Source apportionment of elevated wintertime PAHs by compound-speciﬁc radiocarbon analysis

Natural abundance radiocarbon analysis facilitates distinct source apportionment between contemporary biomass/biofuel ( 14 C “alive”) versus fossil fuel ( 14 C “dead”) combustion. Here, the ﬁrst compound-speciﬁc radiocarbon analysis (CSRA) of atmospheric polycylic aromatic hydrocarbons (PAHs) was demonstrated for a set of samples col- lected in Lycksele, Sweden a small town with frequent episodes of severe atmospheric pollution in the winter. Renewed interest in residential wood combustion means than this type of seasonal pollution is of increasing concern in many areas. Five individ-ual/paired PAH isolates from three pooled fortnight-long ﬁlter collections were analyzed by CSRA: phenanthrene, ﬂuoranthene, pyrene, benzo[b + k]ﬂuoranthene and in- deno[cd]pyrene plus benzo[ghi]perylene; phenanthrene was the only compound also analyzed in the gas phase. The measured ∆ 14 C for PAHs spanned from − 138.3‰ to 58.0‰. A simple isotopic mass balance model was applied to estimate the fraction biomass ( f biomass ) contribution that was constrained to a range of 71% for in-deno[cd]pyrene + benzo[ghi]perylene to 87% for the gas phase phenanthrene and par- ticulate ﬂuoranthene, respectively. Indeno[cd]pyrene plus benzo[ghi]perylene, known to be enhanced in gasoline-powered motor vehicle exhaust compared to diesel exhaust, had the lowest contribution of biomass combustion of the measured PAHs by 9%. The total organic carbon (TOC, deﬁned as carbon remaining after removal of inorganic carbon) f biomass was estimated to be 77%, which falls within the range for PAHs. This CSRA data of atmospheric PAHs demonstrate the non-uniformity of biomass combustion contribution to di ﬀ erent PAHs even in a location with limited local emission sources and illustrates that regulatory e ﬀ orts would not evenly reduce all PAHs.

control PAH emissions.
Apportioning PAHs to a specific combustion source can be difficult (Hays et al., 2003;Schauer et al., 2002Schauer et al., , 2001McDonald et al., 2003). Differences due to biomass or fossil fuel type and to combustion conditions determine PAH ratios (Benner et al., 1995;Yan et al., 2005;Yunker et al., 2002), and can confound efforts to apportion PAHs 10 in an environmental sample; this also limits their utility as generic biomass or fossil fuel combustion tracers as very location-and source-specific profiles are needed for accurate apportionment (Mandalakis et al., 2004a;Lima et al., 2005). Factor-analysisbased source apportionment, such as principal component analysis and positive matrix factorization (PMF), has been used to attempt to define source contributions to PAHs 15 with mixed success (Larsen and Baker, 2003). The advantage to methods like PMF is that it requires no source profile input, however the goal is often to define the factors as specific sources which does require knowledge of PAH distribution in local source emissions. PMF also requires a large dataset to be effective, which is not always feasible. There is interest both in ascertaining the sources of PAHs and in accurately 20 using PAHs as source tracers in the atmosphere to quantify the contribution of emission sources to total atmospheric organic carbon (OC). Specific PAHs may be more valuable as molecular markers than others in source apportionment models for OC when used in conjunction with more unique tracers such as hopanes and levoglucosan (Chow et al., 2007;Jaeckels et al., 2007;. However, more investigation 25 into the sources of individual atmospheric PAHs is warranted to inform current source apportionment techniques. Natural abundance radiocarbon analysis enables determination of the contribution of fossil and non-fossil sources to carbonaceous compounds. Basically, the 14  Interactive Discussion mospheric CO 2 dictates the 14 C in fresh carbon and this then decays with a half-life of 5730 years. Hence, radiocarbon is ideal to distinguish between fossil fuel (void of 14 C) and biomass/biofuel (contemporary 14 C) combustion sources. However, the large sample masses traditionally required for radiocarbon analysis has largely limited its application in atmospheric particulate matter to characterizing total or bulk organic car- 5 bon (Hildemann et al., 1994;Jordan et al., 2006;Szidat et al., 2004a;Bench et al., 2007). For bulk carbon, contemporary sources can include both biogenic emission and biomass combustion sources, so these results can only be used to define contribution of fossil fuel combustion sources including motor vehicles and power generation. Advances in molecular-level radiocarbon analysis (Eglinton et al., 1996) have allowed 10 compound class-specific radiocarbon analysis (CCSRA) of atmospheric PAHs (Mandalakis et al., 2005;Zencak et al., 2007b;Kumata et al., 2006). These studies have revealed geographic differences in fossil vs modern biomass sources of PAH across Europe and in Japan. The study from Japan is compelling as it indicates the importance of isolating individual PAHs for compound-specific radiocarbon analysis (CSRA) 15 by revealing differences in the contribution from biomass burning between pooled low versus high molecular weight PAHs (Kumata et al., 2006). These results combined with emission source profiles suggest that source contributions would not be uniform for all PAHs.
To further consider the source apportionment of individual atmospheric PAHs, a res-20 idential area in a northern Swedish town was chosen for a winter sampling campaign. Lycksele, Sweden has been the site of previous atmospheric studies due to the high levels of ambient particulate matter and PAHs (Johansson et al., 2004;Hedberg and Johansson, 2006) and the high percentage of households using biofuels for space heating during wintertime (Hedberg and Johansson, 2006;Krecl et al., 2007); this is 25 a typical situation for small towns in the northern boreal zone. The contribution of elemental carbon (EC) and OC to the total PM 10  Printer-friendly Version Interactive Discussion and particle number concentrations for Lycksele which were associated with known traffic emission and biomass burning particle size distributions (Krecl et al., 2008b). A large day-to-day and hour-to-hour variability in aerosol concentrations was also observed with evening aerosol concentrations significantly higher on weekends than on weekdays, presumably associated to residential wood combustion (RWC) emissions. 5 Local traffic emissions were identified based upon their similar contribution every day and characteristic peak in the morning and in the evening (Krecl et al., 2008a). RWC and traffic emissions were thus identified as the two major local sources using several techniques, with long-range transport contributing to a lesser degree. In the current study, the relative impact of the RWC and traffic emissions is quantified by perform-   (Krecl et al., 2007;Krecl et al., 2008a). Two custom-built high-volume samplers were employed (Broman et al., 1991), each one consisting of an inverted filter holder which collected total suspended particle matter (TSP) on borosilicate glass fiber 20 filters (GFF, 293 mm diameter, Millipore, USA). Downstream of the GFF filters were two sequentiall polyurethane foam (PUF) traps for collecting volatile compounds. Filter changes were made on a 2-week schedule while PUF were replaced on a 1-week schedule to limit potential breakthrough. During the initial two weeks (23 January-6 February), the sampling was conducted with both GFF and PUFs; sampler A had an  For the final four weeks of the campaign (6 February-8 March), the sampling was performed only in the particle phase with an average flow rate of ∼35 m 3 h −1 for both samplers. In total, three sets of two-week filter samples and two sets of one-week PUF samples were collected. Additionally, one filter blank and two PUF blanks were transported to and from the measurement site together with the other filters and PUFs and 5 were exposed outdoors in the sampling equipment for 30 s. Prior to sampling, filters were baked at 450 • C for 5 h and PUFs were pre-cleaned by washing in a washing machine at 90 • C without detergent, dried at 35 • C for 4 days and Soxhlet extracted first with toluene for two days and then with acetone for one day. PUFs were thereafter dried in a desiccator with vacuum suction for one day and 10 finally wrapped in aluminum foil and sealed in airtight plastic bags until sampling. After sampling, filters were wrapped in aluminum foil, packed into airtight plastic bags and PUFs were wrapped in aluminum foil and sealed in airtight plastic bags. All samples were first stored for several days in a refrigerator (+4 • C) and then in a freezer (−18 • C) prior to analysis.

TOC Analysis
Total organic carbon (TOC) is here defined as the carbon remaining after inorganic carbon is removed by acid treatment (Zencak et al., 2007a). TOC analysis was conducted on 3 filter samples (Table 1). A 2.90 cm 2 portion of each filter sample was treated with 37% HCl fumes in a desiccator for 2 days to remove carbonates. The 20 residual (TOC) was determined with a high-temperature catalytic elemental analyzer (Carlo Erba/Fisons, Italy) and blank corrected.

PAH Analysis
For each 2-week sample period, 75% of each filter from the two samplers were combined for extraction resulting in 3 filter samples. For the PUFs, a combined sample was 25 extracted for each of the two 1-week PUF sample periods from the two samplers result- Interactive Discussion ing in 2 PUF extracts. The filter and PUF samples were extracted with cyclohexane in a Soxhlet apparatus. After extraction, a 5% aliquot of the extract was spiked with deuterated PAHs (phenanthrene-d 10 , fluoranthene-d 10 , pyrene-d 10 , and benzo[a]pyrene-d 12 , benzo[g,h,i ]-perylene-d 12 ). The remainder of the method is based on previously reported work (Mandalakis et al., 2005;Zencak et al., 2007b). Briefly, the cyclohexane extracts were reduced to 2 ml by rotary evaporation. To purify the extracts, each was applied to a deactivated silica gel column (SiO 2 -10% H 2 O, 63-200 µm particle size, 10 cm×1 cm i.d.) and eluted with n-hexane. The samples were further treated with a dimethylformamide (DMF-5% water)-pentane partitioning cleanup procedure (Mandalakis et al., 2004b) to isolate PAHs from interfering aliphatic compounds. Extracts 10 were then applied to a second deactivated silica gel column containing disodium sulfate (Na 2 SO 4 , anh.) to remove water. At this point, the extracts were spiked with a co-injection standard, chrysene-d 12 , and analyzed by gas chromatography/mass spectrometry (Fisons 8060GC interfaced to a Fisons MD 800 mass spectrometer) with the MS operated in selective ion monitoring mode (Mandalakis et al., 2004a). 15

Preparative capillary gas chromatography and accelerator mass spectroscopy
The details of the Gerstel preparative capillary gas chromatography (PCGC) method used for PAH isolation and harvesting have been reported previously (Mandalakis and Gustafsson, 2003;Mandalakis et al., 2004a) and are summarized here. Ninety-five percent of each cyclohexane extract was first purified as described in the PAH analysis 20 section except for the spiking with deuterated PAHs. Each purified extract was then repeatedly injected into the PCGC system (roughly 40 injections per extract) and the PAHs listed in Table 1 were trapped separately. PAHs harvested in the glass traps were then transferred using several rinses of hexane and passed through a small column packed with SiO 2 (4 cm×0.5 cm i.d.) using hexane. A small aliquot from each trap was 25 used to check purity and to quantify the harvested PAHs with GC-MS. Purities ranged from 92-100%, with an average of 97% and yields were low at 24-46% for all but benzo[ghi]perylene which had a yield of 15%. The low yield will not impact the radio-20907 Introduction  (2007c). In order to achieve sufficient mass for the CSRA method, the PAHs from the 3 filter extracts were combined after harvesting. The harvested PAHs from the 2 PUF extracts were likewise combined (Table S1: http://www.atmos-chem-phys-discuss.net/8/ 5 20901/2008/acpd-8-20901-2008-supplement.pdf). This was the final step conducted at Stockholm University before the extracts were shipped out for further analysis. Carbon isotope analysis was performed at the National Ocean Sciences Accelerator Mass Spectrometry (NOSAMS) facility of the Woods Hole Oceanographic Institution (USA). Acid fumigated TOC and PCGC-isolated individual PAH samples were oxidized 10 to CO 2 , purified and quantified by manometry (Pearson et al., 1998). About 10% of CO 2 was kept for δ 13 C analysis by isotope ratio mass spectrometry. The remaining 90% was reduced to graphite and subjected to accelerator mass spectrometry to determine the fraction of modern carbon, f M , which is the 14 C/ 12 C ratio of the sample related to that of the reference year 1950. The reported ∆ 14 C error is the larger of the 15 internal error (statistical error calculated using the number of counts measured from each AMS sample) and external error (error calculated from the reproducibility of individual analyses of a single sample) (Pearson et al., 1998). More details on radiocarbon conventions and metric, the routine AMS sample preparation and analysis procedures are available elsewhere (Zencak et al., 2007b;Klinedinst and Currie, 1999 winter season in Lycksele, Sweden (Fig. 1). Gas phase samples were only collected in the first sampling period, but the cold winter temperatures ensured that the volatile component was important only for phenanthrene, the most volatile of the measured PAHs; other measured PAHs were predominantly present in the particulate phase.
Despite the 14-day sampling integration, the three filter samples had different PAH 10 concentrations and fingerprints. The lower molecular weight PAHs (fluoranthene and pyrene) have relatively higher concentrations in the first sampling period (23 January-6 February) while the higher molecular weight PAHs (chrysene to coronene) were higher in the second sampling period (6-22 February). The particulate TOC concentration for the three sampling periods parallels the trend seen for the higher molecular weight 15 PAHs. To put these PAH concentrations in perspective, Table 2 compiles PAH concentrations from several previously reported studies of areas impacted by biomass burning (Sheesley et al., 2007;Schnelle-Kreis et al., 2007;Chowdhury et al., 2007;Schauer and Cass, 2000

Carbon isotope results
The high ambient PAH concentrations in Lycksele made it an ideal site for this CSRAbased source apportionment study. The five individual/paired PAHs of highest concentration were successfully isolated, harvested and quantified for radiocarbon content ( Table 1). Results of 13 C isotope analysis are presented as δ 13 C values relative to the 5 Vienna Pee Dee Belemnite (VPDB) standard (Coplen, 1996). Radiocarbon data are reported as ∆ 14 C and fraction modern (f M ) relative to NBS Oxalic Acid I and have been corrected for carbon blank and isotopic fractionation both in the environment and during sample preparation using a δ 13 C of −25‰ as described elsewhere (Klinedinst and Currie, 1999;Stuiver and Polach, 1977;Zencak et al., 2007c). Above-ground nuclear 10 testing in the mid-twentieth century nearly doubled atmospheric ∆ 14 C values, but since that time this radiogenic "bomb spike" signal has been consistently decreasing (Levin et al., 2003), which means that biomass which grows over many years, such as trees, will have a higher ∆ 14 C than contemporary CO 2 . Therefore, contemporary CO 2 and freshly produced biomass has a ∆ 14 C value of +70‰ (Levin  (Mandalakis et al., 2005). Additionally, the TOC and all the PAH δ 13 C values are within the reported range for terrestrial C 3 plant combustion (Zencak et al., 2007a). It should be noted that although ∆ 14 C values reported in the literature are commonly corrected for fractionation (Stuiver and Polach, 1977), the same is not true of δ 13 C values. Therefore δ 13 C results derived from samples prepared using PCGC may be biased if the entire peak was not collected (Zencak et al., 2007c).

Radiocarbon-based source apportionment
An isotopic mass balance equation was then applied to the TOC and PAH ∆ 14 C results 15 to calculate fractional contribution of wood burning (f biomass ) and fossil fuel combustion (f fossil =1−f biomass ) (Reddy et al., 2002;Currie et al., 1999;Mandalakis et al., 2004a;Reddy et al., 2003): In the above equation, ∆ 14 C sample represents the radiocarbon results presented in Ta-20 ble 1 for TOC and PAHs and the ∆ 14 C biomass is in the present study set to +218‰ (see above discussion). The resulting f biomass are presented in Fig. 2. The PAH f biomass results represent the full 6-week sampling period for all species except gas-phase phenanthrene (2-week sample), while the TOC average f biomass ±one standard deviation is included in Fig. 2 The f biomass ranges from 71-87% for all PAHs and the average for TOC is 77±3%. Hence, TOC is in the middle of the f biomass range for the measured PAHs. Several previous studies have reported the f biomass for pooled, compound-class specific radiocarbon analysis (CCSRA) of PAHs. Background site studies in Europe found a f biomass for particulate CCSRA-PAHs of 50%, 9% and 7% for Southern Sweden, Croatia and 5 Greece, respectively (Mandalakis et al., 2005). The Croatian and Greek values represent summer conditions whereas the Southern Sweden f biomass is a multi-year average; a winter event at the same Southern Sweden site had an f biomass for TOC of 75-85% (Zencak et al., 2007a), which illustrates the impact of seasonal sources such as RWC. A more comprehensive study in the western Balkans measured f biomass in the range 10 35-65% for a selection of urban, industrial and rural sites (Zencak et al., 2007b). In a study in Tokyo, particulate CCSRA-PAHs in the PM 10 fraction were split into low and high molecular weight fractions by season. The contribution of biomass burning to PM 10 samples was 17-38% (summer) and 24-27% (winter) for low and high molecular weight (MW) PAHs, respectively, with little seasonal differences. Road traffic emissions 15 might be expected to dominate in and around a megacity like Tokyo, and therefore it is surprising to see the high fraction of PAH from biomass combustion in both summer and winter. The f biomass calculated for Lycksele is much higher than all three of these previously reported CCSRA-PAH studies. The f biomass of PAHs from this study plus the f biomass of TOC from Southern Sweden (Zencak et al., 2007a) indicate the dominance 20 of RWC during winter in the boreal zone.
To make a start at evaluating the isotopic (and source) heterogeneity between gas phase and particulate PAHs, this study afforded radiocarbon analysis on phenanthrene in both phases. The f biomass of gaseous and particulate phenanthrene were indistinguishable within three times the error in Lycksele. Although the gas and particulate 25 phase phenanthrene samples for radiocarbon were not synchronous (23 January-6 February for gas phase, and 23 January-6 February and 22 February-8 March for particulate phase, see Table S1: http://www.atmos-chem-phys-discuss. net/8/20901/ 2008/acpd-8-20901-2008-supplement.pdf), the TOC radiocarbon results for the three,

ACPD Introduction
Conclusions References Tables  Figures   Back  Close Full Screen / Esc

Printer-friendly Version
Interactive Discussion two-week periods show stability in the source contributions despite differences in concentration (Table 1). Based on the imperfect comparison in this study, the partitioning of phenanthrene between the gaseous and particulate phase is apparently dominated by ambient conditions under these winter-time conditions and is less impacted by differences in the physical or chemical characteristics of RWC or motor vehicle exhaust.

Fraction biomass and molecular weight
We further sought to investigate isotopic heterogeneity as a function of PAH vapor pressure and/or size, however the number of samples inhibits statistical analysis. In Fig. 2 skewed toward higher masses for high-temperature combustion processes such as fossil fuel combustion. In fact, both indeno[cd]pyrene and benzo[ghi]perylene have been used in organic molecular marker-chemical mass balance models (MM-CMB) to apportion the contribution of gasoline-powered motor vehicle exhaust to atmospheric particulate organic carbon (Sheesley et al., 2007;Chow et al., 2007). Figure S1 (http://www. 20 atmos-chem-phys-discuss.net/8/20901/2008/acpd-8-20901-2008-supplement.pdf) illustrates that the PAH/OC ratio for benzo [ghi]perylene, in particular, is much higher for gasoline-powered motor vehicle exhaust  in comparison to the same ratio for wood smoke emission profiles (Lee et al., 2005;Schauer et al., 2001;Fine et al., 2002Fine et al., , 2004 Schauer et al., 1999).

Source contribution to ambient loadings
The mean contribution of wood combustion to ambient TOC was 2.4 µg m −3 whereas fossil fuel combustion, most likely traffic emissions, accounted for 0.7 µg m −3 in the pe-5 riod 23 January-8 March. The contribution of RWC to phenanthrene was 7.8 and 1.9 ng m −3 , while traffic emissions contributed 1.  (Krecl et al., 2008a), the CSRA could stand alone to separate the biomass burning and local traffic emissions. For Lycksele, the additional information gained from using CSRA instead of CCSRA enabled the attribution of gasoline-powered motor vehicle exhaust over diesel due to the en-    (Sheesley et al., 2007), 4 semi-urban sites, particle phase, PM 2.5 fraction. c (Schauer and Cass, 2000) 2 sites, 2 winter events, particle phase, PM 2.5 fraction. d (Johansson et al., 2004), Norrmalm and Forsdala residential neighborhoods, gas + particle phases, PM 10 fraction. e gas