African dust outbreaks over the Mediterranean Basin during 2001 – 2011 : PM 10 concentrations , phenomenology and trends , and its relation with synoptic and mesoscale meteorology

The occurrence of African dust outbreaks over the whole Mediterranean Basin has been studied on an 11yr period (2001–2011). In order to evaluate the impact of such mineral dust outbreaks on ambient concentrations of particulate matter, PM10 data from regional and suburban background sites across the Mediterranean area were compiled. After identifying the daily influence of African dust, a methodology for the estimation of the natural dust contributions on daily PM10 concentrations was applied. Our findings point out that African dust outbreaks are sensibly more frequent in southern sites across the Mediterranean, from 30 to 37 % of the annual days, whereas they occur less than 20 % of the annual days in northern sites. The central Mediterranean emerges as a transitional area, with slightly higher frequency of dust episodes in its lower extreme when compared to similar latitudinal positions in western and eastern sides of the Basin. A decreasing south to north gradient of African dust contribution to PM 10, driven by the latitudinal position of the monitoring sites at least 25 E westwards across the Basin, is patent across the Mediterranean. As a result of this, an experimental equation for the estimation of annual African dust contributions based on the latitudinal position was obtained. From 25 ◦ E eastwards, higher annual dust contributions are encountered due to the elevated annual occurrence of severe episodes of dust but also because of inputs from Negev and Middle Eastern deserts. The slightly higher frequency of African dust episodes observed over southern sites in the central Mediterranean Basin is compensated by its moderately lower intensity. Concerning seasonality patterns and intensity characteristics, a clear summer prevalence is observed in the western part, with low occurrence of severe episodes (daily dust averages over 100 μg m −3 in PM10); no seasonal trend is detected in the central region, with moderate-intensity episodes; and significantly higher co tributions are common in autumnspring in the eastern side, with occurrence of various severe episodes throughout the year. Overall, African dust emerges as the largest PM10 source in regional background southern sites of the Mediterranean (35–50 % of PM 10), with seasonal peak contributions to PM10 up to 80 % of the total mass. The multi-year study of African dust episodes and their contributions to PM10 concentrations reveals a consistent decreasing trend in the period 2006/2007 to 2011 in 4 of the 17 studied regions, all of them located in the NW of the Mediterranean. Such decrease is almost parallel to that of the NAO (North Atlantic Oscillati n) index for the summer period, progressively more negative since 2006. Therefore, a sharp change in the atmospheric circulation over the last 5 yr (a similar negative NAO period occurred in the 1950 decade) have affected the number of African dust episodes and consequently the annual dust inputs to PM 10 observed in the NW part of the Mediterranean. By investigating mean temperatures and geopotential height maps at 850 hPa it is evident a displacement of warm air masses accomplishing African dust towards the central Mediterranean in the 2007–2008 biennium, and towards the NW African coast and the Canary Islands in the 2009–2011 triennium. Published by Copernicus Publications on behalf of the European Geosciences Union. 1396 J. Pey et al.: African dust outbreaks over the Mediterranean Basin during 2001–2011

istics, a clear summer prevalence is observed in the western part, with low occurrence of severe episodes (daily dust averages over 100 µg m −3 in PM 10 ); no seasonal trend is detected in the central region, with moderate-intensity episodes; and significantly higher contributions are common in autumnspring in the eastern side, with occurrence of various severe episodes throughout the year.Overall, African dust emerges as the largest PM 10 source in regional background southern sites of the Mediterranean (35-50 % of PM 10 ), with seasonal peak contributions to PM 10 up to 80 % of the total mass.
The multi-year study of African dust episodes and their contributions to PM 10 concentrations reveals a consistent decreasing trend in the period 2006/2007 to 2011 in 4 of the 17 studied regions, all of them located in the NW of the Mediterranean.Such decrease is almost parallel to that of the NAO (North Atlantic Oscillation) index for the summer period, progressively more negative since 2006.Therefore, a sharp change in the atmospheric circulation over the last 5 yr (a similar negative NAO period occurred in the 1950 decade) have affected the number of African dust episodes and consequently the annual dust inputs to PM 10 observed in the NW part of the Mediterranean.By investigating mean temperatures and geopotential height maps at 850 hPa it is evident a displacement of warm air masses accomplishing African dust towards the central Mediterranean in the 2007-2008 biennium, and towards the NW African coast and the Canary Islands in the 2009-2011 triennium.

Introduction
On a global scale, most of the atmospheric particles are emitted by natural sources, mineral dust being the second more abundant component after sea-spray derived aerosols (IPCC, 2007).These crustal aerosols are mainly released to the atmosphere from arid and semiarid regions located in subtropical areas in the Northern Hemisphere, with the Sahara-Sahel-Chad dust corridor being the largest source region in north Africa (Prospero et al., 2002;Moreno et al., 2006).
In general, atmospheric circulation over north-western Africa is mainly controlled by the northeast trade winds and by the mid-tropospheric Saharan Air Layer.Southerly of the Saharan deserts, winds are usually mono-directional with a general westward transport.Along the year, the dust plume extension varies accordingly to the displacement of the Inter-Tropical Convergence Zone (ITCZ) (Prospero et al., 1981).In winter, when the ITCZ is at its southernmost position, dust originated in Sahara and Sahel deserts is transported towards the tropical Atlantic Ocean by such northeast trade winds (Alonso- Pérez et al., 2011).In summer, such trade winds are more constrained owing to the northern displacement of the ITCZ (Prospero et al., 1981).Additionally, high insolation and temperatures over Sahara-Sahel area create strong surface winds and large-scale convection processes, which lift dust particles at high atmospheric levels (up to 5 km).Such mineral dust particles, transported at high-altitude by the Saharan air layer, move partially towards the tropical Atlantic above the boundary layer (Bergametti et al., 1989a).A significant amount of the atmospheric dust is transported also towards the western Mediterranean along an anticyclonic gyre over north-western Africa.Regardless of most of the African dust particles are exported westwards over the Atlantic (Viana et al., 2002;Alastuey et al., 2005;Alonso-Pérez et al., 2011), travelling for long distances and impacting very distant areas in the Caribbean and the United States (Arimoto et al., 1997;Prospero et al., 2002), a considerable amount of dust is also released northerly, affecting the Mediterranean region (Ganor and Mamane, 1982;Bergametti, et al., 1989b;Guerzoni and Chester, 1996;Querol et al., 1998;Rodriguez et al., 2001;Escudero et al., 2005, Gerasopoulos et al., 2006;Kallos et al., 2006;Koc ¸ak et al., 2007;Mitsakou et al., 2008;Papadimas et al., 2008) and even other European areas (Klein et al., 2010).cAfrican dust mobilization may be studied from groundbased measurements (which is important to evaluate air quality, health outcomes, ecosystem damages, visibility reduction), or from a wider perspective by studying the atmospheric column (which is necessary to evaluate climatic feedbacks).Ground-based and columnar measurements are not necessarily correlated.From a ground-based monitoring perspective, African dust towards the Mediterranean region is usually mobilized by a number of meteorological scenarios widely described elsewhere (Rodriguez et al., 2001;Escudero et al., 2005;Gkikas et al., 2009Gkikas et al., , 2012)).As summarized in Querol et al. (2009a), dust-storms affecting western and central Mediterranean are caused by low-pressure systems over the Atlantic or north Africa, high pressures over the Mediterranean, or high pressures at upper levels over NW Africa.Dust storms over the eastern Mediterranean are generally originated by cyclones moving eastwards throughout the Mediterranean, but also because of the combination of low pressures over north Africa with high pressures over Middle East.Details on these meteorological scenarios may be found in Escudero et al. (2005) for the western and central scenarios and in Kallos et al. (2006) for the eastern ones.
One of the most important health outcomes of African dust concerns its chemical and biological composition, as highlighted in a recent review study (Karanasiou et al., 2012).The meteorological scenarios favouring the export of African dust to the western or eastern Mediterranean imply that mineral particles emitted in regions located in east Africa, such as the Bodele Depression, hardly reach western areas in the Mediterranean.Conversely, dusts from NW deserts usually affect eastern locations of the Mediterranean because of the sweep effect caused by low-pressures movement.Taking into account that significant differences in natural soil composition are observed from one region to another in north Africa (Moreno et al., 2006), variations in anthropogenic pollutants travelling with mineral dust are also observed (Perrino et al., 2010;Rodríguez et al., 2011) and the content in microorganisms may be different, the potential effects on health (Pérez et al., 2008;Polymenakou et al., 2008;Tobías et al., 2011a), ecosystems (Arimoto, 2001) and climate (IPCC, 2007;Papadimos et al., 2012) may vary notably.
Yearly, variations in mean ambient temperature or rainfall amount are observed.These changes are associated to alterations in the atmospheric circulation.Thus, it is expected that such variations in the atmospheric dynamics affect also other phenomena such as dust mobilization frequency and/or intensity.The study of long data series of dust contributions at multiple points across a wide area may be indicative of periodic or consistent tendencies.As an example, Cusack et al. (2012) have observed a clear decreasing trend in a number of components of PM 2.5 at a regional background site in NE Spain linked to the implantation of abatement measures at regional and continental scales, but also associated to meteorological cycles.In fact, the general decrease in PM levels observed at ground-based monitoring sites north to south in western and central Europe (Barmpadimos et al., 2012;Cusack et al., 2012) is essentially attributed to recent, and probably cyclic, changes in the atmospheric circulation over the northern Atlantic.
Bearing in mind that African dust is an important source of particulate matter pollution in specific areas, causing by itself or contributing to exceed the daily limit values of PM 10 (in Europe being established in 50 µg m −3 by the 2008/50/EC Directive), and exerting negative health outcomes (Middleton et al., 2008;Mitsakou et al., 2008;Pérez et al., 2008;Jiménez et al., 2010;Mallone et al., 2011;Zauli Sajani et al., 2011;Samoli et al., 2011a, b;Tobías et al., 2011a, b), the identification of such episodes and the quantification of daily and annual contributions of desert dust to PM is currently necessary.Moreover, the identification of temporal trends in African dust contributions at ground-based monitoring sites may be indicative of atmospheric circulation changes.
Previously, Querol et al. (2009a) performed a comprehensive work on African dust significance over the Mediterranean from a database of up to 6 yr at 21 sites.In the present work, a much longer database (based on ground-based measurements) of PM concentrations and African dust contributions (up to 11 yr) at multiple locations (17 areas with at least one monitoring site in each one) across the Mediterranean has been interpreted.The aim of this work is to characterize the phenomenology of African dust outbreaks across the Mediterranean, with special interest in identifying the significance of this natural PM source on ambient air PM 10 concentrations by studying daily and seasonal patterns.Moreover, an extended database of African dust episodes across the whole Mediterranean together with their contributions to PM 10 allows a confident study on inter-annual trends.
This study is part of the LIFE Programme European project MED-PARTICLES (Particles size and composition in Mediterranean countries: geographical variability and shortterm health effects).Overall the MED-PARTICLES project aims at quantifying short-term health effects of particulate matter over the Mediterranean region by distinguishing different particle sizes, chemical components and sources, with special emphasis in the effects of African dust.Since the main motivation of the project is in evaluating health effects, the results of this study are crucial from an epidemiological point of view.These results will be used to estimate effects on health distinguishing between PM 10 from north Africa and that from local and/or regional sources.

Data collection
To evaluate African dust contributions across the Mediterranean Basin (not including north African sites) and to assess their impact on PM 10 levels, data from 19 regional background (RB) and sub-urban background (SUB) sites were obtained (Fig. 1): 7 in Spain covering central and eastern Iberia, and the Balearic islands; 2 in southern France; 5 in Italy covering north to south the peninsula, Sardinia and Sicily; 1 in Bulgaria; 3 in Greece, being 2 in the continent and 1 in Crete; and 1 in Cyprus.Thus, these ground-based monitoring sites are distributed west to east and north to south of the Mediterranean region, with less spatial coverage of the area located 25 • E eastwards, as shown in Fig. 1.From all these regional background sites daily PM 10 concentrations have been obtained from 2001 to 2011 when available.All the data used in this study were obtained from public

African dust occurrence
The methodology used for identifying African dust episodes is the same as in previous studies (Rodríguez et al., 2001;Escudero et al., 2005Escudero et al., , 2007;;Querol et al., 2009a;Pey et al., 2010).This procedure assures the identification of almost all the African dust episodes, independently of their intensity, and consists in the interpretation of a couple of tools: meteorological products (NCEP/NCAR : http://www.esrl.noaa.gov/psd/Negev-Middle Eastern deserts.Secondly, aerosol maps and satellite images are evaluated, which usually result in the consideration of additional days impacted by dust.Finally, for specific cases in which some doubts arise, meteorological maps are calculated to verify the existence of favourable scenarios for the transport of dust.It is important to remark that, in some cases when dust air masses are travelling at high altitude, African dust may affect PM levels at ground levels up to 2 days after the episode ends, as discussed in the European Guidelines for demonstration and subtraction of exceedances attributable to natural sources under the Directive 2008/50/EC on ambient air quality and cleaner air for Europe (http://ec.europa.eu/environment/air/quality/legislation/pdf/sec 2011 0208.pdf).Thus, a final evaluation of PM levels at the different RB and SUB sites is conducted, which incorporated these possible delays.

African dust contribution to PM 10 concentrations
In order to ascertain on daily African dust contributions to PM 10 and PM 2.5 concentrations, a statistical methodology applied to the PM data series have been used.This method is based on the application of 30 days moving 40th percentile to the PM 10 or PM 2.5 data series, after excluding those days impacted by African dust.For those days affected by African dust it is obtained a percentile value which is assumed to be the theoretical background concentration of PM if African dust didn't occur.After that, the African dust contribution is obtained by difference between the experimental PM 10 or PM 2.5 concentration value and the calculated 40th percentile value.This methodology was initially published (Escudero et al., 2007) considering the 30th percentile.Subsequently this method was slightly modified by adopting the 40th percentile instead the 30th one only for conservative reasons.Currently, this is one of the official methods adopted by the European Commission for evaluating the occurrence of African dust outbreaks and quantifying its contributions (Commission staff working paper establishing guidelines for demonstration and subtraction of exceedances attributable to natural sources under the Directive 2008/50/EC on ambient air quality and cleaner air for Europe).It is important to remark that the feasibility of this method was demonstrated by comparing experimentally measured concentrations of mineral matter determined at three Spanish RB sites versus the estimated African dust contributions obtained by this procedure.3 Results and discussions

PM levels across the Mediterranean
Annual averages of PM 10 at RB sites across the Mediterranean Basin (note that monitoring sites in north Africa were not studied) reflect a wide spatial variability (Fig. 2), with the highest concentrations in eastern-basin areas (21-24 µg m −3 ), but also nearby Lyon (Genas, 24 µg m −3 ).On the contrary, the lowest PM 10 concentrations (9-11 µg m −3 ) are observed at high-altitude sites west to east of the Mediterranean: San Pablo de los Montes, EMEP site in central Spain at 1241 m a.s.l.; Febbio, RB site in northern Italy at 1020 m a.s.l.; and Rojen Peak, EMEP site in the Rhodopes (Bulgaria) at 1750 m a.s.l.Intermediate PM 10 concentrations are recorded in the rest of RB sites, being sensibly higher in the vicinity of densely populated and/or industrialized areas, and to north Africa.This is the case of: Fontechiari, in close proximity to Rome (20 µg m −3 ); Víznar, near Granada and north Africa (18 µg m −3 ); Lecce, in southern Italy (17 µg m −3 ); Drome Rurale Sud, close to the highly industrialized area of Marseille (16 µg m −3 ); and Montseny, in the vicinity of the Barcelona metropolitan and industrial agglomeration (15 µg m −3 ).Concerning average PM 10 concentrations at the suburban environments used in this study (selected because of the lack of RB sites covering that geographical areas), they are much higher in the Athens and Thessalonica influence areas (27-28 µg m −3 ) than in Sicily or Majorca (18 µg m −3 ).See Fig. A1 to appreciate average PM 10 levels without the influence of African dust contributions.
Overall, there is an increasing PM gradient (occasionally broken as in the vicinity of Barcelona, Marseille and Rome, because of the high influence of anthropogenic emissions) from the NW to the SE of the Mediterranean.This augment coincides partially with that of African dust, but is mainly driven by the increase of the regional pollution towards the eastern part of the Basin.This increment was formerly reported and chemically characterized by Querol et al. (2009a,   b).In those studies, they found higher concentrations of sulphate and carbonaceous aerosols easterly in the Basin, both components with a prevalent anthropogenic origin at these areas.

PM levels across the Mediterranean: seasonal patterns
PM levels show clear seasonal patterns across the Mediterranean (Fig. 3).In general, a summer maximum is observed throughout the basin attributed to a number of factors including: (1) stagnant conditions over the region and occurrence of a more stable planetary boundary layer over the sea in summer (Dayan et al., 1989;Pey et al., 2009); (2) enhanced formation of secondary pollutants owing to intense solar radiation and humidity; (3) high frequency of wildfires in Mediterranean and surrounding areas; (4) increased anthropogenic pressure since the Mediterranean is a common tourist destination; (5) reduced precipitation and aerosol wet removal; (6) higher emissions from maritime traffic (cruises and small boats) during the warm season.Specifically in the western side of the basin, the highest frequency of Saharan dust outbreaks (Querol et al., 1998;Rodríguez et al., 2001) and the effect of recirculation of air masses over that area (Millán et al., 1997;Rodríguez et al., 2002;Pey et al., 2009) contribute to increase the background levels.Likewise, the transport of polluted air masses from eastern Europe accounts for such summer increase in the eastern part (Gerasopoulos et al., 2006;Koulouri et al., 2008).Similarly at both sides, the lowest PM concentrations are observed in December-January, coinciding with well-ventilated conditions, low photochemical activity, higher precipitation amounts and less frequency of Saharan dust episodes.The most noticeable difference between western and eastern sides corresponds to the late winter-early spring period, when PM concentrations in the eastern Mediterranean are at their maximum due to the impact of severe African dust outbreaks, whereas they are low or intermediate in the western part.

African dust outbreaks: frequency
Figure 4a shows the average frequency of African dust outbreaks across the Mediterranean Basin during the period 2001-2010.It is evident the decreasing gradient of African dust outbreaks frequency from south to north of the Mediterranean Basin.Among the investigated areas, the lowest frequencies of African dust events are observed in central and NE Spain, SE France and northern Italy (17-18 %).On the contrary, the highest frequency is recorded in Sicily (37 %), followed by Cyprus (34 %, affected also by dust outbreaks from Negev and Middle Eastern deserts).As a result, a similar frequency of African dust is observed in south-western and eastern Iberia (23-24 %) and in central Italy and northern Greece (24 %).Furthermore, there is a linear relation between mean frequency of African dust outbreaks and latitude (Fig. 4b).The linear relation is almost the same for areas located in the eastern or western part of the Mediterranean, whereas it is slightly different for the central part of the Basin, where slightly higher frequency of dust episodes is observed in southern sites with respect to similar latitudinal points at both extremes of the Basin.This fact is directly related with the phenomenology of African dust episodes.
As reviewed in Querol et al. (2009a), the western Mediterranean is more affected by African dust in summer.In contrast, the eastern Mediterranean is frequently impacted by African dust air masses in the autumn-spring period.Summer episodes usually affect the central part of the Mediterranean, especially from central Italy towards the south.Similarly, autumn-spring episodes impact often south Italy and Sicily.Thus, the central Mediterranean may be considered as a transitional area in terms of dust outbreaks phenomenology, which is impacted in its lower part by African dust all over the year.An additional issue concerns the African dust export towards the Mediterranean.From a ground-based monitoring perspective, the transport of African dust towards the western Mediterranean mostly occurs at relatively high atmospheric levels, exported from NW African deserts (Escudero et al., 2005).Over the eastern Mediterranean, however, most of the African dust affecting ground-based measurements is transported commonly at surface levels (Querol et al., 2009a).These features influence the intensity of the African dust out- breaks at both sides of the basin, as shown in following sections.

Average PM 10 concentrations
After identifying daily occurrence of dust outbreaks, the methodology described in Sect.2.3 has been applied in order to estimate the dust contributions to PM 10 .Figure 5 shows the average African dust contributions to the mean ambient air PM 10 levels calculated for the existing periods (in most cases a 10-11 yr database was available).As seen in Fig. 5a, African dust inputs are considerably higher in the eastern locations of the Mediterranean when compared to those observed in the western side.In general, average annual concentrations were found to be maximum in Cyprus (8.4 µg m −3 ) and Crete (7.3 µg m −3 ), and minimum in NE Spain, SE France and north Italy (1.1-1.5 µg m −3 ).
The slightly higher dust impact (+1 µg m −3 ) observed in Cyprus with respect to that of Crete might be explained by  the contribution of Negev and Middle Eastern deserts.In relative proportions with respect to the total PM 10 (Fig. 5b), African dust is a dominant component in SE Spain, Crete and Cyprus (35-43 % of the PM 10 ), abundant in SW and central Spain, Sardinia and south Greece (19-25 % of the PM 10 ), and less important northerly and/or close to highly populated areas such as Barcelona, Marseille-Lyon, Rome and Thessaloniki (6-10 % of the PM 10 ).
In order to evaluate the factors governing such spatial distribution, a crossover study between average African contributions versus latitude and longitude has been conducted (Fig. 6a and b).As seen in Fig. 6a, there is a clear dependence of African dust contribution in PM 10 with latitude, comparable to that observed during a severe African dust episode over the Iberian Peninsula (Cabello et al., 2012).It is remarkable that this relation is not linear but exponential, being defined by the following experimental Eq. ( 1) where y is the estimated African dust (in µg m −3 ), and x is latitude (in decimal units).Thus, for given latitude across the Mediterranean Basin, the expected average African dust contribution may be calcu- lated by applying this experimental equation.This approach may be relevant in terms of air quality since allows a quick estimation of the contribution of African dust to PM concentrations on an annual basis.In the European context, this approach could be a preliminary step to be adopted by air quality managers in different countries affected by dust outbreaks (but currently not considering such episodes) before implementing some routinely method to justify such natural inputs.
Complementarily, the effect of the longitudinal position with respect to the average frequency of African dust outbreaks has been evaluated (Fig. 6b).The derived relationship is valid either for the regions where stations are found or near them.Despite that slight differences have been found between the position within the Basin and the frequency of African dust (Fig. 4b, Fig. 6b), locations situated at the same latitude and different longitude across the Mediterranean register on average similar African dust concentrations (Fig. 6c) if they are between 10 • W and 25 • E. At more eastern longitudes (> 25 • E, only two ground-based stations) it is evident a higher African dust contribution (Fig. 6c), having in mind that the frequency of African dust outbreaks does not increment.Regardless of the low number of monitoring sites, such observation is strongly related with the severity of some African dust episodes in the eastern part of the Basin, as mentioned in following sections but also with the contribution of dust from Negev and Middle Eastern deserts (Derimian et al., 2006;Basart et al., 2009) in the case of Cyprus.
Overall, mean annual African dust contributions in PM 10 varied with respect to the latitude.A longitudinal effect is patent from 25 • E eastwards.The decreasing gradient of African dust towards the north is not lineal but exponential.

Seasonal patterns
As mentioned previously, African dust transport occurs in different seasons in western and eastern sides of the Mediterranean.These seasonal patterns exert a clear influence in the African dust contributions along the year.
Figure 7 shows the seasonal contributions of African dust on PM 10 , and the partitioning between bulk ambient air PM 10 and African dust.On average, African dust may occur all along the year across the Mediterranean.However, African dust inputs in the western side of the Mediterranean are considerably higher between May and October, and in March, when compared to the rest of the year.On the contrary, such inputs are clearly higher between November and May in the eastern part of the Mediterranean.An intermediate outcome is observed for central locations in the Mediterranean, where only slightly higher summer contributions are detected.
In relative terms, in the south-western part of the Mediterranean, African dust may account for about 50 % of PM 10 mass in mid summer, and less than 10 % in winter.Similarly, in the most southern site of the central Mediterranean Basin, between 35 and 50 % of the PM 10 in summer is mineral dust from north African deserts.More evident is the impact of the African dust in the most south-eastern locations studied, where up to 80 % of the PM 10 recorded in the period February-April is constituted by African dust.By contrast, mineral dust from desert regions account for less than 10 % of PM 10 in mid summer.The mentioned trends are repeated in the northern areas across the basin, although a substantial diminution in the dust contributions on ambient PM 10 concentrations is evident.

Intensity of dust outbreaks
One of the most interesting aspects of African dust episodes concerns their intensity.By analyzing extended databases of African dust events across the Mediterranean it is possible to identify the occurrence and recurrence of intense African dust episodes.Figure 8 represents the percentage of African dust days according to their intensity (mean contributions in µg m −3 in 9 concentration ranges, the highest being 30-44, 44-99 and > 100) for the whole Mediterranean.The occurrence of extreme dust events (with daily PM 10 African dust contributions higher than 100 µg m −3 ) with respect to the total number of episodes is infrequent.However, the occurrence of such severe outbreaks is relatively frequent (2-5 % of the African dust days) in the most south-eastern sites of this study (Greek Islands, south of Greece and Cyprus).As a result, some of these events occur randomly every year.The occurrence of extreme events is unusual (from less than 1 to 1 % of the African dust days) at equivalent latitudes westerly of the Mediterranean (southern Spain, southern Italy, Sicily).Moderate to intense events (30-99 µg m −3 PM 10 dust) are observed both in western, central and eastern areas, accounting for 15 to 25 % of the dust events in southern areas, 10-15 % in intermediate latitudes, and 5-10 % in northern areas.Low-intensity episodes (1-10 µg m −3 ) prevail in northern locations of the western and central Mediterranean, accounting for 50 to 70 % of the African dust episodes.
Overall, the intensity characteristics described in this section are strongly related with the transport patterns of African air masses.Dust transfer over western and eastern sides of the Mediterranean is caused by different transport mechanisms.In this context, the Atlas mountainous barrier, with a 2500 km extension from western Sahara towards Tunisia and peak altitudes up to more than 4000 m a.s.l., plays a dominant role in local and mesoscale atmospheric circulation patterns.As a result, African dust episodes over the western and central part of the Mediterranean are very frequent in summer (Rodríguez et al., 2001;Escudero et al., 2005), although moderate in intensity given the intricate transport processes (dust is travelling at high altitudes).The situation in the eastern part of the Mediterranean is significantly different.African dust transport is typically induced by cyclones moving eastwards across the Mediterranean and/or north Africa, transporting dust at surface levels.These flows may be enhanced during specific scenarios (Moulin et al., 1998) if air masses are canalized southerly of the Atlas Mountains (with a north-eastern prevalent direction), giving to the occurrence of short but intense dust episodes in the eastern side of the Mediterranean.From autumn to spring, when dust episodes are more frequent over the eastern Mediterranean, dust plumes leaving north African deserts eastwards of Tunisia do not reach elevated atmospheric layers (Kalivitis et al., 2007;Nastos, 2012) since they do not find any significant geographical barrier neither convective processes are intense over north Africa.On the contrary, African dust air masses affecting the eastern Mediterranean in summer are encountered at high altitudes, due to the strong vertical convective flows over the dust source regions (Gobbi et al., 2000;Kalivitis et al., 2007;Papayannis et al., 2008).
In order to assess on the relation between African dust occurrence and intensity, Fig. 9 has been elaborated.Most of the patterns observed in Fig. 9 have been already discussed previously.Nevertheless, there are new specific features to be described and interpreted.In general, there is no clear relationship between occurrence and intensity, with significant seasonal variation of dust episodes but flat variation in the intensity along the year.Only in the eastern side of the Mediterranean, intensity and occurrence are varying in parallel.Concerning the intensity of dust occurrences, although limited variation is observed, February-March and October-November episodes over the western and central Mediterranean are usually more intense than those of other months.This fact is due to the transport mechanisms of the dust, always at ground level, and induced by two main scenarios already described in Escudero et al. (2005): (1) cyclone systems moving from western Iberia towards the Mediterranean; (2) widespread anticyclone over the western and central Mediterranean that moves slightly easterly.Commonly, the first scenario is typical of autumn, whereas the second is frequently observed in late autumn and early spring.

African dust outbreaks: inter-annual variability
Long-term series of annual averages of African dust contributions are helpful to recognize cyclic or climatic tendencies, which would be related to fast changes in atmospheric circulation patterns.Also variations in the emissions from source areas could be recognized by exploring these longterm databases.Figure 10 summarizes the annual averages of African dust contributions from 2001 to 2011 for selected areas across the Mediterranean.Although it was not possible to find a very complete PM 10 database for the whole Mediterranean to quantify the African dust impact, usual data coverage ranges between 7 to 11 yr.
In general over central and southern areas (central and south Spain, central Italy and Sicily, Greece, and Cyprus), there are no evident increasing neither decreasing trends of African dust contribution to ambient PM 10 , with sporadic annual peaks as a result of the occurrence of severe episodes.Thus, south Spain in 2004 registered a number of episodes above 150 µg m −3 ; Sicily also recorded a very intense event in February 2004, with a daily concentration of more than 300 µg m −3 of African dust; south Greece experienced also severe African dust episodes in February, May and November 2004, with daily peak concentrations up to 200 µg m   1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005   contributions are observed, with annual peaks caused by intense episodes.From 2006 or 2007, depending of the area, a marked and continuous decreasing trend is observed in these areas.Because this trend is not observed in the southern areas, a change in the emissions of dust from source regions is not apparent.Nevertheless, a modification in the atmospheric circulation over these areas may be possible.Actually, a recent study on evaluating trends of PM and chemical composition across Europe (Cusack et al., 2012) has demonstrated a partial relation between the observed trends and the North Atlantic Oscillation (NAO) index.The NAO index is accounting for the intensity of the westerly circulation over the north Atlantic (Barnston and Livezey, 1987) that is mainly defined by the position of the two leading atmospheric systems, the Azores high and the Icelandic Low low.Currently, the NAO index is widely used to interpret wintertime weather anomalies in temperature and precipitation over northern, central and western areas across Europe and the United States (Hurrell et al., 2003).In the recent past, such index was evaluated to account for the export of dust (Moulin et al., 1997).In view of the trends found in this work, the NAO index for the summer periods (the dust season in the NW Mediterranean) were calculated from the NOAA data center (http://www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/nao.shtml).As seen in Fig. 11a there is a clear change in NAO index from 2006 onwards.In the period 2001-2005, summer NAO indexes were close to 0 whereas there became consistently more negative from 2006-2007 to be at their lowest in 2011 (−0.90).A wider temporal scale of this summer NAO index (Fig. 11b) allows identifying only one persistent negative period in the 50's.Thus, it is evident that the 2006-2011 summer periods were governed by atypical atmospheric patterns.When contrasting the summer NAO index for the period 2006-2011 with respect to the average annual African dust contributions to PM 10 levels for NE Spain, SE France, the Balearic Islands and north Italy, a perfect fitting is observed (Fig. 12).The correlation observed for these areas is higher in the Spanish areas (R 2 = 0.9) than in SE France (R 2 = 0.8) or north Italy (R 2 = 0.5), thus indicating a progressive lesser influence of the summer NAO index towards the east.Negative NAO phases are related to a displacement of the storm trajectories towards the south.Generally in summer, westerly trade winds blow above 45 • in latitude, leaving the Mediterranean region under weak gradient conditions (Millán et al., 1997).During negative NAO phases the track of westerly winds is observed at lower latitude, thus preventing subtropical air masses to reach the NW Mediterranean.In order to address the influence of African dust air masses over summer periods, the geopotential height maps at 850 hPa (Fig. 13a) and the mean temperatures at 850 hPa (Fig. 13b) have been calculated from 2001 to 2011.From 2006 onwards, the Icelandic Low was reinforced and displaced south-easterly (Fig. 13a).Meanwhile, warm air masses (accomplished with mineral dust) emerged from north Africa following atypical trajectories (Fig. 13b).In

Figure 1 .Figure 2 .
Figure 1.Location of regional and suburban background sites providing data for this study.851852

Figure 1 .Figure 2 .
Figure1.Location of regional and suburban background sites providing data for this study.

Figure 5a .Figure 5b .Figure 5a .Figure 5b .
Figure 5a.Mean African dust contributions to PM10 (in µg m -3 ) across the Mediterranean (average values for the periods when data are available, in most cases from 2001-2010).

Fig. 5 .
Fig. 5. (a) Top: mean African dust contributions to PM 10 (in µg m −3 ) across the Mediterranean (average values for the periods when data are available, in most cases from 2001-2010); (b) bottom: percentage of African dust over bulk PM 10 registered in the monitoring sites selected in this study (average values for the periods when data are available, in most cases from 2001-2010).

Figure 11a .
Figure 11a.NAO index calculated for the period June-September from 2001 to 2011.910

Figure 11b .Figure 11a .
Figure 11b.NAO index calculated for the period June-September from 1950 to 2011.912 913 914 Figure 11b.NAO index calculated for the period June-September from 1950 to 2011.

Fig. 11 .
Fig. 11.(a) Top: NAO index calculated for the period June-September from 2001 to 2011; (b) bottom: NAO index calculated for the period June-September from 1950 to 2011.

Figure 12 .Fig. 12 .
Figure 12.Scatter-plots between summer NAO index with respect to annual African dust 916 contribution to PM10 (in µg m -3 ) for NE Spain, SE France, the Balearic Islands and North Italy for 917 many cases from 2001 to 2006 such warm air masses attained the western Mediterranean.However, in 2007 and 2008 towards the central Mediterranean occurred, a displacement of such warm air towards the central Mediterranean, although still affecting in 2007 the NE Spain and the Balearic Islands.From 2009 to 2011 north African there is a PM10 increase of 3-4 µg/m 3 from western towards central and eastern locations across the Mediterranean.Finally, the highest PM10 concentrations (13-22 µg/m 3 ) correspond to those RB sites located in areas subjected to severe anthropogenic influence (Rome, Marseille and Barcelona).

Table 1 .
Location, type (regional background: RB; sub-urban background: SUB), data availability and measurement methods used in the monitoring sites of this study.
−3 ; Cyprus suffered numerous and very intense dust episodes in February-March, August and December 2008, 12 days with dust daily concentrations exceeding 100 µg m −3 .The situation in the northern locations of the western and central Mediterranean (NE Spain, the Balearic Islands, SE France, and north Italy) is clearly different.In these areas, a prominent decreasing trend on ambient PM 10 contributions is patent from 2006 or 2007 onwards.Before 2007, alternate