Widespread release of old carbon across the Siberian Arctic echoed by its large rivers

Over decadal-centennial timescales, only a few mechanisms in the carbon-climate system could cause a massive net redistribution of carbon from land and ocean systems to the atmosphere in response to climate warming. The largest such climate-vulnerable carbon pool is the old organic carbon (OC) stored in Arctic permafrost (perennially frozen) soils. Climate warming, both predicted and now observed to be the strongest globally in the Eurasian Arctic and Alaska, causes thaw-release of old permafrost carbon from local tundra sites. However, a central challenge for the assessment of the general vulnerability of this old OC pool is to deduce any signal integrating its release over larger scales. Here we examine radiocarbon measurements of molecular soil markers exported by the five Great Russian-Arctic Rivers (Ob, Yenisey, Lena, Indigirka and Kolyma), employed as natural integrators of carbon release processes in their watersheds. The signals held in estuarine surface sediments revealed that average radiocarbon ages of n-alkanes increased east-to-west from 6400 yr BP in Kolyma to 11 400 yr BP in Ob. This is consistent with westwards trends of both warmer climate and more degraded organic matter as indicated by the ratio of high molecular weight (HMW)n-alkanoic acids to HMWnalkanes. The dynamics of Siberian permafrost can thus be probed via the molecular-radiocarbon signal as carried by Arctic rivers. Old permafrost carbon is at present vulnerable to mobilization over continental scales. Climate-induced changes in the radiocarbon fingerprint of released permafrost carbon will likely depend on changes in both permafrost coverage and Arctic soil hydraulics. Correspondence to:̈ O. Gustafsson (orjan.gustafsson@itm.su.se)

permafrost carbon release is provided by the realization that permafrost degradation mobilizes thawed-out organic matter (OM) to streams and rivers, ultimately emptying into the coastal Arctic Ocean Guo et al., 2007;van Dongen et al., 2008;Frey and McClelland, 2009). Hence, to overcome the heterogeneity and upscaling challenges posed by the Arctic landscape mosaic, we here examine the 5 five Great Russian-Arctic Rivers (GRARs; Ob, Yenisey, Lena, Indigirka and Kolyma; Fig. 1a), extended by the westward neighboring Kalix River draining sub-Arctic Scandinavia (Fig. 1a), as natural integrators and a means to study the carbon release processes in their watersheds (Bianchi and Allison, 2009). Different riverine carbon forms trace different components of terrestrial OM. Previ- 10 ous studies have reported on the 14 C signal of dissolved OC (DOC) in the GRARs, with generally young 14 C ages in the range of a few hundred years (Benner et al., 2004;Guo and Macdonald, 2006;Neff et al., 2006;Guo et al., 2007;Raymond et al., 2007), reflecting that this component stems from fresh plant litter and thus traces vegetation dynamics in the drainage basin. In contrast, detailed radiocarbon studies 15 on Arctic soil-leaching and river-release of OM imply that permafrost thawing is predominantly manifested in the age of the particulate OC (POC) form in Arctic rivers with reported 14 C ages of several thousand years (Goñi et al., 2005;Guo and Macdonald, 2006;Guo et al., 2007;Vonk et al., 2010a). Unfortunately, bulk POC may stem from multiple sources including peat, mineral soils and plankton (Schuur et al., 2008; Kolyma are predominantly located in the continuous permafrost region with a drier and colder climate and vast amounts of deciduous forest. This contrasts with the Ob, Yenisey and Kalix watersheds, located in the discontinuous or sporadic permafrost zone, which is wetter and thus holds more extensive peat and wetlands (Kremenetski et al., 2003;Tarnocai et al., 2009).
The Arctic surface sediments were collected in 2004 and 2005 using the H/V Ivan Kireev (Archangelsk) from the estuaries of the five GRARs during the second and third Russia-United States cruises. The Kalix estuary sediments were obtained in 2005 using the research vessel "KBV005" from the Umeå Marine Research Center (UMF, Norrbyn, Sweden). The complete sampling details for all six locations are described 15 elsewhere Vonk et al., 2008). The obtained sediments were kept frozen at −20 • C until processed in the laboratory.

Extraction and fractionation
The bulk sediments were thoroughly mixed prior to sub sampling and small amounts of 20 material were used for bulk radiocarbon analyses. Sub samples, typically between 70-160 g, were freeze dried, grinded, solvent extracted, purified and separated into lipid fractions using column chromatography. Aliquots of the hydrocarbon and acid fractions were then analyzed using gas chromatography/mass spectrometry (GC/MS

Preparative capillary gas chromatography and compound-specific radiocarbon analysis
Individual HMW n-alkanoic methyl esters and HMW n-alkanes were isolated from the purified extracts with preparative capillary gas chromatography (pcGC) (Supplement and Fig. S1).

Results and discussion
3.1 Sources of organic matter 20 The estuarine surface sediments provide integrated diagnostics of land-based carbon release as demonstrated by terrestrial signatures of stable carbon isotopes of total OC (Table 1) and the dominance of terrestrial HMW n-alkanes over marine lowmolecular-weight n-alkanes by an average factor 40 (Fig. 2a) the western Eurasian Arctic, that holds the world's largest peatland (Kremenetski et al., 2003).

Degradation status of remobilized organic matter
The degradation status of mobilized SOC is an important property that relates to its propensity to be converted microbially to greenhouse gases. The degradation of OM 5 is associated with a loss of functional groups. The ratio of HMW n-alkanoic acids to HMW n-alkanes is therefore a proxy for degradation status (e.g., Goñi et al., 2005;van Dongen et al., 2008;Vonk et al., 2010a). Here, the increasing contribution of HMW n-alkanoic acids relative to HMW n-alkanes from west to east among the GRARs ( Fig. 2c; Table 1) follows the continent-scale eastward trend of colder climate and more 10 extensive permafrost coverage (Fig. 1a). Simultaneously, an increasing 14 C age of released bulk OC eastward ( Fig. 2d) presumably reflects that more extensive permafrost coverage yields higher reservoir ages. Hence, these observations indicate a continentscale trend of older yet less degraded terrestrial OM being released toward the eastern reaches of Siberia. This combines with previous molecular-based findings from 15 Russian-Arctic Rivers (Guo et al., 2004;van Dongen et al., 2008;Vonk et al., 2010a) to suggest that fresh biomass produced during the short vegetative season is preserved with little alteration in the "deep freezer" of the East Siberian Arctic. The more degraded nature of the OM fluvially released in western Siberia may indicate what may occur with the deep-frozen OM in the east if large-scale thawing were to take place in its water- 20 sheds, now experiencing the largest temperature increase on Earth (Richter-Menge and Overland, 2010).

Age and origin of organic matter
The compound-specific radiocarbon signal of HMW n-alkanes provides a more distinct source-specific picture of mobilized old SOC. permafrost zone (Table 1), were 6000-6800 yr BP (Fig. 2d). Moving west into watersheds with more discontinuous permafrost, the values increased systematically from 8600 yr BP for Yenisey via 11 400 yr BP for Ob to 13 600 yr BP for Kalix (Fig. 2d, Table 2). This ubiquitously depleted 14 C signal demonstrates for the first time that old SOC is now leaking out from across the entire Eurasian Arctic region. Probing this 5 riverborne molecular radiocarbon signal over time thus offers the possibility to monitor climate-warming induced changes in large-scale releases of permafrost carbon.

Continental scale trends in carbon release
While there is close agreement in 14 C signal between bulk OC and HMW n-alkanes in East Siberia, there is an increasing 14 C age fractionation moving westward (Fig. 2d).

10
What system processes give rise to these dichotomous geospatial trends? Fractionating contributions from either planktonic or petrogenic sources are ruled out based on molecular and isotopic compositions (Supplement Text). We hypothesize that the age offset between bulk carbon and molecular SOC markers in the western watersheds, but absence of such an age offset in the east, is reflecting their differences in permafrost 15 characteristics, associated hydrology and resulting carbon releases. A proposed consequence of permafrost thawing is transition from surface-water dominated transport toward groundwater dominated transport with uncertain but potentially substantial implications for fluvial release of permafrost carbon (Frey et al., 2007;Bense et al., 2009;Frey and McClelland, 2009;Lyon and Destouni, 2010). We suggest that the west-east 20 offset between bulk and molecular 14 C signals combined with the continent-scale trend of younger 14 C age of mobilized molecular SOC markers eastward is a manifestation of the biogeochemical implication of differing hydraulic pathways imposed by differences in permafrost distribution ( Fig. 1b and  In contrast, the radiocarbon ages of the HMW n-alkanes, tracing the permafrost SOC (Guo and Macdonald, 2006;Guo et al., 2007), place constraints on wherefrom in the system soil carbon is mobilized. The comparable ages of the river-integrated molecular SOC markers and the peat basal ages in the western catchments suggest that Further support for a role of a deep conduit in the mobilization of old tundra SOC over West Siberia is provided by a six-fold elevation of mineral-weathered inorganic solutes in permafrost-free compared to permafrost-influenced watersheds (Frey et al., 2007). Mobilization of old permafrost OC can also occur more abruptly through thaw slumping caused by thermokarst development, mostly confined to areas of discontinuous  Table 1). Here, the water movement occurs largely in the thin active layer (seasonal thaw) above the permafrost table (Fig. 1c). The spring flood is likely to transport material mostly 25 from the surface of the underlying frozen vegetation (Frey and McClelland, 2009), but as the active layer deepens throughout the summer, the zone of water movement is in the lower reaches of the active layer. The only mechanism whereby any deeper SOC can be fluvially released is via thermokarst, river bank erosion and permafrost cracks. Introduction Hence, we hypothesize that the younger, yet also old, 14 C ages of the recalcitrant SOC of the eastern GRARs (6000-6800 14 C yr), along with a smaller offset (≤1000 yr) between bulk OC and molecular soil markers, largely reflect thaw release of permafrost carbon at the bottom of the active layer (Fig. 1c). The effect of permafrost thaw on old carbon release and net carbon exchange from tundra,  Ingri et al. (2005), GRAR data from Stein and Macdonald (2004) e given as % continuous; % discontinuous; % non-permafrost calculated from Walker (1998) and estimated from Johansson et al. (2006) f TOC, total organic carbon g data from van Dongen et al. (2008) and Vonk et al. (2008), Kalix data are from stations with sample ID C and D h TAR n−alkane , ratio of terrigenous (sum of C 27 , C 29 and C 31 ) to aquatic (sum of C 17 and C 19 ) n-alkanes i ratio of high-molecular weight (HMW; sum of C 20 -C 30 ) n-alkanoic acids to high-molecular weight (sum of C 20 -C 32 ) n-alkanes j according to Vonk et al. (2009)   Carbon isotopic composition and age of bulk surface sedimentary organic carbon, high-molecular-weight (HMW) n-alkanes (C 27 +C 29 +C 31 ) and HMW n-alkanoic acids (C 24 +C 26 +C 28 ) in surface sediment of Eurasian Arctic River Estuaries. δ 13 C and ∆ 14 C results are given in per mil measured relatively to VPDB and NBS Oxalic Acid, respectively. For 14 C isotope analysis the results are also presented as Fm δ 13 C corr (relative to NBS Oxalic Acid I).

Conclusions and broad scale implications
Sample ID δ 13 C (‰) a ∆ 14 C (‰) 14  −32.9 −532 ± 9 0.471 ± 0.009 6000 ± 150 a error of the δ 13 C measurement is ±0.1‰ b Fm is fraction of modern c yr BP is years before present d data obtained from Vonk et al. (2010); average values of sediments at stations with sample ID C and D