Elsevier

Chemosphere

Volume 80, Issue 7, August 2010, Pages 771-778
Chemosphere

Characterization of PM10 sources in a coastal area near Venice (Italy): An application of factor-cluster analysis

https://doi.org/10.1016/j.chemosphere.2010.05.008Get rights and content

Abstract

In this study a factor-cluster analysis (FCA) applied to chemical composition of atmospheric particulate matter was carried out. Relating specific wind data and back-trajectories to the daily samples grouped using FCA can be useful in atmospheric pollution studies to identify polluting sources and better interpret source apportionment results. The elemental composition and water soluble inorganic ions content of PM10 were determined in a coastal site near Venice during the sea/land breeze season. From the factor analysis four sources were identified: mineral dust, road traffic, fossil fuels and marine aerosol. From a hierarchical cluster analysis, applied on the factor scores, samples with a similar source profile were grouped. Five clusters were identified: four with samples highly characterized by one identified source, one interpreted as general background pollution. Finally, by interpreting cluster results with wind direction data and back-trajectory analysis further detailed information was obtained on potential source locations and possible links between meteorological conditions and PM10 chemical composition variations were detected. The proposed approach can be useful for air quality assessment studies and PM10 reduction strategies.

Introduction

The mass of airborne particulate matter ⩽10 μm (PM10) is one of the most important parameters used to assess the air quality in Europe. The identification of the various sources of particulate matter is one of the main goals of atmospheric research and plays a key role in formulating and applying PM10 abatement strategies. The main emission sources at a given sampling site have been identified using statistical approaches such as the factor analysis (FA). FA extracts from a dataset of variables obtained for numerous observations a fewer number of uncorrelated factors explaining most of the variance of the data matrix. Factors are then qualitatively interpreted as possible sources basing on the presence of one or more known tracers. However, the identification of sources by FA is affected by some critical limitations: (i) the selection of variables to be included in the model can significantly influence the results; (ii) the unavailability of some tracers can yield to subjective source interpretations; (iii) ambiguous tracers can be erroneously interpreted because representative for more than one emission source; (iv) covariant sources might not be efficiently separated, resulting in mixed source profiles; (v) some factors may have no physical meaning (Viana et al., 2008). To solve those problems, a detailed study of the micrometeorology of the area can provide valuable information, accounting for the effect of the horizontal displacement of the air masses on the possible origin of the pollutants.

In this study a factor-cluster analysis (FCA) was performed to select group of samples on the basis of their similar chemical composition and origin. The origin of each group was then matched with wind roses and back-trajectories. This approach can be useful in interpreting the results of a multivariate treatment by: (i) facilitating the interpretation of source apportionment results; (ii) obtaining detailed information on potential source locations; (iii) evidencing possible links between meteorological conditions and particulate matter composition fluctuations. For the FCA, elemental and soluble ion composition data of PM10 were collected in a coastal site during the sea/land breeze season. This period was selected for being more complex in terms of atmospheric circulation and therefore more interesting. Most of the coastal regions are affected by this atmospheric circulation. It plays an important role in local weather and climate dynamics and has significant impacts on the formation and transport of air pollution between sea and mainland, allowing the mixing of different air masses.

The used technique is not widely applied and in particular it was never applied to chemical data of particulate matter.

Section snippets

Study area

This study was carried out in a coastal site not far from Venice, in a relatively remote area in the north-eastern part of Italy, located between the Adriatic Sea and the Po Valley (Fig. 1), which is recognized as the most industrialized district of Italy. Weather conditions frequently involve thermal inversion and prevent pollutant dispersion. The city centre of Venice lies in the middle of a coastal lagoon of ∼550 km2, and air quality is affected by the emissions of artistic glass-making

Sampling

PM10 samples were collected at the lighthouse of Punta Sabbioni (45.4227 N, 12.4368 E), located on a ∼300 m-long dam on the Lido inlet, from a flat roof at 12 m above mean sea level (Fig. 1). The sampling period, ranging between May 19 and September 17, 2007, was selected for being representative of the sea/land breeze season as reported in the literature of historical wind data (Camuffo, 1982).

The sampling site was selected for its closeness to the sea and its relative distance from direct

Results and discussion

Table 1 shows some statistics from the dataset. Minimum and maximum concentrations of PM10 were 4.6 and 41.4 μg m−3 respectively, and the 24-h limit of 50 μg m−3 (99/30/EC Directive) was never exceeded. No significant weekly and monthly trends of PM10 or element and ion concentrations were detected. The highest average elemental concentrations followed the order: S > Na > Ca > Cl > Si > Fe > Al > K > Mg > Zn > P > Ti > Mn > V > Cu > Ni > Cr. High sulfate concentrations, are mainly due to various emission sources, e.g. of marine

Conclusions

A FCA was performed on PM10 chemical composition, to identify and characterize the main sources in a coastal area affected by both natural and anthropogenic emissions. Factor analysis pointed out two natural sources (marine aerosol and mineral dust) and two anthropogenic ones (fossil fuels and road traffic) for PM10. Cluster analysis applied to the factor scores sorted samples by source apportion similarity. Five group of samples were identified: group 1 associated with intense marine aerosol

Acknowledgments

The authors would to thank Ente Zona Industriale di Porto Marghera and Comando Zona Fari e dei Segnalamenti Marittimi Venezia for logistics, Prof. P. Mittner (FISAMB-PD) for PIXE facility, Dr. E. Ghedini and Prof. F. Pinna for ion chromatography. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for proving the HYSPLIT transport and dispersion model and READY website (http://www.arl.noaa.gov/ready.html) used in this publication. Gabriel Walton revised the English text.

References (23)

Cited by (55)

View all citing articles on Scopus
View full text