehyde and acetone have shown similar seasonal variations, with an increase of in their concentrations more marked in summer than in winter (Fig. 6), suggesting they were probably mainly of both secondary (anthropogenic/biogenic) and primary biogenic origins. Acetaldehyde is known to be mainly produced through the chemical transformation of anthropogenic and biogenic VOCs (Rottenberger et al., 2004;Schade and Goldstein, 2001;Seco et al., 2007;Wolfe et al., 15 2016), particularly in clean and remote areas. Acetaldehyde can also be released by plants (Jardine et al., 2008;Rottenberger et al., 2008;Winters et al., 2009). Acetone emissions are thought to be globally of biogenic rather than anthropogenic origin (Goldstein and Schade, 2000;Schade and Goldstein, 2006). Acetone can also be resulted of from the oxidation of various VOCs (Goldstein and Schade, 2000;Jacob et al., 2002;Singh et al., 2004) and roughly half of its concentrations measured at diverse urban or rural sites have been assigned to regional background pollution by several studies (e.g., Debevec et al., 2017;20 de Gouw et al., 2005;Legreid et al., 2007) ; with regional contributions at a scale of hundreds of kilometres). Additionally, acetaldehyde and acetone concentration variations in winter (e.g., mean February concentration in 2013s was 0.5 and 2.4 µg m -3 higher 0.5 and 2.4 µg m -3 in 2013 than in 2014, respectively) also pinpointed primary/secondary anthropogenic origins (Sect.

3.4.2).
Glyoxal and MEK showed an increase ofin their concentrations both in summer and winter ( Fig. 6 and Table 4), 25 suggesting they were probably produced by several biogenic and anthropogenic sources. Glyoxal increases Those of glyoxal were in similar proportions ( Fig. 6 and Table 4) while the MEK increase in winter was more marked than in summer, which may indicate that primary/secondary anthropogenic sources primarily contributed to MEK concentrations. Fu et al. (2008) found that the largest global sources for glyoxal were isoprene, and to a lesser extent acetylene and direct emissions. MEK can be emitted from both vegetation and numerous anthropogenic sources, and can also be produced as a result of from the of the year. Moreover, a slight increase in n-nonanal concentrations was noted in August 2013, which is consistent with a biogenic source contribution (Matsunaga et al., 2003;Wild et al, 2003).

Major NMHC sources
In the coming section, major NMHC sources which have impacted primary NMHC concentrations measured at Ersa were identified using a PMF 5-factor solution (from simulations presented in Sect. 2.3S2) and a dataset composed of 14 variables 10 (selected NMHCs measured from steel canisters, see Sect. S2). Figure 7 depicts factor contributions to the species chosen as inputs for the PMF tool along with NMHC contributions to the 5 factors defined by the factorial analysis. Figure 8 and Table   5 show PMF factor contribution time series and their relative contributions to the total concentrations of the selected NMHCs selected in this factorial analysis, respectively. Average factor relative contributions as a function of the measurement sampling times are also indicated in Table 5. In the present section, lifetimes were assessed from kinetic rate constants of the reactions 15 of the selected NMHCs with OH (Atkinson and Arey, 2003) given an average OH concentration of 0.5 10 6 and 2.5 10 6 molecules cm -3 in winter and summer, respectively (Spivakovsky et al., 2000).
As VOC concentrations arised resulted from direct emissions, chemistry, transport and mixing, each individual computed factor cannot be attributed solely to one source category, especially for such a remote receptor site aslike Ersa. A part of them may not be precisely specifically associated with emission profiles but should rather be explained considered as 20 aged profiles originating from several source regions comprising several source categories (Sauvage et al., 2009). The PMF analysis was hence performed in this study to define co-variation factors of primary VOCsNMHCs that were characteristic of aged or local primary emissions along with secondary photochemical transformations taking place during the transport of air masses observed at this remote site (Michoud et al., 2017).
Since the low number of NMHCs considered in the factorial analysis in of this study, PMF result relevance was 25 checked, benefiting from a previous PMF analysis performed with the Ersa VOC summer 2013 dataset (42 variables; Michoud et al., 2017; see Sect. 3.5.6). and eExperimental strategies to characterize VOC concentration variations at a remote background such as Ersa are also discussed in Sect. 3.5.6.

Local biogenic source (factor 1)
The average contribution of factor 1Factor 1 contributions to measured concentrations of the NMHCs selected in the factorial 30 analysis is were of 0.2 ±0.4 µg m -3 on average during the observation period (corresponding to 4% of the sum of theseselected NMHCs - Table 5), and peakeding up at 3.1 µg m -3 on 20 June 2012. In late spring/summer, it was one of the main factors observed (16% on average and up to 53%; Fig. 8). The chemical profile of factor 1 depicts an elevated contribution of isoprene, recognized known as a chemical marker for biogenic emissions, having its variability fully related to this factor. The isoprene relative contribution to factor 1 is 70%. The estimated tropospheric lifetime of isoprene was quite short (winter: 5.6 hours and summer: ~1.1 hour), indicating that this compound was emitted mostly by local vegetation. Consequently, factor 1 is labelled "local biogenic source". Note that factor 1 contribution to selected NMHC concentrations observed at Ersa during the 2-yr 5 period may be slightly influenced by the two different sampling times used during the 2-yr observation period (Table 5) and as well as the number of VOCs and data points considered in the PMF analysis (see Sects. 3.4.1 and S2). However, Michoud et al. (2017) has provided additional information on this local primary biogenic source in summer, such aslike the contributions of additional primary BVOCs (the sum of monoterpenes) and some OVOCs (carboxylic acids, methanol and acetone), and presented the its clear diurnal variations of the local primary biogenic source. 10

Short-lived anthropogenic sources (factor 2)
Factor 2 contributed 19% of to the sum of the selected measured NMHCs was attributed to factor 2 ( Table 5; average contribution of 1.0 ±0.5 µg m -3 ). This factor is mainly consisted of primary anthropogenic compounds, such as toluene (73% of its variability attributed to this factor; Fig. 7), C8 aromatic compounds (93%), ethylene (48%) and propene (83%), typically emitted by combustion processes and with short-to-medium lifetimes (winter: 24 h1-4.1 days; summer: 5-20 hours). The 15 cumulative relative contribution of these VOCs to factor 2 is 66%. Besides road traffic, toluene is also a good marker for solvents generated by industrial sources (Buzcu and Fraser, 2006;Leuchner et al., 2015;Zhang et al., 2014), suggesting that this profile could also be imputed to industrial sources. Additionally, a significant proportion of C5-C6 alkanes, i.e. i-pentane (32% explained), n-pentane (37%) and n-hexane (51%), typically emitted by gasoline evaporation and with medium lifetimes (winter: 4-6 days; summer: 21-31 hours-1.3 day), also contributes 19% to this factor by 19%. Factor 2 did not show a good 20 correlation with CO (Pearson correlation coefficient only of 0.2). As a conclusion, fFactor 2 is hence attributed to the grouping of several short-lived anthropogenic sources, partly related to gasoline combustion and/or evaporation and solvent use. Note that the PMF model did not reconstruct well 5 of the species composing this factor (propene, n-pentane, n-hexane, toluene, and C8 aromatics; -Sect. S2 in the Supplement), especially their concentration variations over short periods. As a result, factor 2 contributions over short periods may be underestimated. However, given the objectives of this study, the examination of 25 factor 2 contribution variations will be limited to seasonal and interannual ones (Sect. 4). Factor 2 contribution variations over a short period was were previously investigated in Michoud et al. (2017).
Regarding factor 2 contributions coupled with air mass clusters origins (Fig. 9), more elevated contributions were noticed at Ersa under the influence of continental air masses coming from France and Europe. The distinction of shorttrajectories from long ones (see Sect. 3.2) highlighted that factor 2 was potentially influenced by relatively close sources when 30 Ersa received air masses from continental France, whereas other continental European sources were probably more distant (Fig. 9). Furthermore, CF analysis applied to factor 2 contributions (see Fig. S3 of the Supplement) confirmed that this factor was influenced by various potential emission areas, either located in Italy (the Po Valley and Central Italy), France (southeast region) or possibly in central Europe (western Hungary, Croatia and Slovenia). Given the CF results, ship emission contribution cannot be discarded as well, as already suggested inby Michoud et al. (2017). Indeed, ship emissions are predominantly composed of light alkenes, aromatic compounds and heavy alkanes (> C6 compounds;Eyring et al., 2005).

Evaporative sources (factor 3)
The average contribution of fFactor 3 contributions to the sum of the selected measured NMHC concentrations is 5 approximately estimated atwere 1.2 ±1.0 µg m -3 on average (22% of their sum; Table 5) during the studied period. The profile of this anthropogenic factor displays an important contribution from alkanes, principally i-/n-butanes (having lifetimes of 10-11 days in winter, and ~2 days in summer) and with more than 69% of their variabilities explained by factor 3 (Fig. 7), along with i-/n-pentanes (50%), n-pentane (59%), n-hexane (42%) and propane (4342-59%; lifetimes of 4-21 days in winter, and 21 hours-4 days in summer). The cumulatedive relative contribution of these alkanes to factor 3 is up to 88%. The C3-C6 alkanes 10 are identified in the gasoline composition and evaporation sources (storage, extraction and distribution of gasoline or liquid petroleum gas; Sauvage et al., 2009 and references therein). Additionally, propane can be viewed as a relevant profile signature of natural gas transport, storage and use (Leuchner et al., 2015). The cumulated relative contribution of these alkanes to factor 3 is up to 88%. As a resultConsequently, this factor can be viewed as "evaporative sources".
Regarding factor contributions as a function of air mass clusters origins (Fig. 9), more elevated contributions were 15 noticed noticed at the Ersa station when aged air masses originated from France and Europe and were probably transported toward Ersa by relatively distant sources. According to the CF analysis applied to factor 3 contributions (Fig. S3 of the Supplement), the Po Valley (especially Emilia-Romagna, an Italian region centre for food and automobile production), central Italy, the southeast of France and the Sardinian region seemed to be identified as main potential emission areas for factor 3, as well as possibly more distant emission areas in central Europe such as western Hungary (i.e. western Transdanubian region 20 specialised in automotive and machinery industries).

Long-lived combustion sources (factor 4)
The average contribution of factor 4 to the total concentration of the selected measured NMHC concentrationss is roughly evaluated at 0.9 ±0.7 µg m -3 (corresponding to 16% of their sum; - Table 5) on average during the observation period. Its profile displays an important contribution from acetylene (100% explained; Fig. 7), benzene (49%) and propane (37%), with 25 lifetimes of 19-26 days in winter and of 4-5 days in summer. and with anThe average cumulated cumulative relative contribution of these VOCs to this factor offactor 4 is 80%. Aromatic compounds and acetylene are generally associated with combustion sources, such as vehicle exhaust (e.g., Badol et al., 2008;Pang et al., 2014). However, factor 4 is characterized by a loading of benzene much more superior to toluene one (49 and 3 %, respectively), suggesting it is more related to a residential heating source than a traffic one (Elbir et al., 2007;Leuchner et al., 2015;Sauvage et al., 2009). Factor 4 profile, mainly 30 composed of long-lived species together with a low contribution of shorter-lived species, may indicate that it is related to partly aged air masses which were advected towards the Ersa station. These suggestions are consistent with the fact that this factor correlated particularly well with CO (Pearson correlation coefficient of 0.8). As a result, this factor can be viewed as "longlived combustion sources", including residential heating.
Furthermore, fFactor 4 showed higher contributions when the Ersa station received European air masses (see Fig. 9), especially by thoseparticularly those having with a long trajectory, which is consistent with factor 4 profile examinationies.
The CF analysis depicted in Fig. S3 of the Supplement only pointed out western Hungary and to a lesser extent the Po Valley, 5 as main potential emission areas for factor 4.

Regional background (factor 5)
Contributing at 39% to the total concentration of the selected measured NMHC concentrationss (average contribution of 2.2 ±1.0 µg m -3 ), factor 5 corresponds to the dominant NMHC source detected at Ersa during the study period. The Its profile of this factor is principally dominated by composed of ethane, (having its variability fully explained by factor 5; ( Fig. 7), and is 10 also composed of propane (18% explained; lifetimes of 21-93 days in winter and 4-19 days in summer). Their relative contribution to factor 5 is up to 96% and they These compounds, with lifetimes of 21-93 days in winter and of 4-19 days in summer, typically result from the transport, storage and use of natural gas and their relative contribution to factor 5 was up to 96%. Additional anthropogenic NMHCs with shorter lifetimes are attributed to this factor, including ethylene (16% explained) and propene (12-16% explained; lifetimes of 21 hours-to 3 days in winter and of 4-13 hours in summer), despite a low 15 contribution to factor 5 mass (~3%). Given tHence, the high abundance of of long-lived species in its profile may result here from aged air masses advected to the Ersa station. Consequently, factor 5 can be viewed as a regional contribution of diverse remote sources of the Mediterranean region, thus indicating the continental regional background (Hellén et al., 2003;Leuchner et al., 2015;Sauvage et al., 2009; regional contributions at a scale of hundreds of kilometres). These sources were advected towards the sampling site by aged air masses, which have not been recently in contact with supplementary anthropogenic 20 sources. Within the time of emission transport from distant sources, atmospheric oxidation depletes a large proportion of the reactive species and the remaining fraction is mainly constituted of the less-reactive VOCs, like ethane and propane. As a resultconclusion, we associated factor 5 with the "regional background".
Factor 5 showed slightly higher contributions when the Ersa station was under the influence of European air masses, (especiallyparticularly those having a long trajectories trajectory and hence( potentially connected to distant emission areas; 25 Fig. 9). As expected, the CF analysis applied to factor 5 contributions did not clearly pinpoint a specific potential emission area (Fig. S3 of the Supplement), apart from maybe western Hungary and to a lesser extent the Po Valley, which are areas experiencing high anthropogenic emissions.

Towards the best experimental strategy to characterize variations in VOC concentrations observed at a remote background site 30
In this study, we decided The choice to restrain the number of VOCs in the factorial analysis, in order to have a better temporal representativeness of factor contributions (see Sect. S2), as well as VOC measurement temporal resolution and collection periods and the number of observations, . This choice have has undoubtedly had some consequences on the PMF solution results of this study (in terms of factor number, chemical profiles and variations). As a resultConsequently, the 2-yr PMF results of this studyto support our results, have been supported by the consideration of the other VOC source apportionment previously performed with the Ersa summer 2013 dataset (Michoud et al., 2017). These two studies provided complementary results. On one hand, the SOP-1b intensive field campaign occurred in summer 2013 and offered good conditions to (i) monitor 5 at a specific period anthropogenic sources, influenced by several geographic origins, along with local biogenic sources and secondary oxygenated sources and (ii) to assess their diurnal variations. On the other hand, the 2-yr monitoring period had the advantage to examine seasonal and interannual variations of main primary sources impacting VOC concentrations observed at the Ersa station. These two studies also helped to discuss on the best experimental strategy to characterize variations in VOC concentrations measured at a remote background site surrounded by vegetation such as Ersa (Sect. 2.1). 10 The two studies on Ersa VOC measurements also helped to discuss here the best experimental strategy to characterize variations in VOC concentrations measured at a remote background site surrounded by vegetation such as Ersa (Sect. 2.1).
Firstly, tFurthermore, the consideration of monoterpenes in the 2-yr PMF analysis would significantly increase the contribution of primary biogenic sources (see Sect. 4.1). A better temporal resolution and coverage of the VOC measurements would have been beneficial in this study for a betterto further assessment of the contribution of local biogenic sources, considering BVOC day-to-day variations (, as noticed inassessed with the summer 2013 dataset; (Sect. 3.4.1). Additional measurements of speciated monoterpenes, sesquiterpenes and BVOC oxidation products would even help identifying different local sources associated with various emitting Mediterranean plants (Owen et al., 2002). This statement is supporting supported by the fact that different interannual BVOC concentration variations in BVOC concentrations were noticed between isoprene and selected 5 speciated monoterpenes in this study (Sect. 3.4.1). Contrarily, cConcentrations of non-speciated monoterpenes and isoprene measured in summer 2013 were mainly attributed to the same primary biogenic source as for isoprene concentrations in Michoud et al. (2017). However, concentration vBut variations ofin the sum of monoterpenes may have reflected mirrored those related to abundant biogenic species such as α-pinene, and hence thus masked those characterizing more-reactive species such as α-terpinene (Kalogridis, 2014;Debevec et al., 2018). 10 The limited numberrange of VOCs included in the our PMF study had the advantage to of a better deconvolution by the PMF model oflve primary sources and a better reconstruction by the PMF model of measured VOCs having low concentrations (Sects. S2 and S6), without the need to minimize the weight of species which hadhaving elevated background levels (unlike for ethane, methanol and acetone of the summer 2013 VOC datasetin; Michoud et al., 2017). These low concentrations thus represented a higher proportion on of the total measured VOC concentrations that the model was supposed 15 to explain. The longer time scale of VOC measurements (i.e. the 2-yr period) presented here even helped to deconvolve longlived combustion sources from regional background (Sect. S5). However, PMF comparison results also raised the importance of the consideration of a finer time resolution and a higher temporal coverage of VOC measurements conducted over a long period at a remote background site such as Ersa to support results from source apportionment results, in terms of deconvolution (Sects. S5 and S6), ). Bbut, it can modify their relative contributions to the total VOC concentration, emphasizing the contribution those of local biogenic/anthropogenic sources and decreasing the contributionthose of regional anthropogenic sources (Michoud et al, 2017). The VOCs observed at Ersa were largely dominated at Ersa by OVOCs (Sect. 3.3), especially in summer (Michoud et al., 2017). OVOC abundance observed at Ersa was further increased by the measurement of alcohols and carboxylic acids, carried out only in summer 2013 withby automatic analysers (Sect. S4). The incorporation of OVOCs in 30 the 2-yr PMF analysis would have helped to better identify co-variations between these species (Sect. 3.4.3) and their proportion associated withresulting from primary sources, but the limited number of samples realized with DNPH cartridges did not favorfavour it (Fig. S1 and Sect. S2). Formaldehyde has shown in this study high concentration levels (Table 3) and clear seasonal variations (Sect. 3.4.3) in this study and would have been of high interest to be considered in the summer 2013 PMF analysis.

Discussions on the seasonal variability of VOC concentrations
In this section, seasonal accumulated concentrations of the 35 VOCs selected in this study (21 NMHCs and 14 OVOCs; Sect. S1) are examined (Fig. 10) so as to identify the prevailing drivers of their variations. Measured concentrations of the 17 5 NMHCs selected in the factorial analysis (Sect. 3.5) have beenwere distributed among their major sources in Fig. 10. Figure   S4 presents separately variations in NMHC factor contributions. As a reminder, anthropogenic sources originated from different locations, including local and more distant ones, and should therefore be considered as more or less aged emissions advected to Ersa. Based on the work of Michoud et al. (2017) and Kalogridis (2014), it has been assumed in this study that the concentrations of the four monoterpenes (Table 1), not included in the PMF analysis, were fully emitted by the local vegetation. 10 Biogenic sources refer here to the grouping of the PMF biogenic source and speciated monoterpenes. The coming discussion has been structured in two steps: on one hand the factors explaining the VOC concentration levels and their variations in spring and summer and on the other hand the factors controlling those in winter fall and fallwinter.

VOC concentration variations in spring and summer
Firstly, the 21 NMHCs selected in this study have shown low concentrations during summer spring and summerpring periods 15 (average seasonal accumulated concentration of 4.6 ±0.1 µg m -3 ; Fig. 10) while the 14 OVOCs exhibited high concentrations (8.0 ±1.8 µg m -3 ). PBL height can be higher in these seasons (von Engeln and Teixeira, 2013), favouring vertical dispersion.
OVOC concentration levels in summer spring and summerpring are in agreement with a higher photochemical decay of VOCs, which were rapidly depleted. Consequently, the most reactive VOCs composing emission profiles of the anthropogenic sources can have reacted and/or been dispersed before reaching the Ersa station. The clear seasonal cycles of these sources, with a 20 decline in their contributions in spring and summer (Fig. S4), confirmed this finding.
Moreover, rRegional background explained contributed in spring and summer from 24 to 53% of to the total NMHC Ersa concentrations of the NMHCs selected in this study. As natural gas sources were attributed to the regional background (Sect. 3.5.5), a decrease in their emissions can presumably occur in the hot season, enhancing the decline in regional background contributions (Fig. S4). These regional background contributions also suggest that aged emissions advected by air 25 masses to the Ersa station significantly influenced VOC concentrations observed during these seasons. This statement is also in agreement with the high acetone concentration levels in summer spring and summerpring (4.8 ±1.5 µg m -3 ; Fig. 10). As a reminder, this carbonyl compound can result from distant sources and/or be formed within polluted air masses before they reaching the Ersa station (Sect. 3.4.3). As pinpointed byshown in Fig. 10, biogenic sources contributed to the total NMHC concentrations especially in summer (up to 41%), in connection with temperature and solar radiation variations (Sect. 3.4.1). 30 This finding is also supported by the high summer concentration levels of formaldehyde, methylglyoxal, acetaldehyde and acetone ( Fig. 10; Sect. 3.4.3). Short-lived anthropogenic sources explained in spring and summer from 16 to 24% of the NMHC concentrations. Despite the high reactivity of the VOCs composing these anthropogenic sources (Sect. 3.5.2), their contributions were not reduced as sharply as those of other anthropogenic sources (Fig. S4), which can probablymay be related to high influences of ship transport and relatively close potential emission areas (Sect. 3.5.2). Evaporative and long lived combustion sources only contributed from 6 to 16% in spring and summer, as they likely originated from emission areas 5 relatively distant from the Ersa station (Sects. 3.5.3 and 3.5.4). Low contributions of long-lived combustion sources can also be partly explained by a lower source strength (especially for residential heating source; Sect. 3.5.4).
Looking now at the interannual variations during summer spring and summerpring periods, Fig. 10 highlights that the total NMHC concentrations of the selected NMHCs were in the same range during the two summer periods as well as the two spring ones (absolute difference was of 0.2 µg m -3 between summers 2012 and 2013 and below 0.1 µg m -3 between springs 10 2013 and 2014). As depicted in Fig. S4, summer spring and summerpring contributions of evaporative sources, long-lived combustion sources, and regional background, were in the same range over the studyied period, as well as MEK concentrations

VOC concentration variations in fall and winter 30
During fall and winter periods, total NMHC concentrations of the selected NMHCs increased (7.8 ±1.2 µg m -3 ) while OVOC concentrations declined (7.1 ±0.8 µg m -3 ). The decline in OVOC concentration levels mainly result from the decrease in acetone and formaldehyde concentrations (3.3 ±0.8 µg m -3 and 1.0 ±0.3 µg m -3 , contributing 34-52% and 10-16% to the total OVOC concentrations of 34-52% and 10-16%, respectively), partly related to a weaker biogenic activity. Chemical reaction rates also dropped in fall and winter as a consequence of decreased the decrease in OH concentrations owing related to lowerthe decline in available UV light and temperatures (Fig. 3). HenceAs a result, VOCs were not removed from the atmosphere as quickly as in summer spring and summerpring,. This statement is in agreement with OVOC concentration levels. PBL height also decreased during these seasons, impacting the pollutant vertical dispersion. These conditions, favouring the pollutant 5 advection and accumulation, explained the higher contributions of anthropogenic sources perceived at Ersa in fall and winter, especially those from sources mainly of regional origins (Fig. S4).
The regional background and evaporative sources were identified as the dominant contributors to the total NMHC concentrations of the selected NMHCs measured at Ersa in fall and winter (contributions of 31-45% and of 22-29%, respectively). Long-lived combustion sources also contributed significantly particularly to NMHC concentrations specifically 10 in winter (explaining 18-24% of the total NMHC concentrations), since the typical low winter ambient temperatures typical of winter (Fig. 3) may involve lead to an increased use of residential heating (Sect. 3.5.4). In order tTo better identify regional influences, average winter and fall contributions of the anthropogenic sources in fall and winter were investigated as a function of air mass origin in Fig. 11. During these seasons, the Ersa station was mostly influenced by continental air masses coming from Europe and France (their cumulated cumulative occurrences were from 36% to 70%). NMHC anthropogenic sources 15 showed higher accumulated contributions when European air masses were advected to Ersa (Fig 11), and they could be attributed to potential emission areas located in Italy and possibly more distant potential emission areasones in central Europe (Sect. 3.5). To a lesser extent, high anthropogenic source accumulated contributions of anthropogenic sources were also noticed when Ersa received air masses originating from continental France (potential emission areas located in the southeast of France) and Corsica-Sardinia. 20  Fig. S4. These findings could be an evidence of homogenous regional background pollution distribution at synoptic scale.
Mean rRegional background winter contributions monitored observed in 2013 and 2014 (absolute difference below 0.1 µg m -3 ; Fig. 10) are also in agreement with this suggestion. Acetone and MEK winter concentrations were lower in 2014 than in 2013 (Fig. 10), which could be related to variations in their anthropogenic primary/secondary sources rather than in their 30 background levels.
Moreover, tThe interannual variations in VOC winter concentrations could be partly explained by occurrences and origins of air masses advected to the Ersa station, their enrichment in different anthropogenic sources (based on potential emissions areas observed in winters 2013 and 2014; Fig. 12) and hence their respective contributions, according to ( Fig. 11).
This statement is also in agreement with results depicted in Fig. 12 which presents potential emissions areas associated with the 4 primary anthropogenic sources in winters 2013 and 2014. On one hand, Ersa was more under the influence of European air masses in winter 2013 than in winter 2014 (occurrences of 37 and 18%, respectively). When continental European and French air masses were advected to Ersa, anthropogenic sources showed higher accumulated winter contributions in 2013 (10.8-and 9.6 µg m -3 , respectively; Fig. 11) than in 2014 (8.2-and 7.8 µg m -3 ). During In winter 2013, the main potential 5 emission areas for the anthropogenic NMHC sources were located in Italy (Tuscan coasts and the Po Valley), central Europe (Slovenia and western Hungary) and to a lesser extent the south of France. Contrariwise, in winter 2014, while Ersa did not seem to be influenced at all by air masses originating from central Europe in winter 2014 (especially from Slovenia and Hungary; - Fig. 12). Surprisingly, and potential emission areas located in the Po Valley and the southeast of France , known to experience high anthropogenic emissions, did not seem to have contributed significantly to VOC concentrations in winter 10 2014. On the other hand, the Ersa station more frequently received air masses originating from Corsica-Sardinia, i.e. relatively close potential emission areas, more frequently in winter winter 2014 than in winter winter 2013 (occurrences of 24 and 0%, respectively). Accumulated The cumulative contribution of anthropogenic sources in winter 2014 contribution associated withunder Corsican-Sardinian air mass influence in winter 2014 was only 7.0 µg m -3 (Fig. 12). These findings can suggest less . As a reminder, winter 2013 was considered rather rigorous, since e.g., French temperatures were lower up to 1-1.5 °C than average value for 1981-2010 according to Météo France (http://www.meteofrance.fr/climat-passe-et-futur/bilansclimatiques/bilan-2013/bilan-de-lhiver-2012-2013, last access: 11/10/2020). On the other hand, winter 2014 was rather mild and temperatures were the hottest of the 1951-2014 period (mean temperature for Europe was ~2.6 °C higher than the pre-25 industrial period; EEA, 2015). As a consequence, the interannual variations in VOC winter concentrations may also result from different influences of meteorological conditions. Indeed, mMeteorological winter conditions were probably more favourable to induce higher OH concentrations in 2014 than in 2013, leading to higher photochemical decay, and so lower NMHC concentrations observed at the Ersa station. Nevertheless, OVOC winter concentrations were higher in 2013 than in 2014, which can therefore highlight they were mainly due to increased contributions from anthropogenic sources of regional   NMHCs with typically medium-to-long lifetimes in the atmosphere, i.e. ethane, propane, n-butane, acetylene and 15 benzene (Sect 3.4.2), were examined here since their concentrationsthey can be significantly influenced by regional contributions, as observed in this study (Sect. 3.5) and elsewhere (Debevec et al., 2017;Michoud et al., 2017;Sauvage et al., 2009). Globally, these selected anthropogenic NMHCs measured at Ersa showed the same seasonal variations as observed at other European stations (Sect 3.4.2), i.e. with an increasing winter trend followed by a decrease in spring/summer, and hence assuring the representativeness of the Ersa station for monitoring regional pollution in Europe. As a reminder, concentrations 20 observed at Ersa were mainly explained in this study by regional background for ethane, by long-lived combustion sources for propane, acetylene and benzene, and by evaporative sources for n-butane (see Sect. 3.5 and Fig. 7). The study of concentration variations of these source tracers can thus help to highlight temporal and spatial variations in source contributions to NMHC concentrations observed in most of continental Europe. As a result, the study of concentration variations of these source tracers may help to highlight temporal and spatial variations in source contributions to NMHC concentrations observed in most of 25 continental Europe. In addition, despite its shorter lifetime compared to other NMHCs of the selection, Fig. 13 also presents ethylene concentration variations were also examined in this studyin order to investigate short-lived anthropogenic source importance and variability in continental Europe. As a result, the study of concentration variations of these source tracers may help to highlight temporal and spatial variations in source contributions to NMHC concentrations observed in most of continental Europe. 30

VOC concentration variations in continental Europe
Globally, concentrations of these selected NMHCs measured at the 17 European stations showed the same seasonal variations as observed at Ersa (Fig. 13 and Sect 3.4.2), hence assuring the representativeness of the Ersa station for monitoring regional pollution in continental Europe. On the one hand, Monthly NMHC monthly concentration levelss in summer were European wide lower and relatively homogeneous from June to August whatever the location and the typology of the station.
T(the highest absolute differences between anthropogenic NMHC summer concentrations measured at two stations in summer was of 0.4-0.7 µg m -3 for ethane, 0.1-0.2 µg m -3 for acetylene, 0.1-0.7 µg m -3 for propane and benzene, 0.2-0.6 µg m -3 for nbutane and 0.3-1.2 µg m -3 for ethylene). It They suggests that the temperature was the main driver in regulating summer 5 concentration levels, linked to photochemistry processes and the vertical dispersion (Sect. 4.1). Ethane concentration levels were still relatively important during summer (mean concentrations >above 1.0 µg m -3 ) suggesting long-range transport (up to intercontinental pollution transport) was among the main parameters governing VOC summer concentrations in continental Europe. On the other hand, anthropogenic NMHC monthly concentration levels in winter appear to be more spatially variable in continental Europe in winter. Indeed, tThe highest absolute differences between VOCNMHC winter concentrations 10 measured at two stations were of 1.3-2.6 µg m -3 for ethane, 0.6-1.6 µg m -3 for propane, acetylene and benzene, 0.6-1. Biogenic VOCs were principally imputed to the local vegetation and exhibited high concentrations and interannual 10 variations in late spring and summer, related to temperature and solar radiation variations. Anthropogenic NMHCs, having atmospheric lifetimes ranging from a few hours to some days, have shown low concentrations in spring and summer since the Ersa station was mostly influenced by aged advected air masses. APrimary anthropogenic sources identified in this study were hence largely influenced by the pollutant depletion (in relation to the oxidizing capacity of the environment) and vertical dispersion before reaching the Ersa station, especially those originating from distant emission areas. As a result, the regional 15 background and short-lived anthropogenic sources (from various locations relatively close to Ersa) mainly contributed to NMHC concentrations in summerspring and springummer. The selected aAnthropogenic NMHCs were also characterized by high concentration levels in fall and winter and winter concentrations higher in 2013 than in 2014. The weaker pollutant depletion and vertical dispersion in winterfall and winterfall have favoured the advection toat Ersa of air masses enriched byin primary anthropogenic sources originating from more distant emission areas (located in France, in Italy and to a lesser extent 20 in Central Europe) than in summer spring and summerpring. As a result, the regional background, evaporative sources and long-lived combustion sources were identified as the dominant contributors to NMHC the total fall and winter concentrations of the selected NMHCs. Interannual variations in their winter contributions were mainly governed by occurrences and origins of air masses advected to the Ersa station as well as by meteorological conditions (temperature and precipitations) occurring before they reached the Ersa station. 25 The OVOCs selected in this study largely prevailed in the VOC abundance during the 2-yr monitoring period. They have also shown different seasonal and interannual variations, suggesting several source contributions. OVOC concentrations measured at Ersa were generally higher in summer, which could be the result of a high contribution of secondary anthropogenic and primary/secondary biogenic sources, whereas their concentrations during fall and winter declined and were potentially more influenced by primary/secondary anthropogenic sources, more specifically for MEK and glyoxal. OVOCs have shown 30 higher winter concentrations in 2013 than in 2014, as a result of the enrichment of air masses advected to Ersa by in primary anthropogenic sources offrom regional origins.
Selected NMHC variabilities in continental Europe supported that Ersa can be considered as a good regional representative station. However, NMHC winter concentration levels can significantly vary temporally and spatially, pointing out spatial variations in anthropogenic source contributions to NMHC concentrations observed in continental Europe, in relation to meteorological conditions. Ersa winter concentration variations were found more representative of central Europe than southern/western or northern Europe. These findings point out the interest in conducting multi-site and multi-year 5 measurements to be sufficiently representative of interannual and spatial variations in regional pollution impacting continental Europe in winter. They also pointed out that the PMF-derived factors controlling VOC concentration variations at remote sites may be mainly controlled by the meteorological conditions that occurred during the study period when issued from short observation periods (i.e. up to two months).
Concurrent datasets of VOC concentrations from 3 summer campaigns performed at Ersa helped to support the 10 representativeness of the 2-yr monitoring period in terms of summer concentration levels, variations, and source apportionment. They also highlight limitations of this study investigating on the investigation of seasonal variations in Ersa VOC concentrations, related toin connection with the VOC range and the temporal resolution of their measurements considered in the PMF analysis. After this work, some questions remain in terms ofregarding the identification and characterization of OVOC sources and origins and of the contribution of local BVOC and OVOC sources at seasonal and interannual scales. It 15 would be interesting to conduct additional long-term VOC measurements at Ersa, including OVOCs and tracers of various primary sources, at a finer time resolution and a higher temporal coverage. This would help to completeing the understanding of determinants governing VOC concentration variations initiated both by Michoud et al. (2017) and this study.

Data availability
Access to EOP the 2-yr and summer 2013 SOP-1b VOC datasets used for this publication is open to registered users following 20 the data and publication policy of the ChArMEx program (http://mistrals.sedoo.fr/ChArMEx/Data-Policy/ChArMEx_DataPolicy.pdf, last access: 11/10/2020). VOC datasets from Ersa summer field campaigns of 2012 and 2014 are available upon request. Please contact Stéphane Sauvage (stephane.sauvage@imt-lille-douai.fr) for further information.

Author contributions 25
SS, NL, JS and FD designed the research and were involved in the logistics and the collection of VOC samples on the field. TS calculated uncertainties related to VOC measurements conducted with DNPH cartridges and steel canisters and validated them following the ACTRIS protocol. CD, SS, VG and NL analyzed analysed VOC data, conducted and interpreted the VOC PMF analysis and examined dominant factors controlling VOC concentrations. CD wrote the manuscript. All co-authors were involved in data discussion and edited the paper. 30 0005-01 and by the Regional Council Nord-Pas de Calais and the European Funds for Regional Economic Development (FEDER). This research was also funded by the European Union Seventh Framework Programme under grant agreement number 293897, the DEFI-VOC project, CARBO-SOR/Primequal and SAF-MED (ANR grant number ANR-12-BS06-0013-02). The authors are thankful for the O3 and CO datasets made available, respectively, by François Gheusi from Laboratoire d'Aérologie and by ICOS-France monitoring network. The authors also want to thank Thierry Leonardis, Emmanuel Tison, 10 Vincent Gaudion, Laurence Depelchin and Isabelle Fronval for their contributions to obtaining of VOC datasets, from the organization and the setting of instruments at Ersa to the analysis of VOC samples at the laboratory, but also Thierry Bourianne and Qualitair Corse team for their involvement on site concerning VOC sample routines as well as the associated logistic procedure. The authors finally thank the ChArMEx project manager Eric Hamonou for his logistical help.    Table 5: Average relative factor contributions ± standard deviations (%) for the whole period and as a function of the measurement sampling times (see Table 1).
5 Figure 9: NMHC factor contributions (µg m -3 ) as a function of air mass origins represented by box plots; the blue solid line, the red marker, and the box represent the median, the mean, and the interquartile range of the values, respectively. The bottom and top of the box depict the first and third quartiles (i.e. Q1 and Q3) and the ends of the whiskers correspond to the first and ninth deciles (i.e. D1 and D9). PMF factors: fFactor 2 -short-lived anthropogenic sources; factor 3 -evaporative sources; factor 4 -long-lived 5 combustion sources; factor 5 -regional background. Air masses originating from France and Europe are subdivided into short and long trajectories to highlight local and more distant contributions (see Sect. 3.2). Note that the NMHC dataset used for the PMF analysis included different sampling time hours (09:00-13:00 or 12:00-16:00) following shifts during the two-year period (see Table  1).  Table 1).