Preservation of terrestrial organic carbon in marine sediments offshore Taiwan: mountain building and atmospheric carbon dioxide sequestration

. Geological sequestration of atmospheric carbon dioxide (CO 2 ) can be achieved by the erosion of organic carbon (OC) from the terrestrial biosphere and its burial in long-lived marine sediments. Rivers on mountain islands of Oceania in the western Paciﬁc have very high rates of OC export to the ocean, yet its preservation o ﬀ shore remains poorly constrained. Here we use the OC content (C org , %), radiocarbon ( ∆ 14 C org ) and stable isotope ( δ 13 C org ) composition of sediments o ﬀ shore Taiwan to assess the fate of terrestrial OC, using surface, sub-surface and Holocene sediments. We account for rock-derived OC to assess the preservation of OC eroded from the terrestrial biosphere and the associated CO 2 sink during ﬂood discharges (hyperpycnal river plumes) and when river inputs are dispersed more widely (hypopycnal). The C org , ∆ 14 C org and δ 13 C org of marine sediment traps and cores indicate that during ﬂood discharges, terrestrial OC can be transferred e ﬃ ciently down submarine canyons to the deep ocean and accumulates o ﬀ shore with little evidence for terrestrial OC loss. In marine sediments fed by dispersive river inputs, the C org , ∆ 14 C org and δ 13 C org are consistent with mixing of terrestrial OC with marine OC and suggest that e ﬃ cient preservation of terrestrial OC ( > 70 %) is also associated with hypopycnal delivery. Sub-surface and Holocene sediments indicate that this preservation is long-lived on millennial timescales. Re-burial of rock-derived OC is pervasive. Our ﬁndings from Taiwan suggest that erosion and o ﬀ shore burial of OC from the terrestrial biosphere may sequester > 8 TgC yr − 1 across Oceania, a signiﬁcant geological CO 2 sink which requires better constraint. We postulate that mountain islands of Oceania provide a strong link between tectonic uplift and the carbon cycle, one moderated by the climatic variability which controls terrestrial OC delivery to the ocean.


Introduction
Photosynthesis sequesters CO 2 within living matter as organic carbon (OC).If a fraction of this productivity escapes respiratory consumption and oxidation, it represents a carbon sink that will reduce greenhouse gas concentrations and influence Earth's radiation energy balance (Sundquist, 1993;Stallard, 1998;Berner, 2006).On geological timescales, the Published by Copernicus Publications on behalf of the European Geosciences Union.
burial of OC in marine sediments is the second largest sink of atmospheric CO 2 after carbonate deposition formed from the products of continental silicate weathering (Gaillardet et al., 1999;Hayes et al., 1999, Burdige 2005).The erosion of terrestrial OC and its delivery by rivers to the ocean along with clastic sediments is thought to contribute approximately half of this oceanic OC burial flux (Schlunz and Schneider, 2000;Burdige, 2005;Blair and Aller, 2012), in part because the efficiency of OC burial is closely related to the accumulation rate of the accompanying sediment (Canfield, 1994;Burdige, 2005;Galy et al., 2007a).Therefore CO 2 -sequestration by OC burial may be sensitive to changes in tectonic and climatic conditions which regulate the erosion and transfer of clastic sediment and terrestrial OC by rivers (Dadson et al., 2003;Hilton et al., 2008Hilton et al., , 2012;;Milliman and Farnsworth, 2011), giving rise to feedbacks in the global carbon cycle (West et al., 2005) which are not represented in current models of the carbon cycle (Berner, 2006).
The Himalayan orogeny is thought to exert tectonic forcing on the carbon cycle (Gaillardet and Galy, 2008), sequestering 3.7 ± 0.4 TgC yr −1 through erosion of recently photosynthesized OC sourced from vegetation and soil in the terrestrial biosphere (OC biosphere ) and its preservation and burial in the distant Bengal Fan (France-Lanord and Derry, 1997; Galy et al., 2007a).On mountain islands of Oceania (Taiwan, Philippines, Indonesia, Papua New Guinea and New Zealand), where land-ocean linkages are strong, small mountain rivers drain a larger combined source area than the Himalaya (∼ 2.7 × 10 6 km 2 vs. 1.6 × 10 6 km 2 ).These rivers transport ∼ 7000 Tg yr −1 of clastic sediment (Milliman and Farnsworth, 2011) and an estimated 20-40 % of the global particulate OC flux to the oceans (Lyons et al., 2002).There, convergent plate margins have steep, high standing topography where erosion of OC biosphere occurs at very high rates (up to ∼ 70 MgC km −2 yr −1 ) and rivers can deliver particulate materials rapidly to the ocean across short floodplains (Dadson et al., 2003;Dadson et al., 2005;Scott et al., 2006;Hilton et al., 2008;Bass et al., 2011;Hilton et al., 2012).These conditions should be conducive to high rates of OC burial and higher OC preservation efficiencies than rivers draining passive margins (Galy et al., 2007a;Bianchi, 2011;Blair and Aller, 2012).However, unlike the Himalayan system, our understanding of the fate of OC biosphere offshore and the resultant CO 2 sequestration around these ocean islands remains incomplete (Eglinton, 2008).
Firstly, the incomplete understanding of OC biosphere burial reflects the challenge of accounting for "petrogenic" OC derived from sedimentary rocks (OC petro ) in river sediments and marine sediments (Blair et al., 2003) which can contribute significantly to the particulate load of mountain rivers (Kao and Liu, 1996;Komada et al., 2004;Leithold et al., 2006;Hilton et al., 2010;Clark et al., 2013).OC petro transfer and re-burial lengthens the residence time of OC in the lithosphere (Galy et al., 2008;Hilton et al., 2011), however the reburial of OC petro does not represent recent atmospheric CO 2 and so must be quantified separately.Secondly, it reflects the difficultly of assessing the range of delivery mechanisms to the ocean by mountain rivers.During floods, high suspended sediment concentrations (> 40 g L −1 ) can cause the density of the river outflow to surpass that of ambient seawater (hyperpycnal) and result in density currents transporting sediment down submarine canyons into the deep ocean (Mulder and Syvitski, 1995).Previous work has postulated that hyperpycnal discharges are essential for the efficient transfer of terrestrial OC into marine deposits offshore mountain islands (Kao et al., 2006;Hilton et al., 2008).However, large amounts of terrestrial OC and sediment are also delivered to the surface ocean by rivers in hypopycnal plumes with a density lower than seawater.Such plumes disperse fluvial materials over a larger region, which may result in re-suspension and reworking of terrestrial OC (Mulder and Syvitski, 1995;Dadson et al., 2005;Kao and Milliman, 2008).In analogy with passive margin settings, this may lower terrestrial OC burial efficiency (Aller et al., 1996;Aller, 1998;Aller and Blair, 2006;Sampere et al., 2008;Blair and Aller, 2012).
In order to shed some light on the fate of OC biosphere eroded from high standing ocean islands, we consider the mountain island of Taiwan (Fig. 1).In Taiwan, the rapid convergence of the Philippine Sea Plate with the Eurasian continental margin combines with a climate characterized by frequent tropical cyclones, driving high rates of fluvial sediment export to the ocean (Dadson et al., 2003;Kao and Milliman, 2008).Findings from Taiwan are of wider relevance because the steep mountain rivers draining this island are common throughout Oceania (Milliman and Farnsworth, 2011).Steep mountain rivers have short transit times (e.g.Hilton et al., 2008) and deliver most of their sediment loads (∼ 60-70 %) under hypopycnal conditions (Dadson et al., 2005;Kao and Milliman, 2008).However, Taiwan's rivers can also produce hyperpycnal plumes (Dadson et al., 2005) allowing us to study terrestrial OC transfer and preservation associated with both modes of fluvial delivery.To assess the offshore transfer of terrestrial OC and its preservation in marine sediments upon deposition, we have collected seafloor sediments and material from sediment traps from (i) the submarine Gaoping Canyon off Southwest Taiwan, which is prone to hyperpycnal inputs; and (ii) the Okinawa Trough, Taiwan Strait and the Gaoping Shelf where hypopycnal inputs are thought to be more important (Fig. 1).To assess longer-term terrestrial OC preservation and burial, we examine sub-surface sediments from these locations and Holocene sediments from the Okinawa Trough.Employing an established approach, we have measured the OC content, stable OC isotopes and radiocarbon content of OC to determine sources of OC in the sediments and to assess the preservation of terrestrial OC biosphere and OC petro offshore (e.g.Komada et al., 2004;Leithold et al., 2006;Galy et al., 2007aGaly et al., , 2008;;Hilton et al., 2008Hilton et al., , 2010;;Blair et al., 2010).The findings from Taiwan are placed in a regional context, and their implications for the global carbon cycle are discussed.S1).Marine samples fed by dispersive fluvial inputs (white symbols) were obtained from box core surface sediments, with the location of the longer piston core MD012403 indicated.Sediments were also acquired from within the Gaoping Canyon (black triangles), which is fed by hyperpycnal river plumes.Sediment traps (indicated by squares within symbols) were deployed in the Okinawa Trough (depths provided) and at 608 m in the channel of the Gaoping Canyon.

River suspended sediment samples
To characterize the composition of OC input into the ocean by Taiwanese rivers, suspended sediment samples were collected for this study from the primary rivers (Fig. 1) under common flow conditions as well as during tropical cycloneinduced floods, covering water discharges ranging from < 1 to ∼ 40 times the long-term average (Table S1), and then complemented with published data (Kao and Liu, 1996;Hilton et al., 2008).For each sample, a known volume (between 250 mL and 1 L) of river water was collected from the surface of the main river channel in a wide-mouthed plastic bottle thoroughly rinsed with river water.The sample was then filtered through 0.7 µm GF/F membrane filters and the contents were dried at 60 • C, weighed to determine total suspended sediment concentration (SSC, g L −1 ) and stored in sealed glass dishes.Water discharge (Q w , m 3 s −1 ) was measured by the Water Resources Agency, Taiwan, and reported here where available (Table S1).

Marine sediment samples
To assess the fate of terrestrial OC delivered to the ocean by hyperpycnal discharges, a sediment trap mooring was deployed at 608 m water depth, 42 m above the seafloor, in the submarine Gaoping Canyon, fed by the Gaoping River (Fig. 1).Full details of the collection methods can be found elsewhere (Huh et al., 2009;Liu et al., 2012Liu et al., , 2013)).Briefly, the sediment trap mooring and an upward-facing long-range acoustic Doppler profiler were moored in the canyon during the 2008 typhoon season.The mooring was configured with a non-sequential sediment trap, consisting of a conical funnel and core liner, and an intervalometer timer capable of inserting Teflon discs into the collected sediment as embedded time markers (Xu et al., 2010).During deployment, Typhoon Kalmaegi impacted Taiwan (17 July 2008) and the Teflon discs inserted before its landfall and after the flood waters had ceased allowed us to constrain sediment associated with the typhoon event (Liu et al., 2012).Conical sediment traps may result in conservative estimates of accumulation rate due to potential re-suspension of sediment in the funnel (Buesseler et al., 2007).Here we do not rely on accumulation rate data, but note that hydrodynamic sorting may result in a lower percentage of smaller, more buoyant particles present in the trap than the sediment plume during high current velocities.To assess longer-term preservation, a box core was collected by R/V Ocean Researcher-1 in September and October 2009 at station K1 (160 m water depth), located at the thalweg of the Gaoping Canyon (Fig. 1).The core was sub-sampled at different depths (Table S2).The sediments are thought to represent deposits associated with hyperpycnal river discharge during Typhoon Morakot in August 2009 (Sparkes, 2012;Liu et al., 2013), whose exceptionally heavy rainfall in Taiwan triggered a very large number of landslides (West et al., 2011) and high rates of sediment delivery offshore (Carter et al., 2012).Together, these marine sediments allow us to assess the transfer and deposition of terrestrial materials by river hyperpycnal flows.
To assess the fate of terrestrial OC delivered by more dispersive events (hypopycnal discharges), sediments were collected from marine trap moorings at 760 and 940 m in the southern Okinawa Trough (Fig. 1), where direct hyperpycnal river discharges are less common (Dadson et al., 2005;Hsu et al., 2006;Kao and Milliman, 2008).In addition, seafloor sediment samples collected between 1994 and 2009 with a box corer on R/V Ocean Researcher-1 and -2 from the Gaoping Shelf, southern Okinawa Trough and Taiwan Strait were selected (Table S3).Samples were collected from the top 2 cm of these cores using a stainless steel spatula and freeze-dried.We have also examined Holocene sediments in the long piston core MD012403 collected by R/V Marion Dufresne in 2001 from a water depth of 1420 m (Kao et al., 2008).The depositional age of these sediments was determined by analysis of the radiocarbon content of planktonic foraminifera so the ∆ 14 C org at time of deposition can be estimated (Kao et al., 2008).The methods of sampling terrestrial and marine sediments mean that the study focuses on the transfer and offshore preservation of sand and finer materials.Large woody debris (e.g.logs and trunks) are not likely to be recovered and so their fate remains a question for future research.

Geochemical methods
All marine samples were rinsed with deionized water (> 18MΩ) to remove salts.All dried sediment samples were homogenized in an agate mortar.Prior to measurement of the OC concentration (C org , %) and analysis of the stable isotopes of OC (δ 13 C org , ‰), samples were treated with 1 N HCl at 20 • C for 16 h to remove carbonate; the residue was centrifuged and freeze-dried (Kao et al., 2008).δ 13 C org analysis was carried out using Carlo-Erba 2100 elemental analyser connected to a Thermo Finnigan Deltaplus Advantage isotope ratio mass spectrometer and reported in δ notation with respect to the PDB standard and renormalized based on working standards (USGS 40 and acetanilide), with reproducibility better than 0.2 ‰.Radiocarbon ( 14 C) was measured on OC by accelerator mass spectrometry after carbonate removal and graphitization at Woods Hole Oceanographic Institution, USA, Institute of Geological and Nuclear Sciences, New Zealand, and Keck-Carbon Cycle AMS Facility at University of California at Irvine, USA. 14 C values are given after correction for 13 C fractionation (normalization to a δ 13 C value of −25 ‰), and expressed as percent modern carbon (pMC) comparative to 95 % of the 14 C activity of the NBS oxalic acid and ∆ 14 C based on established protocols (Stuiver and Polach, 1977), with precision typically better than 10 ‰.Samples from the Liwu River in 2004 were analysed by similar methods described elsewhere (Hilton et al., 2008).Inorganic carbon removal by HCl leaching was preferred over HCl vapour to ensure complete removal of dolomite (Galy et al., 2007b), which may be present in Taiwanese bedrock and river sediments (Hilton et al., 2010).As such, following previous work, all OC isotope measurements refer to the acid-insoluble OC (Galy et al., 2007b;Hilton et al., 2010).The deviation between these two methods for terrestrial materials, δ 13 C org ∼ ± 0.2 ‰ and ∆ 14 C org ∼ ± 10 ‰ (Komada et al., 2008), was similar to the precision of the analyses.

Terminology
Previous work quantifying OC transfers from Taiwan and other orogenic belts has used the term "fossil" OC to define OC derived from sedimentary rocks (Kao and Liu, 2000;Galy et al., 2007a;Hilton et al., 2008Hilton et al., , 2011;;Clark et al., 2013).This is identical to the term OC petro , with "petrogenic" used here because of its unambiguous reference to rock-derived OC (Galy et al., 2008).The term "non-fossil" OC has been used to refer to OC derived from vegetation and soil in the solid load of Taiwanese rivers (Hilton et al., 2008, Table 1.Average composition of fluvial OC in rivers draining the eastern and western flanks of Taiwan (Table S1) and the average input assuming approximately equal input of sediment from both sides of the mountain range (Dadson et al., 2003;Kao and Milliman, 2008).
Average west 20 0.43 0.16 −25.5 0.7 −646 237 Average east 28 0.45 0.27 −24.4 1.1 −677 271 Average 48 0.44 0.22 −24.9 0.9 −661 254 2012) and elsewhere (Clark et al., 2013;Smith et al., 2013) because the output of the mixing analysis is defined as not fossil (Hilton et al., 2010).Here, OC biosphere is used to refer to the same component of OC termed "non-fossil" in previous work (Hilton et al., 2012) because of its clear reference to the source of OC from the terrestrial biosphere.
The range in isotopic composition of terrestrial OC define a triangular domain between ∆ 14 C org and δ 13 C org (Fig. 2a) and the measured values are consistent with previous measurements on suspended sediments from Taiwan (Kao and Liu, 2000;Hilton et al., 2008Hilton et al., , 2010)).

Composition of marine OC
The sediments collected from the trap in the Gaoping Canyon accumulated during Typhoon Kalmaegi (17 July 2008), as constrained by the timing discs deployed by the sediment trap.At that time, SSC in the Gaoping River reached > 20 g L −1 the day after the flood peak.Based on past records of Q w and SSC (Dadson et al., 2005;Kao and Milliman, 2008), it is highly likely that the Gaoping River surpassed SSC = 40 g L −1 necessary for hyperpycnal discharge in this region during Typhoon Kalmaegi.This is consistent with the very high throughput of sediment in the canyon during the event (Liu et al., 2012).Particulate OC samples from the trap have an average C org = 0.5 ± 0. (C org = 0.43 ± 0.16 %, n = 20) and similar to previous measurements in the canyon (Kao et al., 2006).When we include the core samples collected from the canyon following Typhoon Morakot, all of the marine sediments from the Gaoping Canyon (n = 15) have a mean C org = 0.6 ± 0.4 %, which is only slightly higher than the terrestrial OC (Fig. 3b).Both sets of Gaoping Canyon samples were radiocarbon-depleted (Table S2) and their δ 13 C org and ∆ 14 C org values fall into the triangular domain defined by terrestrial OC carried by Taiwanese rivers (Fig. 3a).Away from direct hyperpycnal river inputs, marine particulates collected from sediment traps in the Okinawa Trough (Fig. 1) all had higher δ 13 C org , ∆ 14 C org and C org values than terrestrial OC and Gaoping Canyon samples (Fig. 4, Table S3).When plotted with seafloor sediments collected from the Okinawa Trough, Taiwan Strait and the Gaoping Shelf (Fig. 1), the samples reveal a significant (P < 0.0001, n = 26) positive, linear correlation between ∆ 14 C org and δ 13 C org (Fig. 4a).Samples collected from below the sediment water interface on the Gaoping Shelf (Table S3) also plot on this trend, as do Holocene sediments from the Okinawa Trough (Kao et al., 2008).These seafloor sediments and trap samples also define a significant negative correlation (P = 0.003, n = 26) between 1 / C org and ∆ 14 C org (Fig. 5a).A linear trend between isotopes can result from mixing of two dominant sources (a binary mixture), or may reflect a process by which both sets of isotopes are modified (for example by preferential loss of one component of OC).A linear trend between the inverse of concentration and isotope composition may also result from binary mixing, but OC loss can result in a linear trend which is distinct from mixing.The explanation of these trends will be considered in detail in the Discussion.

Erosion and transfer of terrestrial OC
Physical erosion processes occur at very high rates in mountain landscapes.In Taiwan, suspended sediment yields reach > 10 000 Mg km −2 yr −1 due to a combination of steep slopes and intense precipitation during tropical cyclones (Dadson et al., 2003;Kao and Milliman, 2008).These factors mean that bedrock landslides are common, delivering clastic sediment to mountain rivers (Hovius et al., 2000).Landslides also erode organic matter from the terrestrial biosphere and supply OC, mixed with clastic sediment, to rivers (Hilton et al., 2008;West et al., 2011).In addition, the high runoff intensity promotes mobilisation of soil organic matter by overland flow processes (Hilton et al., 2012).These previous studies have examined the erosion and transport of OC biosphere in detail and estimated rates of OC biosphere transfer which rank amongst the highest in the world (Kao andLiu, 1996, 2000;Stallard, 1998;Hilton et al., 2008Hilton et al., , 2012)).Here we summarize the key findings of this previous work in light of the new 14 C data from Taiwan (Fig. 3) and to inform our assessment of the offshore fate of terrestrial OC (Sects.4.2 and 4.3).
The geomorphic processes occurring in Taiwan can explain the triangular domain in ∆ 14 C org and δ 13 C org values due to mixing of OC petro and OC biosphere (Hilton et al., 2010).Radiocarbon depletion in the samples can be accounted for by the input of OC petro from bedrock.Due to its geological age (> 50 ka), OC petro from Mesozoic-Cenozoic sedimentary rocks has no measurable 14 C and so has ∆ 14 C org ∼ −1000 ‰.
In addition, OC petro input can account for the range in δ 13 C org values at low ∆ 14 C org in Taiwan (Fig. 2a), where metasedimentary bedrocks have δ 13 C org values ranging from −25 ‰ up to −20 ‰ (Hilton et al., 2010).Hilton et al. (2010) reported that rivers draining the east flank of Taiwan can have higher δ 13 C org values ( 13 C-enriched) than those draining the west due to different bedrock geology leading to variable OC petro composition.The 14 C-depleted samples are consistent with this observation (Fig. 2a).In contrast, when values of ∆ 14 C org are higher ( 14 C-enriched), the stable isotope composition is much less variable (Fig. 2a).The δ 13 C org values are similar to those of C3 biomass and soil in Taiwanese mountain forest (Kao and Liu, 2000;Hilton et al., 2013).Previous work has suggested that surface soil horizons in Taiwan have a C org -weighted average ∆ 14 C org ∼ 0 ‰ (Hilton et al., 2008), reflecting the high rates of OC export in the source area which can act to limit soil age (Hilton et al., 2012).Addition of this young OC biospehre can explain the range of ∆ 14 C org values in the suspended sediments (Fig. 2a).However, we note that an expanded soil data set (Hilton et al., 2013) shows that older soil OC biospehre (∆ 14 C org ∼ −100 ‰) may be present in this mountain belt.Its volumetric contribution to the river load is difficult to assess because the 14 Cdepletion in the bulk suspended sediment is dominated by OC petro inputs (Fig. 2a).We suggest that organic-compound specific 14 C measurements may shed light on the erosion and transport of any aged soil OC (e.g.Galy and Eglinton, 2011).
The rapid fluvial transit times in Taiwan, combined with young soils (Hilton et al., 2008), mean that the 14 C content of particulate OC can provide a proxy for OC petro input to the suspended load, in analogy with work from other small mountain river systems (e.g.Komada et al., 2004;Leithold et al., 2006;Clark et al., 2013).Taking ∆ 14 C org values for bedrock (−1000 ‰) and OC biosphere based on the C orgweighted mean of 10 surface soil samples of 0 ‰ (Hilton et al., 2008) we have used an end member mixing model to quantify the fraction of OC biosphere in river suspended sediments.If older soil organic matter (∆ 14 C org < 0 ‰) is more important than the existing soil samples suggest (Hilton et al., 2008(Hilton et al., , 2013)), the estimated OC biosphere component will be conservative.Results indicate that the flux of particulate OC biosphere (g s −1 ) increases with water discharge (Fig. 2b).This confirms a strong climatic control on the erosion and fluvial transfer of OC biosphere highlighted in several Taiwanese catchments using a slightly different method for quantifying the mixing of OC petro and OC biosphere (Hilton et al., 2010;Hilton et al., 2012).It reflects the activation of erosion processes (overland flow, gully incision, landslides) during heavy precipitation and supply of OC biosphere and sediment to rivers when their transport capacity is high (Hilton et al., 2008(Hilton et al., , 2012)).This mechanism is not unique to Taiwan, and has been observed elsewhere when flux and OC source data are both available in tropical (Bass et al., 2011;Clark et al., 2013;Lloret et al., 2013) and temperate mountain forests (Hatten et al., 2012;Smith et al., 2013).A consequence of this behaviour is that flood events can dominate the particulate flux, with 80-90 % of the decadal transfer of OC biosphere by the Liwu River found to occur during cyclonic storms with return times > 1 yr (Hilton et al., 2008).The importance of storm-triggered floods for particulate transfer appears to be a wider feature of small mountain rivers (Townsend-Small et al., 2008;Bass et al., 2011;Lloret et al., 2013).
The suspended sediment samples provide good constraint on the compositional range of terrestrial OC delivered directly to the ocean by hyperpycnal plumes.In addition, because our sample set captures particulate OC from across the mountain belt (Fig. 1), we can assess the likely composition of hypopycnal inputs, which may be expected to be a mixture of sediments sourced from individual river catchments.The average of all river samples (Table 1) is C org = 0.44 ± 0.22 %, δ 13 C org = −24.9± 0.9 ‰, ∆ 14 C org = −661 ± 254 ‰.The ∆ 14 C org value suggests that OC biosphere contributes ∼ 30 % of the total particulate OC on average with OC petro making up the remaining fraction, in agreement with previous estimates from Taiwan (Kao and Liu, 2000;Hilton et al., 2008;Hilton et al., 2010).

Fate of terrestrial OC offshore Taiwan: hyperpycnal inputs
To constrain the transfer of terrestrial OC to the marine environment, we can use the isotopic and elemental composition of samples from marine sediment traps offshore Taiwan.
To assess the longer-term preservation, the composition of seafloor sediments and longer cores can be used.The ∆ 14 C org and δ 13 C org values of terrestrial OC exported by rivers from Taiwan (Fig. 2a) have been characterized more thoroughly than any previous study (Kao and Liu, 1996;Hilton et al., 2008).Recent marine OC can be distinguished from terrestrial OC due to its higher δ 13 C org and ∆ 14 C org values (Hsu et al., 2006).However, the assessment of OC provenance is not the same as quantifying preservation.To do that, we have used C org values of the sediments because C org values are sensitive to changes in the association of OC with clastic particles and can track OC loss (e.g.Galy et al., 2007a;Blair and Aller, 2012;Cathalot et al., 2013).Our investigation of marine samples does not extend to the very deep ocean waters offshore the east coast of Taiwan (Fig. 1).However, O 2 concentrations in the unsampled region, which reach water depths > 4000 m, are low and comparable to those in the Bay of Bengal at 2000 m water depth (Garcia et al., 2010).OC preservation may be higher in these unsampled areas due to the lower oxidation potential of these deep waters (e.g.Cai and Sayles, 1996;Galy et al., 2007a) than the sites which form the focus of our study (Fig. 1).As a result, our estimates of terrestrial OC preservation efficiency may be conservative.
To assess the fate of terrestrial OC delivered to the ocean by rivers during hyperpycnal discharges, we have examined OC collected by the sediment trap moored in the channel thalweg of the Gaoping Canyon, fed by the Gaoping River (Fig. 1).The sediments trapped during the passage of Typhoon Kalmaegi (17 July 2008) had a range in ∆ 14 C org and δ 13 C org values, consistent with the mixture of terrestrial OC petro and OC biosphere observed in river sediments (Fig. 3a), with an average C org very close to the river samples (Fig. 3b).These observations suggest that loss of terrestrial OC during transfer to mesopelagic depths (∼ 600 m) during this hyperpycnal delivery event has been negligible.The trapped sediment included a "young" organic rich subsample (C org = 1.6 %, ∆ 14 C org = −112 ‰, Fig. 3b) with some shredded woody debris visible to the naked eye and up to ∼ 1 cm in size (see also Fig. 4 in Liu et al., 2012).Samples collected from the floor of the Gaoping Canyon after Typhoon Morakot also lie within the terrestrial mixing domain (Fig. 3a).Their C org values also imply little evidence for terrestrial OC loss (Fig. 3b).Thus, we have found that, contrary to a previous study (Kao et al., 2006), loss of OC biosphere (which would systematically lower ∆ 14 C org and C org ) is not consistent with the data.Together, the trap and core samples suggest efficient transfer to surface sediments and preservation in the sub-surface (∼ 100 %) of terrestrial OC (both OC biosphere and OC petro ) in a submarine canyon fed by hyperpycnal flows.While the fate of terrestrial OC transported deeper down the canyon (e.g.Carter et al., 2012) remains to be assessed, the low O 2 levels (Garcia et al., 2009) and high accumulation rates (Huh et al., 2009) are likely to promote longer-term OC burial.
Moreover, it appears that the natural buoyancy of some macro-particles of OC biosphere can be overcome during hyperpycnal flood discharges, as observed in modern sourceto-sink settings elsewhere (Leithold and Hope, 1999) and in the geological record (e.g.Saller et al., 2006).This suggests that the density of the turbid river plume may be high enough to effectively sequester woody debris carried in the sand fraction, while coarser woody material (e.g.logs) float upon discharge to the ocean (West et al., 2011).In addition, water logging of sand-sized woody debris may occur prior to entrainment or during transport, as observed in the sand-sized bedload of larger fluvial systems (Bianchi et al., 2007).In the short mountain rivers of Taiwan, it is unclear whether this mechanism operates; the observation warrants further investigation of the transport of macro-particles of OC biosphere in mountain rivers (e.g.Turowski et al., 2013).

Fate of terrestrial OC offshore Taiwan: hypopycnal inputs
The fate of terrestrial OC away from direct hyperpycnal inputs can be examined using marine samples collected from a wider region around Taiwan (Fig. 1).The core and trap samples from the Okinawa Trough, Taiwan Strait and the Gaoping Shelf displayed significant trends between ∆ 14 C org and δ 13 C org (Fig. 4a), which distinguish them from the Gaoping Canyon samples (Fig. 3a).However, compared to recent marine OC (OC marine ) from the western Pacific (Hsu et al., 2006; Table 2), they were variably depleted in both 13 C and 14 C (Fig. 4a).The values cannot be explained by aging and  Table 2. Compositions used in the mixing and terrestrial OC loss models.a From open marine surface trap samples (Hsu et al., 2006).b Indicative values used to examine the nature of trends in the data (Figs.4b and 5b) informed by the measured compositions (Table 1, Figs. 2a, 4a and 5a).c Measured mean of bulk soils from Taiwan weighted by organic carbon content (Hilton et al., 2008).
re-suspension of OC marine because this 14 C-depletion only results in ∆ 14 C org values of approximately −50 ‰ to −100 ‰ in this setting (Hwang et al., 2010).Instead, the linear trend between ∆ 14 C org and δ 13 C org values may be indicative of binary mixing between end members with distinct compositions (Komada et al., 2004;Clark et al., 2013).The best fit to the data intersects the average of measured riverine OC (itself a mixture of OC petro and OC biosphere ) and the values expected for recent OC marine (Fig. 4a).Thus, the first order pattern in the samples collected away from direct hyperpycnal inputs can be explained by mixing OC marine with terrestrial OC.However, loss of terrestrial OC during marine transfer and deposition may have caused ∆ 14 C org and δ 13 C org values to evolve towards the composition of OC marine .The linear trend suggests that if this loss has occurred, it has done so in a relatively short period of time, because otherwise ∆ 14 C org would vary with time and produce a non-linear relationship with δ 13 C org .To assess the possible loss of terrestrial OC in the marine realm, we model a scenario of instantaneous loss (see Supplement).The results indicate that preferential loss of OC biosphere (e.g.Kao et al., 2006;Cathalot et al., 2013) produces a negative, linear trend between ∆ 14 C org and δ 13 C org (Fig. 4b), which is not consistent with the data.On the other hand, bulk loss of terrestrial OC (both OC biosphere and OC petro ) can produce the observed positive, linear trend between ∆ 14 C org and δ 13 C org (Fig. 4b).
To constrain whether mixing or loss is the dominant control on the isotopic composition of the marine samples, we have turned to C org .The model of bulk terrestrial OC loss, which can explain the ∆ 14 C org and δ 13 C org values (Fig. 4b), cannot reproduce the negative linear relationship between 1 / C org and ∆ 14 C org .The modelled bulk terrestrial OC loss results in a trend which is perpendicular to that observed in the samples (Fig. 5b).Thus, it appears that the patterns in the data are not consistent with either selective (i.e.OC biosphere , Fig. 4b) or pervasive (i.e.OC biosphere and OC petro , Fig. 5b) loss of terrestrial OC.Thus, the only way to account for the first order trends in the measured isotopic and elemental composition of OC (δ 13 C org , ∆ 14 C org and C org ) in the offshore sediments is through a mixture of OC marine and riverine OC (itself a mixture of OC petro and OC biosphere ) (Figs. 4 and 5).
The scatter around the linear trends in the data may reflect second-order temporal (or spatial) variations in the ∆ 14 C org , Earth Surf.Dynam., 2, 127-139, 2014  1) with symbols as in Fig. 4. (a) All samples show a negative relationship between the variables (r = −0.6;P = 0.003) with a linear fit to all samples apart from those in brackets shown by the blank line (r 2 = 0.7; P = 0.0001, grey line is the 95 % confidence interval).(b) Linear relationship displayed in the samples along with the OC content predicted by (i) mixing marine OC and terrestrial OC (black line and dashes with fraction of terrestrial OC); (ii) loss of OC biosphere (starting at fraction terrestrial OC = 0.8, circles with % loss); and (iii) bulk terrestrial OC loss (squares with % loss).
δ 13 C org and C org values of the marine and terrestrial OC end members (Table 1), which can explain the sample compositions.However, the scatter may also reflect some terrestrial OC loss.The instantaneous loss model provides constraint on the percentage of terrestrial OC loss, which is compatible with the variability around the linear trends (Fig. 5).For bulk OC loss, the scatter corresponds to ∼ 20 % loss (Fig. 5b).To acknowledge the uncertainty on this value, and to provide a conservative estimate of terrestrial OC burial efficiency, we suggest < 30 % loss (i.e.preservation efficiency > 70 %) has occurred in these deposits.Note that this sample set is dominated by surface marine sediments (Table S3) and so the preservation efficiency refers to the land-to-ocean transfer and deposition of terrestrial OC biosphere and OC petro .However, sub-surface sediments from the Gaoping Shelf and the Holocene sediments from the Okinawa Trough (Table S3) are consistent with the trends (Figs.4a and 5a).While the limited number of these subsurface samples (n = 7) makes it difficult to draw irrefutable conclusions, it appears that high preservation efficiencies are also a feature of sedimentary burial on longer timescales in this setting (Kao et al., 2008).
Efficient preservation of terrestrial OC biosphere and OC petro in both hyperpycnal and hypopycnal marine sediments (Figs. 3, 4 and 5) is consistent with the high sedimentation rates in the deep ocean basins surrounding Taiwan, which result from the tectonic and climate setting.Sedimentation rates in the southern Okinawa Trough reach > 1 mm yr −1 , sustained by fluvial sediment delivery (Hsu et al., 2006).Indeed, if the total suspended sediment flux of ∼ 380 Tg yr −1 (Dadson et al., 2003) is mostly deposited within ∼ 100 km of the coast (over ∼ 150 000 km 2 ), then the average sedimentation rate would be ∼ 2 mm yr −1 (density of 2.2 g cm −3 ).The rapid accumulation of clastic sediment limits the time over which OC is exposed to O 2 (Burdige, 2005;Galy et al., 2007a).Since the O 2 penetration depth in muddy marine sediments is typically on the order of 1-10 mm (Hedges and Keil, 1995;Cai and Sayles, 1996), OC biosphere deposited offshore Taiwan is probably exposed to O 2 for only a matter of years.
Our findings are consistent with marine sediments undergoing rapid accumulation elsewhere, with bulk OC preservation efficiencies of 70-100 % (Galy et al., 2007a;Blair and Aller, 2012).These tend to exceed terrestrial organic carbon preservation rates in other depositional settings (Burdige, 2005;Blair and Aller, 2012).Unlike steep mountain rivers, large river deltas on passive margins can experience successive resuspension and deposition episodes within mobile muds (e.g. the Amazon Delta).In these systems, marine organic material can be entrained into mobile deposits, driving early diagenetic reactions and facilitating loss of refractory terrestrial OC during sedimentary refluxing and suboxic diagenesis (Aller, 1998;Aller and Blair, 2006).While conditions in large rivers on passive margins can promote rapid export and deposition of marine OC (Bianchi et al., 2006), resuspension and re-working of sediments on shallow-sloping deltas can also promote incineration of terrestrial OC (Aller et al., 1996(Aller et al., , 2008;;Aller and Blair, 2006), even following a rapid sediment accumulation event (Sampere et al., 2008).In contrast, apart from the ∼ 100 m deep Taiwan Strait (Fig. 1), which may be analogous to some of these passive margin www.earth-surf-dynam.net/2/127/2014/Earth Surf.Dynam., 2, 127-139, 2014 settings, rivers export materials to deep basins around the island.The rapid accumulation rates offshore Taiwan and the delivery of terrestrial OC to water depths > 500 m in O 2 poor waters (Garcia et al., 2009) are conditions more analogous to the Bengal Fan system, where terrestrial OC burial efficiencies are also very high (Galy et al., 2007a).

Implications for the global carbon cycle
After accounting for addition of terrestrial OC petro and OC marine to the offshore sediments (e.g.Fig. 4), we can now assess how the erosion of terrestrial biomass (OC biosphere ) drives sequestration of atmospheric CO 2 .Our findings suggest that in the Taiwan sediment routing system, rivers deliver sediments which contain on average 0.15 % of terrestrial OC biosphere (average C org × average fraction of OC biosphere ) in hypo-and hyperpycnal river plumes.Given the suspended sediment flux from Taiwan to the ocean of 384 Tg yr −1 (Dadson et al., 2003;Dadson et al., 2005), this abundance of OC biosphere and the estimated preservation efficiencies of terrestrial OC of > 70 % (see Sect. 4.2), we calculate an OC biosphere burial flux of 0.5-0.6TgC yr −1 in basin fills derived from Taiwan.This may be a lower bound if material coarser than ∼ 500 µm (i.e.large woody debris, whose fate remains poorly constrained) contributes importantly to OC biosphere transfer and burial (West et al., 2011).Normalized over Taiwan's mountain island surface area (35 980 km 2 ), this estimated flux represents an OC biosphere burial yield of 13-16 MgC km −2 yr −1 .CO 2 sequestration associated with physical erosion of Taiwan appears to be seven times more efficient per km 2 than the Himalayan erosion system, which has a burial yield of ∼ 2 MgC km −2 yr −1 (Galy et al., 2007a).Our analysis suggests that the rock-derived OC petro is reburied offshore Taiwan at a rate of 0.9-1.1 TgC yr −1 , similar to the total OC petro buried annually in the Bay of Bengal (Galy et al., 2008).OC biosphere burial from this single mountain island represents ∼ 1 % of the estimated total annual OC burial in the oceans (Schlunz and Schneider, 2000;Burdige 2005) from only 0.02 % of Earth's continental surface.
The islands of Oceania are sediment production hotspots, with suspended sediment yields typically > 1000 Mg km −2 yr −1 (Milliman and Farnsworth, 2011).The transfer of OC biosphere together with clastic sediment may enhance the terrestrial OC burial efficiency (Canfield, 1994;Burdige, 2005;Blair and Aller, 2012), even when materials are delivered by hypopycnal river plumes (see Sect. 4.3).As in Taiwan, terrestrial productivity is high across tropical Oceania and mountain forests contain large stores of OC biosphere in standing biomass and soil (Dixon et al., 1994).This permits us to extrapolate our observations to provide a tentative estimate of the CO 2 sink associated with the oceanic burial of terrestrial OC biosphere .Assuming a linear relationship between sediment yield and OC biosphere yield (e.g.Hilton et al., 2012), the terrestrial OC biosphere content from Taiwan can be combined with the Oceania sediment export of ∼ 7000 Tg yr −1 (Dadson et al., 2003;Milliman and Farnsworth, 2011) and the range of preservation efficiencies obtained here (70-100 %) to estimate a terrestrial OC biosphere burial flux of 8-11 TgC yr −1 from the mountain islands of Oceania.This estimate is likely to be conservative because the high sediment yields in Taiwan of ∼ 9000 Mg km −2 yr −1 result in a lower percent of OC biosphere in sediments when compared to other mountain rivers (Leithold et al., 2006;Hilton et al., 2012;Clark et al., 2013;Smith et al., 2013).Alternatively, it could be assumed that the OC biosphere burial yield from Taiwan (13-16 MgC km −2 yr −1 ) holds over the Oceania area (2.7 × 10 6 km 2 ), giving a terrestrial OC biosphere burial flux of 35-40 TgC yr −1 .This value could be viewed as an upper bound, since the erosion rate of OC biosphere from Taiwan may be toward the high end of global values (Hilton et al., 2012).Both estimates do not consider OC marine burial associated with these sediments, which could increase CO 2 sequestration (Fig. 4a).The lower conservative estimate of OC biosphere burial by the erosion of Oceania is globally significant.It represents ∼ 10 % of estimates of the total OC burial in clastic sediments in the oceans (Schlunz and Schneider, 2000;Burdige 2005;Blair and Aller, 2012).Adjustment of these global estimates is warranted and requires further observational constraint on the processes and magnitude of this significant flux in the global carbon cycle.
Our findings suggest that mountain building in Oceania can result in a globally important geological CO 2 sink through erosion of the terrestrial biosphere, OC biosphere transport by mountain rivers and preservation in marine sediments from hyperpycnal but, importantly, also hypopycnal delivery events (Figs. 4 and 5).This region provides a strong link between tectonic uplift and the carbon cycle, which should influence atmospheric CO 2 concentrations on geological timescales.Importantly, CO 2 sequestration by erosion of OC biosphere should be sensitive to the coverage of terrestrial biomass in the tropics, which is in part moderated by the available supply of CO 2 for productivity (Norby et al., 2005).In addition, the amount and variability of runoff control the erosion and export of terrestrial OC biosphere (Fig. 2b) and clastic sediment by small mountain rivers (Dadson et al., 2003;Hilton et al., 2008Hilton et al., , 2012)).Therefore, islands of Oceania have the potential to introduce stabilizing feedbacks in the carbon cycle on geological timescales, which are presently not considered in Earth System models (Berner, 2006;Archer et al., 2010) and not invoked to explain the evolution of atmospheric CO 2 levels in the Cenozoic (e.g.Pagani et al., 2009).One aspect of this may be the link between warming ocean temperatures and the occurrence of extreme tropical cyclones in the western Pacific (Elsner et al., 2008) which deliver OC biosphere and sediment efficiently to the ocean, as previously hypothesized (Hilton et al., 2008).However, a wider response across Oceania may be felt due to CO 2 fertilization of tropical forests (Norby et al., 2005), while ocean warming also increases runoff and runoff variability in the tropics away from tropical cyclone influence (Milly et al., 2005).The corresponding increase in terrestrial OC biosphere export to the oceans from tropical islands may help to mitigate an increase in greenhouse gas concentration, with enhanced CO 2 sequestration by terrestrial OC biosphere burial in the ocean.These feedbacks may play a significant role in linking tectonics and climate and their impact on the longterm carbon cycle, all of which deserve further attention.

Figure 1 .
Figure 1.Location of terrestrial and marine samples from Taiwan and the surrounding ocean used in this study.River sediments were collected during typhoon floods across the island (black circles) with the sampled river names indicated (TableS1).Marine samples fed by dispersive fluvial inputs (white symbols) were obtained from box core surface sediments, with the location of the longer piston core MD012403 indicated.Sediments were also acquired from within the Gaoping Canyon (black triangles), which is fed by hyperpycnal river plumes.Sediment traps (indicated by squares within symbols) were deployed in the Okinawa Trough (depths provided) and at 608 m in the channel of the Gaoping Canyon.

Figure 3 .
Figure3.Composition of marine sediments in the Gaoping Canyon (Fig.1) fed by periodic hyperpycnal flows.(a) Stable and radioactive isotope compositions of organic carbon (δ 13 C org and ∆ 14 C org , ‰) in marine sediments (sediment trap and box core) from the Gaoping Canyon, along with river suspended sediments of Taiwan (as shown in Fig.2b).(b) Organic carbon concentration (C org , %), with mean ±standard deviation shown by large symbol and whiskers for river suspended sediments and canyon sediments (trap and core) from the Gaoping Canyon.

Figure 4 .
Figure 4. Stable and radioactive isotopic compositions of organic carbon (δ 13 C org and ∆ 14 C org , ‰) in marine sediments offshore Taiwan fed by dispersive terrestrial inputs (Fig.1).(a) White dots denote samples from the trap moorings.The mean terrestrial OC composition delivered by rivers (green circle, whiskers ± SD) and the expected composition of recent marine OC (blue box) are shown.The samples display a positive linear relationship (black line, with 95 % confidence intervals in grey).(b) Linear relationship displayed in the samples along with the isotopic composition of OC predicted by (i) mixing marine OC and terrestrial OC (black line and dashes with fraction of terrestrial OC); (ii) loss of terrestrial OC biosphere starting at fraction terrestrial OC = 0.8 (circles with % loss); and (iii) bulk terrestrial OC loss (squares with % loss).

Figure 5 .
Figure5.Radiocarbon contents of organic carbon (expressed as ∆ 14 C org , ‰) versus the inverse of OC concentration (1 / C org ) in marine sediments offshore Taiwan fed by dispersive terrestrial inputs (Fig.1) with symbols as in Fig.4.(a) All samples show a negative relationship between the variables (r = −0.6;P = 0.003) with a linear fit to all samples apart from those in brackets shown by the blank line (r 2 = 0.7; P = 0.0001, grey line is the 95 % confidence interval).(b) Linear relationship displayed in the samples along with the OC content predicted by (i) mixing marine OC and terrestrial OC (black line and dashes with fraction of terrestrial OC); (ii) loss of OC biosphere (starting at fraction terrestrial OC = 0.8, circles with % loss); and (iii) bulk terrestrial OC loss (squares with % loss).
Stable and radioactive isotope compositions of organic carbon (δ 13 C org and ∆ 14 C org , ‰) of suspended sediments from rivers in Taiwan.Analytical errors are smaller than the point size.Samples from catchments draining the two flanks of the mountain belt define a mixing domain between organic carbon from the terrestrial biosphere (OC biosphere ) and petrogenic OC (OC petro ) from bedrocks.(b) Instantaneous particulate flux of OC biosphere (g s −1 ) as a function of water discharge during floods in the sampled catchments, demonstrating a strong hydrological control on OC biosphere export.