Multi year sun-photometer measurements for aerosol characterization in a Central Mediterranean site

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy Atmospheric Research 104–105 (2012) 98–110

Contents lists available at SciVerse ScienceDirect

Atmospheric Research j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / a t m o s

Multi year sun-photometer measurements for aerosol characterization in a Central Mediterranean site A. Boselli a, b,⁎, R. Caggiano a, C. Cornacchia a, F. Madonna a, L. Mona a, M. Macchiato b, G. Pappalardo a, S. Trippetta a a b

IMAA, Istituto di Metodologie per l'Analisi Ambientale, CNR, C.da S. Loja, Z.I. 85050 Tito Scalo (PZ), Italy DSF, Dipartimento di Scienze Fisiche, CNISM, Università Federico II, Via Cintia, 80126 Napoli, Italy

a r t i c l e

i n f o

Article history: Received 23 March 2011 Received in revised form 4 August 2011 Accepted 9 August 2011 Keywords: Aerosol optical depth Ångström exponent Sun photometer Clustering Particulate matter

a b s t r a c t Aerosol characterization at a Mediterranean site is carried out on the base of 40-month (November 2005–March 2009) measurements of aerosol optical depth (AOD) at 440 nm and Ångstrom coefficient (α) at 440/870 nm collected at the atmospheric observatory of the Istituto di Metodologie per l'Analisi Ambientale of the Italian National Research Council (CNRIMAA). Mean values of 0.161 ± 0.004 and 1.44 ± 0.54 are observed for AOD and α, respectively. Both AOD and α are characterized by a wide range of values from 0.03 to 0.6 and from 0.15 to 3.14, respectively, and a day-to-day variability larger than 100% for AOD b 0.18 and α b 0.95. A seasonal behavior is found with higher AOD and lower α values in spring–summer. Four aerosol populations are found in the count distribution of AOD. The k-means cluster analysis allowed the identification of measurements belonging to each one of the four populations identified in the AOD distribution. Four prevailing aerosol classes are identified by using backtrajectories and model analyses: dust, continental, maritime and mixed aerosols. Dust and continental aerosol are the most common at Central Mediterranean (37.5% and 41% of the cases, respectively), with a wide variability in both AOD and α. Only in about 4% of the cases can aerosol be classified as maritime, and however the mixing with other aerosol is not negligible. A comparative study of cluster results and aerosol type identification reveals that the classification, based on the cluster analysis, is reliable for dust event and continental case, with a confidence level of 85% and 65% respectively. Finally, the Principal Component Analysis (PCA) applied on each cluster's PM1 and trace element daily concentration reveals an influence of dust on PM1 measurements at ground level. © 2011 Elsevier B.V. All rights reserved.

1. Introduction The understanding of atmospheric aerosol role on the Earth system is a subject of growing interest due to its influence on Earth-atmosphere climate system, air quality and human health. The last IPCC scientific report highlighted the large uncertainties on the aerosol impact on radiation budget because of the

⁎ Corresponding author at: IMAA, Istituto di Metodologie per l'Analisi Ambientale, CNR C.da S. Loja, Z.I., 85050 Tito Scalo (PZ), Italy. Tel.: + 39 081 676276; fax: + 39 0971 427271. E-mail address: [email protected] (A. Boselli). 0169-8095/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.atmosres.2011.08.002

variability in their properties and distribution (IPCC, 2007). In particular, large uncertainties are related to secondary effects of aerosol on the radiation budget, i.e. effects related to the crucial role the aerosols play on the cloud formation and properties as well as on the precipitation cycle (IPCC, 2007). In order to reduce the current uncertainties, long-term monitoring of the aerosol properties is essential. Moreover, a coordinated research strategy is needed to quantitatively address the complex aerosol–climate problem. As stressed by several international programs (e.g., GEOSS Global Earth Observation System of Systems), integration of data coming from multiple platforms (e.g., ground-based networks, satellites, ships, and aircrafts) and different sensing techniques (e.g.,

Author's personal copy A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

in-situ measurements and remote sensing observations) should be developed. For this reason, several efforts have been made in recent years to improve and sustain measurements, including: the establishment of networks devoted to the systematic observation of aerosols (e.g., European Aerosol Research LIdar NETwork – EARLINET (Pappalardo et al., 2009), AErosol RObotic NETwork – AERONET (Holben et al., 1998), European Monitoring and Evaluation Programme – EMEP, China Aerosol Remote Sensing NETwork – CARSNET, Canadian Sunphotometer Network – AeroCAN, French network of Sunphotometer – PHOTONS (Goloub et al., 2007), Iberian Network for Aerosol Measurements – RIMA), the development and implementation of new and enhanced satellite sensors and the improvement of numerical models. In particular, global photometer networks such as AERONET, SKYradiometer NETwork - SKYNET (Kim et al., 2004; Takamura et al., 2004), and the Global Atmosphere Watch Precision Filter Radiometer GAW PFR network (Wehrli, 2002), have been developed for the systematic observation of optical, microphysical and radiative aerosol properties from the ground. These measurements provide the basis for a long-term global climatology of the aerosol properties and are also very important to validate satellite observations and evaluate global and regional aerosol models. In this respect, local studies based on long data-sets have been carried out in different areas of the AERONET network (e.g. Dubovik et al., 2002; Perrone et al., 2005; Fan et al., 2006; Saha et al., 2008; Toledano et al., 2009; de Meij and Lelieveld, 2011). In particular, the Mediterranean area is one of the most interesting regions for aerosol study, being often referred to as a natural laboratory for atmospheric and climate study. This area is surrounded by the main sources of natural and anthropogenic aerosols: the Sahara desert on the south, the Mediterranean Sea itself, Europe on the north with highly populated and industrialized countries and finally the heavy pollution sources of the developing countries on the east. Several studies indicate that aerosol radiative forcing is among the highest in the world over the Mediterranean that acts as a crossroad of global pollution (Lelieveld et al., 2002; Hatzianastassiou et al., 2007; Mallet et al., 2011). Due to the complexity of the aerosol content over this area, a characterization of the aerosol over the Mediterranean basin is particularly useful. In the recent literature, there are some reports related to the identification and seasonal behavior of aerosol events (e.g. Gkikas, et al., 2009) as well as studies focused on specific aerosol types, such as dust (Mona et al., 2006; Tafuro et al., 2006; Toledano et al., 2007a; Papayannis et al., 2008; Meloni et al., 2008), forest fires (Pace et al., 2005), and sea-salt aerosol (Marenco et al., 2007). CIAO, the ground-based atmospheric observatory (Madonna et al., 2011) of the Istituto di Metodologie per l'Analisi Ambientale of the Italian National Research Council (CNR-IMAA, Tito Scalo - Southern Italy - 40°36′N, 15°44′ E, 760 m above sea level) is an optimal site for the study of optical and microphysical properties of different types of aerosols. Thanks to its central location in the Mediterranean basin, this site is affected by long-range transport of aerosols both of natural (from Sahara desert and Tyrrhenian, Adriatic and Ionian Seas) and anthropogenic (from the European Continent) origin. In this work, we have reported on long-term observations of column integrated aerosol optical properties carried out at CIAO by means of a sun-photometer operating in the

99

framework of AERONET. Statistical and cluster analyses are performed in order to provide a local aerosol climatology and distinguish the main aerosol types observed over the measurement area. Moreover, the impact on particulate matter measurements at ground level is also investigated. Instrumentations and methods are described in Section 2. Results of climatological and statistical analyses of the aerosol optical depth (AOD) and Ångstrom coefficient (α) values are discussed in Section 3. Moreover, the identification of the main aerosol types observed over the study area is also reported. Finally, the influence of the different aerosol types on PM1 measurements at ground level is discussed. 2. Instrumentation and methods 2.1. CIMEL sun-photometer CIAO is one of the most advanced infrastructures for ground-based remote sensing in Europe (Madonna et al., 2011). CIAO is equipped with active and passive instrumentations whose combined use represents the best strategy for studying the aerosol–clouds–radiation interactions and the chemical, physical and dynamic processes occurring in the atmosphere. Among the instruments operating at CIAO, a CIMEL sun-photometer is operative in the framework of AERONET, measuring aerosol columnar properties. AERONET is a world-wide sun-photometer network which provides near real-time measurements of atmospheric aerosol optical properties and precipitable water (Holben et al., 1998). Data collected at CIAO are automatically transferred to the AERONET processing system through the Internet. Afterwards, data are calibrated, processed with inversion algorithms and then made available to users in near real-time (level 1.0 - unscreened and level 1.5 - automatically cloud screened). The highest quality data (level 2.0 - cloud-screened and quality assured) are provided after field calibration at calibration facility. The sun-photometer has been routinely operational at CIAO since December 2004 and provides direct solar irradiance measurements at 8 different wavelengths from UV to near IR (340, 380, 440, 500, 675, 870, 1020 and 1640 nm). The sun-photometer provides the aerosol optical depth (AOD) at each wavelength from the surface to the top of the atmosphere (Holben et al., 2001), along with the water vapor column content, and the Ångstrom coefficient (α) at different wavelengths. A big amount of information about the optical and microphysical aerosol properties, such as the refractive index and the aerosol size distribution, is also systematically retrieved from the sun-photometer measurements (Dubovik and King, 2000; Holben et al., 2001). In particular, the AOD is proportional to the total column loading of absorbing and scattering particles, whereas α characterizes the AOD spectral dependence and is related to the aerosol size, increasing as the particle size decreases. 2.2. PM1 in-situ measurements At the CNR-IMAA site, a low-volume (16.7 lmin − 1 flow rate) gravimetric sampler equipped with a PM1 cut-off inlet is also operative allowing the collection of PM1 daily samples on filters. During the examined period, the sampling time for each filter was 24 h. Before and after the sampling, each filter

Author's personal copy 100

A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

was humidity-conditioned in a filter-conditioning cabinet (approximately T = 20 ± 2 °C and RH = 50 ± 5%) for 48 h. The PM1 collected mass was determined by applying the gravimetric method. The total trace element content was determined by acid digesting the filters according to the chemical protocol described by Caggiano et al., 2001. Twelve elements (Al, Cd, Cr, Cu, Fe, K, Mg, Mn, Ni, Pb, Ti and Zn) were measured by Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES) or Atomic Absorption Spectrometry (AAS). In particular, Al, Cr, K, Mg, Mn, Na, Ti, and Zn concentrations were measured by means of ICP-OES, while Cd, Cu, Fe, and Pb by means of AAS (Grafite Furnace Atomic Absorption Spectrometry for Cd, Cu, Pb and Flame Atomic Absorption Spectrometry for Fe). 2.3. Methodology Level 2.0 daily data of AOD at 440 nm and Ångström exponent at 870 nm and 440 nm measured at CIAO from November 2005 to March 2009 were analyzed in this study. The 440–870 nm wavelengths band was selected as it provides information on the relative influence of coarse versus fine mode aerosols (Reid et al., 1999). Climatological and statistical analyses of AOD and α daily mean values were performed to characterize the aerosol columnar properties. In particular, the daily and monthly variability of AOD and α values were studied. The k-means classification technique (MacQueen, 1967) was applied to the dataset for identifying possible clusters in there. Source regions of the observed aerosols were investigated by means of analytical backward air trajectories, based on the trajectory code developed at NASA/Goddard and available at the aeronet.gsfc.nasa.gov website (Schoeberl and Newman, 1995). The 7-day back-trajectories corresponding to four different pressure levels (from 950 to 500 hPa) were used. The pressure levels, corresponding to 0.5 to 5 km a.s.l., were chosen to take into account the air mass path close to the surface as well as at free troposphere altitudes. In this way, both the aerosol within the atmospheric boundary layer and that long range transported into high-altitude atmospheric layers were considered. Due to the columnar nature of the aerosol properties considered in this study, the association of a specific origin to AOD and α values was possible only if the back trajectories at all levels had the same origin and path. This reflects the consideration reported by Meloni et al. (2008) in their study focused on the desert dust outbreaks detected at Lampedusa. In fact, if the back trajectories at all levels have the same origin and path both the “entrainment condition“(the trajectory interacts with the mixed layer) and the “permanence condition” (the trajectory spends a large fraction of time over a determinate region) defined by Meloni et al. (2008) were accomplished. Once the air mass path was identified by back-trajectory analysis, information about the emission at potential source region was needed for the aerosol classification. As far as dust is concerned, the back-trajectory analysis was supported by the outcomes of the Dust REgional Atmospheric Model (DREAM) (Nickovic et al., 2001) provided by the Barcelona Supercomputing Center (http://www.bsc.es/projects/earthscience/DREAM) and those of the Navy Aerosol Analysis and Prediction System (NAAPS) model provided by the Naval Research Laboratory

(http://www.nrlmry.navy.mil/aerosol). In particular, DREAM dust concentration profiles and NAAPS surface dust concentration maps were analyzed and their combination with the backtrajectory analysis allowed the identification of dust cases. The contribution of aerosol particles coming from biomass-burning was also considered. To this aim, Fire Information for Resource Management System (FIRMS) maps, from Web Fire Mapper (http://maps.geog.umd.edu), were used to reveal hotspot/fire near the area under study. Finally, the influence of the different aerosol types on PM measurements at ground level was evaluated. An explorative statistical analysis and the Principal Component Analysis (PCA) were performed on the PM1 data. In particular, PCA was used to reduce data and extract a small number of latent factors (principal components, PCs) to further analyze possible relations between the observed variables (Loska and Wiechuya, 2003). PCA was applied to Pearson's correlation coefficient matrix, each variable was normalized to the unit variance so as to contribute equally, and all the principal factors with eigenvalues N1 were retained (Kaiser, 1960). In order to clarify both the interpretation of the loadings and the meaning of the PCs, the retained factors were subsequently subjected to a VARIMAX normalized rotation. 3. Results 3.1. Climatological and statistical analyses Long-term observations of atmospheric aerosol columnar optical properties were carried out at CIAO from December 2004 to March 2009. During this period, level 2.0 data were retrieved from 536 days of measurements corresponding to about 25,000 observations. Fig. 1 shows the temporal pattern of observed AOD (Fig. 1a) and α (Fig. 1b) daily mean values. Gaps in the data are due to the system calibrations and upgrade. AOD and α daily values range from 0.03 to 0.6 and from 0.15 to 3.14, respectively, during the considered period, according to values reported in a previous study by Santese et al., 2008 in the Mediterranean basin. They report aerosol optical depth at 550 nm ranging from 0.1 to 1.0 and α values in the range from 0.1 to 2.2, respectively, at Lecce AERONET station. Different values were instead measured by Pace et al. (2006) in the period 2001–2003 at the island of Lampedusa. They report values of aerosol optical depth ranging from 0.03 to 1.13 and α values ranging from −0.32 to 2.05. This difference is related to the prevalence of marine aerosol at Lampedusa associated to smaller values of α. As Fig. 1 shows, both AOD and α values are characterized by a large day-to-day variability that is found to be larger than 100% for AOD b0.18 and α b0.95. This variability reflects the natural high variability of aerosol concentration, optical and microphysical properties in the Mediterranean area. High day-to-day variability of the AOD data from MODIS was also reported by Kaskaoutis et al. (2007a) in the Eastern Mediterranean region. Monthly averages of AOD and α reported in Fig. 2a and b, evidence a seasonal behavior: higher AOD values and lower α values are mainly observed during spring–summer months. This behavior suggests an increased contribution of aerosol particles in the coarse mode, such as Saharan dust particles, when meteorological synoptic pattern favors northward motions of air masses from the Sahara desert to the Mediterranean area. The standard deviation calculated for

Author's personal copy A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

101

0.8

a AOD

0.6

0.4

0.2

0.0 4

b

α

3

2

1

0 01/07/2005

01/07/2006

01/07/2007

01/07/2008

01/07/2009

Date (dd/mm/yyyy) Fig. 1. Daily values of AOD at 440 nm (a) and α (440 nm/870 nm) (b) derived from sun-photometer observations carried out at CIAO from November 2005 to March 2009.

each month is reported as error bar in Fig. 2a and b for the AOD and α values, respectively. The high standard deviations indicate a large variability for these parameters within each 0.3

a

AOD

0.2

0.1

0.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2.5

b

2.0

α

1.5 1.0 0.5 0.0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Fig. 2. Monthly values of AOD at 440 nm (a) and α (440 nm/870 nm) (b) derived from sun-photometer observations carried out at CIAO from November 2005 to March 2009. The reported error bars represent the monthly standard deviations.

month, without an evident seasonal dependence of this variability. The aerosol seasonal variability over the Mediterranean region was previously investigated by Barnaba and Gobbi (2004) by means of the analysis of one year record of AOD data from MODIS instrument. Their study highlighted an AOD seasonal cycle with maximum values on summer over the whole Mediterranean. According to these results, the seasonal analysis derived from a long-term monitoring (2000–2005) of AOD data from MODIS over Athens revealed a significant AOD variability, with minimum values in winter, and maximum in summer (Kaskaoutis et al., 2007a). Pace et al. (2006) in their study performed in the island of Lampedusa related the observed increase of the AOD in summer both to desert dust cycle in the central Mediterranean and to the stronger convection and the longer permanence time of aerosol in the atmosphere during this season. Anyway, the summer increase in the AOD observed at Lampedusa is generally linked with an increment in the values of α, corresponding to an increased contribution of fine particles. Count distributions of optical depth and Ångström exponent were also analyzed. The frequency distribution of the aerosol optical thickness is better approximated by a lognormal distribution than by a normal law, as showed in the work of O'Neill et al. (2000). On the contrary, α being derived from the ratios of AOD at different bands, it is expected to be normally distributed if the data are from a single population (Knobelspiesse et al., 2004). For this reason the count distributions of the AOD natural logarithm (lnAOD) and α were studied and a multipeak Gaussian fit to both parameters was applied. As Fig. 3a shows, the count distribution of lnAOD is well fitted (quadratic correlation coefficient r 2 = 0.99) by a four-mode normal distribution centered around −2.80 ± 0.06, −2.26 ± 0.02, −1.59 ± 0.03 and − 0.86 ± 0.04 with a

Author's personal copy 102

A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

standard deviation of 0.42, 0.40, 0.58 and 0.24, respectively. The centers of these normal distributions correspond in the AOD space to the values 0.06, 0.10, 0.20, 0.42. These multiple peaks reveal the presence of different levels of atmospheric turbidity (O'Neill et al., 2000). In particular, the mode centered around 0.06 corresponds to almost clear atmospheric conditions. The mode centered around 0.42 is probably related to cases of high aerosol content in the free troposphere, such as Saharan dust, that occurs 1 day out of 10 in the Central Mediterranean area (Mona et al., 2006), and biomass burning and forest fires typical in Southern Europe during the summer. Finally, the two central and more populated modes, peaked around 0.10 and 0.20, are representative of the most typical aerosol conditions. However, the separation into two modes allows one to suppose that they are related to different atmospheric conditions. A further investigation is necessary for the correct identification of the aerosol types associated to these two modes. The α count distribution (Fig. 3b) is well fitted (quadratic correlation coefficient r2 = 0.99) by a bi-modal normal distribution that is the sum of two normal distributions centered around 1.66 ± 0.04 and 0.69 ± 0.14 and with standard deviations of 0.79 and 0.74, respectively. The first mode includes most of the data and corresponds to aerosol whose size distribution is dominated by fine particles which are generally

60

associated with local urban aerosol. This mode could correspond to background conditions. In fact, the mean value of the backscatter-related Ångstrom exponent at 355/532 nm obtained by systematic lidar measurements performed at CIAO was equal to 1.8 in the local PBL (Mona et al., 2008). The second mode corresponds to aerosol whose size distribution is dominated by coarse mode particles such as sea salt and dust. In particular, this mode could mainly include desert dust particles. In fact, a mean value of the backscatter-related Ångstrom exponent at 355/532 nm of about 0.7 was obtained in case of severe Saharan dust events (Mona et al., 2006). Fig. 4 shows the lnAOD count distributions for observations performed in cold (October–March) and warm (April– September) seasons. As Fig. 4 shows, a higher occurrence of small AOD values was observed during the cold season. The largest AOD values instead were mainly observed during the warm season. In particular, the peak at 0.42 observed in the total count distribution of lnAOD (Fig. 3a) is related mainly to the warm season. This is probably due to an increase of the amount of aerosol particles of natural origin (e.g., dust, fires and biomass burning) during this period. Looking at the α distributions (Fig. 5), a little shift towards larger α values is observed in the cold seasons. In particular, values of α larger than 2 were mainly obtained during these seasons. These values are indicative of days during which aerosol particles belong mainly to the fine mode. This could be related to the low frequency of long-range transport events

a Warm season

50 30

Counts

Counts

40 30 20

20

10

10 0 -4

-3

-2

-1

0

0 -4

lnAOD

-3

-2

-1

0

40

140

b

Cold season

120 30

Counts

Counts

100 80 60 40

20

10

20 0

0 0.0

0.5

1.0

1.5

α

2.0

2.5

3.0

3.5

Fig. 3. Count distributions of lnAOD (a) and α (440 nm/870 nm) (b). The bold solid lines represent the multi mode normal distributions that best fit the data distributions.

-4

-3

-2

-1

0

lnAOD Fig. 4. Count distributions of lnAOD corresponding to daily sun photometer observations performed in cold (from October to March) and warm (from April to September) seasons.

Author's personal copy A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

103

3.5

such as dust intrusions in the cold period of the year. Moreover, the two modes observed in the Ångström distribution (Fig. 3b) are less evident in the cold seasons (Fig. 4b), when they seem to be more mixed than in the warm period (Fig. 4a).

3.0 2.5

3 2.0

α

3.2. Aerosol characterization 3.2.1. k-means cluster analysis in the AOD-α relationship In order to group AOD and α data into representative clusters, the k-means classification technique was used. This technique requires normally distributed input values (Knobelspiesse et al., 2004) and a priori knowledge of the number of clusters. Therefore, lnAOD and α were used as input variables and according to the results of the analysis of the lnAOD count distribution the number of clusters was a-priori fixed to four. Fig. 6 shows the results of the k-means classification, while the range of variability of AOD and α values for each obtained cluster is summarized in Table 1. The k-means cluster analysis provides AOD mean values for the 4 clusters in good agreement with the peaks of the multimodal fitting curve reported in Fig. 3a. The cluster indicated as C2, related to AOD around 0.21 and α mean value of 1.62, is the most populated. This cluster reflects the typical conditions at CIAO with local continental aerosol and the absence of coarse-mode aerosols such as dust and sea salt.

2 1.5

4 1.0

1

0.5 0.0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

AOD Fig. 6. AOD-α scatter-plot reporting, with different colors, the four clusters identified by means of the k-means cluster analysis.

The less populated cluster is instead C1 (AOD and α mean values of 0.28 and 0.55, respectively). The quite high AOD as well as the small Ångström exponent further supports the dust origin of the small peak in Fig. 3a. In between, C3 and C4 clusters are populated by cases with small aerosol optical depth. In particular C3 cluster is associated with almost clear atmospheric conditions with fine mode aerosols, while C4, because of the relatively small α, can be related to maritime and continental aerosol.

60

Warm season 50

Counts

40 30 20 10 0 -0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

80

Cold season 70

Counts

60

3.2.2. Aerosol origin In order to identify the aerosol sources and types, backtrajectories were examined for each day of measurements collected at CIAO. Three main sectors are identified as origin of the air masses reaching CIAO: south, northwest/northeast and west directions. This result was in agreement with those reported for other AERONET sites of the Mediterranean area, indicating the representativeness of this site for the Mediterranean area at least as long as the long-range transported aerosol cases are concerned (Pace et al., 2006; Toledano et al., 2007a; Santese et al., 2008). According to air mass origin and path, three aerosol classes were identified. The first, named “dust”, corresponds to air masses coming from the African continent and refers to Saharan dust transport events. The second, named “continental”, corresponds to air masses coming from northwest/northeast directions and passing through the European continent. The

50 40 30 Table 1 Range of variability of AOD and α parameters in the obtained clusters.

20

Cluster

10 0 -0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

α Fig. 5. Count distributions of Ångström exponent corresponding to daily sunphotometer observations performed in cold (from October to March) and warm (from April to September) seasons.

C1 C2 C3 C4

N

84 191 123 138

α

AOD Δ

δ

m

M

Δ

δ

m

M

0.28 0.21 0.08 0.10

0.11 0.07 0.02 0.02

0.13 0.11 0.03 0.03

0.61 0.52 0.12 0.16

0.55 1.62 2.02 1.22

0.26 0.25 0.29 0.28

0.15 1.05 1.51 0.35

1.17 2.14 3.14 1.66

Legend: N = number of data, Δ = mean value, δ = standard deviation, m = minimum value, M = maximum value.

Author's personal copy 104

A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

third, named “maritime”, corresponds to air masses coming from west direction and crossing the Mediterranean Sea in the last part of their atmospheric path. Finally, a fourth class called “mixed” collects all those cases in which back-trajectories at different levels have different origin and/or path. This takes into account the different aerosol types probably present at different altitude levels within the total atmospheric column investigated by the sun-photometer. The variability of α with respect to the AOD and the statistical parameters for the different identified aerosol types are shown in Fig. 7 and Table 2, respectively. Cases with a dominant aerosol type were identified in the Mediterranean in previous studies by using the MODISderived AOD data and its respective ratio attributed to finemode particles (Barnaba and Gobbi, 2004; Kaskaoutis et al., 2007a; Kosmopoulos et al., 2008) or combining AOD and α data derived from sun-photometer observations (Pace et al., 2006; Kaskaoutis et al., 2007b). The results of these studies highlighted that the different aerosol sources and the mixing processes in the atmosphere influence the aerosol optical properties in this area making difficult to discriminate a single typology in most of the examined cases that are characterized as mixed aerosol type. The classes identified in this study fit well with the previous cataloging. As Fig. 7 shows, a large variability of AOD was observed for dust cases which showed largest AOD and lowest α (mean values of AOD and α are 0.21 and 1.12, respectively). The observed variability could depend on the different aerosol properties at the origin and the residence time of the aerosol particles over the areas crossed along their path (see for example Müller et al., 2009). In fact, the aerosol columnar optical properties can change due to enrichment with continental polluted aerosol or sea-salt aerosol. A further separation within dust cases is shown in Fig. 7, between cases characterized by α values larger than 1.0 and those with a large content of coarse mode particles and α values around 0.5. These latter cases correspond to DREAM forecast dust concentration higher than 50 μg/m3 and NAAPS optical depth at wavelength of 550 nm and dust mass mixing ratio (μg/m3) at surface related to a large dust component, and can be classified as strong dust events (mean observed AOD of about 0.3). AOD

3.5 Mixed Dust Continental Maritime

3.0 2.5

α

2.0 1.5 1.0 0.5 0.0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

AOD Fig. 7. AOD-α scatter-plot reporting, with different colors, the identified aerosol classes.

Table 2 Range of variability of AOD and α parameters in the defined aerosol classes. Class

Dust Continental Maritime Mixed

N

201 220 21 94

α

AOD Δ

δ

m

M

Δ

δ

m

M

0.21 0.13 0.09 0.15

0.11 0.07 0.03 0.08

0.04 0.03 0.03 0.05

0.61 0.45 0.15 0.45

1.12 1.67 1.41 1.60

0.55 0.46 0.36 0.40

0.15 0.39 0.63 0.49

2.21 3.14 2.04 2.45

Legend: N = number of data, Δ = mean value, δ = standard deviation, m = minimum value, M = maximum value.

values related to these more intense dust events resulted in agreement with the mean values obtained for desert dust type by Barnaba and Gobbi (2004) in the Mediterranean sector corresponding to the measurement area (AOD~0.4) and with the value obtained by Kaskaoutis et al. (2007a) and Kosmopoulos et al. (2008) over Athens (AOD ~0.5). Moreover, α values corresponding to strong dust cases over the study area are in agreement with values reported for desert dust at Ispra AERONET station (Kaskaoutis et al., 2007b). For the other dust cases, characterized by α values larger than 1.0, a bimodal nature of the size distribution (Fig. 9a) is retrieved from the sun-photometer indicating the presence of fine as well as coarse mode particles in the atmospheric column. A difference in the characteristic of dust particles was also found by a climatological analysis of Saharan dust intrusions performed at CIAO during multi-year systematic lidar measurements. In that study, the authors reported for dust cases a three-modal distribution of the lidar ratio, an optical parameter depending on the aerosol microphysical properties. The shown distribution takes into account: a few cases of high aerosol load with desert dust/maritime aerosols contamination, a wide mode with mixing between Saharan dust and tropospheric aerosols, and the last mode representative of pure Saharan dust (Mona et al., 2006). In the region of the AOD–α plot characterized by AOD values N0.3, isolated points do not correspond to Saharan dust transport events. Among these cases, those corresponding to large α values could be related to fine particles deriving from biomass-burning. In fact, these biomass-burning-related aerosol particles are typically characterized by high AOD associated with α values above 1.5 (Balis et al., 2003; Eck et al., 2003; Toledano et al., 2007b). The close-by biomass burning origin for these cases was confirmed by FIRMS maps that revealed hotspot/fire in northeast direction from the measurement area. Fig. 8 shows the FIRM hotspot/fire location to respect to CIAO location for a specific event. In the figure, FIRM result was combined with 24 h air mass backtrajectories ending over the measurement area at an altitude between 500 and 3000 m. On the other hand, the isolated cases characterized by α values b1 could be related to low cloud contamination, as revealed by ceilometer measurements performed at CIAO on these days. The continental cases are characterized by AOD values typically less than 0.15 and α values typically larger than 1.5 (pink circles in Fig. 7). In particular, mean values of AOD and α are 0.13 and 1.67, respectively. According to our results, continental cases discriminated by Barnaba and Gobbi (2004) over the measurement area refer on average to AOD of about 0.17. The large α values observed indicate the prevalence of

Author's personal copy A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

105

Fig. 8. FIRM hotspot/fire location to respect to CIAO location for the July 27, 2008 selected case. FIRM result was showed with 24 h air mass back-trajectories ending over the measurement area at 16.00 UT between 500 and 3000 m of altitude.

fine particles. In fact, as Fig. 9b shows, the aerosol size distribution retrieved from the sun-photometer observations on these days typically corresponds to a larger content of fine mode particles (0.1 μm b r b 1 μm). However, also for this class of aerosol there is a large variability both in AOD and α, corresponding to cases with small to relatively high aerosol load (0.03–0.45) and Ångstrom ranging between 0.22 and 2.21. This is related to the many possible scenarios in the aerosol content over continental Europe from small aerosol content of fine particles, to high polluted aerosol load (air masses passing over industrialized sites in Europe). Very few cases correspond to back-trajectories crossing the Mediterranean Sea (about 4%). These cases correspond to AOD mean values of 0.09 and α mean values of 1.41. The size distribution on these days indicates the presence of fine as well as coarse mode particles over the measurement area (Fig. 9c). These results indicated maritime particles mixed with anthropogenic or dust aerosols as expected in the Mediterranean basin, generally not considered representative for clean marine aerosols (Smirnov et al., 2002; Pace et al., 2006), because of the influence of surrounding regions and Mediterranean Sea limited dimension. Finally, about 17.5% of the cases belonged to the mixed class. Even though both the variability range and mean value of AOD measured in these cases are very close to those obtained for the continental cases, α values are on average lower, corresponding to a larger contribution of coarse particles. In fact, as Fig. 9d shows, the aerosol size distribution derived from the sun-photometer on these days corresponds

typically to a mixed content of particles in fine and coarse modes. This is in agreement with the presence in the atmospheric column of aerosols coming from different regions and therefore with different mean aerosol sizes.

3.2.3. Cluster representativeness The cluster analysis allowed a first characterization of aerosol particles over the study area on the base of AOD-α values, whereas the back-trajectories allowed a separation in typical aerosol classes through the path and the origin of the air masses ending over the measured area. To evaluate how much each cluster represents an aerosol typology, a comparison between the two cataloging was performed and the percentage of the different aerosol classes was analyzed for each cluster. As Fig. 10 shows a dominant aerosol typology can be identified in C1 and C3 clusters. In fact, about 84% of the data included in C1 cluster correspond to dust cases, and 70% of these are related to strong Saharan dust episodes. Moreover, around 66% of the data included in this cluster corresponds to measurements performed on warm seasons when the greatest number of Saharan dust outbreaks has been observed over the Mediterranean area (Papayannis et al., 2008). The C3 cluster includes about 65% of the data, corresponding to air masses crossing the European continent. Moreover, around 91% of the data included in this cluster correspond to measurements performed in winter. Therefore, the C3 cluster is characterized by particles mainly in the fine mode, that can be associated both with aerosol advected from urbanized regions

Author's personal copy 106

A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

0.6

0.012 September 6, 2008 05:52:53 06:18:34 07:15:44 07:50:43 07:57:52 13:57:49 14:39:40 16:03:17

0.5 0.4 0.3 0.2 0.1

January 6, 2007

b dV(r)/dln(r) (μm3/μm2)

dV(r)/dln(r) (μm3/μm2)

a

08:04:35 09:04:35 10:04:36 13:04:37 14:04:34

0.010 0.008 0.006 0.004 0.002 0.000

0.0 0.1

1

10

100

0.1

Radius (μm)

c

November 25, 2006 09:45:54 10:45:55 11:45:54 12:45:55 13:45:55 13:59:51

0.005 0.004 0.003 0.002 0.001

0.007

dV(r)/dln(r) (μm3/μm2)

dV(r)/dln(r) (μm3/μm2)

0.006

1

10

100

Radius (μm)

d

November 8, 2008 07:42:51 08:42:51 09:42:51 10:43:06 12:42:49

0.006 0.005 0.004 0.003 0.002 0.001 0.000

0.000 0.1

1

10

100

0.1

Radius (μm)

1

10

Radius (μm)

Fig. 9. Aerosol volume size distribution corresponding to dust (a), continental (b), maritime (c) and mixed (c) cases.

Continental Mixed Maritime Dust

Cluster1 84%

Continental Mixed Maritime Dust

Cluster2 1%

41%

20%

0% 5%

11%

38%

b

a Continental Mixed Maritime Dust

Cluster3 20%

10% 19% 29%

5%

65%

Continental Mixed Maritime Dust

Cluster4

10% 42%

c

d Fig. 10. Percentage of the data corresponding to the different aerosol typologies in C1, C2, C3 and C4 clusters.

100

Author's personal copy A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

located in the northwest/northeast of Europe and with local aerosol that is generally dominant in winter months. By analyzing the percentage of data included in the C2 and C4 clusters, one can observe that these clusters include a similar percentage of data corresponding to continental and mixed cases, but dust cases are more common in the C2 cluster (about 41%), while the C4 cluster contains the largest number of maritime cases (about 10%). 3.2.4. Evaluation of the influence of the aerosol types on PM1 measurements In order to evaluate the potential influence of the different aerosol types on PM measurements at ground level, PM1 and trace element daily concentrations were assigned to the corresponding AOD-α cluster, and both an explorative statistical analysis and the Principal Component Analysis (PCA) were performed on the PM1 data groups obtained. PM1 and trace element concentration mean values calculated for each of the four clusters identified are shown in Table 3. As can be seen from Table 3, the PM1 concentrations measured in the four clusters are comparable. Regarding the trace element concentrations Al and Na show higher mean values in the C1 and C2 clusters. Significantly higher mean values of Fe and Mn concentrations were found in the C1 cluster. In contrast, K and Ti did not seem to show significant differences between the four clusters. Concerning Cd, Cr, Cu Pb and Zn, elements of prevailing anthropogenic origin, no significant change could be found in the four clusters, except for Pb and Zn that showed higher mean values in the C1 cluster. As far as PCA is concerned, the C1 cluster was not taken into account since the related number of trace element concentration data is not sufficient for this kind of analysis. As Table 4 shows, four principal components (PCs) explaining more than 70% of the data variance were pointed out in all the three cases examined. By analyzing the VARIMAX rotated component matrix obtained for the C2 cluster (Table 4a), it can be observed that the first component (PC1), which explains most of the variance (26%), includes Al (0.88), Fe (0.53), K (0.87), Mg (0.65) and Ti (0.92). All these elements mainly originate from sources of crustal type (Na and Cocker, 2009; Saliba et al., Table 3 Mean value ± standard error of PM1 and trace element daily concentrations for the four obtained clusters. N indicates the number of PM1 and trace element data for each the four identified clusters. PM1 and trace element concentrations are expressed in μg m− 3 and ng m− 3, respectively.

Al Cd Cr Cu Fe K Mg Mn Na Pb Ti Zn PM1

C1 cluster

C2 cluster

C3 cluster

C4 cluster

N = 12

N = 67

N = 56

N = 40

103.7 ± 21.3 3.5 ± 0.2 39.8 ± 2.2 6.2 ± 1.3 167.5 ± 34.0 50.0 ± 9.8 23.9 ± 2.9 4.7 ± 0.9 233.1 ± 36.9 26.3 ± 7.5 4.3 ± 0.5 9.8 ± 2.4 8.5 ± 1.8

110.1 ± 6.6 3.5 ± 0.1 38.5 ± 0.7 3.7 ± 0.3 91.9 ± 5.0 49.5 ± 3.4 23.6 ± 1.2 3.3 ± 0.2 242.8 ± 13.1 12.3 ± 1.9 4.9 ± 0.2 6.6 ± 0.5 7.0 ± 0.5

81.5 ± 5.7 4.3 ± 0.1 43.2 ± 0.9 4.1 ± 0.3 97.7 ± 7.1 45.0 ± 9.1 26.7 ± 1.7 2.9 ± 0.2 159.0 ± 12.1 13.4 ± 2.4 4.4 ± 0.3 7.4 ± 0.9 6.1 ± 0.7

79.4 ± 8.2 3.9 ± 0.1 41.6 ± 1.4 4.3 ± 0.4 100.0 ± 7.3 59.1 ± 9.0 27.6 ± 1.8 2.6 ± 0.3 201.7 ± 17.5 9.6 ± 2.4 5.3 ± 0.4 4.5 ± 0.8 9.5 ± 1.3

107

Table 4 VARIMAX rotated component matrix for trace element concentration data included in C2 cluster (a), C3 cluster (b) and C4 cluster (c). Loadings, percentage of explained variance (P%) and cumulative percentage (CP%) are reported. PC1

PC2

PC3

PC4

(a) Al Cd Cr Cu Fe K Mg Mn Pb Ti Zn P% CP%

0.88 0.00 − 0.14 0.04 0.53 0.87 0.65 0.13 − 0.01 0.92 − 0.14 26 26

− 0.02 0.02 0.08 0.70 0.45 0.08 0.13 0.67 0.92 − 0.13 0.80 24 50

0.14 0.06 0.80 − 0.41 0.13 − 0.20 0.40 0.05 0.12 − 0.01 0.31 13 63

− 0.06 0.86 − 0.04 0.06 0.39 0.01 0.32 − 0.50 0.04 0.01 0.08 11 74

(b) Al Cd Cr Cu Fe K Mg Mn Pb Ti Zn P% CP%

0.06 0.15 − 0.05 0.73 0.42 0.12 0.24 0.78 0.87 − 0.15 0.76 25 25

0.13 − 0.13 0.02 0.52 0.08 0.88 0.60 0.16 − 0.10 0.78 − 0.05 19 44

0.83 − 0.17 0.49 0.02 0.49 − 0.06 0.58 0.41 − 0.15 0.31 0.45 18 62

− 0.04 0.86 0.72 0.12 0.09 − 0.08 0.19 0.19 0.01 − 0.15 − 0.04 13 75

(c) Al Cd Cr Cu Fe K Mg Mn Pb Ti Zn P% CP%

0.01 − 0.27 0.09 0.56 0.21 0.10 0.01 0.85 0.84 − 0.11 0.88 24 24

0.24 − 0.18 − 0.10 0.48 0.03 0.93 0.84 0.19 − 0.16 0.69 0.01 22 46

0.86 − 0.29 0.32 − 0.31 0.74 0.02 0.17 0.23 0.23 0.54 − 0.08 18 64

0.02 0.70 0.81 0.40 0.04 − 0.19 − 0.03 0.19 − 0.17 − 0.09 − 0.17 13 77

2010) and in particular are jointly found in the minerals forming the Saharan dust (Formenti et al., 2003; Kandler et al., 2007; Dall'Osto et al., 2010). Since the C2 cluster contains the highest percentage of dust cases among those analyzed this points out a possible influence of Saharan dust on the PM1 measurements. Moreover, these elements are in the component which explains the highest data variance thus highlighting a significant Saharan contribution. The second component (PC2) explains 24% of data variance and is characterized by Cu (0.70), Mn (0.67), Pb (0.92) and Zn (0.80). These chemical elements, when found together in the same component, are indicative of road traffic emissions (Saliba et al., 2010). In fact, Pb and Zn are considered as good tracers of emission from fossil fuel combustion processes, including vehicle exhausts, Cu is also emitted from worn tires and brakes (Saliba et al., 2007), while Mn, an element of prevailing natural origin, could also be found in re-suspended

Author's personal copy 108

A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

road dust (Guo et al., 2009; Hellebust et al., 2010) which includes soil dust mixed with traffic related particles. Therefore, PC2 is mostly related to traffic-derived local emissions rather than long-range transport impact on PM1 measurements. Finally, the third (PC3) and fourth (PC4) components explain 13% and 11% of the data variance, respectively. PC4 consists of Cr (0.80), while PC5 includes Cd (0.86). These elements may be related to local anthropogenic activities such as non-ferrous metal industries, chromium plating or the use of fertilizer (Mohanraj et al., 2004; Chandra Mouli et al., 2006; Gioda et al., 2006). Regarding the C3 cluster (Table 4b), PC1, which explains 25% of the data variance, includes Cu (0.73), Mn (0.78), Pb (0.87), and Zn (0.76). In this case, the highest percentage of variance is explained by the component which includes the chemical elements identified as traffic tracers. As found in the previous cluster, these elements are grouped together supporting their local origin. This is also in agreement with the optical properties of the particles included in the C3 cluster. In fact, this cluster is characterized by particles that can be associated with local aerosol. PC2 and PC3 explain 19% and 18% of the data variance, respectively. Both are characterized by chemical elements of natural origin that could be mainly related to local soil derived emissions. In particular, PC2 is characterized by K (0.88), Mg (0.60) and Ti (0.78) while PC3 is characterized by Al (0.83), Fe (0.49) and Mg (0.58). Finally, PC4 explains 13% of the data variance and consists of Cd (0.86) and Cr (0.72), identifying local anthropogenic activities such as industries. As to the C4 cluster (Table 4c), the results are very similar to those obtained for the C3 cluster supporting the prevailing influence of local sources rather than long-range transport related ones. In fact, also in this case the highest variance (24%) is explained by Cu (0.56), Mn (0.85), Pb (0.84), and Zn (0.88) that is the traffic source. PC2 and PC3 explain 22% and 19% of the data variance, respectively. PC2 is characterized by K (0.93), Mg (0.84) and Ti (0.69). PC3 is characterized by Al (0.86) and Fe (0.74). Both are characterised by chemical elements of natural origin that could be related to local soil. PC4 explains 13% the data variance and consists of Cd (0.70) and Cr (0.81). No relationship with the marine aerosol was found due to the lack of concentration measurements of chemical elements (e.g., Na and Cl) which could reveal its influence on PM1 measurements at ground. In conclusion, given the small size of the particles included in the PM1 fraction, a prevailing influence of local aerosol on PM1 measurements was found. As to long-range transport related aerosol, only the influence of Saharan dust was pointed out probably due to both the closeness of the African Continent to the study area and the amount of aerosol transported during this type of events. In contrast, no influence of continental or maritime aerosols on PM1 measurements was found likely due also to a lack of specific chemical elements (e.g., sulfate and nitrate for continental aerosol and Na and Cl for maritime aerosol) that could reveal their influence on PM1 measurements at ground level. 4. Conclusions Multi-year records of aerosol optical depth (AOD at 440 nm) and Ångstrom coefficient (α at 440/870 nm) data from a sun-photometer of the AERONET global network,

operating at CIAO atmospheric observatory in Southern Italy, was used to characterize the aerosol columnar properties and to quantify the contribution of different aerosol types in the Central Mediterranean. A total amount of 536 cloud screened and quality assured daily data were analyzed. The multi-modal nature of the AOD and α daily mean values count distributions revealed the presence of different aerosol populations. The k-means cluster analysis allowed the separation of the AOD–α space into four classes. The centers of these clusters were in good agreement with the modes identified in the AOD count distribution. The clusters' composition in terms of aerosol types was studied taking advantage of the possibility to associate each measurement to one cluster, and with the support of back-trajectories analysis and aerosol transport models. Three prevailing aerosol types classified as dust (37.5%), continental (41%), and maritime (4%) aerosol were identified examining back-trajectories in terms of air masses path and origin. A fourth class, named mixed aerosols class, included all cases (17.5%) in which back-trajectories related to different altitude levels do not show the same path and origin. It was shown that all measurements with AOD larger than 0.3 were related to dust observations, with few isolated exceptions identified as close-by forest fires. The percentage of occurrences of the 4 aerosol type cases was analyzed for each cluster identified by the k-means analysis, in order to invest its representativeness. Finally, the influence of the different aerosol types on PM measurements at ground level was also evaluated. The k-means cluster analysis performed in the AOD–α space provided a tool for the identification of the aerosol types, that was representative for dust aerosol (at 84% confidence level), for which in addition an impact on PM1 measurements was found. The classification as continental aerosol was similarly possible with a high confidence level (65%) with the k-means cluster analysis. It was also clear from this study that for all cases in which back-trajectories at different levels did not have the same origin and path, namely the mixed class, information about aerosol optical properties vertical profiles are necessary. In order to better address these points, a synergic use of model outputs and in situ, sunphotometric and lidar data available at CIAO will be performed. Moreover, a wider PM chemical characterization will be carried out in order to better identify the influence of the different long-range transported aerosols on PM measurements at ground level. Acknowledgments The authors would like to thank T. L. Kucsera and A. M. Thompson at NASA/Goddard for back-trajectories available at the aeronet.gsfc.nasa.gov website. The authors are grateful to the Navy Research Laboratory-USA for the contribution given by the NAAPS aerosol maps. Data and/or images from the BSCDREAM8b (Dust REgional Atmospheric Model) model were operated by the Barcelona Supercomputing Center (http:// www.bsc.es/projects/earthscience/DREAM/). Finally, MODIS active fire products were supplied by Fire Information for Resource Management System FIRMS (NASA Information for Research Management System) at the University of Maryland. The authors gratefully acknowledge the NOAA Air Resources

Author's personal copy A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and/or READY website (http://www.arl.noaa. gov/ready.html) used in this publication. References Balis, D.S., Amiridis, V., Zerefos, C., Gerasopoulos, E., Andreae, M., Zanis, P., Kazantzidis, A., Kazadzis, S., Papayannis, A., 2003. Raman lidar and sunphotometric measurements of aerosol optical properties over Thessaloniki, Greece during a biomass burning episode. Atmos. Environ. 37, 4529–4538. Barnaba, F., Gobbi, G.P., 2004. Aerosol seasonal variability over the Mediterranean region and relative impact of maritime, continental and Saharan dust particles over the basin from MODIS data in the year 2001. Atmos. Chem. Phys. 4, 2367–2391. Caggiano, R., D'Emilio, M., Macchiato, M., Ragosta, M., 2001. Experimental and statistical investigation on atmospheric heavy metal concentrations in an industrial area of Southern Italy. Nuovo Cimento C 24, 391–406. Chandra Mouli, P., Mohan, S.V., Balaram, V., Praveen Kumar, M., Reddy, S.J., 2006. A study on trace elemental composition of atmospheric aerosols at a semi-arid urban site using ICP-MS technique. Atmos. Environ. 40, 136–146. Dall'Osto, M., Harrison, R.M., Highwood, E.J., O'Dowd, C., Ceburnis, D., Querol, X., Achterberg, E.P., 2010. Variation of the mixing state of Saharan dust particles with atmospheric transport. Atmos. Environ. 44 (26), 3135–3146. de Meij, A., Lelieveld, J., 2011. Evaluating aerosol optical properties observed by ground-based and satellite remote sensing over the Mediterranean and the Middle East in 2006. Atmos. Res. 99, 415–433. Dubovik, O., King, M.D., 2000. A flexible inversion algorithm for the retrieval of aerosol optical properties from sun and sky radiance measurements. J. Geophys. Res. 105 (D16), 20,673–20,696. doi:10.1029/2000JD900282. Dubovik, O., Holben, B.N., Eck, T.F., Smirnov, A., Kaufman, Y.J., King, M.D., Tanré, D., Slutsker, I., 2002. Variability of absorption and optical properties of key aerosol types observed in worldwide locations. J. Atmos. Sci. 59, 590–608. Eck, T.F., Holben, B.N., Reid, J.S., O'Neill, N.T., Schafer, J.S., Dubovik, O., Smirnov, A., Yamasoe, M.A., Artaxo, P., 2003. High aerosol optical depth biomass burning events: a comparison of optical properties for different source regions. Geophys. Res. Lett. 30 (20), 2035. doi:10.1029/ 2003GL017861. Fan, X., Chen, H., Goloub, P., Xia, X., Zhang, W., Chatenet, B., 2006. Analysis of column-integrated aerosol optical thickness in Beijing from Aeronet observations. China Particuology 4, 330–335. Formenti, P., Elbert, W., Maenhaut, W., Haywood, J., Andreae, M.O., 2003. Chemical composition of mineral dust aerosol during the Saharan Dust Experiment (SHADE) airborne campaign in the Cape Verde region, September 2000. J. Geophys. Res. 108 (D18), 8576. doi:10.1029/ 2002JD002648. Gioda, A., Perez, U., Rosa, Z., Velez, B.J., 2006. Concentration of trace elements in airborne PM10 from Jobos Bay National Estuary, Puerto Rico. Water Air Soil Pollut. 174, 141–159. Gkikas, A., Hatzianastassiou, N., Mihalopoulos, N., 2009. Aerosol events in the broader Mediterranean basin based on 7-year (2000–2007) MODIS C005 data. Ann. Geophys. 27, 3509–3522. doi:10.5194/angeo-27-3509-2009. Goloub, P., Li, Z., Dubovik, O., Blarel, L., Podvin, T., Jankowiak, I., Lecoq, R., Deroo, C., Chatenet, B., Morel, J.P., Cuevas, E., Ramos, R., 2007. PHOTONS/ AERONET sunphotometer network overview: description, activities, results. Proc. SPIE 6936, 69360V. doi:10.1117/12.783171. Guo, H., Ding, A.J., So, K.L., Ayoko, G., Li, Y.S., Hung, W.T., 2009. Receptor modeling of source apportionment of Hong Kong aerosols and the implication of urban and regional contribution. Atmos. Environ. 43, 1159–1169. Hatzianastassiou, N., Matsoukas, C., Drakakis, E., Stackhouse Jr., P.W., Koepke, P., Fotiadi, A., Pavlakis, K.G., Vardavas, I., 2007. The direct effect of aerosols on solar radiation based on satellite observations, reanalysis datasets, and spectral aerosol optical properties from Global Aerosol Data Set (GADS). Atmos. Chem. Phys. 7, 2585–2599. doi:10.5194/acp-72585-2007. Hellebust, S., Allanic, A., O'Connor, I.P., Jourdan, C., Healy, D., Sodeau, J.R., 2010. Sources of ambient concentrations and chemical composition of PM2.5–0.1 in Cork Harbour, Ireland. Atmos. Res. 95, 136–149. Holben, B.N., Eck, T.F., Slutsker, I., Tanré, D., Buis, J.P., Setzer, A., Vermote, E., Reagan, J.A., Kaufman, Y.J., Nakajima, T., Lavenu, F., Jankowiak, I., Smirnov, A., 1998. AERONET—a federated instrument network and data archive for aerosol characterization. Remote. Sens. Environ. 66, 1–16.

109

Holben, B.N., Tanré, D., Smirnov, A., Eck, T.F., Slutsker, I., Abuhassan, N., Newcomb, W.W., Schafer, J., Chatenet, B., Lavenue, F., Kaufman, Y.J., Vande Castle, J., Setzer, A., Markham, B., Clark, D., Frouin, R., Halthore, R., Karnieli, A., O'Neill, N.T., Pietras, C., Pinker, R.T., Voss, K., Zibordi, G., 2001. An emerging ground-based aerosol climatology: aerosol optical depth from AERONET. J. Geophys. Res. 106 (D11), 12,067–12,097. doi:10.1029/ 2001JD900014. IPCC, Climate Change, 2007. The physical science basis. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, p. 996. Kaiser, H.F., 1960. The application of electronic computers to factor analysis. Educ. Psychol. Meas. 20, 141–151. Kandler, K., Benker, N., Bundke, U., Cuevas, E., Ebert, M., Knippertz, P., Rodríguez, S., Schütz, L., Weinbruch, S., 2007. Chemical composition and complex refractive index of Saharan Mineral Dust at Izaña, Tenerife (Spain) derived by electron microscopy. Atmos. Environ. 41 (37), 8058–8074. Kaskaoutis, D.G., Kosmopoulos, P., Kambezidis, H.D., Nastos, P.T., 2007a. Aerosol climatology and discrimination of different types over Athens, Greece, based on MODIS data. Atmos. Environ. 41, 7315–7329. Kaskaoutis, D.G., Kambezidis, H.D., Hatzianastassiou, N., Kosmopoulos, P.G., Badarinath, K.V.S., 2007b. Aerosol climatology: On the discrimination of aerosol types over four AERONET sites. Atmos. Chem. Phys. Disc. 7, 6357–6411. Kim, D.-H., Sohn, B.-J., Nakajima, T., Takemura, T., Takemura, T., Choi, B.-C., Yoon, S.-C., 2004. Aerosol optical properties over East Asia determined from groundbased sky radiation measurements. J. Geophys. Res. 109, D02209. doi:10.1029/2003JD003387. Knobelspiesse, K.D., Pietras, C., Fargion, G.S., Wang, M., Frouin, R., Miller, M.A., Subramaniam, A., Balch, W.M., 2004. Maritime aerosol optical thickness measured by handheld sun photometers. Remote. Sens. Environ. 93, 87–106. Kosmopoulos, P.G., Kaskaoutis, D.G., Nastos, P.T., Kambezidis, H.D., 2008. Seasonal variation of columnar aerosol optical properties over Athens, Greece, based on MODIS data. Remote. Sens. Environ. 112, 2354–2366. Lelieveld, J., Berresheim, H., Borrmann, S., Crutzen, P.J., Dentener, F.J., Fischer, H., Feichter, J., Flatau, P.J., Heland, J., Holzinger, R., Korrmann, R., Lawrence, M.G., Levin, Z., Markowicz, K.M., Mihalopoulos, N., Minikin, A., Ramanathan, V., de Reus, M., Roelofs, G.J., Scheeren, H.A., Sciare, J., Schlager, H., Schultz, M., Siegmund, P., Steil, B., Stephanou, E.G., Stier, P., Traub, M., Warneke, C., Williams, J., Ziereis, H., 2002. Global air pollution crossroads over the Mediterranean. Science 298, 794–799. doi:10.1126/ science.1075457. Loska, K., Wiechuya, D., 2003. Application of principle component analysis for the estimation of source of heavy metal contamination in surface sediments from the Rybnik Reservoir. Chemosphere 51, 723–733. MacQueen, J.B., 1967. Some methods for classification and analysis of multivariate observations. Proceedings of the fifth Berkeley Symposium on Mathematical Statistics and Probability, Vol. 1. University of California Press. Madonna, F., Amodeo, A., Boselli, A., Cornacchia, C., D'Amico, G., Giunta, A., Mona, L., Pappalardo, G., Cuomo, V., 2011. CIAO: the CNR-IMAA advanced observatory for atmospheric research. Atmos. Meas. Tech. 4, 1191–1208www.atmos-meas-tech.net/4/1191/2011/. doi:10.5194/amt4-1191-2011. Mallet, M., Gomes, L., Solomon, F., Sellegri, K., Pont, V., Roger, J.C., Missamou, T., Piazzola, J., 2011. Calculation of key optical properties of the main anthropogenic aerosols over the Western French coastal Mediterranean Sea. Atmos. Res. 101, 396–411. Marenco, F., Mazzei, F., Prati, P., Gatti, M., 2007. Aerosol advection and sea salt events in Genoa, Italy, during the second half of 2005. Sci. Total Environ. 377, 396–407. Meloni, D., di Sarra, A., Monteleone, F., Pace, G., Piacentino, S., Sferlazzo, D.M., 2008. Seasonal transport patterns of intense Saharan dust events at the Mediterranean island of Lampedusa. Atmos. Res. 88, 134–148. Mohanraj, R., Azeez, P.A., Priscilla, T., 2004. Heavy metals in airborne particulate matter of urban Coimbatore. Arch. Environ. Contam. Toxicol. 47, 162–167. Mona, L., Amodeo, A., Pandolfi, M., Pappalardo, G., 2006. Saharan dust intrusions in the Mediterranean area: three years of Raman lidar measurements. J. Geophys. Res. 111, D16203. doi:10.1029/2005JD006569. Mona, L., Amodeo, A., Boselli, A., D'Amico, G., Madonna, F., Pappalardo, G., 2008. Systematic multi-wavelength Raman measurements: a reference for aerosol study. Reviewed and Revised Papers Presented at the 24th International Laser Radar Conference, 23–27 June 2008, Boulder, Colorado, USA, vol. II, pp. 893–896. Müller, D., Heinold, B., Tesche, M., Tegen, I., Althausen, D., Amiridis, V., Amodeo, A., Ansmann, A., Arboledas, L., Balis, D., Comeron, A., D'Amico, G., Gerasopoulos, E., Freudenthaler, V., Giannakaki, E., Heese, B., Iarlori,

Author's personal copy 110

A. Boselli et al. / Atmospheric Research 104–105 (2012) 98–110

M., Mamouri, R.E., Mona, L., Papayannis, A., Pappalardo, G., Perrone, R.M., Pisani, G., Rizi, V., Sicard, M., Spinelli, N., Tafuro, A., 2009. EARLINET observations of the 14–22-May long-range dust transport event during SAMUM 2006: validation of results from dust transport modelling. Tellus B Special Issue: Results of the Saharan Mineral Dust Experiment (SAMUM-1), Volume 61, pp. 325–339. Issue 1. Na, K., Cocker III, D.R., 2009. Characterization and source identification of trace elements in PM2.5 from Mira Loma, Southern California. Atmos. Res. 93, 793–800. Nickovic, S., Kallos, G., Papadopoulos, A., Kakaliagou, O., 2001. A model for prediction of desert dust cycle in the atmosphere. J. Geophys. Res. 106 (D16), 18,113–18,129. O'Neill, N.T., Ignatov, A., Holben, B.N., Eck, T.F., 2000. The lognormal distribution as a reference for reporting aerosol optical depth statistics; empirical tests using multi-year, multi-site AERONET sunphotometer data. Geophys. Res. Lett. 27 (20), 3333–3336. doi:10.1029/2000GL011581. Pace, G., Meloni, D., di Sarra, A., 2005. Forest fire aerosol over the Mediterranean basin during summer 2003. J. Geophys. Res. 110, D21202. doi:10.1029/2005JD005986. Pace, G., di Sarra, A., Meloni, D., Piacentino, S., Chamard, P., 2006. Aerosol optical properties at Lampedusa (Central Mediterranean).1. Influence of transport and identification of different aerosol types. Atmos. Chem. Phys. 6, 697–713. Papayannis, A., Amiridis, V., Mona, L., Tsaknakis, G., Balis, D., Bösenberg, J., Chaikovski, A., De Tomasi, F., Grigorov, I., Mattis, I., Mitev, V., Müller, D., Nickovic, S., Pérez, C., Pietruczuk, A., Pisani, G., Ravetta, F., Rizi, V., Sicard, M., Trickl, T., Wiegner, M., Gerding, M., Mamouri, R.E., D'Amico, G., Pappalardo, G., 2008. Systematic lidar observations of Saharan dust over Europe in the frame of EARLINET (2000–2002). J. Geophys. Res. 113, D10204. doi:10.1029/2007JD009028. Pappalardo, G., Papayannis, A., Bösenberg, J., Ansmann, A., Apituley, A., Alados Arboledas, L., Balis, D., Böckmann, C., Chaikovsky, A., Comeron, A., Gustafsson, O., Hansen, G., Mitev, V., Mona, L., Nicolae, D., Perrone, M.R., Pietruczuk, A., Pujadas, M., Putaud, J.-P., Ravetta, F., Rizi, V., Simeonov, V., Spinelli, N., Stoyanov, D., Trickl, T., Wiegner, M., 2009. EARLINET coordinated lidar observations of Saharan dust events on continental scale. 2009 IOP Conf. Ser. Earth Environ. Sci. 7, 012002. Perrone, M.R., Santese, M., Tafuro, A.M., Holben, B., Smirnov, A., 2005. Aerosol load characterization over south-east Italy for one year of AERONET sunphotometer measurements. Atmos. Res. 75, 111–133. Reid, J.S., Eck, T.F., Christopher, S.A., Hobbs, P.V., Holben, B.N., 1999. Use of the Ångström exponent to estimate the variability of optical and physical

properties of aging smoke particles in Brazil. J. Geophys. Res. 104 (D22), 27,473–27,489. doi:10.1029/1999JD900833. Saha, A., Mallet, M., Roger, J.C., Dubuisoon, P., Piazzola, J., Despiau, S., 2008. One year measurements of aerosol optical properties over an urban coastal site: effect on local direct radiative forcing. Atmos. Res. 90, 195–202. Saliba, N.A., Kouyoumdjian, H., Roumié, M., 2007. Effect of local and longrange transport emissions on the elemental composition of PM10–2.5 and PM2.5 in Beirut. Atmos. Environ. 41, 6497–6509. Saliba, N.A., El Jam, F., El Tayar, G., Obeid, W., Roumie, M., 2010. Origin and variability of particulate matter (PM10 and PM2.5) mass concentrations over an eastern Mediterranean city. Atmos. Res. 97, 106–114. Santese, M., De Tomasi, F., Perrone, M.R., 2008. Advection patterns and aerosol optical and microphysical properties by AERONET over southeast Italy in the central Mediterranean. Atmos. Chem. Phys. 8, 1881–1896. Schoeberl, M.R., Newman, P.A., 1995. A multiple-level trajectory analysis of vortex filaments. J. Geophys. Res. 100 (D12), 25,801–25,816. Smirnov, A., Holben, B.N., Kaufman, Y.J., Dubovik, O., Eck, T.F., Slutsker, I., Pietras, C., Halthore, R., 2002. Optical properties of atmospheric aerosol in maritime environments. J. Atmos. Sci. 59, 501–523. Tafuro, A.M., Barnaba, F., De Tomasi, F., Perrone, M.R., Gobbi, G.P., 2006. Saharan dust particle properties over the central Mediterranean. Atmos. Res. 81, 67–93. Takamura, T., Nakajima, T., SKYNET community group, 2004. Overview of SKYNET and its activities. Proceedings of AERONET workshop, El Arenosillo. Opt. Pura Apl. 37, 3303–3308. Toledano, C., Cachorro, V.E., de Frutos, A.M., Sorribas, M., Prats, N., de la Morena, B.A., 2007a. Inventory of African desert dust events over the southwestern Iberian Peninsula in 2000–2005 with an AERONET Cimel Sun photometer. J. Geophys. Res. 112, D21201. doi:10.1029/ 2006JD008307. Toledano, C., Cachorro, V.E., Berjon, A., de Frutos, A.M., Sorribas, M., de la Morena, B.A., Goloub, P., 2007b. Aerosol optical depth and Ångström exponent climatology at El Arenosillo AERONET site (Huelva, Spain). Q. J. R. Meteorol. Soc. 133, 795–807. Toledano, C., Cachorro, V.E., de Frutos, A.M., Torres, B., Berjon, A., Sorribas, M., Stone, R.S., 2009. Airmass classification and analysis of aerosol types at El Arenosillo (Spain). J. Appl. Meteorol. Clim. 48 (5), 962–981. Wehrli, C., 2002. Calibration of filter radiometers for the GAW Aerosol Optical Depth network at Jungfraujoch and Mauna Loa. Proceedings of ARJ Workshop, SANW Congress, Davos, Switzerland, pp. 70–71.