The Arctic Ocean is undergoing rapid loss in ice cover with yet unknown consequences for the cycling of organic material. Here we examine persistent terrigenous (land-based) alkane and polycyclic aromatic hydrocarbons with vascular plant, combustion, and petrogenic sources in seven cores collected from all major basins of the Arctic Ocean for insight as to how organic cycling at the Last Glacial Maximum compares to the present day. We find only modest changes between the glacial and postglacial sediments for atmospherically transported hydrocarbon biomarkers, demonstrating that glacial sea ice was not a barrier to atmospheric inputs. In stark contrast, particle-associated biomarkers were captured strongly at basin edges during the glacial period and much more evenly transported across basins during the postglacial period. Evidently the capture of organic matter shifted from the slopes to the shelves as the latter flooded during the Holocene, and the Transpolar Drift and Beaufort Gyre evolved from minor carriers of plant detritus from the glacial ocean margins to major modern transporters of shelf sediment to the basins. This suggests that changes in organic transport currently accompanying the loss of sea ice are likely to be very different from those that occurred at the end of the last glacial period.
 Recent dramatic changes observed in the Arctic Ocean's ice climate [Stroeve et al., 2007] have stimulated interest in Arctic Ocean processes that must have accompanied the large changes in sea level, ice and glacier cover since the Last Glacial Maximum (LGM) [Stein and Macdonald, 2004; Stein, 2008]. During the LGM the Arctic Ocean had heavy, perennial sea ice cover, low amounts of meltwater or riverine inputs and dry shelves covered with glaciers and vast, arid grasslands (Figure 1) [Nørgaard-Pedersen et al., 1998; Polyak et al., 2004; Svendsen et al., 2004; Sher et al., 2005; Darby and Zimmerman, 2008; Darby, 2008]. The low sediment and ice-rafted debris deposition rates during the LGM suggest a dramatic reduction in ice rafting, with a much reduced Transpolar Drift (TPD) and Beaufort Gyre (BG), and transport of terrigenous material to the basin edges primarily by turbidity currents and glacier activity at the shelf edges [Stein and Macdonald, 2004; Backman et al., 2004; Stein, 2008]. In contrast, the present-day Arctic Ocean has ∼120 m greater sea level, input from the Pacific Ocean, and the highest proportion of terrestrial input of freshwater and organic matter of any ocean [Dittmar and Kattner, 2003]. The current massive riverine discharge (3120 × 109 m3/a of water and ∼250 × 106 t/a of suspended solids) onto the broad shelves of the Arctic Ocean drives the export of terrigenous particulate frozen into sea ice to the central basins via the BG and TPD [Rigor et al., 2002; Stein and Macdonald, 2004; Stein, 2008]. The BG plays a major role in feeding particulate into the TPD, particularly during a positive Arctic Oscillation when the TPD swings closer to North America [Rigor et al., 2002]. Longer term (century-to-millennial-scale) shifts of the TPD in the Holocene also have been documented using driftwood delivery patterns to the Canadian Arctic archipelago [Dyke et al., 1997]. It is likely that the TPD, and possibly the BG, have existed in some form for the last ∼14 Ma, with the TPD providing a major route for the export of ice through Fram Strait [Stokes et al., 2005; Darby, 2008].
 PAH (polycyclic aromatic hydrocarbon) and alkane hydrocarbons with vascular plant and petrogenic (coal, bitumen) sources are more stable than their oxygenated precursors (e.g., alcohols, fatty acids, and sterols) and have been widely applied as historical proxies of terrigenous and petrogenic inputs in lacustrine sediments [e.g., Meyers and Ishiwatari, 1993; Yunker and Macdonald, 2003]. Multiple ring vascular plant and petrogenic biomarkers have seen only limited application in the Arctic Ocean [Yunker et al., 2002a; Weller and Stein, 2008; Yamamoto et al., 2008] and there are few (if any) geochemical studies of PAHs in ocean basin sediments. One reason for the paucity of deep ocean studies using PAHs can be ascribed to early work on the global distribution of PAHs, which concluded that anthropogenic activity is the major source of PAHs in ocean sediments, including the deep basins [Laflamme and Hites, 1978]. It is surprising that this early interpretation has endured so long since there is now a broad appreciation of the value of PAH, hopane, and sterane hydrocarbons as tracers as a result of detailed work following the Exxon Valdez oil spill [e.g., Bence et al., 1996; Short et al., 2007] and PAHs can be persistent enough to retain a source signature over geological time scales [Venkatesan and Dahl, 1989; Killops and Massoud, 1992].
 Our previous work has established the presence of a strong natural PAH, hopane and sterane signal in the Mackenzie River/Beaufort Sea [Yunker et al., 1991, 1993, 2002a] that differs from the Barents Sea [Yunker et al., 1996]. With an expanded sample set covering an appreciable fraction of the Arctic Ocean, we can now apply persistent hydrocarbon tracers to other arctic shelves and the deep basins. Here we apply detailed hydrocarbon biomarker analyses to seven box cores collected on the Arctic Ocean Section 1994 (AOS-94) grand transect of the major Arctic basins (Figure 1). These samples were collected before the dramatic changes in ice climate manifested themselves during the past few years [Macdonald, 1996; Stroeve et al., 2007; Stein, 2008]. The application of proxies with sufficient stability and specificity to this sample set allows us to address the key question of what do terrigenous biomarkers tell us about change in the margin-basin processes from the LGM to the present.
 Sediments from the Canadian, Eurasian and Greenland Sea basins (Figure 1) were collected in 1994 at depths of 2265 to 4230 m with a Pouliot box corer (0.06 m2, 50–55 cm deep) using rigorous sampling protocols for trace hydrocarbon and organic contaminant analysis. Samples included two cores from the Canada Basin (station 11 at 76.650°N, 173.383°W, 2265 m and station 18 at 80.150°N, 173.333°W, 2860 m), one each from the Makarov Basin (station 26 at 84.067°N, 175.083°E, 3130 m) and Amundsen Basin (station 35 at 90.000°N, 4230 m), two from the Nansen Basin (station 36 at 85.733°N, 37.750°E, 3605 m and station 37 at 84.250°N, 35.083°E, 4000 m) and one from the Greenland Sea Basin (station 39 at 75.000°N, 6.050°W, 3550 m). Immediately after collection, sediment cores were sectioned in the ship's laboratory using stainless steel tools that were cleaned between samples (tap water, distilled water, acetone). One wall of the removable stainless steel liner was lowered to allow subsampling of each core with minimal disturbance and sediment from the outer 5 cm of the box was discarded. Subsamples for organic determinations were placed in stringently precleaned (detergent rinse, 4-h soak in 2% RBS, distilled water rinse and baked overnight at 350°C) glass Mason jars with Teflonlined lids and frozen [Yunker et al., 1999].
 Sediments were combined with anhydrous sodium sulphate, spiked with perdeuterated surrogate standards (four n-alkanes and 10 PAHs covering the full range of compounds quantified), extracted by elution in a glass column (3 cm i.d.) with methanol (100 mL) followed by dichloromethane (250 mL) and concentrated using Kuderna-Danish apparatus. Following separation into fractions onto a silica gel column (Biosil, 10 g; 5% deactivated) using pentane (25–35 mL, alkane fraction) and then dichloromethane (30–35 mL, PAHs), samples were spiked with recovery standards (three PAH surrogates) and analyzed using selective ion monitoring GC/MS using a minimum of two ions per analyte. Full method details are provided by Yunker et al. .
 Samples from the 0–1, 3–4, 10–12, 20–25 and bottom ∼5 cm core sections (core bottom mean 46 cm; range 40–50 cm) were analyzed for alkanes and PAHs (Figures 2a–2c and 2e) by Axys Analytical Ltd. of Sidney, B.C., Canada, and for steranes, tricyclic terpanes and hopanes (Figure 2d) by the Geological Survey of Canada laboratories in Calgary, Alberta, Canada. Perdeuterated n-alkane and PAH standards were used for quantification in all cases. Sample duplicates and recoveries from spiked samples were within the QA/QC criteria of ±20% and 70–130%, respectively. Procedural blanks demonstrated no hydrocarbon interferences and samples exhibited none of the contamination typical of contact with plastics [Grosjean and Logan, 2007; Yamamoto et al., 2008].
 Alkanes quantified were the n-alkanes from C13–C36, and isoprenoids from 2,6-dimethylundecane to phytane. Parent PAHs quantified were those of low mass (molecular weight 128–228: naphthalene, acenaphthylene, acenaphthene, fluorene, phenanthrene, anthracene, fluoranthene, acephenanthrylene, pyrene, benz[a]anthracene, chrysene) and those of high mass (molecular weight 252–278: benzo[b/j/k]fluoranthene, benzo[e]pyrene, benzo[a]pyrene, indeno[7,1,2,3-cdef]chrysene, indeno[1,2,3-cd]pyrene, benzo[ghi]perylene, anthanthrene, dibenz[a,j]anthracene, dibenz[a,h]anthracene, pentaphene, benzo[b]chrysene, picene) ranges plus perylene. Saturated petroleum biomarkers quantified include a comprehensive suite of tricyclic terpanes (C23-C25 tricyclics plus C24 tetracyclic), diagenetic hopanes (C27 18α(H)- and 17α(H)-hopanes, C29–C33 17α(H),21β(H)-hopanes (R and S for C31–C33) and C29 and C31 17β(H),21α(H)-moretanes), and C27–C30 steranes and diasteranes, while the bacterial hopenes include trisnorhop-17(21)-ene, neohop-13(18)-ene and diploptene. Octahydrochrysenes (OHC) quantified were the 3,3,7,12a- and 3,4,7,12a-isomers of 1,2,3,4,4a,11,12,12a-octahydrochrysene, plus one other isomer eluting after the 3,4,7,12a-isomer, tetrahydrochrysenes (THC) were the 3,4,7- and 3,3,7-isomers of 1,2,3,4-tetrahydrochrysene and the tetrahydropicenes (THP) were the 1,2,9- and 2,2,9-isomers of 1,2,3,4-tetrahydropicene.
 Analytical replicates were averaged and undetectable values were replaced by a random number between zero and the limit of detection before Principal Components Analysis (PCA). Samples were normalized to the concentration total and the centered log ratio (or log contrast) transformation (division by the geometric mean of the concentration-normalized sample followed by log transformation) was then applied to this compositional data set to produce a PCA data set that was unaffected by negative bias or closure. Data were then autoscaled (biomarker variables were scaled by subtracting the variable mean and dividing by the variable standard deviation) to yield a data set where each variable had equal weight and the average concentration and concentration total were identical for every sample [Yunker et al., 2005].
 Dates for sediments were inferred from dates published for cores collected either along or close to the AOS-94 transect (Figure 1). Sediment core dates used station 92BC17 for station 11, 94BC16 for 18, 94BC20 for 26, PS2177–1 for 35, PS2195–4 for 36, OD-041-4 for 37 and PS1906–1/2 for 39 [Darby et al., 1997; Nørgaard-Pedersen et al., 1998, 2003]. Obvious changes in sedimentation rate within individual cores were interpreted as core segments with different age versus core depth linear regression slopes (Figure 1, inset). Hydrocarbon fluxes (ng/cm2/ka) for each core segment were obtained by multiplying the hydrocarbon concentration by the cumulative mass (g/cm2; calculated from the porosity profiles of each core) and dividing by the age of each core segment in ka.
3. Results and Discussion
3.1. Age Chronology and Biomarker Fluxes
 The dates for sediments inferred from surrogate cores [Darby et al., 1997; Nørgaard-Pedersen et al., 1998, 2003] reveal a coherent chronology (Figure 1, inset). Surrogate cores from the Arctic Ocean basin edges bordering the continental slopes (stations 11 and 37) and Greenland Sea (station 39) exhibit sedimentation rates of 1.5 to 2.1 cm/ka and a linear (or nearly linear) increase in age with core depth. Cores from the central basins (stations 18, 26, 35 and 36) exhibit a decrease in sedimentation rate at the approximate base of the Holocene (∼10 cm) with rates between 0.70 and 1.6 cm/ka for postglacial sediments and 0.25 and 0.40 cm/ka for glacial-deglacial sediments. This chronology, with a postglacial sedimentation rate of ∼1 cm/ka and rates lower than ∼1 cm/ka for glacial stages, is in agreement with the current consensus on Arctic Ocean basin sedimentation rates [Backman et al., 2004].
 To compare Arctic Ocean basin sediments from the last glacial-deglacial period (referred to as glacial) with the postglacial (modern) period, we have used hydrocarbon fluxes for core sections that are definitely pre-Holocene (the 20–25 and bottom ∼5 cm sections; >12 ka) for glacial, and primarily in the surface mixed layer (SML; the 0–1 and 3–4 cm sections; ∼5 ka) for postglacial (Figure 2). Analytical replicates have been averaged to give one flux measurement per core segment, and undetectable concentrations have been taken as a zero flux. Including or not including the 10–12 cm section with the 20–25 and bottom ∼5 cm sections had little effect on the calculated glacial fluxes, and the 10–12 cm section has been used for the hopane, hopene and tricyclic terpane glacial fluxes for stations 11, 18 and 26 because of the absence of data for deeper sections. On the basis of 210Pb and 137Cs profiles and the penetration depth of 137Cs and 239,240Pu bomb radionuclides [Gobeil et al., 1999; Smith et al., 2003], the SML comprised the top 2–3 cm which, given the sedimentation rates, comprises sediment deposited during the past ∼5 ka.
 The hydrocarbon measurements used to calculate these fluxes show little concentration variation within the glacial (>20 cm) and postglacial (0–1 and 3–4 cm) segments of the cores (e.g., relative standard deviation averages are 18% and 10%, respectively, for the n-C27, n-C29 and n-C31 alkanes; not shown). Stations 11 and 37 exhibit generally uniform concentrations throughout the cores, while cores from other stations have low concentrations >20 cm, generally higher concentrations 10–12 cm, and the highest concentrations in the two near-surface sections (not shown). Hydrocarbon fluxes show much greater variation (average relative standard variation for all parameters in Figure 2 is 51% for glacial and 82% for postglacial segments), because of large differences in porosity between core segments (average 1.8 × increase between the 20–25 and bottom ∼5 cm segments and 8.2 × between the 0–1 and 3–4 cm segments). Nevertheless, comparisons based on single core intervals (e.g., 20–25 cm for glacial and 0–1 cm for postglacial) exhibit the same trends as in Figure 2, and the averages do provide a valid comparison of glacial and postglacial hydrocarbon fluxes.
3.2. Hydrocarbon Sources and Transport
 The persistent hydrocarbon biomarker indicators presented in Figure 2 derive from vascular plant (n-alkanes, hydrochrysenes and hydropicenes), combustion (mainly parent PAH) and petrogenic (mainly alkyl PAH, tricyclic terpanes and hopanes) sources. Differences in volatility and particle association between hydrocarbon groups determine whether they are delivered to the Arctic Ocean basins by windborne aerosols, by sediment particulates transported by ice or turbidity currents, or by both aerosols and sediments. The constant presence of ice-rafted debris in the central Arctic Ocean since ∼200 ka [Spielhagen et al., 2004; Yamamoto et al., 2008] indicates that the heavy glacial sea ice pack is a semipermeable barrier capable of shedding its particulate load of aerosols and sediments [cf. Andrews, 2000]. Given the slow sediment accumulation rates (∼1 cm/ka; Figure 1, inset) [Darby et al., 1997; Nørgaard-Pedersen et al., 1998, 2003; Backman et al., 2004] and ∼50 cm core depths, the biomarker hydrocarbons must be largely protected from degradation in stable matrices such as organic rich shales, wood and grass chars, bitumens and coals [Yunker et al., 2002b; Yunker and Macdonald, 2003].
 The plant n-alkanes transport in aerosols as waxes and smoke from biomass burning [Oros and Simoneit, 2001a, 2001b; Oros et al., 2006; Bendle et al., 2007], and in the water column as plant detritus in suspended sediment delivered to the shelves by rivers and shoreline erosion [Yunker et al., 2002a]. The n-C27, n-C29 and n-C31 alkanes (Figure 2a) are well established as persistent tracers of vascular plants in marine sediments [Yunker et al., 2005; Fahl and Stein, 2007; Yamamoto et al., 2008]. The vascular plant alkane signal also can often be preserved in immature coals, which can exhibit a marked dominance of the higher odd alkanes with well defined odd-even alkane predominance [Kalkreuth et al., 1998; Tuo et al., 2003]. In all Arctic Ocean sediment samples the C23–C33 odd n-alkanes (with n-C27, n-C29 and n-C31 as the major components) dominate a profile with quantifiable n-alkanes from C13–C36, isoprenoids from 2,6-dimethylundecane to phytane and no unresolved complex mixture (not shown).
 Combustion PAH emissions from forest and prairie fires are dominated by parent PAHs [Yunker et al., 2002b] (Figure 2b), which transport atmospherically and by sediment and ice. Smaller combustion particulates transport easily in air and water, while larger particles, like wood chars, undergo limited atmospheric transport but can move extended distances with ice or water [Yunker et al., 2002b]. In particular, the chars provide a solid matrix that traps and stabilizes PAHs, allowing a combustion PAH fingerprint to survive over geological time scales [Venkatesan and Dahl, 1989]. In Arctic Ocean basin sediments, the reactive and low molecular weight combustion derived PAHs have been degraded in the SML (e.g., anthracene and benz[a]anthracene are low and benzo[a]pyrene is almost entirely depleted), likely because of the very long exposure to oxic conditions, and the PAH composition is unlikely to change further below the mixed layer.
 The vast glacial grasslands that thrived on the dry Siberian shelves (Figure 1) during the Pleistocene provide a hydrocarbon source that is unique to the Arctic in glacial periods. Continuous pollen, plant macrofossil and fossil insect records from >46 to 12.5 ka at a well preserved site on the Laptev Sea near the present Lena River delta indicate that this ecosystem consisted of grass-sedge dominated vegetation in a very dry to arid, strongly continental environment [Sher et al., 2005]. Summer growing season temperatures for most of this time were higher than those observed today (even at the LGM, the coldest period in the record), while winter temperatures were constantly much lower than at present. 14C ages for mammal bones demonstrate that mammoths and horses foraged to at least 75°N throughout the period (including the LGM) on the Taimyr Peninsula and New Siberian Islands (Figure 1), suggesting that the grasslands covered all the shelf lowlands during the Pleistocene [Sher et al., 2005]. Accordingly, these grasslands should supply plant n-alkanes from grass detritus across most of the dry eastern Siberian shelves and, given the arid conditions and warm temperatures, grassland fires are likely to provide an additional source of n-alkanes and parent PAHs as well (Figures 2a and 2b). To our knowledge, there are no published charcoal records in sediment (or better, ice cores) from the Arctic to support our inference of grassland fires in this arid environment. Nevertheless, where sedimentary records do exist, elevated charcoal levels are observed worldwide for the glacial period where the glacial climate was more arid than the present [Power et al., 2008]. Given the low sedimentation rates in basin locations and bioturbation in the SML, episodic fires (years to decades) would suffice to contribute to the sedimentary record.
 For petrogenic alkyl PAHs (Figure 2c), free oil from oil seeps would rapidly disperse and degrade, and is unlikely to be detectable far from its source. More likely, organic rich shales, bitumens and coals provide the stable matrix required for the PAH and saturated petroleum biomarkers (Figures 2c and 2d) to transport long distances by sediment and ice and be preserved. Hydrocarbons in circum-Arctic petroleum basins close to the coast or adjacent to major rivers are predominately derived from organic-rich material deposited under shallow marine or deltaic regimes [Leith et al., 1992; Peters et al., 2007; Gautier et al., 2009]. Deposition of high organic carbon (OC) sediments became increasingly seasonal as the Arctic Ocean shifted to high paleolatitudes during the late Mesozoic, and became predominantly focused at the ocean margins in the Cretaceous and Tertiary [Leith et al., 1992]. Deposition of OC rich deposits capable of hydrocarbon generation likely ceased in the late Eocene [Snowdon et al., 2004; Weller and Stein, 2008; Stein, 2008], before seafloor spreading created the Eurasian Basin. Accordingly, the margins (including the Lomonosov ridge) appear to be the only likely source of petrogenic material today.
 The Mackenzie River delivers petrogenic hydrocarbons in the form of shales or bitumens to the Beaufort Sea and Canadian Basin [Yunker et al., 2002a]. In contrast, the diagenetic hopanes are negligible in the five large Siberian rivers [van Dongen et al., 2008], and Yamamoto et al.  have concluded that petrogenic material in the Eurasian Basin is more likely to take the form of eroded coals. Schubert and Stein  cite a number of unpublished reports describing the presence of coal fragments in sediments of the Eurasian Basin and published work has documented coals in sediments in the eastern Arctic Ocean, Fram Strait and Greenland Sea, and the south central Canadian Arctic Archipelago [Bischof et al., 1990; Bischof and Darby, 1999]. Coals in the archipelago likely have a local source [Kalkreuth et al., 1998], while coal deposits from an area between the northern Taimyr Peninsula and the Lena River district are a possible source for the Eurasian Basin, Fram Strait and Greenland Sea [Bischof et al., 1990; Stein, 2008].
Figure 2 shows dramatically different patterns between hydrocarbon categories (compare downward), between basin edges and central basins (compare within each of Figures 2a–2e), and between core sections that are glacial (>12 ka) and postglacial (∼5 ka; compare left and right portions of Figures 2a–2e). Glacial fluxes of the plant alkanes (Figure 2a) suggest a relatively uniform transport across basins, mediated by aerosols, with markedly higher fluxes at the shelf edge stations 11 and 37 and at station 35 by the Lomonosov Ridge (Figure 1). The high-alkane fluxes with n-C31 predominant at station 11 imply input from grasses either as detritus or smoke [Meyers and Ishiwatari, 1993; Oros et al., 2006] from the nearby extensive East Siberian Sea arid, glacial grasslands [Sher et al., 2005]. For station 37, the increased flux of the three alkanes can be attributed to enhanced ice transport of soils/aerosols accumulated in glaciers on the Barents Sea shelf [Knies and Stein, 1998; Stein et al., 2001] and the influence of relatively warm and saline surface and near-surface water in the southwestern Eurasian Basin flaw leads during the LGM [Nørgaard-Pedersen et al., 2003; Spielhagen et al., 2004]. For station 35 near the Lomonosov Ridge, the enhanced alkane fluxes show a robust transport of plant material from the Laptev Sea implying a viable TPD pathway [Schubert and Stein, 1997] (Figure 1). In postglacial sediments, the n-alkanes exhibit relatively uniform distributions in which all three alkanes contribute equally, but almost all of the contrast between basin edges and central basins has disappeared. These distributions imply that transport from the sides of the basin in, for example, resuspended sediments, is less important than during the glacial period, but that ice and aerosol remain strong primary transport mechanisms.
 The elevated plant alkane fluxes (Figure 2a) and concentrations (not shown) at station 35 postglacial are consistent with the results of Schubert and Stein , who also observed elevated concentrations of these alkanes in the Lomonosov Ridge region relative to other locations in the Eurasian basin. While these alkane inputs were clearly related to vascular plant input, as evidenced by the pronounced odd-even predominance, the elevated alkane concentrations at the Lomonosov Ridge did not coincide with higher overall terrigenous organic matter supply, as suggested by hydrogen indices and C/N ratios [Schubert and Stein, 1996, 1997]. The authors concluded that the most probable source for the terrigenous alkanes locations was sea ice transport from the Laptev Sea and attributed the higher concentrations to enhanced sediment release from ice due to frequent leads and crushed sea ice generated by currents and ice drift in vicinity of Lomonosov Ridge. Unpublished alkane flux data from Schubert's thesis for three locations in the Nansen, Amundsen and Makarov basins report fluxes of 0.25–0.37 μg/cm2/ka in the last glacial period increasing to 0.57–0.78 μg/cm2/ka postglacial [Schubert, 1995]. By comparison, in this study fluxes at stations 18, 16 and 36 range from 1.2–2.1. μg/cm2/ka in the glacial period and 1.5–2.0 μg/cm2/ka postglacial.
 Parent PAHs (Figure 2b) differ substantially from the n-alkanes. High-mass parent PAHs (252–278) predominate somewhat in the glacial sediments, low mass (128–228) PAHs in postglacial sediments and all samples have only minor amounts of perylene (i.e., diagenetic production is minimal or absent) [Meyers and Ishiwatari, 1993; Yunker and Macdonald, 2003]. Glacial PAH fluxes are very high at the basin edges and drop precipitously in the central basins. Furthermore, station 35 along the Lomonosov Ridge shows no enhanced fluxes. Postglacial fluxes are slightly higher at the basin edges and along the Lomonosov Ridge, and the center of the Eurasian Basin is slightly higher than the Canadian Basin. As with the plant alkanes, distributions are consistent with ice and aerosol transport augmented by sediment resuspension at the margins. The small flux in the central basins suggests a constant aerosol input of combustion PAHs supplemented by inputs at the basin edges from the adjacent shelves.
 All of the particle-bound biomarkers (combustion, petrogenic or plant) are strongly constrained to the basin edges during the glacial period (compare Figures 2b through 2e). The leading candidate to explain the common distributions seen in all these indicators is the absence of water over shelves. Dry shelves then lead to the more efficient transport of soils from the edge directly to the basin slopes either by rivers or by erosion of soils and sediments along the exposed continental edge. Accordingly, although the compound classes have differing predominant sources, their transport must have involved a common component in soil erosion. The high parent PAH fluxes at station 11 could be augmented during the glacial by biomass burning in the East Siberian Shelf grasslands [Sher et al., 2005]. The absence of a glacial flux increase in PAH at station 35 suggests that the enhanced plant alkane fluxes in Figure 2a are due to transport of plant detritus by ice rather than sediment containing both alkanes and PAHs.
 Alkyl PAHs (Figures 2c and 2e), with predominant sources in organic-rich shales, bitumens, coals and vascular plant debris, have glacial distributions like the parent PAHs. OHC and THC at station 35 are exceptions, and are presumably transported in plant components by the TPD along with the n-alkanes (see above and compare Figures 2a and 2e). The similarity in glacial flux distributions for petrogenic and plant alkyl PAHs also indicates that there is very little aerosol transport of alkyl PAHs. Postglacial fluxes are higher at the edges and along the Lomonosov Ridge than in the central basins (Figures 2c and 2e) indicating supply from the shelves or land. Because these compounds are largely particle bound, the difference in fluxes between the central basins and basin edges likely reflects uniform transport of particulates to the basins by ice transport with enhanced inputs at the edges and beneath the TPD along the Lomonosov Ridge.
 For the petrogenic alkanes (Figure 2d), the diagenetic hopanes predominate over the tricyclic terpanes and usually have the highest fluxes, while the bacterial sourced hopenes are major constituents for a few Eurasian Basin stations (station 35 glacial, and stations 35 and 36 postglacial). Fluxes of hopanes and hopenes are minor relative to the alkyl PAHs for postglacial samples and glacial samples from stations 11, 37 and 39, while glacial samples from the central basins have fluxes of hopanes and hopenes that are 4–19 times higher than the alkyl PAHs (Figures 2c and 2d). These increased glacial fluxes of the petrogenic alkanes could derive from natural coal fires [Oros and Simoneit, 2000] or they may have petrogenic sources that are local or transported from the margins by ice. Postglacial fluxes of the petrogenic alkanes are higher in the Eurasian Basin than in the Canadian Basin. Higher fluxes of both petrogenic alkanes and PAHs in the Eurasian Basin postglacial are likely due to a combination of factors, including more sediment incorporation into ice in the Laptev Sea, the generally lighter ice conditions and greater melt rates in the Eurasian Basin, and a more direct path to the basins via the TPD [Rigor et al., 2002; Stein and Macdonald, 2004; Stein, 2008].
3.4. Composition Pattern Changes Between Glacial and Postglacial Sediments
 A principal components analysis provides decisive evidence that the change from glacial to postglacial regimes in the Arctic Ocean central basins is very different from the basin edges and Greenland Sea. The PCA model shows three distinct variable end-members (circled in Figure 3a), with the separation between alkyl PAHs and n-alkanes in PC1 (principle component 1) providing the major trend between samples and the higher-mass parent PAHs (mass 276 and 278) dominating PC2 (principle component 2). Sediments from the basin edges and Greenland Sea all project on the left side of axis center (Figure 3b), indicating a dominance of alkyl PAHs throughout the core (Figure 2c). Accordingly, these locations have received an uninterrupted supply of petrogenic material from glacial times continuing right through the Holocene. There are small differences in alkyl PAH content between the three cores but the proportion changes little, and the major determinant of sample projections is a decrease in the proportion of the combustion-related, high-mass parent PAHs toward the core surface (Figure 2b), implying a decrease in biomass burning with time (see following). In contrast, sediments from the central basins show large changes in composition between glacial and postglacial intervals as evidenced by the wide range in PC1 (Figure 3c). These projections indicate a glacial setting rich in n-alkanes, shifting to a modern setting rich in alkyl PAHs, indicating an increase in particle-bound transport from the edges during the Holocene. Station 36 is exceptional in showing a gradual, monotonic decrease in PC2 and a decrease in biomass-burning (mass 276 and 278) PAHs with time.
 The predominance of high-mass PAHs in glacial samples (Figures 2b and 3) is accompanied by an increase in the proportion of the thermodynamically less stable mass 276 and 278 PAHs that is consistent with biomass combustion [Yunker et al., 2002b]. PAH ratios of IP/(IP + BgP), where IP is indeno[1,2,3-cd]pyrene and BgP is benzo[ghi]perylene, show a small decrease between glacial (mean ± SE: 0.42 ± 0.03, n = 14; sections below 10 cm) and postglacial (0.37 ± 0.01, n = 12; sections to 4 cm) Arctic Ocean basin sediments, suggesting more biomass combustion under glacial regimes [Yunker et al., 2002b]. These IP/(IP + BgP) ratios are consistent with mixed petrogenic (bitumen, coal etc.) and biomass combustion sources, given that ratios of 0.17 ± 0.01 (n = 33) are observed for petrogenic-dominated sediments in the Mackenzie River/Beaufort Sea [Yunker et al., 2002a] and ratios >0.50 indicate biomass combustion [Yunker et al., 2002b]. Present-day liquid fossil fuel combustion (typical ratios 0.40–0.50) can be ruled out as a significant source because of the uniformity of composition of PAHs accumulated over millennia in near-surface core sections (Figure 3) and the lack of correlation of PAHs with the Pb isotope data measured on these cores by Gobeil et al. . In particular, PCA models including Pb isotope concentrations and parent PAHs, with and without the alkyl PAHs, indicate that Pb makes its greatest relative contribution for late glacial, 10–12 cm sediment samples (the deepest section with Pb data; not shown) [Gobeil et al., 2001]. PAH ratios for IC/(IC + BgP) and DjA/(DjA + DhA), where IC is indeno[7,1,2,3-cdef]chrysene, DjA is dibenz[a,j]anthracene and DhA is dibenz[a,h]anthracene, show clearer decreases of the thermodynamically less stable isomer [Yunker et al., 2002b] between glacial (0.28 ± 0.04, n = 11 and 0.56 ± 0.05, n = 14, respectively) and postglacial (0.16 ± 0.02, n = 11 and 0.41 ± 0.03, n = 12) sediments, strongly supporting the inference of higher proportions of biomass combustion PAHs pre-Holocene. The close proximity of the arid, glacial grasslands to the glacial shoreline makes grass fires a likely candidate for this biomass combustion.
 Throughout the core at station 35 (Figure 2d), the high flux of biogenic hopanes is accompanied by high ratios (0.26 ± 0.01, n = 4) of 28H/(28H + 30αβ), where 28H is 17α(H),21β(H)-28,30-bisnorhopane and 30αβ is 17α(H),21β(H)-hopane. Ratios in glacial and postglacial sediments from other basin locations are much lower (0.09 ± 0.01, n = 18). In a larger data set of Arctic sediment samples there are significant correlations (p ≪ 0.001, v ≈ 50) between the biogenic hopanes (C27–C31 17β(H),21β(H) substitution) and the three hopenes in Figure 2d and two C30 pentacyclic triterpenes (elution at 54% of the way between the 27β and 29βα hopanes, and coeluting with S-31αβ hopane) [Venkatesan and Kaplan, 1982; Yunker et al., 1993]. Both triterpanes are major constituents of Beaufort Sea coastal peat samples, suggesting that the triterpenes either may have a source in vascular plants or be related to bacterial decomposition of plant material. Elevated 28H/(28H + 30αβ) ratios are observed in sediments from the Laptev Sea edge (0.21 ± 0.07, n = 7) and in the Barents Sea (0.29 ± 0.10, n = 11), but the Laptev Sea is more likely as a source given the lower ratios at stations 36 and 37 and the flow of the TPD [Stein and Macdonald, 2004]. The C28 hopane can be affected by both maturity and source, but the relatively narrow range of maturities from basin sediments means that it is most likely a source indicator in the Arctic Ocean [Nytoft et al., 2000]. Taken together, the correlations for the biogenic hopanes and the 28H/(28H + 30αβ) ratios suggest that the biogenic hopanes at station 35 are transported in soils or sediments by the TPD along with OHC and THC (discussed in section 3.3).
 Petroleum maturity assessments can be inferred from the relative enrichment of the more thermally stable 20S isomer relative to the biologically derived 20R stereochemistry for the 5α(H),14α(H),17α(H) C29 steranes (29ααα), and the greater thermal stability of the 29αββ steranes relative to the 29ααα steranes [Farrimond et al., 1998]. The 20S/(20S + 20R) and αββ/(αββ + ααα) ratios for the C29 steranes both increase from glacial (0.27 ± 0.01 and 0.31 ± 0.01, respectively) to postglacial (0.34 ± 0.01 and 0.37 ± 0.01, n = 11 in all cases) sediments. All ratios are less than the equilibrium endpoints (∼0.55 and ∼0.60, respectively) indicating petroleum maturity [Farrimond et al., 1998] and petrogenic inputs are less mature overall in the glacial regime. Ratios of 27dβS/(27dβS + 27ααα), where 27dβS is the C27 diasterane 13β(H),17α(H),20(S)-cholestane and 27ααα is the 20R sterane, also echo the glacial (0.42 ± 0.02) to post-glacial (0.51 ± 0.02, n = 11 for both) increase.
 The increased maturity in postglacial samples implies an increase in the proportion of the more stable components, and consequently must represent a change in average maturity of source, not degradation. Given the degradation of the reactive and small ring, combustion derived PAHs in the SML, the proportional increase in low-mass parent PAHs and alkyl naphthalenes postglacial (Figures 2b and 2c) is likely due to additional low-mass PAHs introduced with the petrogenic material postglacial. Even if some PAH degradation is occurring below the mixed layer, PCA indicates that change in the low-mass parent and alkyl PAH compositions is a minor process (Figure 3a) that makes only a small contribution to the differences in core sample projections.
3.5. Glacial to Postglacial Switch in Sediment Transport
 The predominant changes occurring between glacial and postglacial regimes outlined above include (1) increased basin sedimentation rates (Figure 1, inset), (2) substantially increased fluxes of particle-associated hydrocarbons in the central basins (Figure 2), and (3) a clear separation in the composition of hydrocarbons depending on location (central basin versus basin edge; Figure 3). These changes collectively can be understood as consequences primarily of sea level rise, inundating the enormous shelves, and secondarily as a change in ice regime including lower thickness and higher mobility (Figure 4) [Stein and Macdonald, 2004].
 The most important consequence of flooding the Arctic shelves is a shift in the zone of dominant deposition of terrigenous material from the shelf edge/slope (Figure 4a) up onto the shelf (Figure 4b). In other words, the marginal filter [Lisitzin, 1995], which presently captures most riverine sediment in the inner shelf zone where freshwater and salt water mix (salinities of ∼2–10) and fine-grained suspended particulate accumulates rapidly [Stein et al., 2004; Stein, 2008], has been turned on over the shelf. The flooding of the shelves together with the shift in deposition then allows the suspension freezing process (sediment incorporation into sea ice through frazil and anchor ice) to operate midshelf on the Eurasian shelves during winter [Rachold et al., 2004, and references therein]. The subsequent offshore transport of this sediment laden sea ice facilitates the direct delivery of terrigenous material from the shelves to the basins by the TPD and BG.
 The reduced flux of hydrocarbons observed at the basin edges postglacial (Figure 2) is consistent with a dramatic reduction of sediment supply directly to the slopes, with reduced deposition at the basin edges and likely general importance of turbidite transport. Nevertheless, beyond an expected broad reduction of sediment supply directly to the slopes, the processes controlling sediment deposition on the Arctic Ocean continental slopes are still poorly understood. For example, cores collected from the continental slope of Alaska and the Chukchi Sea show extremely high sedimentation rates (>1.5 m/ka) at the edges of underwater canyons, likely in response to localized high currents [Darby et al., 2009]. For this reason the mid continental slope is shown as having low to high deposition in Figure 4b and, while not shown in Figure 4a, the uncertainty in mid slope deposition extends to the glacial period as well.
 The Arctic Ocean, which presently is almost 50% continental shelf, is more susceptible than any other ocean to change in the balance of its basins and marginal seas. While clearly the switch of the shelves from terrigenous to marine ecosystems has been an important change, other changes are likely to occur in a system that goes from a basin ocean to an ocean with 50% continental shelf.
 The trends in hydrocarbon biomarkers illustrate that the Arctic Ocean witnessed large changes in transport pathways for terrigenous material in response to the large rise in sea level, extensive deglaciation and loss of sea ice after the LGM (Figure 4). These changes differ strongly between central basins and basin edges. During the glacial regime, substantial fluxes of particle-associated biomarkers were introduced directly at the edges, but the central basins received only atmospheric inputs and plant material with small amounts of sediment/soil carried by a less vigorous TPD and BG. Combustion inputs to glacial sediments were dominated by biomass burning, and petrogenic material was of low maturity, likely coming from nearshore locations. Postglacial and glacial atmospheric inputs differ little in the central basins, but larger amounts of particle-associated biomarkers are found in the central basins postglacial, likely carried by ice. We infer that the absence of seawater on shelves was the primary mechanism restricting the delivery of particle-bound biomarkers to central Arctic Ocean sediments.
 During the glacial period, the arid grassland plains on the dry Siberian shelves [Sher et al., 2005] were evidenced by inputs of plant alkanes and biomass burning products to the basin edges, while shelf-based processes such as suspension freezing that presently entrain sediment into the TPD were turned off. Even with no shelf sediment entrainment into ice, the TPD, which was then confined to the basin, still carried plant material likely supplied by rivers directly to the edge of the continental slope. Not only would these rivers carry interior continental organic carbon, but they would also carry terrigenous products from extensive grassland plains that are now under water. Inundation of the shelves during the Holocene initiated a robust TPD (Figure 1) supplying substantial sediment from the Laptev and Kara seas [Stein and Macdonald, 2004], and less direct delivery of sediment to the basin edges. During this transition, the marginal filter [Lisitzin, 1995] evidently has been moved from the outer edge of the continental shelves, involving a horizontal displacement of up to 600 km in the case of the East Siberian Sea. We conclude that the changes in organic transport consequent to ongoing loss of summer sea-ice cover and the meltout of multiyear ice will differ greatly from those arising from deglaciation and sea level rise.
 We wish to acknowledge the assistance of the officers and crew of the CCGS Louis S. St. Laurent, David Paton for help with core sectioning, Sneh Achal and Marg Northcott for petroleum biomarker analyses, and Katherine Kaye, Rachel Robinson, Brenda Dunn, and Brian Fowler of Axys Analytical Services Ltd. for alkane and PAH analyses. The detailed constructive comments of reviewers Dennis Darby and Kristen Fahl improved the paper and are much appreciated. This work has been supported by the Northern Oil and Gas Action Program (NOGAP) and Canadian International Polar Year (IPY) programs through the Canadian Department of Fisheries and Oceans.