Environ. Sci. Technol. 2007, 41, 3821-3828
Semi Volatile Organic Compounds in Ambient PM2.5. Seasonal Trends and Daily Resolved Source Contributions JU ¨ R G E N S C H N E L L E - K R E I S , * ,†,‡ M A R T I N S K L O R Z , ‡,⊥ J U ¨ RGEN ORASCHE,† MATTHIAS STO ¨ LZEL,§ ANNETTE PETERS,§ AND R A L F Z I M M E R M A N N †,‡,⊥ Department of Chemistry, Bavarian Institute of Applied Environmental Research and Technology, BIfA GmbH, Am Mittleren Moos 46, 86167 Augsburg, Germany, Institute of Ecological Chemistry, GSFsNational Research Centre for Environment and Health, Ingolsta¨dter Landstrasse 1, 86573 Neuherberg, Germany, Institute of Epidemiology, GSFs National Research Centre for Environment and Health, Ingolsta¨dter Landstrasse 1, 86573 Neuherberg, Germany, and Department of Analytical Chemistry, University of Augsburg, Universita¨tsstrasse 1, 86159 Augsburg, Germany
Concentrations of ambient semivolatile organic compounds (SVOC) in the PM2.5 fraction of Augsburg, Germany, have been monitored on a daily basis from January 2003 through December 2004. Samples were taken in a large garden in the city center. Quantitative analysis of n-alkanes, alkanones, alkanoic acid methylesters, long chain linear alkyl benzenes and toluenes, hopanes, polycyclic aromatic hydrocarbons (PAH) and oxidized PAH, and some abietan type diterpenes was done. All compounds showed distinct seasonal variations in concentration. Most compounds showed highest concentrations during the cold seasons, but some n-alkanones and 6,10,14-trimethylpentadecanone showed maximum concentration during summer. Changes in patterns between and within compound classes were obvious, e.g., the hopane pattern exhibited a strong seasonal variation. The main source related contributions to changes observed were discussed. Using positive matrix factorization (PMF) for the statistical investigation of the data set, five factors have been separated. These factors are dominated by the pattern of single sources or groups of similar sources: factor 1, lubricating oil; factor 2, emissions of unburned diesel and heating oil consumption; factor 3 , wood combustion; factor 4, brown coal combustion; and factor 5, biogenic emissions and transport components. Like the SVOC, the factors showed strong seasonality with highest values in winter for factors 1-4 and in summer for factor 5. * Corresponding author phone: +49 (0) 821 7000228; fax +49 (0) 821 7000100. e-mail:
[email protected]. † Bavarian Institute of Applied Environmental Research and Technology. ‡ Institute of Ecological Chemistry, GSFsNational Research Centre for Environment and Health. § Institute of Epidemiology, GSFsNational Research Centre for Environment and Health. ⊥ University of Augsburg. 10.1021/es060666e CCC: $37.00 Published on Web 04/27/2007
2007 American Chemical Society
Introduction Epidemiological studies have shown the association of ambient particulate matter and various health endpoints (1). Several aerosol properties responsible for health effects under discussion include transition metals (2), organic compounds (3), endotoxins (4), and fine (0.9 and homohopane
FIGURE 1. Monthly mean concentrations of selected n-alkanes (A), alkanones (B), n-alkanoic acid methylesters (C) and molecular markers related to wood combustion (D). For abbreviations see Table 1.
FIGURE 2. Monthly mean values of the hopane index (30ab/(30ab+30ba) and the homohopane index (31abS/31abS+31abR). index in the range of 0.54-0.67. The cold periods on the other hand are dominated by hopane pattern which most likely derive from solid fossil fuel combustion. This winter pattern is similar to the hopane pattern found in brown coal smoke by Oros and Simoneit (33). Ten PAH and retene were quantified in the samples (Table 1). Like all other combustion-source-derived compounds, PAH-concentrations showed a strong seasonality with maximum concentrations in the winter months and lowest concentration during summer. The ratio of the highest and lowest concentrations (monthly averages) were in the range of 30-50 for the more stable PAH like BBKF; BEP, BGH and up to 110 for the more reactive PAH like BAA and BAP. The reasons for the changes in PAH pattern detected were changes in source contributions, (photo-) oxidation of PAH during atmospheric transport and losses of reactive PAH during sampling due to oxidation reactions (49, 50). PAH-emission concentrations and pattern strongly depend on combustion source and conditions (e.g., type and age of engine or boiler, temperature, engine or boiler load, temperature, fuel, and PAH content in fuel, e.g., ref 51). For wood combustion, for instance, PM emission factors in the range of 0.1-3 mg kg-1 wood with rations of BEP/BEP+BAP in the range of 0.3-0.9 were reported (34, 35, 52 ,53). Therefore, we made no attempts to correlate the changes in PAH concentration or pattern with PAH sources. That applies to oxidized PAH (O-PAH), too. There’s only sparse information on O-PAH in emissions. Additionally, they can degrade or be formed from precursor PAH during atmospheric transport and sampling.
The molecular markers for biomass combustion dehydroabietan (DHA) and dehydroabietic acid methylester (DHAM) were quantified in the samples. The diterpenoids abietan and dehydroabietic acid methylester are specific for emissions from wood combustion (54, 55). Dehydroabietic acid is predominant in the resin of conifer wood; during combustion, the methylester is partially formed by esterification of dehydroabietic acid with wood alcohol (34). Therefore, DHAM is a unique marker for (conifer) wood combustion. Abietan is a degradation product of the resin; therefore, it can originate from (conifer) wood combustion (34) or degradative/microbial emissions from trees (55). Diterpenoids have also been found in emissions from low rank coal (33). Both markers for biomass combustion showed high concentrations in winter and low concentrations in summer (Figure 1). The ratios of monthly average winter and summer concentrations for DHAM was 40. DHA showed higher summer concentrations, supporting degradative/ microbial emissions as a possible source. Factor Analysis. Positive matrix factorization (PMF) was used to identify the emission sources which govern the concentrations of the SVOC investigated. Five factors could be separated. The interpretation of these factors is based on the compound pattern (Figure 3, top and middle) and the seasonal pattern (Figure 3, bottom) of each factor. It was evident that no factor could be separated which exclusively defines a single source. But all five factors are clearly dominated by a single source or a group of similar sources. VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3825
FIGURE 3. Results of PMF analysis. Top and middle: proportions of source factors on average concentrations of SVOC used in PMF. Bottom: monthly mean proportions of source factors on ambient concentrations of the SVOC used in PMF. In the presence of two or more factors with very similar variation in time (as, for instance, domestic heating with different fuels in the cold season) statistical approaches like PMF may have difficulty in the separation of these sources in the time series analysis. Another difficulty with receptor modeling approaches is that they are based on the assumption of constant source patterns over time. This is not true in the real world. However, from emission measurements, it is known that patterns from one source type may vary with the conditions (e.g., for combustion sources: type and age of engine or boiler, temperatures, engine or boiler load, etc.) even within a class of compounds emitted. The variation between compound classes is even higher. Despite these inherent problems the five factors separated by means of PMF could be assigned to dominating sources based on the basic pattern of the analyzed compound classes discussed above. The first factor is characterized by a compound pattern very similar to lubricating oil emissions from vehicles. High amounts of n-alkanes with Cmax at C25 (28) dominate this pattern. They are accompanied by LAB with the same range of boiling points and a hopane pattern, which is characteristic for mineral-oil-based sources (46). This factor showed the least variation in time. Lowest values were found in summer and highest values were found in the cold season. The ratio of highest/lowest monthly average (November 2003/August 2003) was 14. The average winter/summer ratio was 2. These ratios are in agreement with the observed higher emissions from vehicles under cold start and driving conditions (56, 28). The second factor is dominated by n-alkanes in the lower boiling range of the analyzed compounds with Cmax at C22. All n-alkanes C19) n-alkanoic acid methylesters with a pronounced preference for even numbered chain lengths (CPI ∼ 7) and Cmax at C22C24 (40). The n-alkane pattern found in this factor in principle can be explained by emissions from wood combustion, too (34, 35). On the other hand, the LAB and hopanes in this factor cannot be assigned to wood combustion. They most likely originate from other fuel used for domestic heating. About 80-95% of the PAH and 40-90% of the O-PAH are associated with this factor. As wood combustion is expected to be used mainly for domestic heating in winter and only used to a minor extent to heat water in summer, the summer/ winter ratio of >50 supports the interpretation of factor 3 as dominated by emissions from wood combustion. In factor 4, a pronounced hopane and LAB pattern, very similar to the pattern found in brown coal combustion emissions (33), was found. The interpretation of this factor as brown coal combustion dominated is supported by the high concentra-
tions of long chain methylesters (>C23) with Cmax at C26. The distinct fraction of long chain n-alkanones with Cmax at C23C25 can be explained by coal combustion (33), too, but they also can originate from wood combustion (34, 35). Brown coal is used mainly for domestic heating in old (single room) stoves in old buildings. Therefore, only very low values of factor 4 in summer were expected. In a monthly average, factor 4 had a ratio >400 between highest (November 2003) and lowest (July 2003) values. The avererage winter/ summer ratio was >100. These very high ratios also support the interpretation of factor 4 as brown coal combustion dominated. In factor 5, the influences of two sources/processes are mixed. On the one hand there are clear indications for vegetative emissions, (due to abrasion or evaporation) to be relevant in this factor. This interpretation, first of all, is supported by 6,10,14-trimethylpentadecan-2-one, which is nearly solely concentrated in this factor. The n-alkane pattern, which is dominated by even numbered alkanes (CPI > 3) and Cmax at C29 and C31 is an indicator for plant abrasion (30) as one possible source in this factor, too. On the other hand, there are several compounds which could not be explained by plant abrasion. Especially hopanes with a similar pattern as in mineral-oil-based sources (46), high concentrations of methylesters and alkanones with chain lengths
