Variability of Atmospheric Aerosol and Ozone Concentrations at Marine, Urban, and High-Altitude Monitoring Stations in Southern Italy during the 2007 Summer Saharan Dust Outbreaks and Wildfire Episodes

ABSTRACT

In order to evaluate the spatial variation of aerosol (particulate matter with aerodynamic diameter ≤10 μm [PM₁₀]) and ozone (O₃) concentrations and characterize the atmospheric conditions that lead to O₃ and PM₁₀-rich episodes in southern Italy during summer 2007, an intensive sampling campaign was simultaneously performed, from middle of July to the end of August, at three ground-based sites (marine, urban, and high-altitude monitoring stations) in Calabria region. A cluster analysis, based on the prevailing air mass backward trajectories, was performed, allowing to discriminate the contribution of different air masses origin and paths. Results showed that both PM₁₀ and O₃ levels reached similar high values when air masses originated from the industrialized continental Europe as well as under the influence of wildfire emissions. Among natural sources, dust intrusion and wildfire events seem to involve a marked impact on the recorded data. Typical fair weather of Mediterranean summer and persisting anticyclone system at synoptic scale were indeed favorable conditions to the arrival of heavily dust-loaded air masses over three periods of consecutive days and more than half of the observed PM₁₀ daily exceedances have been attributed to Saharan dust events. During the identified dust outbreaks, a consistent increase in PM₁₀ levels with a concurrent decrease in O₃ values was also observed and discussed.

IMPLICATIONS

In the summertime, the central-southern Mediterranean Basin is heavily affected by Saharan dust outbreaks and wildfire events. A focus on their significant influence on either oxidizing capacity of the atmosphere and air quality over Calabria, southern Italy, was here presented. Similar studies for most regions surrounding the Mediterranean Basin are needed to implement effective emission reduction measures, to prevent apparent air quality parameter exceedances and to define an appropriate health alert system. Because the frequency of these events is expected to increase due to climate change, these studies could even be a valid effort to better understand and characterize such atmospheric variations.

INTRODUCTION

During the last decade, air pollution in the Mediterranean region has received considerable attention due to the high level of industrialization and dense population around the coastal zone of the Mediterranean Basin. Several research studies have shown that the Mediterranean atmosphere is influenced by air masses originating from Central-Eastern Europe, North Africa, and sometimes from southeast Asia; for this reason the Mediterranean has been described as an air pollution crossroads.1 In addition to anthropogenic emissions from industrialized population centers surrounding the basin itself,2 emission in busy shipping lanes and large port cities are important sources of pollution plumes containing a wide range of particulate and gaseous pollutants.3–5 Natural emissions occur through Saharan dust outbreaks,6 geological processes, including volcano and geothermal activities, as well as through wildfire, marine emissions,7,8and evasion from terrestrial and aquatic surfaces.9 The relative contribution of anthropogenic and natural emissions within the Mediterranean region vary greatly with location, season, and meteorological conditions. Air quality in the Mediterranean region is therefore closely related to the unique geographical and meteorological features which characterize the sea basin.10 Particularly in summer, the high levels of solar radiation and temperatures combined with thermally induced recirculation near densely populated coastal areas11 and frequent high-pressure situations (i.e., the Azores high pressure) leading to subsidence, stability, and clear sky12–14 favor photochemical ozone production15–17 and particulate matter accumulation.18,19 Particulate matter with aerodynamic diameter ≥10 μm (PM₁₀) and O₃ pollution events occurring particularly from April to October6,15,16 have been shown to cause an additional violations per year of both the European health standard for O₃ (maximum daily 8-hr mean ≤120 μg m⁻³ or about 60 ppb)20,21 and the European daily PM₁₀ target value (daily PM₁₀ concentration ≤50 μg m⁻³).6 Owing to this context, during summer 2007 in Calabria region, southern Italy, an intensive sampling campaign was simultaneously performed at a marine, an urban, and a high-altitude ground-based monitoring station in order to assess the spatial variation of PM₁₀ and O₃ concentrations and characterize the atmospheric conditions that lead to O₃- and PM₁₀-rich episodes. The obtained results were then analyzed using both backward trajectories and satellite derived products which were useful to understand the aerosol sources and transport pathways. Findings were confirmed by soundings of PM₁₀, O₃, humidity, and temperature recorded at other existing monitoring stations over Calabria area. Fire hotspots were identified by the MODIS Rapid Response System (http://rapidfire.sci.gsfc.nasa.gov/), which provides for specific locations and daily frequency of wildfire events. Finally, a cluster analysis, based on the prevailing air mass influence, was performed allowing to discriminate the contribution of different air mass origin and paths and to better characterize the correlation between O₃ and PM₁₀ concentrations. In order to assess the Saharan dust contribution in exceeding PM₁₀ target values, a statistically based methodology was applied to the three PM₁₀ data sets obtained. Spatial variability between observations performed at the three sampling locations and the comparison of the obtained results with measurements previously performed across the Mediterranean Basin were furthermore presented.

MATERIALS AND METHODS

Monitoring Sites

From 18 July to 30 August 2007, ground-based air measurements were simultaneously performed at three monitoring stations located in Calabria, southern Italy. Calabria region is a mountainous area of 15,230 km² with a coastal perimeter of 738 km. Ninety percent of the regional territory consists of mountains with maximum elevations of 2266 m above sea level (a.s.l.) and only 10% is represented by coastal zones. The complex topography of the region is characterized by microclimate conditions representing disturbance factors within the Mediterranean climate regime (i.e., hot dry summer and mild winter) and three rainfall subzones could be well recognized: the Tyrrhenian, Central, and Ionian subzones.22 All three monitoring stations involved in this study are located in the province of Cosenza (Figure 1), and their main characteristics are summarized in Table 1. According to the European Environmental Agency,23 air quality monitoring stations involved in this work has been classified as urban (UB), suburban (SB), and rural (RB) background stations with regard to Rende (N39°20′0″ E16°11′0″), San Lucido (N39°19′22″ E16°02′44″), and Longobucco (N39°23′39″ E16°36′49″) sites, respectively.

Figure 1. Monitoring site locations: suburban background (SB), urban background (UB), and rural background (RB) sites.

Table 1. Coordinates and monitoring station classification for each sampling site

Site	Town	Altitude (m a.s.l.)	Latitude	Longitude	Monitoring Station Classification*
SB	San Lucido	49	39° 19′ 22″	16° 02′ 44″	Suburban background
UB	Rende	250	39° 20′ 0″	16° 11′ 0″	Urban background
RB	Longobucco	1379	39° 23′ 39″	16° 36′ 49″	Rural background
Note: *According to Decision 2001/752/CE.

The urban background (UB) station is located inside the campus of the University of Calabria in Rende. This town is about 4 km far from the major urban area of Cosenza even if the two conurbations effectively run into each other, constituting one continuous urbanized area of roughly 100,000 inhabitants. The increasing urbanization of the area led to a 4% increase in fuel consumption between 2002 and 2006 and the traffic flow is the major local source of air pollution.19 The A3 motorway crosses the area at about 1km east of the measurements site with a considerable traffic flow of about 25,000 vehicles per day.24

The suburban background (SB) site is a marine station located on a small headland 49 m a.s.l. on the Thyrrenian coast near San Lucido, a village with 5906 inhabitants. During the summer season, particularly in July and August, this seaside area attracts a large number of tourists, with a resulting rapid increase in population and traffic flow on the nearest road (about 3000 cars per day25), the SS 18, located at about 600 m east from the sampling site.

The rural background (RB) station is an inland mountainous site located at 1379 m a.s.l., near Longobucco village in the Sila Massif (∼1900 m a.s.l.), which is part of a north-south mountain chain between the Tyrrhenian and Ionian seas. This area is often characterized by cooler summers with some precipitations22,26,27 and the mountain chain itself represents an important climate regulator for downwind zones,28,29 with the function of a topographic barrier for air masses blowing eastward respect to our RB site. According to the European Environmental Agency,23 this monitoring station is further classified as “remote” site, the influence of local emission sources on this sampling area being very small. Longobucco village has in fact a population of about 4000, located in a valley some kilometers away from our sampling site.

Experimental Setup

PM₁₀ samples were collected on preweighed and preconditioned 47-mm Teflon filters over a 24-hr sampling period, using a dual-channel high-volume particulate matter sampler with a flow rate of 38 L min⁻¹. A total of 44 samples were obtained at each monitoring site from 18 July to 30 August 2007. O₃ measurements, available for both UB and RB sites, were performed by automatic Teledyne ultraviolet (UV) absorption ozone analyzers (model 400E). The O₃ analyzers were automatically calibrated every 24 hr by an internal permeation source. A sampling flow rate of 0.8 L min⁻¹ was used to obtain O₃ concentrations every 5 min. CO measurements were carried out by an automatic CO analyzer Teledyne (model 300E) only at the SB site. Meteorological parameters were continuously measured at all three monitoring stations.

Data Analysis Treatment

For each monitoring site the data set consists of PM₁₀ daily mass concentrations (μg m⁻³) and daily meteorological parameters average values (temperature [T; °C], relative humidity [RH; %], precipitation [mm]; daily mean wind vector along with component wind speeds [WS; m s⁻¹], and wind direction [φ; deg]). In addition, 5-min O₃ average values were available for UB and RB stations. whereas 5-min carbon monoxide (CO) levels were recorded only at the SB site. For each measured variable, explorative statistical parameters of data have been performed. The investigation on the origin of the long-range transported air masses was supported by backward-trajectory analysis.30,31 Five-days backward-trajectories were calculated at different altitudes (500, 1000, and 1500 m a.g.l.) for each sampling day by the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model,32 available at the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (ARL).33 Several tools, such meteorological maps, numerical models, and satellite images were also analyzed as useful sources of information. In particular, the interpretation of the aerosol variability34–36 was supported by the Navy Aerosol Analysis and Prediction System (NAAPS), developed by the Naval Research Laboratory (NRL).37 MODIS true-color images (http://rapidfire.sci.gsfc.nasa.gov/) were used to identify the origins and the extension of fire events. Specific locations and daily total fire hotspots over the Cosenza province were also obtained from the Fire Information for Resource Management System (FIRMS) (http://maps.geog.umd.edu/firms), which integrates remote sensing and Geographical Information System (GIS) technologies to deliver global Moderate-resolution Imaging Spectroradiometer (MODIS) hotspot/fire locations.

Identification of Saharan Dust Outbreaks in Calabria Region

The identification of Saharan dust outbreaks (SDOs) was performed applying the methodology well described in Meloni et al.38 As shown in the adopted methodology, the basic assumption is that the air mass is loaded with desert dust, which is expected to become the main aerosol type when the trajectory interacts with the mixed layer (entrainment condition: we assume that the interaction takes place if the air mass passes within 500 m of the mixed layer top altitude), or spends a large fraction of time over the Sahara (permanence condition). The source region is identified as the one where the distance between the trajectory and the mixed layer is the lowest, or where the intrusion into the mixed layer is the deepest. The entrainment condition was checked for the trajectory ending at 4000 m first, and then for that ending at 2000 m. In order to account wet deposition events, as also reported in the adopted methodology, the entrainment and the permanence conditions were accepted only if it did not rain after the dust loading.38 The methodology reported in Escudero et al.,39 based on the statistical data treatment of PM₁₀ time series, has been also adopted in order to quantify the daily African PM₁₀ load. After a prior extraction of the days with Saharan dust influence, the daily dust load was obtained by applying a monthly moving 30th percentile to the PM₁₀ time series at the reference rural station located at Firmo (N39°43′22″ E16°10′15″, 370 m a.s.l.) (Cosenza province). The recorded PM₁₀ time series at the Firmo rural station are available since 2004 and can be easily downloaded by the Banca dati Relazionale Aria Clima Emissione (BRACE) data set (http://www.brace.sinanet.apat.it/web/struttura.html).

RESULTS AND DISCUSSION

PM₁₀ and O₃ Levels Recorded during Summer 2007

The sampling campaign was performed during typical fair weather conditions of the Mediterranean summer. No precipitations were generally recorded with the exception of one event occurred at RB during 23 August. The hourly maximum of air temperature reached values of 37, 42.4, and 31.7 °C at SB, UB, and RB sites, respectively. Local winds were quite weak, whereas a persisting anticyclone system often occurred at synoptic scale. Figure 2 shows the temporal series of PM₁₀ concentrations recorded at the three sites. A common pattern has been observed with particular evidence in the raising values, even though the second and third peak values, reached at both SB and UB sites, did not appear at RB. PM₁₀ daily levels ranged widely at each station, with concentrations from 11.8 to 134.0, from 1.1 to 99.3, and from 7.7 to 86.4 μg m⁻³ at SB, UB, and RB site, respectively, with mean values of 39.7 ± 23.8 (SB), 35.5 ± 22.7 (UB), and 24.7±16.0 (RB) μg m⁻³. Such high standard deviations are not unusual for data obtained in the Mediterranean atmosphere,40–42 suggesting that ambient air concentrations are often influenced by nonlocal sources.43 PM₁₀ mean values were generally lower than the yearly limit value of 40 μg m⁻³; however, a number of 9, 10, and 2 days exceeding the daily PM₁₀ limit value44 was recorded at SB, UB, and RB sites, respectively. The observed PM₁₀ data have been compared with PM₁₀ mass levels previously recorded across the Mediterranean Basin.6 In this area, as reported in Querol et al.,6 PM₁₀ concentrations reach the maximum values during summertime as result of concurrent factors, such as higher frequency of African dust outbreaks, lower precipitation, higher (local/regional) resuspension, dryness of soils, and increased formation of secondary aerosols caused by the maximum solar radiation. Summer PM₁₀ mean values reported in Querol et al.6 highlights a clear west–east and north–south increasing trends ranging from 12.7 μg m⁻³ in northern Italy to 28.8 μg m⁻³ in southern Italy and from 15.0 μg m⁻³ in the western Mediterranean Basin to 34.1 μg m⁻³ in the eastern Mediterranean Basin. PM₁₀ mean values reported in literature agree rather well with our results also with regard to the sites' location. PM₁₀ trend observed during this study (SB > UB > RB) seems in fact to be in line with PM₁₀ trends measured in the Mediterranean Basin, with lower PM₁₀ concentrations observed over high mountain areas compared to those measured at island and coastal sites (probably due to the sea spray contribution).6

Figure 2. Time evolution of PM₁₀ levels recorded at (a) SB, (b) UB, and (c) RB sites.

Table 2 shows the O₃ statistic results obtained during the sampling campaign along with maximum 1-hr mean and maximum daily 8-hr mean values. Daily maximum 8-hr average concentration over 15 days at the UB site and 31 days at the RB site exceeded the ozone long-term objective for the protection of human health (>120 μg m⁻³ or approximately 60 ppb), whereas hourly mean values at both stations were less than the information threshold.45 The O₃ concentrations observed during summer were quite higher than those recorded at the same site locations during wintertime,46 thus corroborating the tropospheric seasonal cycle of background O₃. Over the Mediterranean region and in southern Europe, O₃ levels are, in fact, usually dominated by the presence of a broad spring–summer peak,47–51 due to the superposition of the hemispheric-scale spring maximum (April–May) and the increased photochemical production of O₃ characterizing the lower troposphere during summertime.1,52 Similar O₃ behaviors were pointed out by Vecchi and Valli53 for two measurement sites located in the southern Italian Alps and Pre-Alps (Alpe Motta, 1800 m, and Passo S. Marco, 1900 m) where a winter minimum (from 26 to 32 ppb) was followed by a spring peak (46–49 ppb) and subsequent summer peak (47–55 ppb). Previous measurements performed at Mt. Cimone (northern Italian Apennines; 2165 m), during the period 1996–2004,54 have also revealed the presence of a seasonal O₃ cycle (a yearly mean value of 54 ppb) with two peaks in spring (average 59 ppb) and in summer (average 63 ppb). These O₃ values notably agree with those observed during this study at our sampling stations considering that tropospheric O₃ concentrations increase with altitude.55 This phenomenon, owing to the influence on O₃ concentrations of air mass transport from the stratosphere and to the high efficiency of O₃ dry deposition and removal at the surface,56 could also partially explain the difference in ozone levels observed at RB and UB stations. The close location of the motorway to the UB site leads otherwise to lower measured ozone concentrations, most likely due to titration by NO,46 compared to those observed at the RB site.

Table 2. Descriptive statistics for ozone levels

	O3 (ppb)
	Max 24 hr	Mean 24 hr	SD 24 hr	Max 1 hr	Max 8 hr
SB	—	—	—	—	—
UB	57.5	43.8	6.8	82.3	71.2
RB	63.6	50.9	6.2	82.9	80.4

Saharan Dust Outbreak Influence on PM₁₀ and O₃ Levels

Dust mobilization is a complex process that is a function of many atmospheric, soil, and terrain properties, therefore, the detachment, entrainment, and transport of dust are highly variable both spatially and temporally.57,58 The presence of mineral dust in the atmosphere may contribute to climatic variations and influence the behavior of some tropospheric trace gases, changing the oxidizing capacity of the atmosphere.59,60 In particular, aerosols can either increase or decrease the actinic flux (depending on scattering and absorption properties at UV wavelengths), affecting the photochemical ozone changes.61 In particular, HNO₃ and NO₃ depletion on dust particulate can remove a fraction of O₃ precursors (i.e., NO_x), favoring a reduction of photochemical O₃ production efficiency.62,63

Data from the space-borne total ozone monitoring spectrometer (TOMS) show that the Bodélé Depression of southern central Sahara is the most intense source in the world, with aerosol index (AI) values that exceed 3.0.64 In the Southern Hemisphere there are not major areas of dust production but southern Africa has a large area of arid terrain (Kalahari and Karoo), with AI values ranging from 0.8 to 1.1,64 that gives rise to dust plumes blowing westwards towards the South Atlantic.65 In the Northern Hemisphere, the bulk of dust is mostly produced by the Sahara desert in Northwest Africa and the Gobi desert in East Asia.66 Owing to the above implications, in order to assess the impact of Saharan dust outbreaks on measurements recorded at the three sampling sites, a deep investigation on their dynamic evolution has been carried out. After 18 July, the establishment of an anticyclonic circulation in the central Mediterranean led to typical weather conditions that climatologically characterize the Italian summer.46,67 As the subtropical anticyclone developed, a warm advection was observed to proceed over Italy, causing a constant increase in temperature at all the stations. The eastward movement of the high-pressure system favored the transport of air masses from North Africa to the central Mediterranean and Italy, leading to dust outbreaks at all monitoring stations.

The intrusion of SDOs recorded at our monitoring stations has been observed for periods of 4 (from 22 to 25 July), 5 (from 22 to 26 August), and 2 (from 29 to 30 August) days over the whole sampling period (see Figure 2). The strength impact of the highlighted SDO events on PM₁₀ concentrations recorded at the three monitoring sites showed some differences. In particular, the three SDO events have been observed at both SB and UB sites with PM₁₀ levels reaching similar peak values between 23 and 25 July (first event), between 23 and 25 August (second event), and on 30 August (third event). Concerning PM₁₀ observations at the RB site, only the first SDO event has been recorded, whereas the second and the third SDO events did not occur (see Figure 2). Regarding the second SDO, a rainfall event checked along the trajectory followed by the air masses prior to ending at RB (at 00:00 of 23 August) was the most likely reason causing the difference between PM₁₀ levels observed at SB and UB sites and those measured at RB site. The related outcome was just a slight increase in PM₁₀ mass concentrations (15.6 μg m⁻³) at RB station in contrast to those observed at SB and UB sites (peak values of 134.0 and 99.3 μg m⁻³), respectively. According to Meloni et al.,38 air masses are loaded with Saharan dust through entrainment in the mixed layer; this condition requires that the air mass trajectory approaches the mixed layer within 500 m of the top layer or enters the mixed layer itself. Unlike the previous two SDO events, during the third one, the entrainment condition was met along both trajectories ending at SB and UB sites (see Figure 3a, b), whereas the entrainment of the trajectory ending at RB site did not occur and consequently the loading with Saharan dust did not take place (Figure 3c).

Figure 3. Trajectory and mixed layer altitudes along the path followed by backward trajectory before their arrival at (a) SB, (b) UB, and (c) RB sites, respectively, on 30 August.

Figure 4 shows both PM₁₀ and O₃ time series observed at both UB and RB (left panel) sites and air temperature (T; °C) and relative humidity (RH; (%) behaviors at all monitoring stations. A decrease of O₃ concentrations in conjunction with an increase of PM₁₀ values has been observed during each identified SDO event (see gray bars). Similar trends were also detected for RH (%) and T (°C) values (right panels). In addition, a selection of the days with and without SDO influence on PM₁₀ and O₃ concentrations has been performed, obtaining an explorative statistical analysis, summarized in Table 3. During the identified SDO days, an average increase in PM₁₀ concentrations of 135.5% at the SB site, 167.4% at the UB site, and 103.6% at the RB site and an average decrease in O₃ levels of 17.7% and 13.9% at the UB and RB sites, respectively, have been observed in terms of percentage variation (Δ%) over days without SDO influence. These outcomes agree well with previously studies68–70 as well as observations focused directly on this issue.71–74 Available PM₁₀ and O₃ concentrations recorded at two existing monitoring stations located in Calabria were furthermore checked. During the same sampling period, ground-based air measurements were performed, in particular, at two stations, both classified as rural, one located at Saracena (N39°46′44″ E16°09′28″, 606 m a.s.l.) and the other one at Firmo (N39°43′22″ E16°10′15″, 370 m a.s.l.), 68 km and 61 km NNE, respectively, far from our site in Rende (UB); data available at the BRACE database (http://www.brace.sinanet.apat.it/web/struttura.html). The reported values in Figure 5 show clear increases in PM₁₀ in conjunction with O₃ levels decreasing during the same SDO events detected at our sampling stations. This occurrence corroborates that Saharan dust transport is a large-scale phenomena that could influence significantly both PM₁₀ and O₃ concentrations.

Figure 4. Time series of PM₁₀ (µg m⁻³) and O₃ (ppb) levels (left panels) and of T (°C) and RH (%) values (right panels) for (a, b) SB, (c, d) UB, and (e, f) RB sites, respectively. Grey bars highlight the identified SDO periods.

Table 3. Comparison of PM₁₀, O₃, T, and RH explorative statistic values recorded during days with and without SDO

	During SDO			Without SDO
	Min	Max	Mean	Min	Max	Mean	Δ%
PM10 (μg m^−3)
SB	29.0	134.0	69.8	11.8	64.6	29.6	135.5%
UB	42.9	99.3	66.9	1.1	45.6	25.0	167.4%
RB	29.6	86.4	55.1	8.6	86.4	27.0	103.6%
Firmo	28.6	57.6	41.8	6.3	35.5	16.7	151.1%
Saracena	14.1	33.4	25.6	1.5	21.6	9.5	170.1%
O3 (ppb)
SB	—	—	—	—'	—	—	—
UB	23.0	51.5	36.9	37.0	57.5	44.8	−17.7%
RB	37.9	54.4	44.4	35.1	63.6	51.5	−13.9%
Firmo	33.4	51.3	42.0	37.5	58.3	45.8	−7.5%
Saracena	38.7	55.2	46.0	37.9	62.7	97.8	−5.9%
T (°C)
SB	26.0	31.3	28.5	23.0	26.4	24.8	14.9%
UB	29.0	33.8	30.8	21.9	30.7	25.9	19.0%
RB	19.8	22.9	21.6	11.9	23.0	16.7	28.9%
RH (%)
SB	32.5	72.9	51.9	36.5	70.2	60.3	−13.9%
UB	20.7	44.6	32.9	30.9	59.8	47.2	−30.3%
RB	32.4	40.2	36.1	38.4	77.1	57.1	−36.8%
Note: Percent variation (Δ%) between mean values calculated over days with and without SDO is also reported.

Figure 5. Time series of PM₁₀ and O₃ levels recorded at (a) Firmo and (b) Saracena rural sites.

Wildfire Events

Among the major sources of atmospheric pollutants and climate altering species, an important role is played by wildfire emissions.75,76 Elevated concentrations of trace gases (i.e., O₃, CO) and aerosol particles have been observed in various regions of the world during wildfire events.75–83 The dominant fraction of wildfire emissions consists of CO₂ and CO, whereas less than 5% of the carbon is emitted as particulate matter.78,84 Previous studies have shown that the atmospheric compounds directly emitted by wildfire or produced by photochemical processes within the plumes can be transported over long distances, thus affecting both air quality and climate on the regional and global scales.85 The Mediterranean Basin is affected by large wildfire events, especially during summer.86,87 Calabria region represents an interesting area of study for wildfire occurrences within the Mediterranean Basin, with an high level of wildfire density that characterizes this area as one of the most severely affected in Europe.88 From 1996 to 2007, on average 1150 forest fires occurred in Calabria region, which has a total burned surface of circa 7000 ha and during the year 2007 only, the region reported the largest wooded area in Italy (∼9608 ha) affected by wildfire events, with a total number of 1614 fires.89 During the sampling period, an high number of fires was recorded over Cosenza province by the MODIS Rapid Response System (http://rapidfire.sci.gsfc.nasa.gov/), which provides the daily frequency of wildfire events for selected locations. Figure 6 highlights the frequency of the daily total fire hotspots over Cosenza province, showing a strong correlation with PM₁₀ time series observed at our sampling sites. An apportionment of the fire hotspots number affecting each monitoring station has been furthermore performed, taking into account only those falling in an area extending less than 5 km from each sampling location. The outcomes, reported as dots in Figure 6, showed that the highest frequency of the daily fire hotspots occurred in conjunction with PM₁₀ peak values recorded at each sampling site. It was thus likely to suppose that a fraction of the high PM₁₀ values could also originate from aerosols production related to wildfire events occurred in Cosenza province. The NAAPS-based smoke maps related to the study period (not shown here) highlighted a plume characterized by an aerosol optical depth of 0.1–0.2, thus suggesting large smoke emissions from Calabria region during 23–25 July and 23–30 August.90 The effective influence of wildfire events over measurements performed at SB site was confirmed by CO levels that reached two peak values, both of about 180 ppb, on 24 July and 23 August. Values of 45 ppb in CO concentrations were instead recorded during days with a lower frequency in fire hotspots. At UB and RB sites, for which CO measurements, unfortunately, were not available over the studied period, a thorough investigation in the daily prevalence of local winds was carried out in order to check if the air masses arriving at both UB and SB sites were also influenced by the identified fire hotspots.

Figure 6. Daily total hotspot fires (HSFs) over Cosenza province (Calabria region), from the MODIS Rapid Response System-Web Fire Mapper (http://maps.geog.umd.edu). Dots represent the number of hotspot fires singularly affecting our sampling sites.

Air Mass Back-Trajectory Analysis

In order to evaluate the contribution of different air mass origins and paths on PM₁₀ and O₃ levels recorded at the three sampling stations, a back-trajectory analysis was carried out. Sampling days were classified in six different clusters:

1.	WMED—Western MEDiterranean Basin: Air masses originating and/or passing over West Mediterranean.
2.	NATL—North ATLantic: Air masses having 5-day origin over the North Atlantic at altitudes above 3000 m a.s.l. and descending lower before reaching the measurement sites.
3.	CEU—Continental EUrope: Air masses passing over European highly industrialized and densely populated area (Central-Eastern Europe).
4.	SDO—Saharan Dust Outbreak: Air masses showing to be loaded with Saharan dust over North Africa prior to arriving at the sampling sites.
5.	WF—WildFire events: Air masses traveling in the lower troposphere over the areas where widespread forest and brush fires occurred.
6.	SDO&WF—Saharan Dust Outbreaks and WildFire events: Air masses affected by both these events.

The occurrence frequencies of the above described air mass classes are reported in Figure 7 (right panels). The weight of air mass influences was quite similar for both SB and UB sites, with 45% and 48% of each sampling campaign that was respectively affected by natural events. In particular, WF and SDO&WF clusters prevailed over the influence of SDO, which accounted only for 5% (SB) and 2% (UB). Overall, the NATL class showed the largest weight, whereas the CEU showed the lowest. Influence of natural events at RB site has been shown to account poorly, with a total frequency occurrence of 9% (2% for SDO and 7% for SDO&WF). The NATL cluster prevailed also for RB, followed by WMED and CEU. Within the obtained clusters, basic statistics for PM₁₀ and O₃ levels as well as for T (°C) and RH (%) values were furthermore calculated for each sampling station. The average values are reported in Figure 7 (left panels). Relevant differences were observed among the different air mass classes; in particular, the industrial and densely populated continental Europe (CEU) represented the major sources of O₃ concentrations during the sampling period. With air masses coming from these regions, the mean O₃ level was the highest at both UB and RB sites: 47.9 and 54.3 ppb, respectively (see Figure 7a, c). As reported by EEA,17 during summer 2007, O₃ target value for human health protection was exceeded in different parts of Europe, with the most significant ozone episode occurring from 14 to 21 July. From 18 to 21 July, a CEU influence was detected at our sampling stations, with hourly maximum values of 82.2 ppb at UB and 82.8 ppb at RB recorded on 21 July. In contrast, when air masses arrived from western Mediterranean Basin (WMED), an area characterized by lower anthropogenic contributions, mean O₃ concentrations of 43.7 ppb for UB and 46.3 ppb for RB have been observed. As expected, lowest O₃ concentrations were recorded in conjunction with SDO events and air masses coming from North Africa. Due to the relative lack of pollution sources, North Africa is in fact not considered a source of tropospheric O₃ for Europe.73,91,92 It is relevant to point out that during the WF-alone influence (occurred only at UB), O₃ levels reached an average value of 46.4 ppb quite close to those measured over CEU circulation class (see Figure 7c). Therefore, as observed at UB site, even if the highest O₃ concentrations were related with polluted air masses traveling over continental Europe, the contribution of wildfires on tropospheric O₃ concentrations was not negligible. Otherwise, in concurrence of WF with SDO, the role played by WF events on the enhancement of O₃ levels was not evident at both UB and RB sites. This could be most likely explained by the presence of mineral dust that could have partially hindered O₃ production. In fact, the average PM₁₀ concentrations during the SDO&WF influence at UB was about 3 times higher than that calculated during WF occurrence and 1.3–2.1 times higher than those obtained under the SDO-alone influence at UB and RB sites, respectively. Overall, at each sampling stations, even at SB site, the SDO&WF contribution to PM₁₀ concentrations broadly prevailed, whereas the lowest PM₁₀ values at SB and UB sites were observed with air masses coming from the North Atlantic (NTAL) and at RB under the WMED influence.

Figure 7. Cluster analysis based on the prevailing air mass trajectories and paths. Average values for PM₁₀, O₃, T, and RH over each kind of influence (left panels) and occurrence frequencies of the identified air mass classes (right panels) are reported for (a, b) SB, (c, d) UB, and (e, f) RB sites, respectively.

PM₁₀ Exceedances Evaluation

Because contributions from natural sources can be assessed but cannot be controlled, the new Directive 2008/50/EC⁴⁴ gives the possibility of skipping exceedances due to natural sources with regard to the amount of their contribution. Therefore, where the exceedances observed seem to be due prevalently to natural contributions, these latter could be subtracted when assessing compliance with air quality limit values. In southern Europe, natural episodes with the greatest impact on the PM₁₀ levels are represented by the intrusions of African air masses with high dust loads.34,93–95 In order to assess the daily African dust contribution to PM₁₀ mass levels during Saharan dust outbreaks at the sampling sites, the methodology presented by Escudero et al.,39 based on statistical data treatment of time series of PM₁₀ levels, was used. The results are presented in Figure 8 (left panels) in which the absolute values of dust contribution are presented as black bars. The net dust contribution was quite large, with a weight accounting for 29.3–93.3% at SB site, 32.9–72.1% at UB site, and 38.9–77.0% at RB site during the days affected by Saharan transport. Skipping the obtained net dust load from the measured PM₁₀ levels, a relevant reduction in the number of daily excedancees has been observed, getting down from 9 to 3 at SB site, from 10 to 2 at UB site, and from 2 to 1 at RB site. In Figure 8 (right panels), for each monitoring site the number of exceedances as absolute and percentage values, with and without Saharan Dust contribution, are reported. Referring to the above reported cluster analysis, the days with observed PM₁₀ exceedances have been all classified as SDO and SDO&WF. Therefore, skipping the Saharan dust load, the remaining PM₁₀ exceedances could be mostly attributed to wildfire influence, whose contribution to PM₁₀ daily limit exceedances seems not to be negligible.

Figure 8. Quantification of the net African dust load at (a) SB, (b) UB, and (c) RB sites. Left panels report time series of PM₁₀ concentrations, with the black bars indicating the absolute dust load calculated for days with SDO influence. Absolute and percent values of number of exceedances are showed on the right panels with and without Saharan dust contribution.

CONCLUSION

During summer 2007, an intensive sampling campaign was simultaneously performed, from July to the end of August, at three background monitoring stations in Calabria region (southern Italy) in order to evaluate the PM₁₀ and O₃ spatial variation and characterize the atmospheric conditions that lead to O₃- and PM₁₀-rich episodes. The results presented in this study showed that such variations were probably related to air masses coming from continental Europe and connected with events of mineral dust transport and wildfire products. Because Calabria region is located in the southern Mediterranean Basin, the observations carried out in this area can provide useful information for better understanding the role played by different transport processes in modifying the tropospheric background conditions. Five-day backward trajectories used to determine air mass origin highlighted that continental Europe was the major pollution source area for O₃, whereas the North African desert region was the most important source area for high PM₁₀ and low O₃ concentrations. During dust events, in fact, the PM₁₀ mean concentrations at SB, UB, and RB sites were 69.8, 66.9 and 55.1 μg m⁻³, respectively, compared to 29.6, 25.0, and 27.0 μg m⁻³ PM₁₀ values averaged over days without Saharan dust intrusions. Conversely, during the greater number of dust events, a significant decrease in O₃ levels occurred in Calabria region, with a percent values of 18% and 14% at UB and RB sites, respectively. Furthermore, even if further analysis will be carried out to better evaluate the specific amount of O₃ concentrations directly due to wildfire emissions, our observations suggest that in presence of favourable weather conditions and during specific periods of summertime, the contribution of wildfires on tropospheric O₃ concentrations cannot be neglected. Because in the future the strength and frequency of such phenomena, such as Saharan dust and wildfire emissions, could be affected by climate change,96 investigations on their impact on background O₃ in the southern Europe and the Mediterranean area still appears necessary. Recently, there is a growing interest in quantifying the background PM₁₀ and O₃ concentrations, as well as their variability and trends, in order to support policy makers in defining an effective air quality management.97 In assessing compliance with EU legislation, a statistically based methodology, such as that adopted in the present study for PM₁₀ data sets, is useful for computing the amount of Saharan dust contribution. This application showed that more than half of PM₁₀ daily exceedances could be attributed to Saharan dust. Because natural sources influence on PM₁₀ levels can be assessed but not controlled, in dealing with the scope of air quality management, their contribution can be removed. However, their impact, as reported in several epidemiological studies, raises concerns about adverse health effects and appropriate interventions by health authorities. A know-ledge of such natural events over the southern Mediterranean Basin thus results of interest even to their correct parameterization into air quality models could then support an adequate health alert system. Results presented in this work could be considered as an exploratory study that needs further improvements. In particular, chemical analysis of aerosol appears as a necessary step to identify unambiguously the presence of mineral dust and wildfire products at the measurements sites. Overall, in the absence of a single, recognized standard procedure, more studies will be conducted to assess the specific impact of these two natural sources on PM₁₀ concentrations and their influence over O₃ levels in the Mediterranean area, with a special regard to the southern regions.

ACKNOWLEDGMENTS

The authors acknowledge the NOAA Air Resources Laboratory (ARL) for providing the HYSPLIT transport and dispersion model, and of the READY Web site (http://www.arl.noaa.gov/ready.html), whose results are used in this publication. The NCEP-based images are provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado, USA, from their Web site at http://www.cdc.noaa.gov/, whereas the MODIS “Hotspot” data are from the MODIS Rapid Response System-Web Fire Mapper (http://maps.geog.umd.edu). The authors would like to express their gratitude also to the NASA/Goddard Space Flight Center, NOAA Air Resources Laboratory (ARL), and to the Naval Research Laboratory for the providing the TOMS-AI maps and the NAAPS aerosol maps, respectively.