Fluorescent water-soluble organic aerosols in the High Arctic atmosphere

Organic aerosols are ubiquitous in the earth’s atmosphere. They have been extensively studied in urban, rural and marine environments. However, little is known about the fluorescence properties of water-soluble organic carbon (WSOC) or their transport to and distribution in the polar regions. Here, we present evidence that fluorescent WSOC is a substantial component of High Arctic aerosols. The ratios of fluorescence intensity of protein-like peak to humic-like peak generally increased from dark winter to early summer, indicating an enhanced contribution of protein-like organics from the ocean to Arctic aerosols after the polar sunrise. Such a seasonal pattern is in agreement with an increase of stable carbon isotope ratios of total carbon (δ13CTC) from −26.8‰ to −22.5‰. Our results suggest that Arctic aerosols are derived from a combination of the long-range transport of terrestrial organics and local sea-to-air emission of marine organics, with an estimated contribution from the latter of 8.7–77% (mean 45%).

A tmospheric aerosols have been recognized to play an important role in regulating regional and global climate 1 . For example, the enrichment of aerosol particles with organic compounds can make the aerosol surfaces more hydrophilic or hydrophobic depending on the composition and mixing state, leading to alterations in the cloud condensation nuclei (CCN) activity of the particles. When deposited onto the earth's surface, atmospheric aerosols can darken snow albedo, and modify both land and ocean biogeochemistry 2 . Thus, the long-range atmospheric transport of aerosol particulate matter from continents to open oceans 3 or to the polar regions 4,5 is an active area of research.
Water-soluble organic carbon (WSOC) is a major fraction of atmospheric aerosols. WSOC contains aromatic rings or aliphatic structures with carboxyl, hydroxyl, carbonyl or methoxy functional groups. The chemical nature of 10-20% (by mass) of WSOC has been resolved at a molecular level, and found to include mono-and dicarboxylic acids, sugar compounds, amino acids, and polar organic marker compounds (e.g. 2-methyltetrols and pinic acid) from the photooxidation of biogenic volatile organic compounds (BVOCs) [5][6][7][8][9] . The majority of WSOC is currently thought to be the water-soluble fraction of HUmic LIke Substances (HULIS), which are similar to terrestrial and aquatic humic and fulvic acids 10,11 .
Fluorescence techniques (e.g. synchronous scan and excitation-emission matrix spectroscopy) have been widely used to investigate the sources and optical properties of dissolved organic matter (DOM) or humic substances in aquatic environments [12][13][14][15] . Such techniques are widely applied to study the sources and chemical nature of chromophoric DOM (CDOM) in oceans 16,17 . For example, the use of excitation-emission matrix (EEM) spectroscopy permits discrimination of CDOM sources based on the relative abundances of different fluorophores, including humic-like, protein-like, and pigment-like fluorescence 12,17 . Many studies have suggested that the optical properties of chromophoric WSOC in the atmosphere may be similar to CDOM in aquatic environments 18,19 . Furthermore, fluorescence techniques have been used to study photo-degradation effects on CDOM in rainwater 20 and WSOC in aerosol particles 11,19,[21][22][23] . During the past few years, on-line fluorescence instruments, such as the ultraviolet aerodynamic particle sizer (UV-APS) and the wideband integrated bioaerosol sensor (WIBS) have been successfully used to measure emission spectra of single bioaerosol particles in ambient air 24,25 . Recent field and laboratory studies have also demonstrated that non-biological particles such as biogenic SOA derived from the photooxidation of isoprene and monoterpenes, can emit a strong fluorescence 26,27 .
The Arctic atmosphere was once considered as one of the most pristine environments on earth. However in the 1950s, two pilots flying over the North American Arctic noticed a widespread haze that was subsequently found to be observable every winter and early spring 4 . Since then, there has been much research into Arctic haze, which has now been elucidated as a mixture of sulfate, ammonium, nitrate, black carbon, organic carbon, that also contains relatively high levels of ozone precursors such as nitrogen oxides (particularly in the form of peroxyacetylnitrate, PAN) and volatile organic compounds (VOCs) 4,8 . During the polar sunrise season, the Arctic atmosphere is considered to serve as a unique photochemical reactor influenced by continent-derived particles and their precursors from the mid-latitudes in Eurasia or North America and marine-derived particles from the Arctic Ocean 4 . However, little is known about the changes in optical characteristics and sources of bulk organic aerosols before and after polar sunrise.
The objective of this study was to investigate the EEM properties of WSOC in Arctic aerosols collected at Alert, in the Canadian High Arctic (Figure 1a), in order to understand the influence of marine organics to organic aerosols in the Arctic atmosphere. The temporal trends of humic-like and protein-like fluorescence were compared with molecular marker compounds including methanesulfonic acid (MSA), a specific oxidation product of dimethyl sulfide (DMS) of marine origin 28 , and other polar organic species from biomass burning and biogenic VOC oxidation. In addition, stable carbon isotope ratios (d 13 C) of High Arctic aerosols were determined to obtain new insights into the influence of sea-to-air emissions of marine organics to Arctic aerosol composition from dark winter to light spring.

Results
OC and WSOC concentrations. The concentrations of organic carbon (OC) and WSOC in the Alert aerosols ranged from 73.4-387ng m 23 (mean 253ng m 23 ) and 40.7-300ng m 23 (186ng m 23 ), respectively ( Table 1). The aerosol mass concentrations ranged from 2540-9130ng m 23 (5250ng m 23 ). OC and WSOC made up only a small fraction of the aerosol mass (2.1-4.2% and 1.2-3.3%, respectively), suggesting that mineral sea salts dominate High Arctic aerosols. Their fractions were higher in February and late March, and lower in early summer. The decrease in the concentrations of OC and WSOC (Figure 2a-b) from February to early March is consistent with previous studies 29,30 .
Stable carbon isotope ratios. Stable carbon isotopic composition (d 13 C) has been successfully used to understand the contributions of marine and continental sources to aerosol carbon 29,31,32 . Based on d 13 C measurements, Chesselet et al. 33 reported that terrestrial organic matter is more important (.80%) in the remote marine aerosols than marine-derived organic matter. In the present study, the stable carbon isotope ratios, d 13 C, in total carbon (TC) fall in a relatively narrow range of 222.5 to 226.8% (Table 1). The d 13 C values before the polar sunrise ranged from 225.3 to 226.8%, lower than those (222.5% to 225.0%) during light spring to early summer.
Fluorescence properties of WSOC. Figure 1 presents typical EEM spectra of WSOC in High Arctic aerosols. Before the polar sunrise, strong humic-like fluorophores were observed at ex/em5240-250/ 416-426nm and ex/em5335-345/436-446nm, which are traditionally defined as Peak A and Peak C, respectively 12 . After the polar sunrise, an obvious peak was also found at ex/em5270-280/310-320nm (Peak B), which is generally considered to correspond to protein-like fluorophores 12 . The fluorescence intensities of peaks A (terrestrial humic-like), C (terrestrial humiclike), M (marine humic-like), and B (protein-like) obtained from the excitation/emission pairs are summarized in Table 1   Air masses transport characteristics. In order to understand the general transport characteristics of air masses recorded at the sampling site, we conducted a ten-day backward Lagrangian Particle Dispersion Modeling (LPDM) analysis using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model 34 . This method has been shown to perform well in the simulation of long-living species such as CO, and has been widely used to understand the transport and origins of air pollutants 35,36 . Based on the backward particle release simulation for each hour during the sampling period, the average distribution of retroplumes for each sample collected at Alert in the High Arctic is shown in Figure 3. Before the polar sunrise, the air masses originated from the Arctic Ocean, Canadian High Arctic and Greenland. From late March to mid-May, the retroplumes show that the air masses were mainly delivered from the Arctic Ocean to the sampling site with a significant air mass transport from North America in late March. After late May, the air mass contribution from North America was enhanced.

Discussion
The time series measurements of OC and WSOC (Figure 2a-b) showed a decrease from February toward mid-March. After the polar sunrise, their concentrations increased rapidly and remained high till mid-April. They then decreased in mid-May and slightly increased toward early June. Further, both the humic-like and protein-like fluorescence peaks (Figure 2c-f) showed a similar pattern, consistent with those observed for molecular marker compounds for primary organic aerosols (POA) that are directly emitted from sources such as plant material, soil dust, biomass/biofuel and fossil fuel combustion 5 . For example, during the same campaign, similar patterns were also observed for levoglucosan and polycyclic aromatic hydrocarbons (PAHs) (Figure 2g-h). Levoglucosan, a biomass-burning tracer that is produced in large quantities during the pyrolysis of cellulose 3 , was found in Arctic aerosols with large variations (3-1080pg m 23 ) but higher concentrations in dark winter. It is not surprising given that biomass-burning emissions are one of the most significant sources   for atmospheric HULIS 10,37 . PAHs are mainly representative of anthropogenic aerosols, which are produced from fossil fuel combustion and biomass burning. The winter maxima of OC, WSOC, and organic species can be explained by the concept of the Arctic acting as a cold sink during winter to receive long-range transported aerosols and their precursors transported from their emission regions in the mid-latitudes in Eurasia and North America. Under stagnant conditions with the lack of solar radiation in winter, aerosol removal rates are minimized 38 . After the polar sunrise, the increased concentrations of OC and WSOC are a result of the enhanced sea-to-air emissions of marine organic matter, and the production of secondary organic aerosols (SOA) from the photooxidation of anthropogenic and biogenic VOCs.
Fluorescent aerosol particles may also contain primary biological aerosol particles, which are ubiquitous and comprise a variety of particle types such as bacteria, algae, pollen, fungal spores, plant debris and biopolymers (protein, DNA, chitin, cellulose and other polysaccharides) 24,25,39 . In the present study, the fungal spore tracers, arabitol and mannitol 40 , showed a similar pattern to levoglucosan and PAHs from February to May. However, a significant increase was then observed at the end of May when the air masses mainly originated in North America (Figure 3m-p). Sucrose, a tracer for airborne pollen 41 , also peaked in late spring. Such an enhancement of protein-like fluorescence was also observed during late May to early June (Figure 2c). Thus, the similarity in temporal variations between fluorescence peaks and POA tracers indicates that biomass burning, fossil fuel combustion and primary biological aerosols are the main sources of fluorescent organics in the Arctic regions, and that they may be co-transported long distances from the mid-latitudinal regions. For example, Rousseau et al. 42 reported the long-distance transport of pollen grains from boreal forests in northeastern North America to Greenland in spring.
In the Earth's atmosphere, SOA formed by the photooxidation of BVOCs is believed to be more abundant than directly emitted POA.
Recent field and laboratory studies have demonstrated that nonbiological particles such as biogenic SOA from the photooxidation of isoprene and monoterpenes can also emit a strong fluorescence 26,27 . In a previous study, Fu et al. 8 reported the temporal variations of biogenic SOA tracers, which show that monoterpene SOA tracers such as pinic acid (Figure 2l) exhibited temporal patterns similar to those of OC and WSOC. b-caryophyllinic acid, a specific tracer for the photooxidation of b-caryophyllene 43 (a sesquiterpene known to have a high aerosol yield), also followed such a trend 8 . Positive correlations between fluorescence peaks and monoterpene and b-caryophyllene oxidation products indicate an intrinsic relationship between biogenic SOA and these fluorescent components.
The temporal variations of 2-methyltetrols (Figure 2k), the isoprene oxidation products, were substantially different from those of OC, WSOC, monoterpene-and sesquiterpene-SOA tracers, and fluorescent peak intensities. The general increase of isoprene oxidation products during February-May is most likely caused by local and/or regional emissions of isoprene from the open ocean followed by photooxidation. Isoprene emitted from marine algae could be a significant precursor of marine SOA 44 . The biologically active waters of the adjacent seas and open oceans in the Arctic have been proposed as a source of organic aerosols in summer 45 . The sharp increase of 2-methyltetrols together with molecular markers of fungal spores and pollen in late May (Figure 2i-k) may be caused by the shift in source regions of Arctic aerosols. The LPDM simulations have shown that in mid-May the air masses mainly originated from the Arctic Ocean and north Greenland, whereas in late May they mostly came from North America across the Baffin Bay (Figure 3m-p).
Furthermore, an increase in the relative enhancement of proteinlike fluorescence (peak B) to humic-like fluorescence (peak A) in the WSOC fraction (Figure 4a) from dark winter to light spring may indicate the enhancement of sea-to-air emission of marine biotaderived organics in spring, although such an increase of peak B / peak A may also partly result from the different photodegradation rates of humic-like substances and protein-like substances (humic type organics are considered to be more photoreactive than phytoplankton derived organics 46 ). However, the interpretation of the sea-to-air emission of marine organics as the cause of the enhancement is supported by an increase in methanesulfonic acid (MSA) (Figure 4b) in the same aerosol samples 6 . MSA is a photooxidation product of dimethylsulfide of marine algal origin 28 . In fact, marine biological activities can emit a large amount of primary organic aerosols, together with biogenic VOCs, into the marine atmosphere. The enhanced contribution of marine sources is also supported by the concentration ratios of lower molecular weight (LMW) to higher molecular weight (HMW) fatty acids. LMW fatty acids (i.e. those with carbon numbers ,C 20:0 ) are a major fraction of marine aerosols , while HMW fatty acids (.C 20:0 ) mainly originate from terrestrial higher plants. The concentration ratios of LMW to HMW fatty acids in High Arctic aerosols showed an increased trend from February to May (Figure 4c), ranging from 1.0 to 7.0 with an average of 2.3 (Table 1); the ratio decreased under the influence of the air masses transported from North America in late May (Figure 3).
It has been recognized that in the marine atmosphere, organic enrichment is strongly associated with decreasing size of particles 47 . For example, size distributions of organic carbon in marine aerosols have shown that both WSOC and WIOC (water-insoluble organic carbon) substantially existed in the fine submicron fraction 47 . In the High Arctic region, the photochemically produced components of SOA should be enhanced after the polar sunrise. This is especially true for the photooxidation products of oxalic acid and other dicarboxylic acids in Arctic aerosols with an enrichment in the fine fraction 6 . SOA is generally considered to be a major fraction of WSOC. Thus, the contribution of WSOC to total OC should be enhanced in spring-summer in the High Arctic under strong sunlight irradiation. However, a decrease in the contribution of WSOC to OC was observed between February to June (Figure 4d), suggesting that the contribution of WIOC (WIOC 5 OC -WSOC) increased from dark winter to light spring, which indicates that primary marine organic aerosols generated from sea spray constitute a large fraction of WIOC.
In fact, ocean surface waters are enriched with small particulate materials including phytoplankton, algae, bacteria, viruses, fragments of larger organisms and organic detritus 48 , which are known to be the main sources of marine microgels. Phytoplankton exudates include exopolymer gels consisting of polysaccharides that bind together with LMW organics such as amino acids and peptides. Exopolymer gels are insoluble, thermally stable, highly surface active and hydrated 48 . Thus, the pool of organic matter in the marine surface layer contains not only dissolved organic matter but also water-insoluble organics 48 that are emitted into the marine boundary layer (MBL) as sea spray via bubble bursting and breaking waves. For example, Hawkins and Russell 49 reported that freshly emitted organics from the ocean likely have low water solubility. Facchini et al. 50 also reported that primary organic aerosols emitted with sea sprays in a more biologically active ocean are comprised of water-insoluble organics (up to 77 6 5%), which contain phytoplankton-derived detritus. The strong positive correlation between WIOC contribution to OC and the ratio of protein-like fluorescence to humic-like fluorescence (peak B / peak A) (Figure 5a) suggests an obvious increase of WIOC materials in the High Arctic atmosphere due to the sea-to-air emission of marine organics. In a previous study, marine microgels were abundantly identified in aerosols, fog and cloud waters in the MBL over the Arctic Ocean in summer 51 .
Based on ionic composition and stable carbon isotope ratios (d 13 C) of Arctic aerosols collected during the ALERT2000 campaign, Narukawa et al. 29 postulated an increase in biogenic emissions from the surrounding open ocean in late spring to early summer in the Arctic. The enhanced sea-to-air emission of marine organics from February to June was also supported by the temporal trend of d 13 C values in High Arctic aerosols (Figure 4e), ranging from 226.8% to 222.5% (mean 224.5%) ( Table 1). The isotopic compositions of marine and continental carbonaceous aerosols have been well studied [31][32][33] . In general, d 13 C values of TC for continental aerosols vary significantly between pollution sources. For example, d 13 C values of primary aerosol particles emitted from fossil fuel combustion or biomass burning of C 3 plants range from 224 to 237% 29,31-33,52,53 ; those of biogenic SOA from b-pinene ozonolysis have been reported to be 229.6 6 0.2% 54 ; and of marine-derived particles are 220 to 222% 33 . In order to estimate the relative contribution of marine and continental influences on the Arctic aerosols, the d 13 C values in Figure 4e were applied to the following isotopic mass balance equations: where f marine and f continental are the fractions of marine and continental carbon, and d 13 C marine and d 13 C continental are the reported isotopic values for marine and continental carbon. Based on the previous studies 29,52 , the representative d 13 C values of continental and marine aerosols were taken to be 227.3% and 221.0%, respectively. The contributions of marine carbon to High Arctic aerosols are estimated to be 8.7-77% with an average of 45%, which are consistent with those (0-67%, mean 45%) during the Alert2000 campaign 29 . Our estimates of marine carbon contribution for light spring aerosols (37-77%, 53%) are higher than that (38%) reported at Bermuda 52 , where the marine aerosols are more affected by anthropogenic sources than the Arctic. In a coastal urban site in Chennai, south India, the estimated contributions of marine carbon to urban aerosols were 1-37% (mean 19%) 55 , which are similar to those (8.7-32%, 22%) in the High Arctic during dark winter when the marine carbon can be expected to be transported from lower latitudinal marine regions. Figure 6a shows the concentrations of aerosol carbon from marine and continental sources calculated using the two end-member model (Eqns 1 and 2). The estimated marine carbon concentration ranged from 34.5 to 231ngC m 23 (mean 133ngC m 23 ) and continental carbon concentration from 21.7 to 510ngC m 23 (210ngC m 23 ). The relative contribution of marine carbon clearly increased from winter to early summer (Figure 6b), indicating an enhanced sea-to-air emission of marine organics in the Arctic.
Ratios of fluorescence intensity from specific spectral regions of an EEM are often used as indicators for the relative contribution of organic matter derived from terrestrial or microbial sources in natural waters 13,14,56,57 . The humification index (HIX) was introduced by Zsolnay et al. 56 to estimate the degree of maturation of DOM in soil. During the humification process, the aromaticity of organic matter increases and its microbial availability decreases. Thus, high HIX values (.10) correspond to strongly humified or aromatic organics, principally of terrestrial origin, while low values (,4) are indicative of autochthonous or microbial origin 14,15,57 . From February to June, a decrease of HIX for WSOC was observed, ranging from 5.2 to 0.7 (Figure 4f). In addition, a strong linear correlation (R 2 5 0.60) between HIX and stable carbon isotope ratios was found in the Arctic samples (Figure 5b). These results again demonstrate a decreased contribution of terrestrial organics and an increased contribution of freshly emitted marine organics to Arctic aerosols from dark winter to light spring.
McKnight et al. 13 proposed the fluorescence index (FI) as a proxy for the relative amount of DOM derived from terrestrial and  microbial/algal sources in surface waters. FI values of 1.4 or less correspond to terrestrially derived organics and higher aromaticity, while values of 1.9 or higher correspond to microbial sources and lower aromatic carbon content 13 . Another index, the biological index (BIX) 57 , also allows an estimation of the contribution of autochthonous biological activity. An increase in BIX is related to an increase in the contribution of microbially derived organics. For example, high values (.1) have been shown to correspond to a predominantly biological or microbial origin of DOM and to the presence of OM freshly released into water, whereas values of ,0.6 contain little biological material 57 .
The FI and BIX values obtained from the Arctic aerosols are summarized in Table 1. The FI values ranged from 1.19 to 1.58 with an average of 1.47, indicating that the fluorophores in High Arctic aerosols are representative of both terrestrially and microbially derived organic matter. The BIX values (0.6-0.96, mean 0.72) are within the extreme values for the predominance of humic-or protein-like fluorophores. Interestingly, Lee et al. 26 reported that the BIX values of SOA samples were on average 0.6. In High Arctic aerosols, the presence of marine emitted organics enhances the values of BIX, while biogenic SOA formed from the photooxidation of biogenic VOCs after the polar sunrise may contribute to the lower BIX values. Figure 7 plots the HIX data from this study as a function of FI and BIX, together with literature data for soil and aquatic humic substances, DOM in river and oceanic waters, as well as rain and fogwater samples 14,15 . The fluorescence indices of the High Arctic aerosol samples fall between those of terrestrially and microbially derived organics, which again suggests that High Arctic aerosols are a mixture of organics from continental aerosols brought to the region through long-range transport and marine organics emitted locally from the ocean surface layer.

Methods
Sample collection. Total suspended particulate matter (TSP) was collected weekly by filtering air through precombusted (450uC, 3h) quartz fiber filters using a high-volume sampler in the Canadian High Arctic at Station Alert (82.5uN, 62.3uW) from February 19 to June 10, 1991. The surface air temperature (mean) ranged from 233.5uC in February to 21.9uC in June during the sampling period. Each filter was stored in a precombusted glass bottle with a Teflon-lined screw cap in darkness at 220uC until analysis. Based on the reanalysis for dicarboxylic acids in 2010 of one aerosol filter sample (QFF 298) collected in mid-May during this campaign, and a comparison with the results obtained in 1994 for the same sample, Kawamura et al. 6 found that the concentrations of major organic acids are equivalent within an analytical error of 15%. This suggests no serious degradation of organics during the sample storage. The experiments, including fluorescence and UV-Vis absorbance analysis, were finished in 2010.
Air mass transport features. We conducted backward particle release simulations using the Lagrangian dispersion model Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT), which was developed in the Air Resource Laboratory of the National Oceanic and Atmospheric Administration (NOAA), USA 34 . Here, we used ten-day backward Lagrangian Particle Dispersion Modeling (LPDM), as developed by Ding et al. 36 . In brief, for each hour during the sampling period, the model was run backward for ten days with 3000 particles released 100 m above ground level over the sampling site. The hourly position of each particle was calculated using a 3-D particle method. The spatiotemporal distributions of these particles were used to identify ''footprint'' retroplume and to calculate the potential source contributions 36 .
WSOC extraction and analysis. An aliquot (ca. 10cm 2 ) of filter samples were extracted by 20mL of organic-free Milli-Q water under ultrasonication for 20 min.  The water extracts were then passed through a syringe filter (Millex-GV, 0.22mm, Millipore) and analyzed for water-soluble organic carbon (WSOC) using a TOC analyzer (Model TOC-Vcsh, Shimadzu, Japan).
Fluorescence characterization. An aliquot (ca. 7cm 2 ) of the filter samples were extracted by organic-free Milli-Q water under ultrasonication for 15 min. The water extracts were then passed through a syringe filter (Millex-GV, 0.22mm, Millipore). Excitation-emission matrix (EEM) fluorescence was measured using a fluorometer (Fluoromax-4, Horiba). All fluorescence spectra were acquired in the S/R mode with instrumental bias correction. Inner filter correction 13 was carried out using absorbance spectrum measured with a spectrophotometer (Hitachi U-2000, Japan). After this procedure, the EEM of Milli-Q water was subtracted from the sample EEM. Finally, each EEM was calibrated to the water Raman signal 58 , and the fluorescence is in Raman units (RU, nm 21 ). Based on the ranges of Coble 12 , the fluorescence intensities of peaks A (terrestrial humic-like), C (terrestrial humic-like), M (marine humic-like), and B (protein-like) obtained from excitation/emission pairs are given in Table 1. Here the fluorescence intensity was further corrected to the amount of Milli-Q water used for the extraction and the air volume of each filter sample. Thus, the fluorescence is reported with a unit of RU L 21 m 23 . In addition, the humification index (HIX, the ratio of H/L, where H is the fluorescence intensity recorded at Ex 5 255nm for Em spectrum integrated from 434 to 480nm, and L is the fluorescence intensity recorded at Ex 5 255nm for Em spectrum integrated from 300 to 434nm), index of recent autochthonous contribution (BIX, ratio of emission intensities at Em 5 380nm and 430nm with excitation at Ex 5 310nm) and fluorescence index (FI) (ratio of Ex/ Em 5 370/470nm to Ex/Em 5 370/520nm) were calculated from each EEM 14,57 . The analytical uncertainty in the calculation of HIX, FI and BIX did not exceed 2%.
Stable carbon isotope ratio measurement. Small punches (3.14cm 2 ) of each quartz fiber filter sample were analyzed for total aerosol carbon content using an elemental analyzer (EA) (Carlo Erba, NA-1500). The filter punch was rounded using a pair of flat tipped tweezers, placed into a tin cup, and then caked into a ball. Before use, the tin cup was washed with acetone under ultrasonication to remove organic and other contaminants. Carbon was converted to CO 2 in the combustion furnace (1020uC). Stable carbon isotopic analyses were conducted using the same EA interfaced (ThermoQuest, ConFlo II) to isotope ratio mass spectrometer (IRMS) (ThermoQuest, Delta Plus). The isotopic compositions (d 13 C) were determined using the standard isotopic conversion: where R is the ratio of 13 C/ 12 C. Vienna Pee Dee Belemnite ( 13 C) was used as the isotope standard. Samples were analyzed in duplicate and averaged concentrations and isotopic ratios are reported here after the blank correction. Reproducibility of TC measurement was within 2%. Analytical error in the carbon isotope ratios was within 0.2%.
Organic molecular marker measurement. The dataset of molecular marker compounds and detailed measurement methods have been reported elsewhere 5,8 . Briefly, a filter aliquot (ca. 20cm 2 ) was cut into pieces and extracted three times with dichloromethane/methanol (2:1; v/v) under ultrasonication for 10 min. The solvent extracts were concentrated by a rotary evaporator, and then blown down to dryness with pure nitrogen gas. The extracts were then reacted with 50mL of N,O-bis-(trimethylsilyl)trifluoroacetamide (BSTFA) with 1% trimethylsilyl chloride and 10mL of pyridine at 70uC for 3 h. After reaction, the derivatives were diluted using 140mL of n-hexane with 1.43ng mL 21 of the internal standard (C 13 n-alkane) prior to gas chromatography/mass spectrometry (GC/MS) injection. GC/MS analyses were performed on a Hewlett-Packard model 6890 GC coupled to Hewlett-Packard model 5973 mass-selective detector (MSD). The GC was equipped with a split/splitless injector and a DB-5MS fused silica capillary column (30m 3 0.25mm i.d., 0.25mm film thickness). The GC oven temperature was programmed to increase from 50uC (2 min) to 120uC at 15uC min 21 and then to 300uC at 5uC min 21 with a final isotherm hold at 300uC for 16 min. Helium was used as the carrier gas at a flow rate of 1.0mL min 21 . The sample was injected on a splitless mode with the injector temperature at 280uC. The mass spectrometer was operated on the electron impact (EI) mode at 70eV and scanned from 50 to 650 Da. Data were processed with the Chemstation software. Individual compounds were identified by comparison of mass spectra with those of authenticated standards or literature data. Field and laboratory blank filters were treated as the real samples for quality assurance. Target compounds reported here were not detected in the blanks. Relative standard deviations of the concentrations based on duplicate analysis were ,10%.