Source apportionment of PM10, organic carbon and elemental carbon at Swiss sites: An intercomparison of different approaches

Abstract

In this study, the results of source apportionment of particulate matter (PM10), organic carbon (OC), and elemental carbon (EC) — as obtained through different approaches at different types of sites (urban background, urban roadside, and two rural sites in Switzerland) — are compared. The methods included in this intercomparison are positive matrix factorisation modelling (PMF, applied to chemical composition data including trace elements, inorganic ions, OC, and EC), molecular marker chemical mass balance modelling (MM-CMB), and the aethalometer model (AeM).

At all sites, the agreement of the obtained source contributions was reasonable for OC, EC, and PM10. Based on an annual average, and at most of the considered sites, secondary organic carbon (SOC) is the component with the largest contribution to total OC; the most important primary source of OC is wood combustion, followed by road traffic. Secondary aerosols predominate in PM10. All considered techniques identified road traffic as the dominant source of EC, while wood combustion emissions are of minor importance for this constituent.

The intercomparison of different source apportionment approaches is helpful to identify the strengths and the weaknesses of the different methods. Application of PMF has limitations when source emissions have a strong temporal correlation, or when meteorology has a strong impact on PM variability. In these cases, the use of PMF can result in mixed source profiles and consequently in the under- or overestimation of the real-world sources. The application of CMB models can be hampered by the unavailability of source profiles and the non-representativeness of the available profiles for local source emissions.

This study also underlines that chemical transformations of molecular markers in the atmosphere can lead to the underestimation of contributions from primary sources, in particular during the summer period or when emission sources are far away from the receptor sites.

Introduction

Atmospheric particular matter (PM) has significant negative effects on the environment and on human health. It was recognised that atmospheric aerosols play an important role in the radiative balance of the Earth (IPCC, 2007), reduce visibility (Horvath, 1993), and cause acidification and eutrophication of ecosystems (UNECE, 2004). Moreover, elevated PM levels have been associated with an increase in respiratory and cardiovascular diseases (Nel, 2005) and allergies (Monn, 2001). In order to reduce the concentration of atmospheric PM, a detailed and quantitative knowledge of the sources of PM is required. For these reasons, source apportionment of atmospheric PM, and important PM constituents such as organic carbon (OC) and elemental carbon (EC), is an important research field in the atmospheric sciences.

Source apportionment by receptor modelling is a well established technique that is being increasingly applied in aerosol science (Viana et al., 2008a). The results of source apportionment studies are used for the development of effective and efficient air quality management plans (Hopke, 2008). Among the wide range of receptor models based on different statistical approaches, chemical mass balance (CMB) (Friedlander, 1973) and positive matrix factorisation (PMF) (Paatero and Tapper, 1994) belong to the most commonly applied techniques (Viana et al., 2008b). The choice between CMB and PMF depends primarily on the available information about the main sources of PM. While CMB models require detailed a priori knowledge of the emission sources and of their emission profiles, PMF only requires qualitative or semi-quantitative a posteriori information about the source emission profiles. PMF and CMB models have been applied in numerous source apportionment studies based on different types of tracers. CMB models based on organic (e.g. Schauer et al. (1996) and El Haddad et al. (2011)) or elemental tracers (e.g. Watson et al. (1994) and Pandolfi et al. (2008)) have been successfully applied in the past. Similarly, PMF has been successfully applied using both elemental (e.g. Lee et al. (1999)) and organic tracers (Jaeckels et al., 2007, Zhang et al., 2009), with further developments to highly time-resolved aerosol mass spectra in order to identify sources and components of organic aerosols (Lanz et al., 2007).

In addition to these well-established receptor models, another source apportionment method for carbonaceous matter (CM) has been recently developed (Sandradewi et al., 2008). This model, referred as the aethalometer model (AeM), takes advantage of the different optical properties of CM from wood and fossil fuel combustions for the estimation of the respective contributions of these two specific sources to CM. The AeM was applied successfully in various short-term source apportionment studies (Favez et al., 2009, Favez et al., 2010, Sandradewi et al., 2008), and has also been adapted for long-term source apportionment studies for the apportionment of black carbon (BC) (Herich et al., 2011).

In the past, several intercomparison studies have been performed to evaluate the performances of different source apportionment methods. These studies were primarily focused on the modelling of the same dataset by different receptor models (Bullock et al., 2008, Hopke et al., 2006, Lee et al., 2008, Pandolfi et al., 2008, Rizzo and Scheff, 2007, Viana et al., 2008b), while only few intercomparison studies were based on the analysis of complementary datasets (e.g. parallel measurements of trace elements and organic molecular markers) (see e.g. Ke et al. (2008) and Shrivastava et al. (2007)). Recently, Favez et al. (2010) performed an intercomparison of different methods for the determination of the contribution of wood combustion to organic matter (OM) during the winter season. The authors investigated the optical properties of atmospheric aerosols using an aethalometer, characterised aerosol chemical composition with an aerosol mass spectrometer (AMS), and determined the concentration of selected organic markers. The collected data were successively modelled using three source apportionment models (AeM, CMB, and PMF). This study demonstrated the limitations of the different approaches, but was limited to a two-week campaign performed at an urban site in winter.

In the present study, we compare source apportionment results for PM10, OC, and EC — as achieved through the application of the aethalometer model, CMB based on organic tracers, and PMF based on elemental tracers — for extensive data sets collected at different types of sites during a one-year time period (from August 2008 to July 2009). On the one hand, the possibility to compare results from different types of sites and different seasons is exploited to gain additional information on the performances of the different models under different ambient conditions. Specifically, the difference between rural and urban sites and between the winter and summer seasons are investigated, while aiming to identify the weaknesses and strengths of the source apportionment methods for different site types and seasons. On the other hand, the model intercomparisons are used for a mutual validation of the model outputs. This is of importance because the overall uncertainty of the applied source apportionment methods is unknown and their results should whenever possible be verified by comparison of different methods.