Particle optical properties at a Central Mediterranean site: Impact of advection routes and local meteorology

Particle optical properties at a Central Mediterranean site: Impact of advection routes and local meteorology

Atmospheric Research 145–146 (2014) 152–167 Contents lists available at ScienceDirect Atmospheric Research journal homepage: www.elsevier.com/locate...

2MB Sizes 0 Downloads 10 Views

Atmospheric Research 145–146 (2014) 152–167

Contents lists available at ScienceDirect

Atmospheric Research journal homepage: www.elsevier.com/locate/atmos

Particle optical properties at a Central Mediterranean site: Impact of advection routes and local meteorology M.R. Perrone a,⁎, S. Romano a, J.A.G. Orza b a b

Mathematical and Physical Department, Universita' del Salento, 73100 Lecce, Italy SCOLAb, Física Aplicada, Universidad Miguel Hernandez, 03202 Elche, Spain

a r t i c l e

i n f o

Article history: Received 21 January 2014 Received in revised form 6 March 2014 Accepted 13 March 2014 Available online 12 April 2014 Keywords: Atmospheric aerosol Light scattering Integrating nephelometer Analytical back trajectories Cluster analysis Meteorological parameters

a b s t r a c t A multiwavelength integrating nephelometer and a PM10 sampler have been used to continuously measure optical properties and mass concentrations of particles at the ground level with the main aim of determining airflow and local meteorology effects on particle optical properties and PM10 mass concentrations. Measurements have been performed at Lecce (Italy), a coastal site in the Central Mediterranean and included scattering (σp) and hemispheric backscattering (βp) coefficients, and hemispheric backscattering fraction (βp/σp) at three wavelengths (450, 525, and 635 nm), in addition to PM10 concentrations. The scattering Ångström exponent (å (λ1, λ2)) for different wavelength pairs (λ1, λ2), and the scattering Ångström exponent difference (Δå = å (450 nm, 525 nm) − å (525 nm, 635 nm)) have been calculated to estimate the airflow impact on the relative weight of fine and coarse mode particles. The yearly mean values ± 1 standard deviation of σp (450 nm), βp (450 nm), βp/σp (450 nm), å (450, 635 nm) and Δå are equal to 100 ± 50 Mm−1, 12 ± 6 Mm−1, 0.13 ± 0.02, 1.1 ± 0.4, and 0.2 ± 0.2, respectively. σp is well correlated to PM10 mass concentrations (r = 0.96) and the PM10 mass scattering cross section is equal to 3.6 ± 0.1 m2 g−1. The back trajectory cluster analysis has identified 5, 5, 7 and 7 distinct airflow types reaching the study site at 271, 500, 1500 and 3000 m above the sea level, respectively, with only slight differences in airflow type among the four arrival heights. We have found that σp and βp values and their respective dependence on wavelength are strongly dependent on airflows. Therefore, we have shown that the in situ particle properties and the local meteorological parameters vary with advection routes. Given the dependence of å, Δå, and βp/σp on particle size and shape, their strong association with airflows has indicated that the mean size distribution of sampled particles varies with air mass history and it has been shown that å, Δå, and βp/σp values allow a satisfactory differentiation of the particle properties associated with different advection routes. More specifically, å and Δå values have allowed distinguishing the airflows mainly responsible for the advection of fine mode particles from the ones which are mainly responsible for the advection of coarse mode particles. These results have provided a satisfactory understanding of the dependence on air flows of the mass scattering cross section values which vary from 2.9 ± 0.2 m2 g−1 to 4.3 ± 0.1 m2 g−1 with air flows. The airflow analysis has also allowed understanding the seasonality of the particle optical properties being linked to the airflow seasonality. © 2014 Elsevier B.V. All rights reserved.

1. Introduction

⁎ Corresponding author. Tel.: +39 3475511893, +39 0832297505. E-mail address: [email protected] (M.R. Perrone).

http://dx.doi.org/10.1016/j.atmosres.2014.03.029 0169-8095/© 2014 Elsevier B.V. All rights reserved.

Aerosol composition and sources in the Mediterranean Basin are quite complex, including not only anthropogenic aerosols from human activities of the industrialised surrounding regions,

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

but also natural aerosol from desert areas (e.g. northern Africa) and the Mediterranean Sea (e.g. Perrone et al., 2013). Several studies indicated that the aerosol radiative forcing is among the highest in the world during Mediterranean summer (e.g. Lelieveld et al., 2002). In situ (e.g. Perrone et al., 2011a; Collaud Coen et al., 2013), ground- and satellite-based remote sensing measurements (e.g. Santese et al., 2008; Perrone et al., 2012; Perrone et al., 2011b) are regularly performed worldwide to gain a deeper understanding of aerosol effects on climate and the environment. Aerosol effects depend on particle size and composition and the knowledge of the aerosol optical properties (e.g. scattering and absorption coefficients) is required to understand their effects on climate. Integrating nephelometers and absorption photometers are currently used to characterise aerosol optical properties at the ground level. Lyamani et al. (2010) reported measurements of aerosol optical properties obtained from December 2005 to November 2007 at Granada, an urban site in south-eastern Spain. They found that both aerosol scattering and absorption coefficients were characterised by a marked seasonal dependence with maxima in winter, and that the fine mode particles were dominant mainly in winter. A clear diurnal pattern of the aerosol scattering and absorption coefficients was also observed in all seasons, which was attributed to the diurnal evolution of the planetary boundary layer (PBL) and to local anthropogenic activities. Esteve et al. (2012a) which have reported aerosol scattering measurements performed at Valencia (Spain) also found that the daily variation of the aerosol scattering properties was due to the traffic and the evolution of the PBL. The influence of the back trajectory pathways on the in situ measurements of aerosol properties was also investigated by Esteve et al. (2012b). They found that the aerosol scattering properties were sensitive to the back trajectory pathways. The highest scattering coefficients were obtained under the influence of air masses from North Africa and the European continent. The lowest scattering coefficients were obtained for arctic-type air masses. Measurements of aerosol scattering and absorption performed at Montseny, a regional background site in the Western Mediterranean Basin, were reported by Pandolfi et al. (2011). They found that on average, mean values of aerosol scattering and absorption coefficients were quite low compared with values reported in literature in more industrialised areas around the Mediterranean Basin. Moreover, a high level of variability was also observed as a function of the origin of the air masses. Surface aerosol scattering properties measured during a period of seven years (2002– 2008) at Évora, Portugal were reported by Pereira et al. (2011). They show in the paper that both seasonal and daily cycles of the scattering and backscattering coefficients were related to local production and transport of particles from elsewhere. Long-term trends of in-situ aerosol optical properties measured within the framework of the WMO-GAW program at Finokalia (Greece) were recently analysed by Collaud Coen et al. (2013). They report that Finokalia was the only European station for which a statistically significant decreasing trend in the aerosol scattering coefficient was found. As mentioned, continuous measurements across the world are required to establish a comprehensive picture of the aerosol properties and their impact on climate. In fact, the European Commission has strengthened the networking of different sites by founding, as an example, the ACTRIS (Aerosol, Cloud and Trace gases

153

Research InfraStructure networks) Project to improve data quality and access (Collaud Coen et al., 2013). Multiwavelength nephelometer measurements performed at a coastal site of south eastern Italy, from December 2011 to November 2012, are analysed in this study in order to contribute to the characterisation of the aerosol optical properties in the Central Mediterranean. More specifically, one year results of scattering (σp) and hemispheric backscattering (βp) coefficients, hemispheric backscattering fractions (βp/σp) at three wavelengths (450, 525, and 635 nm), scattering Ångström exponents (å (λ1, λ2)) for different wavelength pairs (λ1, λ2) and scattering Ångström exponent differences (Δå = å (450 nm, 525 nm) − å (525 nm, 635 nm)) are reported. It will be shown that å and Δå values allow to estimate the airflow impacts on the relative weight of fine and coarse mode particles (Schuster et al., 2006 and references therein). Daily PM10 measurements performed by the Regional Air Quality Agency (http://www.arpa.puglia.it/web/guest/qariainq) at a site that is ~0.5 km away from this study have been used to investigate the relationship of aerosol scattering properties with simultaneous PM10 mass concentrations. Back trajectories combined with statistical analyses have then been used to examine the long-range transport impact on σp, βp, (βp/σp), å, Δå, and PM10 mass concentrations, as well as on the local meteorology. The impact of the main advection routes on the particle optical properties at the ground level represents one of the main topics of this study. Atmospheric particles are quite affected by long-range-transport contributions at the study area and all over the Central Mediterranean, as several previous studies have already revealed (e.g. Santese et al., 2008; Pavese et al., 2009; Santese et al., 2010; Perrone et al., 2013). Sampling site, instruments and methodology are presented in Section 2. Results and discussion are presented in Sections 3. Summary and conclusion are reported in Section 4. 2. Sampling site, instruments and methods 2.1. Site description Nephelometer measurements were performed from December 2011 to November 2012 on the roof of the Mathematical and Physics Department of the University of Salento, at ~10 m above the ground level (a.g.l.). PM10 mass measurements were performed 0.5 km away from it, at ~2 m from the ground. The Mathematical and Physics Department is in a flat peninsular site (40.33°N; 18.11°E), 6 km away from the town of Lecce (~95,000 inhabitants), and ~20 km away from both the Ionic and Adriatic Seas (Fig. 1). A coal power plant and a large industrial area are about 35 and 100 km away from it, respectively. The monitoring site of this study can be categorised as rural background according to Larssen et al. (1999). Therefore, it may be considered as representative of coastal sites of the Central Mediterranean away from large sources of local pollution (Perrone et al., 2013; Basart et al., 2009). The Balkan and northern Africa coasts are ~100 and 800 km away from it, respectively. 2.2. Instruments A LED-based integrating nephelometer (model Aurora 3000, ECOTECH, Australia) was used to measure particle scattering

154

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

Fig. 1. Geographical location of the monitoring site (full dot).

and hemispheric backscattering coefficients at 450, 525 and 635 nm with a temporal resolution of 5 min. The hemispheric backscattering fraction is given by the hemispheric backscattering to the scattering coefficient ratio: βp/σp. Air sampling was obtained from the top of a stainless steel tube, 15 mm internal diameter and about 1.5 m length. The inlet was fitted with a funnel covered by an insect screen to prevent rain drops and insect from getting into the sample line. Hence, no aerosol size cut-off was applied to the sampled air. To prevent the hygroscopic effects from enhancing the particle scattering properties, a relative humidity threshold of 60% was set by using a processor-controlled automatic heater inside the nephelometer. A full calibration of the device was carried out every two months using CO2 as high span gas and internally filtered free air (particle with a diameter b0.1 μm) as low span gas. Precision checks were executed weekly to ensure the data quality. The nephelometer includes backscatter measurements that allow 9–170° standard integrating measurements and hemispheric measurements. A detailed description of the instrument is given by Müller et al. (2011). Correction factors as a function of the Ångström exponent were also provided by Müller et al. (2011) to correct systematic uncertainties due to angular truncation and non-Lambertian illumination. A beta gauge particulate monitor (SWAM 5A Monitor, FAI Instrument, Italy; http://www.fai-instruments.com/index.php/ it/main-page-ita/prodotti/swam) with PM10 inlet was used by the Regional Air Quality Agency (http://www.arpa.puglia.it/ web/guest/qariainq) to continuously measure the PM10 mass concentration at a flow-rate of 2.5 m3 h−1.

2.3. Back trajectory clustering The HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model version 4.8, from NOAA/ARL (Draxler and Hess, 1998), was used to compute daily 96-hour backward trajectories starting at four heights (271, 500, 1500 and 3000 m

above sea level (asl)), six times a day (00, 04, 08, 12, 16, and 20 UTC) from December 2011 to November 2012. The lowest height was selected as 10 m above the ground level of the model. Meteorological fields from the NCEP/NCAR Global Reanalysis dataset were the input data for the model runs, with 2.5° latitude–longitude resolution, 17 pressure levels from 1000 to 10 hPa, and 6 hourly data. The vertical movement of the air parcels was calculated using the vertical velocity field of the meteorological data. Trajectories were classified into homogeneous groups by a robust cluster procedure based on the k-means algorithm, with hourly longitude and latitude as input variables (Moody and Galloway, 1988) and metrics based on great-circle distances. A detailed description of the clustering procedure is given in Perrone et al. (2013).

3. Results and discussion 3.1. Main features of the particle scattering properties Time series of daily average PM10 mass concentrations, scattering and hemispheric backscattering coefficients, and hemispheric backscattering fractions at 450 nm, 525 nm, and 635 nm are plotted in Fig. 2. Plot gaps in Fig. 2 represent the missing data due to instrument failure. Means, standard deviations (SDs), median, minimum, maximum, skewness, and percentiles of PM10, σp, βp, and βp/σp, respectively are given in Table 1. PM10, σp, βp, and βp/σp are characterised by a large day-to-day variability as a consequence of the high variability of the particle types and concentrations due to local and long-range-transport contributions, as it will be shown in the following. Daily PM10 mass concentrations and scattering coefficients at 450 nm vary between 5 and 97 μg m−3 and 17 and 344 Mm−1, respectively. The PM10 yearly mean ± SD that is equal to 30 ± 10 μg m−3 is lower than the yearly limit value for PM10 (40 μg m−3) established by the European Union (http://ec.europa.eu/environment/

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

155

Fig. 2. Time series of daily average (a) PM10 mass concentrations, (b) scattering and (c) hemispheric backscattering coefficients, and (d) hemispheric backscattering fractions at 450 nm, 525 nm, and 635 nm, respectively.

air/quality/standards.htm). The σp yearly mean that is equal to 100 ± 50 Mm−1 at 450 nm decreases by 34% at 635 nm. The sensitivity of βp to wavelength is smaller than that of σp (Table 1). In fact, the βp yearly mean ± SD is 12 ± 6, 11 ± 5, and 10 ± 4 Mm−1 at 450, 525, and 635 nm, respectively. Hemispheric backscattering fraction (βp/σp) values increase with the wavelength as a consequence of the large σp decrease with λ. The βp/σp yearly mean is equal to 0.13 ± 0.09 and 0.16 ± 0.08 at 450 and 635 nm, respectively. βp/σp can be used to estimate the asymmetry parameter g of air particles, even if there is not in general a one-to-one relationship

between the two parameters (Andrews et al., 2006). The g parameter is one of the main parameters required in radiative transfer simulations since it provides a measure of the angular distribution of the scattering radiation (e.g. Tafuro et al., 2007). g ranges from − 1 to 1. In particular, g is equal to zero (corresponding to βp/σp = 0.5) for symmetric (e.g. Rayleigh) scattering light and to 1 (corresponding to βp/σp = 0) for entirely forward scattering light (e.g. Bergamo et al., 2008). Wiscombe and Grams (1976) found a relationship between βp/σp and g from Mie calculations as shown in Fig. 3 of their paper. The Arnott's empirical formula based on the plot

156

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

Table 1 Statistics of daily-averaged aerosol parameters retrieved from measurements performed from December 2011 to November 2012. PM10 mass concentrations, scattering (σp), hemispheric backscattering (βp) and hemispheric backscattering fractions (βp/σp) at 450 nm, 525 nm and 635 nm, respectively are reported, in addition to the Ångström exponent calculated for different wavelength pairs (å (λ1, λ2)) and the Ångström exponent difference Δå = å (450 nm, 525 nm) − å (525 nm, 635 nm). λ (nm)

Parameters

PM10 (μg m σp (Mm−1)

−3

)

βp (Mm−1)

βp/σp

å

450 525 635 450 525 635 450 525 635 450–525 450–635 525–635

Δå

Mean

30 100 80 60 12 11 10 0.13 0.14 0.16 1.3 1.1 1.0 0.2

SD

Median

10 50 40 30 6 5 4 0.02 0.02 0.03 0.3 0.4 0.5 0.2

Min

26 84 68 55 11 10 9 0.13 0.14 0.16 1.34 1.20 1.09 0.20

5 17 16 15 2 2 2 0.09 0.09 0.08 0.02 −0.18 −0.46 −0.27

97 344 268 199 39 34 29 0.23 0.27 0.33 1.90 1.93 1.98 0.85

Skewness

1.66 1.41 1.37 1.27 1.41 1.37 1.30 0.93 0.97 0.77 −0.97 −0.75 −0.57 0.48

Percentiles 10

25

75

17 39 33 29 6 5 5 0.11 0.12 0.12 0.77 0.61 0.41 0.01

19 54 45 37 8 7 7 0.12 0.13 0.13 1.06 0.90 0.74 0.08

33 119 95 76 15 13 12 0.14 0.16 0.18 1.51 1.40 1.35 0.37

(2011) also found that σp and βp were characterised by a positively skewed distribution. Scattering, hemispheric backscattering, and hemispheric backscattering fractions of this study are compared in Table 2 with the corresponding parameters obtained at different sites of southern Europe to highlight the dependence of aerosol scattering properties on the location in the Mediterranean Basin. σp and βp mean values are in agreement within ±1 SD with all sites of Table 2. σp and βp mean values typically 1 order of magnitude smaller than the ones of Table 2 were measured in clean-air Northern European sites (Aaltonen et al., 2006). It is interesting to note from Table 2 that the smallest σp and βp mean values were found at Montseny (Spain) since it represents the typical regional background conditions of the Western Mediterranean Basin. On the contrary, rather high σp and βp mean values were found at the monitoring site of this study (Table 2). σp mean values rather close or even larger than the ones of the study area have been found at Valencia (Spain) that is on a region subject to the influence of long-range transported particles, and at Sde Boker, a remote site in the Negev Desert (Israel) and at Erdemli (Turkey). βp/σp means at 525–550 nm were equal to 0.14 at the western and central European sites of Table 2. Smaller mean values were found at the eastern European sites of Mt. Athos (Greece) and Sde Boker (Israel). This last result is likely due to a significant

relating βp/σp and g in the solar spectral range is (Andrews et al., 2006):  3 g ¼ −7:143889 βp =σ p  2   þ 7:464439 βp =σ p −3:96356 βp =σ p þ 0:9893

Max

ð1Þ

Eq.(1) indicates that g decreases regularly with βp/σp and we found that its yearly mean value is equal to 0.59 ± 0.05 at 450 nm. It is interesting to observe that the g(450 nm) yearly mean is in agreement within ±1 SD with the columnar mean value of the asymmetry parameter gcol (440 nm) = 0.67 ± 0.03 retrieved from sun-sky photometer measurements (Santese et al., 2008; Bergamo et al., 2008) performed within the AErosol RObotic NETwork (AERONET, Holben et al., 1998) and co-located in space. This last result is likely due to the size distribution similarity of ground and aloft particles (Andreae et al., 2002). PM10, σp and βp mean values are larger than the corresponding median values (Table 1) since they are all characterised by a positively skewed distribution. The contrary applies in case the skewness is negative. The skewness is a measure of the lack of symmetry of a data set: the higher is the skewness, the higher is the probability of measuring values higher than the mean for the considered data set. Pandolfi et al.

Table 2 Scattering (σp), hemispheric backscattering (βp), hemispheric backscattering fractions (βp/σp) and Ångström exponents (å) for different wavelength pairs at southern European sites. Location, site type, observation (Obs.) time and mean values ± 1 standard deviation are provided. Location

Site type

Obs. time

λ (nm)

σp (Mm−1)

βp (Mm−1)

βp/σp

Lecce, Italy Évora, Portugal Granada, Spain Montseny, Spain

2011–2012 2002–2008 2005–2007 2009–2010

525 550 550 525

80 40 60 30

11 ± 5 6±6 9±4 5±3

0.14 0.14 0.14 0.14

Valencia, Spain M. Athos, Greece Finokalia, Greece Erdemli, Turkey

Rural Urban Urban Regional background Urban Rural Remote coastal Remote coastal

2006–2010 1999–2002 2001–2002 1999–2000

550 550 532 532

8±5 6±4 – –

– 0.10 ± 0.02 – –

1.6 ± 0.3 1.7 ± 0.4 – –

450, 700 450, 700 – –

Esteve et al. (2012a) Gerasopoulos et al. (2003) Vrekoussis et al. (2005) Vrekoussis et al. (2005)

Sde Boker, Israel

Desert

1996–1997

550

80 ± 50 65 ± 40 50 ± 20 90 ± 160 (45 ± 20)a 90 ± 50



0.12 ± 0.02

1.4 ± 0.4

450, 700

Andreae et al. (2002)

a

Mean value when April's 2000 dust event was excluded.

± ± ± ±

40 50 30 30

± ± ± ±

0.02 0.03 0.02 0.02

å (λ1, λ2)

λ1, λ2 (nm)

Reference

1.1 1.4 1.5 1.3

450, 450, 450, 450,

This study Pereira et al. (2011) Lyamani et al. (2010) Pandolfi et al. (2011)

± ± ± ±

0.4 0.5 0.3 0.5

635 700 700 635

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

contribution of crustal non-spherical particles, since Mt. Athos is in a rural area and Sde Boker is in the northern part of the Negev Desert. 3.2. Main features of the scattering Ångström exponent and its spectral curvature The Ångström exponent Å describes the spectral dependence of the extinction by particles and is commonly used as a qualitative indicator of the dominant particle size (Schuster et al., 2006 and references therein). The scattering related Ångström exponent (å) which describes the wavelength dependence of the scattering coefficient is used in this study as an indicator of the dominant particle size. å is calculated from the following relation åðλ1 ; λ2 Þ ¼ − ln

h  i σ p ðλ1 Þ= σ p ðλ2 Þ =½ ln ðλ1 =λ2 Þ

ð2Þ

where σp (λ1) and σp (λ2) represent the scattering coefficient at the wavelengths λ1 and λ2, respectively, and we can assume that å owns the same properties of Å (Andreae et al., 2002). In fact, Schuster et al. (2006) have found that the Å sensitivity to the bulk absorption coefficient is small. Hence, particles larger than few microns in diameter as dust and sea-salt aerosols are characterised by å values close to zero, while particles of less than 1 μm diameter as anthropogenic particles have å values larger than 1 (Seinfeld and Pandis, 1998). å-Value decreases (increases) are indicative of a shift to coarse (fine) particles by convention. However, å alone does not provide unambiguous information on the relative weight of coarse and fine mode particles if different aerosol types or mixtures of fine and coarse mode particles are present in the sampled air. Large fine mode particles can have the same å as mixtures of coarse mode and small fine mode particles as Schuster et al. (2006) have clearly

157

shown in Fig. 3 of their paper. Schuster et al. (2006) pointed out that it is important to consider the wavelength pair used to calculate the Ångström exponent when making qualitative assessments about the corresponding aerosol size distribution. They found that Å values calculated from longer wavelength pairs (e.g. λ = 670, 870 μm) were sensitive to the fine mode volume fraction of aerosols but not the fine mode effective radius. Conversely, shorter wavelength pairs (e.g. λ = 380, 440 μm) were sensitive to the fine mode effective radius but not the fine mode volume fraction. Then, some authors (e.g. Kaufman, 1993; O'Neill et al., 2003; Schuster et al., 2006) have demonstrated that the spectral variation of Å can provide further information about the aerosol size distribution. Kaufman (1993) pointed out that negative values of the difference Δå = å (440 nm, 613 nm) − å (440 nm, 1003 nm) indicate a dominance of a single small particle mode, while positive differences indicate the effect of two separate particle modes. In fact, a positive curvature requires the presence of enough coarse mode particles to reduce the wavelength dependence of extinctions at longer wavelengths (Schuster et al., 2006). In this study we have defined the scattering Ångström exponent difference Δå ¼ åð450 nm; 525 nmÞ–åð525 nm; 635 nmÞ

ð3Þ

as a measure of the scattering Ångström exponent curvature to obtain some hints on the relative weight of coarse and fine mode particles from å-Δå values. More specifically, we have assumed that the values of å and Δå can allow estimating the relative weight of fine and coarse modes as it has been suggested by Schuster et al. (2006) and Gobbi et al. (2007) using the å and Δå values. The time series of å (for different wavelength pairs) and Δå daily means are plotted in Fig. 3a and b, respectively. å (for different wavelength pairs) and Δå daily means are characterised by a large day-to-day and seasonal variability as a consequence of the high variability of the size

2.0

a

1.5

å

1.0 0.5 (450,635 nm) (525,635 nm) (450,525 nm)

0.0 -0.5

Δå

0.8

0.4

0.0

b -0.4 01/01/12

01/03/12

01/05/12

01/07/12

01/09/12

01/11/12

Date (dd/mm/yy) Fig. 3. Time series of the daily mean values of the (a) scattering Ångström exponents calculated for different wavelength pairs (å (λ1, λ2)) and the (b) scattering Ångström exponent difference Δå = å (450 nm, 525 nm) − å (525 nm, 635 nm).

158

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

distribution of the sampled particles. Fig. 3a shows that å (450 nm, 525 nm) daily means are on average, and mainly in summer, greater than the å (525 nm, 635 nm) daily means. As a consequence Δå daily means are on average positive (Fig. 3b) and greater Δå values have on average been reached during summer. These results combined with the discussion on the Ångström exponent and its curvature reported by Schuster et al. (2006) indicate that the particle size distribution at the ground level of the study area is on average characterised by two separate modes and reveals the significant contribution of coarse mode particles in summer. As mentioned, a significant contribution of coarse mode particle is required to reduce the wavelength dependence of scattering at longer wavelength and obtain positive Δå values, in accordance with Eq. (3). Main statistics of å and Δå values are given in Table 1. å mean values are smaller than the corresponding median values for all wavelength pairs since they are characterised by a negatively skewed distribution. å (450 nm, 635 nm) values range between −0.18 and 1.93 with mean value and standard deviation of 1.1 ± 0.4. Δå values ranged between −0.27 and 0.85 with mean value and standard deviation of 0.2 ± 0.2. Scattering Ångström exponents of this study are compared in Table 2 with the corresponding parameters calculated at different sites of southern Europe: å mean values larger than 1 were found at all sites. The study area (Lecce) is the site with the lowest å mean value. The greater contribution of coarse mode particles at Lecce than at the other sites of Table 2 is likely responsible for this last result as the positive value of the Δå yearly mean also indicates. 3.3. Relationships between optical parameters Fig. 4 shows the scatterplot of σp (450 nm) daily means versus daily means of PM10 mass concentrations. Linear correlation coefficient (r) and slope of the fitting regression line which are provided in the figure show that σp values are well correlated (r = 0.96) to PM10 mass concentrations. The slope of the fitting line represents the PM10 mass scattering cross section σPM10 = 3.6 ± 0.1 m2 g−1 at 450 nm. The high daily variability of the optical and microphysical properties of the atmospheric particles is responsible for the spread of data around the regression line, as it will be shown in the following. Both, scattering coefficients and particular matter (PM) mass concentrations depend on number concentration

and size of air particles, and as a consequence σp could also be used to define PM mass concentrations. However, one must be aware that the scattering efficiency is also dependent on the shape, size, and absorption of the sampled particles, and on average it decreases with particle size as the particle size becomes much larger than the wavelength of the incident light. On the contrary, PM mass concentrations on average increase with particle size. Very good correlations (R ≥ 0.92) between σp and PM1 mass concentrations were also reported by Pandolfi et al. (2011). More specifically, they found that the PM1 mass scattering cross section was equal to 4.5 ± 0.8 m2 g−1 at 450 nm. Andreae et al. (2002) found at a remote site in the Negev Desert (Israel) that the mass scattering cross section at 450 nm was 7.6 ± 0.3 and 0.4 ± 0.4 m2 g−1 for particles with the aerodynamic diameter b2 μm and within 2–10 μm, respectively. The decrease of σp with particle size was responsible for the smaller massscattering-cross-section values of coarse mode particles (Andreae et al., 2002). Hence, both the PM fraction and the particle type, which depend on monitoring site, affect the mass scattering cross section values, as the results reported in Section 3.6 of this paper will clearly reveal. We have found that σp and βp are strongly correlated. The hemispheric backscattering fraction is also dependent on σp and we have found that βp/σp decreases gradually with the increase of σp. This result which is in agreement with the one reported by Pandolfi et al. (2011) is due to the increasing importance of the scattering compared to the backscattering when σp increases. Fig. 5 (open dots) shows the scatterplot of σp (450 nm) versus å (450 nm, 635 nm). Full dots and error bars in Fig. 5 represent σp mean values and the 95% confidence interval when considering equal number of data points in each bin. σp increases gradually with å (450 nm, 635 nm) and hence, with the contribution of fine mode particles. Note that σp varies with particle size and concentration while, å is only dependent on particle size. Then, the spread of data around the σp mean values which increases with å (Fig. 5) is likely due to the concentration variability of particles having equal å value. As mentioned, different mixtures of fine and coarse mode particles can have the same å but likely different σp values. Hence, this last effect may also have contributed to the data spread of Fig. 5. βp/σp is dependent on the size and shape of the scattering particles as mentioned. Nevertheless, it is worth noting that there is no correlation between βp/σp and å. Similar

Fig. 4. Scatterplot of σp (450 nm) daily-means versus daily-means PM10 mass concentrations. Linear correlation coefficient (r) and slope (±1 SD) of the fitting regression line (dotted line) are also provided in the figure.

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

159

Fig. 5. Scatterplot of σp (450 nm) daily-means versus å (450 nm, 635 nm) daily-means (open dots). Full dots and error bars represent σp mean values and the corresponding 95% confidence intervals when equal number of data points in each bin is considered.

results have been reported from Andreae et al. (2002). The lack of the expected negative correlation between βp/σp and å is likely due to the different sensitivity of the two parameters when sampled particles are characterised by bimodal size distributions. Different mixtures of fine and coarse mode particles can have the same å, as mentioned but, this last comment likely does not apply to βp/σp. Work is on progress to analytically investigate the sensitivity of βp/σp and å on bimodal size distributions. The presence of non-spherical particles may also contribute to the lack of correlation between βp/σp and å. Fig. 5 shows that 69% of the å (450 nm, 635 nm) daily means are larger than 1 for the significant contribution of fine mode particles throughout the year (Fig. 3a), if we assume that the particle size distribution is bimodal. Fig. 6 (open dots) shows the scatter plot of å (450 nm, 635 nm) versus Δå. Full dots and error bars represent the Δå mean values and the 95% confidence interval when considering equal number of data points in each bin. å decreases gradually with the increase of Δå. For bimodal size distributions, this result indicates that the coarse mode contribution increases as å decreases. As mentioned, positive Δå values require the presence of enough coarse mode particles to decrease the wavelength dependence of scattering at longer wavelengths (Eq. (3)). More specifically, the particle size distribution is expected to be made by two separate particle modes with a significant coarse mode contribution in the days with å b 1 and Δå N 0, according to Schuster et al., 2006 and Gobbi et al., 2007. Fine mode particles are likely dominant in the days with å N 1.5

whatever the value of Δå even if the contribution of coarse mode particle increases with the shift of Δå towards greater values. 3.4. Main transport patterns The cluster analysis of back trajectories identified 5, 5, 7 and 7 distinct airflow types reaching the study site at 271, 500, 1500 and 3000 m above the sea level (asl), respectively. Each group of trajectories is represented by its centroid in Fig. 7a–d, which also provides the percent frequency of occurrence and the mean length of the trajectories within each cluster. The airflow seasonality is shown in Fig. 7e–h. Airflow types were labelled according to their overall direction and length. Short trajectories were indicative of slow-moving air masses while extremely long trajectories corresponded to fast flows, a difference with direct implications for air quality. There were only slight differences in airflow type among the four arrival heights: west (W), south (S), north-east (NE), north-north-west (NNW) and slow north-west (slowNW) flows were found to be common to all heights. The advection of fast polar air masses from North America (fastNW) and those coming from north-west Africa (SW) were only identified at 1500 and 3000 m. The air flows reaching the study site at higher levels travelled longer distances than those arriving at lower levels as wind speed increases with increasing height above the ground due to the reduced influence of the friction with the surface. Also the

Fig. 6. Scatter plot of å (450 nm, 635 nm) daily-means versus Δå daily-means (open dots). Full dots and error bars represent å mean values and the corresponding 95% confidence intervals when equal number of data points in each bin is considered.

160

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

Fig. 7. (a–d) Cluster centroids retrieved from 4-day analytical back trajectories. 5, 5, 7, and 7 distinct centroids starting at 271, 500, 1500, and 3000 m were identified by the cluster analysis. The percent of occurrence of trajectories within each cluster and the centroid length are also provided in each plot. (e–h) Airflow seasonality. Airflow types are labelled in accordance with their overall direction and length.

frequency of occurrence of each airflow type varied with height. slowNW flows, which reached south-eastern Italy after passing on average along the Adriatic Sea, are the most frequent flows with over 30% of the trajectories arriving at the lower heights (Fig. 7a–b) and 21–25% at the higher ones (Fig. 7c–d). They are

dominant in summertime (Fig. 7e–h), when the Azores or the north-African anticyclone extended to the west and north of the area in blocking situations that lead to weak westerlies. The next more common paths were NE flows, which passed over Ukraine and the Balkans, and were found in 25% and about 18% of the

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

trajectories arriving at low (271 and 500 m asl) and high (1500 and 3000 m asl) levels, respectively. NE advections were found in association with a depression (often just a relative low) located to the east somewhere between the Black Sea and the Gulf of Sidra in the Mediterranean. Surrounding high pressure systems with some influence were either the Siberian high extended to Finland and drawing air to eastern Europe (winter), an anticyclone located over Romania (spring) or a long and tilted ridge extended northeast from northern Africa to eastern Europe (late spring to autumn). S flows crossed the Mediterranean from Libya/Tunisia, in about half the cases as dust-loaded air-masses according to the columnar loading product of the BSC-DREAM model (www.bsc.es). They were primarily associated with low pressure over the Western/Central Mediterranean extended towards northern Africa and/or to the presence of a high pressure area between Libya and Turkey. Such low pressure situations corresponded frequently to depressions travelling eastward (e.g., north-African depressions, Genoa lows) or lows stalled in southern Italy. SW flows reached the study site at 3000 and 1500 m in April, summer and autumn, after passing over Algeria/Tunisia. In all the cases, the BSC-DREAM model forecasted the presence of dust plumes over southeastern Italy. Note that dust particles from Sahara are on average lifted from the ground up to few kilometres in height before being transported to Europe (Perrone and Bergamo, 2011). W flows at the lowest heights were found in April and late autumn (Fig. 7e–h), mainly associated either with travelling frontal systems that entered into the Western Mediterranean across the Gulf of Lions or to advection in the southern flank of a deep Genoa low, with high pressures over northern Africa. In summertime, they were found only at 3000 and 1500 m. W flows eventually swept African dust already suspended in the Western Mediterranean to the study area. fastNW flows reaching the study site at 3000 and 1500 m were associated with the polar-front jet stream located at upper levels, in situations with a ridge on the eastern Atlantic and a long-wave trough over Europe. They were mainly found between November and April with W flows at low heights (as mistral winds surging across the coast of southern France into the study area). NNW flows corresponded to continental polar and arctic air that reached the study site from the eastern coast of the Adriatic Sea in January–March. These air masses were previously advected from the north, usually by the combination of a high pressure area to the west of the study site (in situations as diverse as the presence of a strong anticyclone northwest of the Alps and the north-African high extended to south-western Italy) and a depression located to the east (somewhere between Turkey and western Russia). In a few cases, the passage of cold fronts over the area was found as well. At low levels NNW flows were associated with bora winds over the study area. 3.5. Local meteorological parameters and advection patterns Measurements from a local meteorological station are used in this study to feature temperature (T), relative humidity (RH), wind speed (V), and rainfall (Rn) at the surface of the study area and to show the dependence of T, RH, V, and Rn on airflows. Time series of daily average temperature, relative humidity, wind speed, and rain are plotted in Fig. 8. Mean values ± 1 SD, median, minimum, and maximum values,

161

skewness, and percentiles of T, RH, V, and Rn are given in Table 3. Temperature and relative humidity daily-means vary within 2–31 °C and 36–97%, respectively. Both parameters are characterised by a marked seasonality: T N 20 °C and RH b 60% are on average found in summertime. The T and RH yearlymeans are 17 ± 7 °C and 70 ± 10%, respectively. The yearlymean values of V and Rn are 3 ± 2 ms−1 and 2 ± 8 mm, respectively. Fig. 9 shows by box plots the dependence of T, RH, V, and Rn on advection routes for different arrival heights. Horizontal bold lines and dots represent in each box the median and the mean values, respectively. The lower and upper boundaries of the box indicate the 25th and the 75th percentiles, respectively, while the error bars represent the 5th and the 95th percentiles. The percentage of rainy days for each airflow type is also provided in the rainfall plots (Fig. 9d). Each monitoring day was assigned to a trajectory group if at least four of the six day-trajectories belonged to that group in order to consider days with a well defined advection pattern. Statistically significant differences among the values of each parameter grouped by airflow type were examined with pair wise comparisons using Mann–Whitney tests, with confidence levels adjusted for multiple comparisons by the Dunn–Sidàk correction (Brankov et al., 1998). Results are given at 95% overall confidence level. Fig. 9a reveals that the smallest mean temperature value is associated with NNW airflows and that the highest mean temperature is associated with both S and slowNW airflows. The relative humidity is on average larger when S and NE air flows reach the study site (Fig. 9b). The highest mean wind speed is associated with NNW airflows at all starting heights. 46% of the days associated with S and NE airflows (271 m arrival height) have been affected by rainfall. Note that the high mean rainfall values have been due to some relatively intense rainfall episodes (Fig. 8d). 3.6. Airflow and local meteorology effects on PM10, σp, βp, βp/σp, å, Δå, and σp/PM10 The box plots depicted in Fig. 10 show the dependence of PM10 mass concentrations, σp, βp, βp/σp, å, Δå, and σp/PM10 on airflows for different arrival heights. As mentioned, each monitoring day was assigned to a trajectory group if at least four of the six day-trajectories belonged to that group. Changes on aerosol sources and atmospheric physical and chemical processes occurring during the transport were likely responsible for the data spread represented in Fig. 10 by the error bars. Fig. 10 reveals that the investigated particle parameters are quite dependent on advection routes and the dependence is similar at all arrival heights. This last result may indicate that each identified air flow has on average been responsible for the advection of atmospheric particles of similar optical and microphysical properties at all the arrival heights. Note that 3-wavelength lidar measurements have revealed that fine particles mainly due to anthropogenic pollution and coarse particles of natural or anthropogenic origin could be found at any altitude sounded by the lidar from 500 up to about 6000 m asl (Perrone et al., 2014), in satisfactory accordance with the above comment. PM10, σp, and βp vary with the size and concentration of the sampled particles. Conversely å, Δå, βp/σp and σp/PM10

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

Temperature [°C]

162

a

30

20

10

0

Relative Humidity [%]

100

80

60

40

b

Wind Speed [m s-1]

10

c

8 6 4 2 0 100

d Rainfall [mm]

80 60 40 20 0 01/01/12

01/03/12

01/05/12

01/07/12

01/09/12

01/11/12

Date (dd/mm/yy) Fig. 8. Times series of the daily mean values of (a) temperature, (b) relative humidity, (c) wind speed, and (d) rainfall at the ground level from a local meteorological station.

depend only on the size and shape of the sampled particles. Then, their strong dependence on airflows (Fig. 10d–g) indicates that the size distribution of the sampled particles was

quite dependent on air flows: particles of different types and/or from different sources have likely been transported by different advection routes.

Table 3 Statistics of daily-averaged meteorological parameters (temperature T, relative humidity RH, wind speed v and rainfall Rn) retrieved from measurements performed from December 2011 to November 2012. Parameters

Monitoring days

Mean

SD

Median

Min

Max

Skewness

T (°C) RH (%) V (m/s) Rn (mm)

366 366 357 357

17 70 3 2.0

7 10 2 8

17 71 2.5 0.1

2 36 0.0 0.0

31 97 8.8 96.7

0.03 −0.19 1.05 7.26

Percentiles 10

25

75

7 52 0.9 0.1

11 60 1.6 0.1

24 80 3.6 0.2

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

163

Fig. 9. (a) Temperature, (b) relative humidity, (c) wind speed, and (d) rainfall box plots versus airflows arriving at 271, 500, 1500, and 3000 m asl, respectively. Horizontal bold lines and dots represent in each box the median and the mean values, respectively. The lower and upper boundaries of the box indicate the 25th and the 75th percentiles, respectively. Error bars represent the 5th and the 95th percentiles, respectively. sNW and fNW indicate slowNW and fastNW, respectively.

The 271 m arrival-height results (Fig. 10a–f) show that the most polluted air conditions are on average associated with NE airflows which passed over Ukraine and the Balkans before reaching south eastern Italy. The greater PM10, σp (450 nm), βp (450 nm), å (450, 635 nm), and σp (450 nm)/PM10 mean values and the smallest Δå and βp/σp mean values are associated with NE airflows. The small wind speed and the large RH associated with NE airflows (Fig. 9b–c) have likely favoured the occurrence of pollution conditions. It is well known that the wind speed governs the concentration of pollutants and the inverse relationship between wind speed and pollutants concentrations up to some threshold has been found in many studies. We have found that σp (450 nm) increases gradually with relative humidity at RH N 60%. Chemical reactions leading to the formation of new atmospheric particles that are favoured by large RH values have likely contributed to this last result (Seinfeld and Pandis, 1998). The rather large å (450, 635 nm) mean value and the rather small mean value of Δå = 0.17 (Fig. 10e–f) indicate that fine mode particles were dominant

during NE airflows, in accordance with the discussion reported in Section 3.2. The small βp/σp mean value and the rather large σp (450 nm)/PM10 mean value support the last comment. Note that Perrone et al. (2013) found that fine particles as those due to sulphate and traffic sources (Fig. 6 of their paper) are mainly transported to the study area from NE airflows. Lelieveld et al. (2002) also revealed the important contribution of fine pollution aerosols transported from Central and Eastern Europe to south eastern Italy. For the transport routes reaching the site at 271 m, statistically significant differences were found between the high PM10, σp, and βp mean values associated with NE airflows and the rather small PM10, σp, and βp mean values associated with W airflows. These results indicate that the air conditions associated with W airflows are cleaner than the ones associated with NE airflows. The small and large RH and wind speed, respectively associated to W flows have likely contributed to this last result. W flows at the lowest heights were found mainly associated with travelling frontal systems

164

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

that entered into the Western Mediterranean across the Gulf of Lions. So, they have eventually been responsible for the transport to the study area of sea salt particles and African dust already suspended in the Western Mediterranean. Note that the small mean value of å (450, 635 nm) = 0.8 and the positive Δå mean value (=0.26) indicate that W airflows have been responsible for the advection of particles characterised by size distributions with a significant contribution of coarse mode particles. Both the hemispheric backscattering fraction and the mass scattering cross section mean values support the last comment. Relatively small PM10, σp, and βp mean values are also associated with NNW airflows. The low relative humidity and the rather high mean wind speed associated with NNW airflows have likely contributed to this result. NNW flows correspond to continental polar and arctic air that reached the study site from the eastern coast of the Adriatic Sea after crossing industrialised regions of north Western Europe. So they have eventually swept anthropogenic pollution to the study area (Perrone et al., 2013). The NNW å (450, 635 nm) mean value that is equal to 1.2 (Fig. 10e) supports the last comment. It is significantly larger than the one associated to W airflows (=0.8). Hence, NNW airflows have been responsible for the advection of particles characterised by a size distribution with a larger contribution of fine mode particles than the ones associated to W airflows being the Δå mean value associated with both air flows rather similar. (Schuster et al., 2006; Gobbi et al., 2007). Polluted air conditions are associated with slowNW airflows. In fact, PM10, σp (450 nm), βp (450 nm), and å (450, 635 nm) mean values are slightly smaller than the corresponding ones associated with NE advections mainly at 271 m asl. slowNW flows are the most frequent (30%) and are dominant in summertime when the lack of rainy days, occurring over the Mediterranean Basin, favours the ageing of polluted air masses, enhances natural and anthropogenic dust resuspension, and limits the removal of atmospheric particles by wet deposition. In addition, slowNW flows reach south eastern Italy after passing on average along the Adriatic Sea, and as a consequence they have likely been responsible for the advection of sea salt particles and polluted particles emitted from the ships crossing the Adriatic Sea (Perrone et al., 2013). The å and Δå mean values associated with slowNW flows (Fig. 10e–f) indicate that the contribution of coarse mode particles associated with slowNW flows is expected to be on average greater than the one associated with NE flows. The comparison of mass scattering cross section values (Fig. 10g) supports the last comment, as well as the slowNW-flow pathway and seasonality. Note that the mean PM10 mass concentration associated with S flows is nearly equal to the one associated with NE advections (Fig. 10a). On the contrary, σp (Fig. 10b), βp (Fig. 10c), å (Fig. 10e), and σp/PM10 (Fig. 10g) mean values are smaller than the corresponding values associated with NE flows. In particular, the å mean value determined by NE airflows is equal to 1.4, while the one associated with the S advection is equal to 0.9. S flows cross the Mediterranean from Libya/Tunisia, in about half the cases as dust-loaded air-masses. Hence, the advection of coarse marine and/or desert dust particle is responsible for the

165

small σp, βp, å, and σp/PM10 mean values. Δå mean values support the last comment: the largest and the smallest Δå mean values are associated with S and NE airflows, respectively. The sensitivity of all parameters to the airflows decreases with increasing starting-height (Fig. 10). This result is due to the fact that air masses at high altitudes have on average a lower impact at the surface. In addition, high-altitude airflows are on average characterised by longer path lengths (Fig. 7a–d) and as a consequence, they are likely more affected by the atmospheric physical and chemical processes occurring during the transport and hence, they are less dependent on source properties. Small å mean values and, relatively large PM10 mass concentrations, σp, βp, and Δå mean values are associated with the SW airflow identified at 1500 and 3000 m arrival heights, respectively. The transport to the study area of desert dust particles by SW airflows (Fig. 7c–d) has been responsible for these results. In fact, the BSC-DREAM model forecasted the presence of dust plumes over southeastern Italy in all the cases, as mentioned in Section 3.4. 4. Summary and conclusion Multiwavelength nephelometer measurements and PM10 mass concentrations have been analysed to characterise the particle optical properties at a coastal site of south eastern Italy, in the Central Mediterranean. The impact of the main advection routes on the scattering properties of ground level particles represents the main topic of the study. The cluster analysis of four-day HYSPLIT back trajectories has identified 5, 5, 7 and 7 distinct advection patterns reaching the study site at 271, 500, 1500 and 3000 m asl, respectively. West (W), south (S), north-east (NE), north-north-west (NNW) and slow north-west (slowNW) flows have been found to be common to all heights. slowNW flows which reached south-eastern Italy after passing on average along the Adriatic Sea are the most frequent with over 30% of the trajectories arriving at the lower heights and 21–25% at the higher ones. The next more common are NE flows, which passed over Ukraine and the Balkans, and are found in 25% and about 18% of the trajectories arriving at low and high levels, respectively. We have found that σp and βp and their respective dependence on wavelength are strongly dependent on airflows. As a consequence, βp/σp, å, Δå, and σp/PM10 which depend on particle size and shape also are strongly affected by airflows. The scattering Ångström exponent (å) for different wavelength pairs and its spectral difference Δå have been calculated to obtain some hints on the relative weight of coarse and fine mode particles associated with different airflows. Their dependence on airflows has indicated that the particle size distribution and hence, the sampled particle type vary with advection routes. Then, we have found that greater σp/PM10 mean values are associated with air flows responsible for the advection of fine mode particles. • We have found that the most polluted air conditions are associated with NE airflows. The large å (450, 635 nm)

Fig. 10. (a) PM10 mass concentrations, (b) σp, (c) βp, (d) βp/σp, (e) å, (f) Δå, and (g) σp/PM10 box plots versus airflows arriving at 271, 500, 1500, and 3000 m asl, respectively. Horizontal bold lines and dots represent in each box the median and the mean values, respectively. The lower and upper boundaries of the box indicate the 25th and the 75th percentiles, respectively. Error bars represent the 5th and the 95th percentiles, respectively. sNW and fNW indicate slowNW and fastNW, respectively.

166









M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167

mean value (1.4 ± 0.3) and the small Δå mean value (0.2 ± 0.2) associated with NE airflows arriving at 271 m asl are due to the prevailing contribution of fine mode particles as those due to anthropogenic pollution. The small and high mean values of βp/σp (450 nm) and σp (450 nm)/ PM10, respectively support the last comment. The large relative humidity and the small wind speeds associated with NE flows have likely favoured the occurrence of pollution conditions. Polluted air condition is also associated with slowNW airflows. The large PM10, σp, and βp (450 nm) mean values associated with slowNW airflows are due to the fact that slowNW airflows are more frequent in summer when the lack of rainy days occurring all over the Mediterranean favours the ageing of polluted air masses, enhances natural and anthropogenic dust resuspension, and limits the removal of atmospheric particles by wet deposition. The βp/σp and Δå mean values which are on average larger than the ones determined by NE airflows indicate that slowNW advections have likely been responsible for the advection of coarse sea-salt particles from the Adriatic Sea. Clean air conditions are associated with W and NNW airflows: the smallest PM10, σp and βp mean values are associated with both advections. The high wind-speed associated with NNW advection routes has contributed to this result. The mean values of å and Δå associated with NNW and W airflows arriving at 271 m asl, respectively have indicated that the contribution of coarse mode particles determined by W advections is larger than that determined by NNW air flows. We have obtained that the mean PM10 mass concentration associated with S airflows arriving at 271 m asl is nearly equal to the one associated with NE advections. On the contrary, σp and βp mean values are smaller than the corresponding values associated with NE flows. The transport of coarse marine and/or desert dust particle by S advections is responsible for this last result, as å, Δå, and σp/PM10 mean values clearly indicate. Relatively large PM10 mass concentrations, σp, βp, and Δå values and small å values are associated with the SW airflows identified at 1500 and 3000 m arrival heights, respectively, for the transport over southeastern Italy of desert dust.

In conclusion, the airflow effects on the scattering properties of ground level particles at a Central Mediterranean site have for the first time been reported and analysed. The strong dependence of the particle scattering properties on airflows has furthermore demonstrated that the atmospheric particles over south eastern Italy are significantly affected by long-range transported particles. We believe that the results of this study can be considered representative of costal sites of the Central Mediterranean away from large sources of local pollution. Therefore, the paper's results are of general interest since they have contributed to the characterisation of the optical properties of the Central Mediterranean aerosol. It has also been demonstrated that å alone does not provide unambiguous information on the relative weight of coarse and fine mode particles if different aerosol types or mixtures of fine and coarse mode particles are present in the sampled air. Δå which depends on the spectral variation of the scattering-related Ångström exponent can provide additional information about

the weight of fine and coarse mode particles. In fact, å and Δå values have allowed to obtain a better differentiation of the aerosol properties associated with different airflows. Note that the scattering Ångström exponent difference has for the first time been used in a paper to better understand the scattering properties of ground collected particles. Work is on progress to analytically investigate the sensitivity of å, Δå, and βp/σp on bimodal size distributions characterised by different mixtures of fine and coarse mode particles. Acknowledgements This study was supported by the European Community through the ACTRIS Research Infrastructure Action under the 7th Framework Programme under ACTRIS grant agreement no. 262254. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT model used in this publication and the Regional Air Quality Agency (ARPA Puglia) for the provision of the PM10 mass concentrations. References Aaltonen, V., Lihavainen, H., Kerminen, V.-M., Komppula, M., Hatakka, J., Eneroth, K., Kulmala, M., Visanen, Y., 2006. Measurements of optical properties of atmospheric aerosols, in Northern Finland. Atmos. Chem. Phys. 6, 1155–1164. Andreae, T.W., Andreae, M.O., Ichoku, C., Maenhaut, W., Cafmeyer, J., Karnieli, A., Orlovsky, L., 2002. Light scattering by dust and anthropogenic aerosol at a remote site in the Negev Desert, Israel. J. Geophys. Res. 107 (D2), 4008. http://dx.doi.org/10.1029/2001JD900252. Andrews, E., Sheridan, P.J., Fiebig, M., McComiskey, A., Ogren, J.A., Arnott, P., Covert, D., Elleman, R., Gasparini, R., Collins, D., Jonsson, H., Schmid, B., Wang, J., 2006. Comparison of methods for deriving aerosol asymmetry parameter. J. Geophys. Res. 111, D05S04. http://dx.doi.org/10.1029/ 2004JD005734. Basart, S., Perez, C., Cuevas, E., Baldasano, J.M., Gobbi, G.P., 2009. Aerosol characterization in Northern Africa, Northeastern Atlantic, Mediterranean Basin and Middle East from direct-sun AERONET observations. Atmos. Chem. Phys. 9, 8265–8282. Bergamo, A., Tafuro, A.M., Kinne, S., De Tomasi, F., Perrone, M.R., 2008. Monthly-averaged anthropogenic aerosol direct radiative forcing over the Mediterranean from AERONET derived aerosol properties. Atmos. Chem. Phys. 8, 6995–7014. Brankov, E., Rao, S.T., Porter, P.S., 1998. A trajectory-clustering correlation methodology for examining the long-range transport of air pollutants. Atmos. Environ. 32, 1525–1534. Collaud Coen, M., Andrews, E., Asmi, A., Baltensperger, U., Bukowiecki, N., Day, D., Fiebig, M., Fjaeraa, A.-M., Flentje, H., Jefferson, A., Jenning, S.G., Kouvarakis, G., Lihavainen, H., Myhre, C.L., Malm, W.C., Mihapopoulos, N., Molenar, J.V., O'Dowd, C., Ogren, J., Schichtel, B.A., Sheridan, P., Virkkula, A., Weingartner, E., Weller, R., Laj, P., 2013. Aerosol decadal trends — part 1: in-situ optical measurements at GAW and IMPROVE stations. Atmos. Chem. Phys. 13, 869–894. Draxler, R.R., Hess, G.D., 1998. An overview of the HYSPLIT_4 modeling system of trajectories, dispersion, and deposition. Aust. Meteorol. Mag. 47, 295–308. Esteve, A.R., Estelles, V., Utrillas, M.P., Martınez-Lozano, J.A., 2012a. In-situ integrating nephelometer measurements of the scattering properties of atmospheric aerosols at an urban coastal site in western Mediterranean. Atmos. Environ. 47, 43–50. http://dx.doi.org/10.1016/j.atmosenv.2011. 11.043. Esteve, A.R., Estelles, V., Segura, S., Utrillas, M.P., Martınez-Lozano, J.A., 2012b. Influence of the trajectory o fair masses on the in situ measurements of aerosol properties in Valencia. Tethys 9, 35–41. Gerasopoulos, E., Andreae, M.O., Zerefos, C.S., Andreae, T.W., Balis, D., Formenti, P., Merlet, P., Amiridis, V., Papastefanou, C., 2003. Climatological aspects of aerosol optical properties in Northern Greece. Atmos. Chem. Phys. 3, 2025–2041. http://dx.doi.org/10.5194/acp-3-2025-2003. Gobbi, G.P., Kaufman, Y.J., Koren, I., Eck, T.F., 2007. Classification of aerosol properties derived from AERONET direct sun data. Atmos. Chem. Phys. 7, 453–458.

M.R. Perrone et al. / Atmospheric Research 145–146 (2014) 152–167 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., Smirnof, A., 1998. AERONET — a federated instrument network and data archive for aerosol characterization. Remote Sens. Environ. 66 (1), 1–16. Kaufman, Y.J., 1993. Aerosol optical thickness and atmospheric path radiance. J. Geophys. Res. 98 (D2), 2677–2692. Larssen, S., Sluyter, R., Helmis, C., 1999. Criteria for EUROAIRNET, the EEA Air Quality Monitoring and Information, Network. http://reports.eea.eu.int/ TEC12/en. Lelieveld, J., Berresheim, H., Bormann, S., Crutzen, P.J., Dentener, F.J., coauthors, 2002. Global air pollution crossroads over the Mediterranean. Science 298, 794–799. Lyamani, H., Olmo, F.J., Alados-Arboledas, L., 2010. Physical and optical properties of aerosols over an urban location in Spain: seasonal and diurnal variability. Atmos. Chem. Phys. 10, 239–254. Moody, J.L., Galloway, J.N., 1988. Quantifying the relationship between atmospheric transport and the chemical composition of precipitation on Bermuda. Tellus 40B, 463–479. Müller, T., Henzing, J.S., de Leeuw, G., Wiedensohler, A., 2011. Design and performance of a three-wavelength LED-based total scatter and backscatter integrating nephelometer. Atmos. Meas. Tech. 4, 1291–1303. O'Neill, N.T., Eck, T.F., Smirnov, A., Holben, B.N., Thulasiraman, S., 2003. Spectral discrimination of coarse and fine mode optical depth. J. Geophys. Res. 108, 4559. http://dx.doi.org/10.1029/2002JD002975. Pandolfi, M., Cusack, M., Alastuey, A., Querol, X., 2011. Variability of aerosol optical properties in the Western Mediterranean Basin. Atmos. Chem. Phys. 11, 8189–8203. Pavese, G., De Tomasi, F., Calvello, M., Esposito, F., Perrone, M.R., 2009. Detection of Sahara dust intrusions during mixed advection patterns over south-east Italy: a case study. Atmos. Res. 92, 489–504. Pereira, S.N., Wagner, F., Silva, A.M., 2011. Seven year of measurements of aerosol scattering properties, near the surface, in the southwestern Iberia Peninsula. Atmos. Chem. Phys. 11, 17–29. Perrone, M.R., Bergamo, A., 2011. Direct radiative forcing during Sahara dust intrusions at a site in the Central Mediterranean: anthropogenic particle contribution. Atmos. Res. 101, 783–798.

167

Perrone, M.R., Piazzalunga, A., Prato, M., Carofalo, I., 2011a. Composition of fine and coarse particles in a coastal site of the central Mediterranean: Carbonaceous specie contributions. Atmos. Environ. 45, 7470–7477. Perrone, M.R., De Tomasi, F., Burlizzi, P., 2011b. Aerosol products by CALIOP at 532 nm and by a ground-based Raman lidar at 355 nm: intercomparison methodology. Atmos. Res. 101, 438–449. Perrone, M.R., Tafuro, A.M., Kinne, S., 2012. Dust layer effects on the atmospheric radiative budget and heating rate profiles. Atmos. Environ. 59, 344–354. Perrone, M.R., Becagli, S., Orza, J.A.G., Vecchi, R., Dinoi, A., Udisti, R., Cabello, M., 2013. The impact of long-range-transport on PM1 and PM2.5 at a Central Mediterranean site. Atmos. Environ. 71, 176–186. Perrone, M.R., De Tomasi, F., Gobbi, G.P., 2014. Vertically resolved aerosol properties by multi wavelength lidar measurements. Atmos. Chem. Phys. 14, 1185–1204. http://dx.doi.org/10.5194/acp-14-1185-2014. 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. Santese, M., Perrone, M.R., Zakey, A.S., De Tomasi, F., Giorgi, F., 2010. Modeling of Sahara dust outbreaks over the Mediterranean by RegCM3: case studies. Atmos. Chem. Phys. 10, 1–24. Schuster, G.L., Dubovick, O., Holben, B.N., 2006. Ångström exponent and bimodal aerosol size distributions. J. Geophys. Res. 111, D07207. http:// dx.doi.org/10.1029/2005JD006328. Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Physics: From Air Pollution Climate Change. J. Wiley &Sons, INC. Tafuro, A.M., Kinne, S., De Tomasi, F., Perrone, M.R., 2007. Annual cycle of aerosol direct radiative effect over southeast Italy and sensitivity studies. J. Geophys. Res. 112, D20202. http://dx.doi.org/10.1029/2006JD008265. Vrekoussis, M., Liakakou, E., Koc¸ak, M., Kubilay, N., Oikonomou, K., Sciare, J., Mihalopoulos, N., 2005. Seasonal variability of optical properties of aerosols in the eastern Mediterranean. Atmos. Environ. 39, 7083–7094. Wiscombe, W.J., Grams, G.W., 1976. The backscattered fraction in twostream approximations. J. Atmos. Sci. 33, 2440–2451.