Permafrost Carbon and CO2 Pathways Differ at Contrasting Coastal Erosion Sites in the Canadian Arctic

Warming air and sea temperatures, longer open-water seasons and sea-level rise collectively promote the erosion of permafrost coasts in the Arctic, which profoundly impacts organic matter pathways. Although estimates on organic carbon (OC) fluxes from erosion exist for some parts of the Arctic, little is known about how much OC is transformed into greenhouse gases (GHGs). In this study we investigated two different coastal erosion scenarios on Qikiqtaruk – Herschel Island (Canada) and estimate the potential for GHG formation. We distinguished between a delayed release represented by mud debris draining a coastal thermoerosional feature and a direct release represented by cliff debris at a low collapsing bluff. Carbon dioxide (CO2) production was measured during incubations at 4°C under aerobic conditions for two months and were modeled for four months and a full year. Our incubation results show that mud debris and cliff debris lost a considerable amount of OC as CO2 (2.5 ± 0.2 and 1.6 ± 0.3% of OC, respectively). Although relative OC losses were highest in mineral mud debris, higher initial OC content and fresh organic matter in cliff debris resulted in a ∼three times higher cumulative CO2 release (4.0 ± 0.9 compared to 1.4 ± 0.1 mg CO2 gdw–1), which was further increased by the addition of seawater. After four months, modeled OC losses were 4.9 ± 0.1 and 3.2 ± 0.3% in set-ups without seawater and 14.3 ± 0.1 and 7.3 ± 0.8% in set-ups with seawater. The results indicate that a delayed release may support substantial cycling of OC at relatively low CO2 production rates during long transit times onshore during the Arctic warm season. By contrast, direct erosion may result in a single CO2 pulse and less substantial OC cycling onshore as transfer times are short. Once eroded sediments are deposited in the nearshore, highest OC losses can be expected. We conclude that the release of CO2 from eroding permafrost coasts varies considerably between erosion types and residence time onshore. We emphasize the importance of a more comprehensive understanding of OC degradation during the coastal erosion process to improve thawed carbon trajectories and models.


INTRODUCTION
The Arctic is outpacing the global warming trend, which has severe impacts on ecosystems and biogeochemical pathways (Ciais et al., 2013;AMAP, 2017). With increasing temperatures, a cascade of environmental responses is initiated. Permafrost is warming and rapidly thawing, sea ice is declining in thickness and extent, seawater temperature is rising and wave and storm intensity are increasing (Manson and Solomon, 2007;Overeem et al., 2011;Schuur et al., 2015;Biskaborn et al., 2019;Dai et al., 2019). Arctic coastlines are particularly vulnerable to thaw and erosion due to the combined effects of environmental change (Barnhart et al., 2014;Fritz et al., 2017;Nielsen et al., 2020). Along with permafrost coasts, Arctic river deltas, estuaries and wetlands are increasingly impacted by sea level rise and glacier melt (Bendixen et al., 2017;Ward, 2020). These systems provide important ecosystem services (Forbes, 2019) but are currently facing changes in their sedimentation dynamics, which impacts organic matter pathways in the coastal zone (Bendixen et al., 2017;Ward, 2020). Permafrost coasts are the dominant interface between the northern landmass and the Arctic Ocean. Therefore, coastal erosion is a key driver for carbon and sediment transport into the Arctic Ocean (Stein and Macdonald, 2004;Macdonald et al., 2015;Wegner et al., 2015). Approximately two third of the Arctic coastline is composed of ice-rich and unconsolidated permafrost sediments prone to degradation, resulting in average erosion rates of 0.57 m yr −1 and extremes of > 25 m yr −1 Jones et al., 2009Jones et al., , 2018Lantuit et al., 2012;Günther et al., 2013Günther et al., , 2015. With erosion, the massive reservoir of seasonally unfrozen and permafrost organic carbon (OC) in the Northern Hemisphere that stores ∼1,300 Pg (10 15 g) OC is accessed at its coastal margins Hugelius et al., 2014;McGuire et al., 2018). Recent studies estimate that the erosion of permafrost coasts transfers up to 430 Tg yr −1 of sediment and 14 Tg yr −1 of OC into the Arctic Ocean (Wegner et al., 2015). In the current understanding of Arctic carbon cycles, coastal erosion is considered to play a modest role in the transfer of OC from land to the Arctic Ocean (McGuire et al., 2009;Stein and Macdonald, 2004;Macdonald et al., 2015). Recent studies on abrupt permafrost erosion on land, however, suggest that OC is already fundamentally altered upon permafrost thaw, erosion and transport in aquatic systems on land, during which greenhouse gases (GHGs) can be released to the atmosphere Mann et al., 2015;Spencer et al., 2015;Heslop et al., 2017). During these erosional processes, the composition of OC is altered substantially with a substantial fraction potentially transformed into carbon dioxide (CO 2 ) and methane (CH 4 ), depending on oxygen availability (Vonk et al., 2012;Semiletov et al., 2016;Tanski et al., 2019). Yet, the degradation of OC from thawing permafrost coastlines has barely been considered (Vonk and Gustafsson, 2013;Fritz et al., 2017).
Coastal erosion releases large amounts of readily degradable permafrost OC, which is transported on relatively short transit times from land to sea with a high potential for GHG release (Vonk et al., 2012;Tanski et al., 2019). Once it enters the ocean, OC can be either further degraded within sediments and the water column or buried in marine sediments Bröder et al., 2018;Couture et al., 2018;Grotheer et al., 2020;Jong et al., 2020). Yet, little is known about the potential GHG formation during erosion as well as within the coastal nearshore corridor that receives the eroded sediments before entering the marine domain. While more comprehensive information on GHG production exist for terrestrial permafrost sites and the Arctic shelf, the coast itself remains sparsely investigated with GHG measurements only available for selected locations (Vonk et al., 2012;Shakhova et al., 2010Shakhova et al., , 2015Karlsson et al., 2015;Overduin et al., 2015;Tanski et al., 2019). Furthermore, the biogeochemical response of permafrost OC upon erosion may depend on coastal morphology and organic matter composition. Along coastlines with thermoerosional features and high bluffs, OC may encounter long transfer times onshore resulting in a delayed release to the sea with GHG formation already taking place onshore (Figures 1A,B). In contrast, at sites with low bluffs (Figures 1C,D), block-failure, active layer detachment slides or cliff collapse, OC may follow a more direct transfer pathway to the ocean. Thus, the type of erosion is strongly related to coastal morphology and may play a substantial role for the magnitude, timing and locus of GHG release. Local bathymetry and shelf morphology further determine the fate of OC and its potential for degradation offshore with rapid burial and sequestration limiting degradation but resuspension facilitating OC turnover in the nearshore (Solomon, 2005;Vonk et al., 2012;Hilton et al., 2015;Spencer et al., 2015;Bröder et al., 2018;Grotheer et al., 2020;Jong et al., 2020).
Although geomorphologic studies on coastal erosion processes are manifold (e.g., Barnhart et al., 2014;Günther et al., 2015;Jones et al., 2018), these studies do not target the OC fraction or relate morphology to the understanding of OC degradation and GHG dynamics. Bridging this information void on OC transformation during land-ocean-transit and the geochemical response during coastal erosion is key to better constrain biogeochemical pathways of OC in a warming FIGURE 1 | Examples of coastal erosion types along the Yukon coast (Canadian Arctic, Beaufort Sea). Thermoerosional features and high coastal bluffs may lead to organic matter degradation onshore due to longer transport times and a delayed release into the ocean (A,B). Along low bluffs, permafrost debris collapses on the beach and is removed relatively quickly by waves via a more direct pathway into the nearshore zone (C,D). Photographs were taken by A. Irrgang and G. Tanski in August 2015 on Qikiqtaruk -Herschel Island (A,B) and the western Yukon coast (C,D).
Arctic and support Earth climate modeling approaches. The objective of this study is to investigate the CO 2 production at two contrasting erosion sites along Arctic permafrost coasts to better determine the pathways and turnover of OC during transport from land to sea.

STUDY AREA
The study area Qikiqtaruk -Herschel Island is situated ∼2 km off the Yukon coast on the narrow Beaufort Sea Shelf (Figure 2). The local climate conditions are polar continental with mean annual air temperatures of −8.4 • C (1995 to 2016) (Coch et al., 2018) and precipitation of ∼170 mm yr −1 (Burn, 2012). During the peak of the warm season in July temperatures are on average 7.8 • C with extremes of 30 • C possible (Burn, 2012). The dominant wind directions are northwest and east with storms being frequent in August and September (Dunton et al., 2006). The island is a push moraine ridge formed ∼18,000 to 23,000 years ago during the Wisconsin Glaciation with a maximum elevation of 183 m (Rampton, 1982;Pollard, 1990;Fritz et al., 2012). The sediments are composed of unconsolidated and fine-grained material of marine and glacigenic origin with glacier ice buried below the island (Blasco et al., 1990;Fritz et al., 2012). A rolling landscape incised by valleys with a lowland tundra vegetation dominated by graminoids and dwarf shrubs is typical for the island (Smith et al., 1989;Kennedy et al., 2001). The island is underlain with continuous permafrost with a shallow seasonal active layer of ∼40-60 cm in thickness, which has deepened by 15-25 cm since the 1980s (Kokelj et al., 2002;Burn and Zhang, 2009). The surface sediments (up to ∼1 m depth) on the island store ∼35 kg OC m −2 with OC storage being lower in areas of mass wasting and higher at sites of accumulation (Obu et al., 2015;Ramage et al., 2019). Permafrost sediments mainly consist of mineral soils with elevated but strongly varying OC contents when cryoturbation occurs (average 6.0 ± 4.7 wt% TOC; Tanski et al., 2017). The ground is very ice-rich with mean volumetric ground ice contents of 30-60% and maximum volumes exceeding 90%, where massive ice beds and wedge ice are present (Couture and Pollard, 2015). The open water season lasts approximately from late June to early October with landfast ice being persistent in Herschel Basin and Thetis Bay (Dunton et al., 2006). The mean tides are 0.15 m and are superimposed to a 0.66 m annual tidal cycle in late July (Huggett et al., 1975;Barnhart et al., 2014). Regional sea-level rise proceeds at rates of 1.1-3.5 mm yr −1 (Manson et al., 2005;James et al., 2014). The prevailing wind directions can influence water levels with northwest winds promoting a positive (higher water level) and eastern winds a negative storm surge (Hequette and Barnes, 1990;Héquette et al., 1995). The Mackenzie River plume has a strong influence on the coastal waters and transports warm water masses towards Qikiqtaruk -Herschel island, which is amplified with easterly winds (Dunton et al., 2006). Qikiqtaruk -Herschel Island is strongly impacted by thermoerosional processes and shoreline retreat. Retrogressive thaw slumps are ubiquitous and doubled in area since the 1950s (Lantuit and Pollard, 2005). On average, the island erodes at 0.68 m yr −1 (2000 to 2011), which is comparable to the average rates of the Yukon coast with 0.7 m yr −1 and the circum-Arctic with 0.57 m yr −1 Obu et al., 2016;Irrgang et al., 2018). Extreme erosion rates of 14 and 22 m yr −1 were observed at low cliffs and active thaw slump systems, respectively (Obu et al., 2015;Cunliffe et al., 2019).

Field Work and Sampling
Permafrost erosion debris was sampled at two active coastal erosion sites on Qikiqtaruk -Herschel Island in July and August 2017 (Figure 2). Mud debris was sampled from a mud lobe draining a retrogressive thaw slump system (Supplementary Figure 1) representing the delayed release scenario, where eroded material has relatively long residence times onshore before release into the ocean. We targeted a retrogressive thaw slump since these features are ubiquitous along the coastline of the western Canadian Arctic (Figures 1A, 2; Pelletier and Medioli, 2014;Ramage et al., 2017). This specific thaw slump was chosen because CO 2 degradation rates of permafrost from the thaw slumps headwall are known  and are used as a permafrost reference for CO 2 release onshore. The mud lobe draining the thaw slump is a mix of thawed permafrost, collapsed active layer and melted massive ice (e.g., buried glacier ice, wedge ice), and can be either transported directly towards the ocean or stored temporarily in mud pools on annual or decadal time scales before release into the ocean (Cray and Pollard, 2015;Tanski et al., 2017). Five samples have been taken from the mud lobe at random locations and one composite sample, mud debris (MUD), mixed in equal amounts from the mud lobe samples under frozen conditions in the cold lab (Table 1).
FIGURE 2 | Study area Qikiqtaruk -Herschel Island off the Yukon coast in the western Canadian Arctic. The sampling sites mud debris (delayed release) and cliff debris (direct release) are situated in the south-eastern part of the island (close-up: middle panel). The mud debris was taken from a mud lobe draining a retrogressive thaw slump system (lower left panel) and the cliff debris from an eroding cliff at the rim of an alluvial fan (lower right panel). Exact sampling locations are displayed in Supplementary Figure 1. Plumes and a turbid nearshore water column are visible and indicate substantial input and transport of eroded sediments. Satellite imagery: Sentinel 2, RGB (27 July 2017).
Frontiers in Earth Science | www.frontiersin.org TABLE 1 | Summary of bulk organic matter and lipid biomarkers (n-alkanes) for initial (T = 0) and incubated mud debris (MUD) and cliff debris (CD) samples after two months (T = 1).

Mud lobe
MUD (T = 0) MUD (T = 1) MUD* (T = 1) Cliff CD (T = 0) CD (T = 1) CD* (T = 1) Sedimentology n = 5 n = 1 n = 5 n = 3 n = 5 n = 1 n = 1 n = 3 Mineral SA (m 2 g −1 ) 21. Values are given as mean with standard deviation (±) and sample size (n). The asterisk (*) indicates incubation set-ups with seawater. "Mud lobe" and "cliff" samples are in situ samples and were not incubated. The second site was a low permafrost cliff (< 5 m height) located at an alluvial fan (Figure 2), where cliff debris was collected. We chose this site to represent the direct release scenario with short residence time of eroded material onshore. This site was chosen since the erosion type is common along coasts with lower coastal bluffs (Figures 1C,D, 2), where eroded material has shorter residence times onshore due to fast removal by wave action at high tides and storm surges (Radosavljevic et al., 2016;Cunliffe et al., 2019). We took five samples from the cliff face (Cliff , Table 1) and one sample from the cliff debris (CD) collected at the cliff toe assuming it to be an integrated signal of the cliff ( Table 1). In contrast to MUD, no composite mix sample was used as we assumed that the cliff material was much more heterogeneous than the mud lobe. Thus, a mixing would have introduced an artificial bias as the mixing ratio was unclear. Surface seawater (SW) was collected approximately 2 km offshore, where water was clear and outside the range of visible erosion plumes.
All non-frozen samples from the mud debris and cliff debris (∼250-500 g wet-weight) were taken with a pre-cleaned sampling spoon and stored in Whirl-pak bags. Samples were taken from approximately 5-10 cm below the surface, at locations where material seemed to be well-mixed during the post-erosive transport process and therefore considered representative. All (frozen) permafrost samples from the cliff face were taken with a drill hammer (Hilti, Liechtenstein) after scraping off the potentially thawed material from the cliff surface. Samples (∼500 g wet-weight) were stored in Whirl-pak bags and stored frozen in a cooler immediately to avoid thaw. Seawater was taken from the sea surface in pre-rinsed (three times) 1L Nalgene bottles. All sampled materials, including seawater, were kept frozen at −20 • C until further processing in the laboratories.

Incubation Set-Ups
We simulated the erosion of mud debris and cliff debris onshore as well as in the nearshore in a laboratory incubation experiment at GFZ Potsdam to assess potential carbon pathways in this coastal transition zone. Onshore refers to incubation set-ups without seawater, mimicking erosion on land, while nearshore set-ups with seawater added were designed to mimic erosion into the seawater column. Frozen mud debris and cliff debris samples were homogenized and subsamples of ∼20 g placed into 120 ml glass vials (autoclaved at 125 • C) at −15 • C in a cold lab. The frozen seawater (SW) sample was melted slowly at 7 • C in a cool (dark) room and shaken thoroughly. Just before the start of the incubation ∼30 ml of unfiltered seawater was added to the glass vials with subsamples for the nearshore incubation setups. Afterwards, the vials were sealed with a rubber septum and the headspace (ca. 90 to 100 ml) was flushed with synthetic air (Linde, Germany) consisting of 20% oxygen (O 2 ) and 80% nitrogen (N 2 ). Vials were then placed in an incubator at 4 • C shielded from light. Aerobic conditions were applied by flushing the headspace with synthetic air after each CO 2 measurement to represent venting and associated oxygen supply likely occurring during the erosion process onshore as well as in the immediate nearshore. The incubation period was set to two months to capture the initial period of highest CO 2 production  and conducted at 4 • C as it is close to the ambient Arctic sea surface temperatures and comparable to most published permafrost incubation studies (Dutta et al., 2006;Lee et al., 2012;Schädel et al., 2014).

Hydrochemistry
Basic hydrochemical parameters pH and electrical conductivity were measured with a 3430 Digital Multiparameter meter (WTW, Germany) at room temperature at AWI Potsdam, Germany. Seawater was filtered with a 0.7 µm GF/F filter (Whatman, United Kingdom) and acidified with HCl (suprapur grade 30%) to pH 2. Dissolved organic carbon (DOC) and dissolved nitrogen (DN) concentrations were analyzed with a Shimadzu analyzer (Shimadzu, Japan) at AWI Bremerhaven, Germany.

Gas Measurements and Mass Calculation
CO 2 and CH 4 concentrations were measured with a 7890A GC (Agilent, United States) at GFZ Potsdam, Germany. Gas samples were taken on a regular basis (every other day during the first two weeks and two times a week afterwards) with a gastight syringe from the headspace of the vials and were immediately injected into the gas chromatograph. Afterwards, the headspace was flushed with synthetic air to reset aerobic conditions. After each measurement, glass vials were shaken to simulate mixing of sediments during mass movement onshore and by wave action and currents in the nearshore. The gas production was measured in parts per million (ppm) and normalized to gram dry weight (gdw −1 ) of initial sediment. The total amount of CO 2 and CH 4 (µmol) was calculated following the methodology by Knoblauch et al. (2018) using gas concentration, headspace volume, water volume, pH, temperature and solubility including carbonate and bicarbonate concentrations for CO 2 calculations (Millero et al., 2007). All gas that remained in the glass vials after flushing was accounted for by assuming that CO 2 in the headspace was completely removed with flushing but dissolved inorganic carbon (DIC) remained in the glass vials. The derived amount of CO 2 in µmol was normalized to gdw −1 and recalculated into mg CO 2 gdw −1 using the molar mass (M) of CO 2 (44). In this study, we only report CO 2 as incubations conditions were aerobic.

Bulk Elemental and Isotopic Analyses
Sediments were weighed, freeze-dried and homogenized before chemical analysis. Elemental and stable carbon isotope analysis was conducted at AWI Potsdam, Germany, following established methodology (e.g., Schirrmeister et al., 2018). Total carbon (TC) and total nitrogen (TN) contents were measured with a VARIO EL III element analyzer (Elementar, Germany) and total organic carbon (TOC) with a VARIO MAX C analyzer (Elementar, Germany), both with a device-specific accuracy of ± 0.1wt %. For stable carbon isotope (δ 13 C-TOC), carbonates were removed from sediments with 1.3 molar hydrochloric acid (HCl) at 95 • C for three hours. Stable carbon isotopes were then measured at the Isotope Facility at AWI Potsdam with a Delta-V-Advantage mass spectrometer (MS) (Thermo Fisher Scientific, Germany) coupled with a CONFLO IV gas mixing system. Stable carbon isotope contents are given in per mil ( ) relative to the standard Vienna Pee Dee Belemnite (VPDB) with an external error better than ± 0.15 . Radiocarbon ( 14 C-TOC) measurements were performed at the Biogeoscience laboratories and the Laboratory for Ion Beam Physics at ETH Zürich, Switzerland. Carbonates were removed by fumigation with HCl (37%, trace-metal purity) at 60 • C for 72 h and samples measured on a coupled elemental analyzer-accelerator mass spectrometer (EA-AMS) system with a vario MICRO cube (Elementar, Germany) coupled to a Mini Carbon Dating System MICADAS (Ionplus, Switzerland) (e.g., Ruff et al., 2010). Radiocarbon values are reported in 14 C ( ) and 14 C-age as uncalibrated years ( 14 C yr).

Grain Size Analysis
Grain size of sediment particles was analyzed at AWI Potsdam following an established methodology (Schirrmeister et al., 2020). Organic material was removed by adding 100ml of 3% H 2 O 2 over 3 to 6 weeks while on a shaking table. Organic-free sediments were then dispersed in 1L of water with dispersing agent (NH 4 OH) in an overhead shaker to de-cluster particles.
Subsamples were then measured with a laser diffraction particle analyzer (Malvern Mastersizer 3000) with grain sizes measured between 0.375 to 1000 µm applying the Frauenhofer optical model. The cut-off value at 1000 µm was pre-set due to technical limitations of the particle analyzer.

Surface Area Analysis
Mineral surface area (SA) was measured to characterize the sediment structure of coastal debris. Approximately 1 g of subsampled sediment was combusted at 450 • C for 12 h and rinsed with ultrapure water (MilliQ) to remove OC and salt, respectively. Afterwards, samples were freeze dried, degassed at 300 • C with a vacuum applied for two hours and SA measured with a Nova 4200e SA analyzer (Quantachrome, United States) using the 6-point Brunauer-Emmett-Teller method (Brunauer et al., 1938).

Extraction and Analysis of Lipid Biomarkers
Lipid biomarkers (long-chain high molecular weight n-alkanes) were used to investigate the molecular composition of organic matter and its state of degradation (Eglinton and Hamilton, 1967). Lipid biomarkers were extracted at Vrije Universiteit Amsterdam, The Netherlands. Sediments (∼1 to 5 g) were extracted with a microwave digestion system MARS 6 (CEM, Germany) using dichloromethane (DCM) and methanol (MeOH) (9:1-volume ratio) at 100 • C for 15 min. Microwave extracts were saponified with 10 ml potassium hydroxide (KOH, 0.5 M) in MeOH at 70 • C for two hours. The neutral fraction was separated by wet-extraction with hexane following addition of 5 ml ultrapure water (MilliQ) with sodium chloride (NaCl). The neutral fraction was further separated into a-polar (containing the n-alkanes) and polar fractions by column chromatography using 5% deactivated silica (SiO 2 ). The n-alkanes were analyzed with a gas chromatograph mass spectrometer (GC-MS, Agilent, United States) at the Institute of Polar Sciences in Bologna, Italy. The GC temperature program started at 55 • C followed by a ramp up of 4 to 5 • C min −1 to a maximum temperature 300 to 310 • C for 15 to 20 min. Lipid biomarker concentrations are given in µg g −1 TOC. In this study concentrations are reported in µg gdw −1 to use the same normalization as for cumulative CO 2 production.

Bulk and Molecular Degradation Proxies
We used several established proxies to assess the degradation state of organic matter. Molar TOC/TN-ratios (atomic ratio) and δ 13 C-TOC values were used as general indicators for bulk organic matter origin, preservation and degradation state (Kuhry and Vitt, 1996;Lamb et al., 2006;Gundelwein et al., 2007).
A low TOC/TN-ratio or relative decrease indicates degradation with a preferential mineralization of TOC and immobilization of TN by microbes coupled to CO 2 release under aerobic conditions (Sollins et al., 1984;Stevenson, 1994;Strauss et al., 2015). High δ 13 C-TOC values (enriched in 13 C, i.e., less negative values) indicate decomposition as organic matter degradation discriminates against the lighter carbon component ( 12 C), which results in a more negative δ 13 C signature in the degrading fraction (Heyer et al., 1976;Strauss et al., 2015).
On a molecular level, we used the carbon preference index (CPI) for n-alkanes (ratio of odd over even-numbered C chain length homologues C 22−34 ) as indicator for the degree of microbial alteration of organic matter (Bray and Evans, 1961;Killops and Killops, 2004;Strauss et al., 2015). Living plants normally have CPI values greater 5 and decrease towards 1 with increasing maturity (Rielley et al., 1991;Vonk et al., 2010). High CPI values therefore indicate the presence of fresh organic matter, while mature or degraded organic matter displays lower CPI values. We further used high-molecular weight (HMW) to low-molecular weight (LMW) ratios of n-alkanes ( C 23−35 / C 14−19 ) as indicators for degradation since with ongoing degradation HMW compounds are broken down into LMW compounds (Killops and Killops, 2004). Additionally, we used the LMW-even to HMW-odd n-alkane ratio (C 16+18+20 /C 27+29+31 ) with an increasing ratio indicating stronger microbial activity and degradation (Grimalt and Albaigés, 1987;Sánchez-García et al., 2014).

CO 2 Production and Model
The total CO 2 production (cum. CO 2 gdw −1 ) over 60 days and daily production rates (µg CO 2 gdw −1 day −1 ) give information on the magnitude and rates of initial aerobic TOC decomposition in a given sample and are commonly used as degradability indicators (e.g., Schädel et al., 2014). With the known total amount of CO 2 released we calculated the TOC loss in %.
CO 2 production was modeled four months and one year using an applied carbon decomposition model (Andrén and Kätterer, 1997;Knoblauch et al., 2013). The model is based on a firstorder kinetic function, which reflects changes in OC content over time and considers how fast OC pools are being decomposed. Degradation rates were estimated using a non-linear least squares approach with an integrated trust-region-reflective algorithm in MATLAB (MathWorks, United States). The cumulative CO 2 production measured during the incubation was used to calibrate the model. A logarithmic transformation was applied to estimate decomposition time of OC for four months and an entire year.

Organic Matter Origin and Composition
The composition of organic matter at the sampling sites was considerably different for most of the parameters considered ( Table 1). The sediments at the investigated erosion sites mud lobe (n = 5) and cliff (n = 5) had TOC contents (1.4 ± 0.03 and 2.0 ± 0.4%), TOC/TN-ratios (10.2 ± 0.2 and 11.5 ± 0.5) and δ 13 C concentrations (−25.8 ± 0.1 and −26.0 ± 0.2 ) typical for sediments containing terrestrial organic matter. Grain size structure was similar in both with the bulk of material (∼70 vol.%) composed of silt but was much more variable in the cliff (Supplementary Figure 2). The generally better sorting of mud lobe sediments may reflect hydrodynamic sorting during transport on land. The relatively high HMW/LMW n-alkane ratios ( C 23−35 / C 14−19 ) in mud lobe (17.7 ± 5.3) and cliff (20.6 ± 9.3) samples indicated good preservation of wax lipids from terrestrial plant matter (Karlsson et al., 2015;Strauss et al., 2015). This is supported by a relatively high (close to 5) CPI for long-chain n-alkanes (C 22−34 ) in mud lobe and cliff sediments (4.6 ± 0.2 and 6.3 ± 0.9, respectively). The lower TOC/TNratios and δ 13 C values in the mud lobe in combination with lower HMW/LMW ratios and CPI for n-alkanes indicated slightly more matured organic matter in the mud lobe and better preserved organic matter in the cliff. The 14 C-age of OC varied considerably between the sites with relatively old organic matter in the mud lobe (13,332 ± 2685 14 C yr), most likely originating from deposition after the Wisconsin glaciation (Pollard, 1990;Fritz et al., 2012) and younger material in the cliff (5679 ± 1896 14 C yr), which was potentially deposited during Holocene warm periods (Burn, 1997;Kaufman et al., 2009;Fritz et al., 2012).
The mud debris (MUD, n = 1) was a mixed sample of the mud lobe deposits and thus showed the same organic matter composition and 14 C-age (14,352 14 C yr). Compared to the mud lobe this indicates its origin from thawed old permafrost, which was likely augmented by active layer material, melted massive ice and old slump floor deposits during transport (Lantuit and Pollard, 2005;Cray and Pollard, 2015;Tanski et al., 2017). Mineral SA of mud debris was 21.5 m −2 gdw −1 , reflecting the presence of fine-grained sediments in the silt and clay size fraction making up ∼72 and 17 vol.% (Table 1), respectively. In combination with relatively low TOC contents (1.5 wt%), the OC loading of mud debris was 0.7 mg OC m −2 . Collapsed active layer chunks from the slump headwall, which contained more modern and OC-rich sediments are usually (temporarily) deposited within the slump floor and mud lobe if not already transported offshore with the thaw stream draining the slump system (Kokelj et al., 2013;Tanski et al., 2017). During transport and mass wasting within the slump, a substantial portion of the organic matter was likely subject to degradation, resulting in the low TOC contents of the mud lobe and mud debris that was used in the incubation (Cray and Pollard, 2015;Tanski et al., 2017). This was supported by the much lower TOC content and TOC/TN-ratio (10.0) of the mud debris than the thaw slump headwall sediments (∼5 wt% and > 15, respectively; Tanski et al., 2017). These slump sediments were also characterized by the presence of fresh organic matter, reflected by CPI values for n-alkanes of close to 30 (Tanski et al., 2017).
The cliff debris (CD, n = 1) collected from the permafrost cliff toe was markedly different in its organic matter composition, both compared to the cliff itself and the mud debris. The relatively high TOC content (7.0 wt%) and TOC/TN-ratio (14.9) in combination with its younger 14 C-age of 2365 14 C yr and higher δ 13 C values of −26.0 indicate a strong contribution from relatively fresh and modern sources and C 3 -plant material that accumulated in the cliff debris (Lamb et al., 2006). The mineral SA of cliff debris was 32.4 m 2 g −1 and silt and clay volumes ∼70 and 20 vol.%, respectively, indicating the presence of very fine-grained sediment fractions which in combination with high OC contents result in a high OC loading (2.1 mg OC m −2 ). The relatively high CPI (5.5) as well as relatively high HMW/LMW ratios for n-alkanes (13.8) support the presence of relatively fresh organic matter in the cliff debris (Rielley et al., 1991;Strauss et al., 2015).

Constant vs. Rapid CO 2 Release
Considerable amounts of CO 2 were produced from mud debris and cliff debris over the two-months incubation period (Figure 3 and Table 2). The total amount of CO 2 produced was significantly different (ANOVA, p < 0.05) between both, mud debris ( Figure 3A) and cliff debris (Figure 3B), alongside with CO 2 production rates (Supplementary Figure 3 and Table 2). Mud debris continuously released CO 2 at a relatively low rate with a maximum production of 50.6 ± 4.5 µg CO 2 gdw −1 day −1 at the beginning, which decreased to 14.0 ± 1.8 µg CO 2 gdw −1 day −1 at the end of the incubation, yielding a total of 1.4 ± 0.1 mg CO 2 gdw −1 over a period of two months ( Figure 3A). Yet, when normalized to TOC content, the cumulative CO 2 production was almost three times higher from mud debris (93.4 ± 7.6 mg CO 2 g TOC −1 ) compared to cliff debris (36.1 ± 12.7 mg CO 2 g TOC −1 ). Outgassing of CO 2 from inorganic or petrogenic sources might have contributed to the observed CO 2 release as petrogenic OC is common in the study area and reactions of carbonates with sulphuric acid producing CO 2 were observed in thaw slump systems (Goñi et al., 2005;Drenzek et al., 2007;Zolkos et al., 2018). In contrast to the mud debris, cliff debris produced ca. three times more CO 2 , adding up to a total amount of 4.0 ± 0.9 mg CO 2 gdw −1 over the two months of incubation, which was likely caused by the higher initial TOC content (7.0 wt%) and OC loading (2.1 mg OC m −2 ; Table 2 and Figure 4). CO 2 production rates were fairly high within the first month. Maximum rates observed were 154 ± 25.7 µg CO 2 gdw −1 day −1 and decreased to 34.2 ± 12.5 µg CO 2 gdw −1 day −1 after two months of incubation.
The total amount of CO 2 released from mud debris mixed with seawater was virtually the same per gram sediment (1.4 ± 0.1 mg CO 2 gdw −1 ) and similar per g OC (96.4 ± 2.3 mg CO 2 g TOC −1 ) as without seawater added. The CO 2 production rates were also similar with a maximum of 48.5 ± 4.9 µg CO 2 gdw −1 day −1 at the beginning (day 7) and 23.7 ± 1.4 µg CO 2 gdw −1 day −1 at the end of the incubation. The most labile OC compounds may have been mineralized and released already as CO 2 during transport onshore before transfer with the mud lobe to the ocean (Vonk et al., 2012;Abbott et al., 2014;Sánchez-García et al., 2014). For cliff debris mixed with seawater, however, the total amount of CO 2 produced was significantly higher (5.7 ± 0.4 mg CO 2 gdw −1 , ANOVA, p < 0.05), corresponding to ∼1.5 times more CO 2 released compared to without seawater added. This pattern was similar but less pronounced when normalized to g OC with 51.0 ± 6.0 mg CO 2 g TOC −1 . The set-ups with seawater resulted in elevated CO 2 production rates with a maximum of 328 ± 99.3 µg CO 2 gdw −1 day −1 . The production was particularly high during the first month, which likely indicates the rapid decomposition of freshly exposed and labile OC (Knoblauch et al., 2013;Schädel et al., 2014). The production rate then decreased to 29.5 ± 6.5 µg CO 2 gdw −1 day −1 , a similar value as for the set-up without seawater added. An increase in CO 2 production with addition of seawater was also observed in earlier studies for thawing permafrost from the thaw slump FIGURE 3 | Cumulative aerobic CO 2 production from incubated mud debris (A) and cliff debris (B). Mud debris (MUD) and cliff debris (CD) were incubated without (black symbols) and with seawater (SW, blue symbols) for a duration of two months at 4 • C under aerobic and dark conditions. Cumulative CO 2 production from permafrost (PF) is displayed as a local reference for CO 2 production from in situ permafrost without seawater . The PF samples originate from the headwall of the thaw slump system from which the mud debris is released. The error bars show the standard deviation of the replicates used.
headwall, which might indicate a priming or co-metabolization effect (Bianchi, 2011;Tanski et al., 2019). Initial CO 2 production rates of mud debris were lower than those of in situ permafrost situated close to the sampling site in the thaw slump headwall (Supplementary Figure 3; Tanski et al., 2019). Yet, baseline production rates at the end of the two-month incubation period are slightly higher than from in situ permafrost. For cliff debris, initial as well as final CO 2 production rates were higher than for the permafrost reference site.
Following the CO 2 production in the first two months, the modeled CO 2 production and OC loss for the approximate length of the Arctic open-water season (∼four months) and an entire year, assuming that eroded sediments in the nearshore zone do not freeze-up anymore and that degradation is not only limited to the warm season, are substantially higher ( Table 2 and Supplementary Figures 4, 5). Model results explain ∼99% of the observed variations in CO 2 production for each individual sample and the residuals are usually unbiased and randomly scatter around zero. The only exceptions are residuals of cliff debris samples PI-14, PI-15 and PI-16 (Supplementary Figure 5), which show some pattern and hence should be dealt with cautiously. After four months, 2.7 ± 0.05 mg CO 2 gdw −1 would be released from mud debris onshore, corresponding to a TOC loss of 4.9 ± 0.1%, which is slightly lower for set-ups with seawater added. The CO 2 release from cliff debris onshore would be 8.3 ± 0.8 mg CO 2 gdw −1 with TOC losses of 3.2 ± 0.3%. This is also slightly lower when seawater is added. After one year upon erosion into the nearshore, mud debris would produce 6.1 ± 0.5 mg CO 2 gdw −1 with 11.1 ± 0.9% of the initial OC being lost. For cliff debris the CO 2 production would be much higher with 17.1 mg CO 2 gdw −1 corresponding to an OC loss of 6.9 ± 1.7%. Model results based on set-ups without seawater show that the CO 2 release would be generally higher onshore. Yet, degradation is then limited to only ∼four months per year, making the erosion and year-round degradation of OC in the nearshore much more effective. On decadal timescales microbes would decompose the more refractory carbon pools and thus support further CO 2 production at low rates, adding to the cumulative release of CO 2 (Knoblauch et al., 2013;Schädel et al., 2014).
Experimental incubations that account for environmental conditions along eroding permafrost coastlines or continuous measurements of CO 2 or CH 4 from erosion sites are scarce in the Arctic, and our data are therefore challenging to put into context. GHG measurements are available from eroding permafrost cliffs at Muostakh Island (Siberian Arctic) with daily production rates ranging from 0.1 to 19.4 mg CO 2 m −2 day −1 , indicating intensive CO 2 release to the atmosphere in the warm season during erosion onshore (Vonk et al., 2012). For terrestrial permafrost sites elevated CO 2 concentrations of > 1000 ppm were observed at thermoerosional features such as active layer detachment slides, gullies and thaw slumps, which also released substantial amounts of CH 4 . These features might closely resemble the biogeochemical response of permafrost and modern OC to coastal erosion onshore as these processes are common along eroding coastlines as well (Figures 1A,B). Once eroded into coastal waters, permafrost OC rapidly settles due to ballasting within mineral matrices but may further degrade during resuspension and lateral transport, thus further producing CO 2 (Vonk et al., 2012Semiletov et al., 2016;Bröder et al., 2018;Tanski et al., 2019;Jong et al., 2020).

Organic Matter Degradation vs. CO 2 Production
The release of CO 2 during the course of only two months indicates that organic matter is quickly degraded during the coastal erosion process. The production of CO 2 was positively correlated to TOC content and was higher in the OC-rich cliff debris than mud debris (Figure 4A), which implies that CO 2 release is likely a function of the availability of fresh and degradable organic matter. This is supported by strong 2 | Summary of observed and modeled cumulative CO 2 production and corresponding TOC loss in % as CO 2 -C after two months (60 days), regional open-water season of four months (120 days) and a full year (365 days) for incubation set-ups with mud debris (MUD), MUD with seawater (SW), cliff debris (CD) and CD with SW.
MUD MUD + SW CD CD + SW n = 5 n = 5 n = 3 n = 3 Incubation (60 days Maximum and end daily CO 2 production rates are reported additionally. All values are given as mean ± standard deviation with the sample size (n).
correlations between CO 2 production and initial n-alkane CPI values and TOC/TN-ratio ( Figure 4E, Supplementary Figure 6 and Table 1). Yet, relative TOC losses (% as CO 2 -C) were significantly smaller (ANOVA, p < 0.05) in cliff debris (1.6 ± 0.3%) compared to mud debris (2.5 ± 0.2%) ( Figure 4B). In general, the lower initial TOC content seems to correlate weakly with a higher (relative) TOC loss, suggesting that the old and OC-poor mud debris might be more prone to degradation than the relatively young and OC-rich cliff debris despite the lower total CO 2 release. This is expressed by the higher CO 2 production per g OC in mud debris (93.4 ± 7.6 mg CO 2 g TOC −1 ) compared to cliff debris (31.1 ± 12.7 mg CO 2 g TOC −1 ) (Figures 4, 4C). In contrast, CO 2 production is higher from cliff debris when normalized to SA (119.8 ± 26.2 µg CO 2 m −2 ) compared to mud debris (62.9 ± 7.0 µg CO 2 m −2 ) ( Figure 4D) as the mineral OC-loading for cliff debris is higher than for mud debris. The production is even higher in set-ups with seawater added, particularly for cliff debris (195.3 ± 15.0 µg CO 2 m −2 ) compared to mud debris (119.8 ± 26.2 µg CO 2 m −2 ). The cliff debris site seems to be accumulating OC-rich fine-grained sediments with high OC content, mineral SA and clay portions, which finally result in the higher CO 2 release, whereas in the coarser mud debris this material may already have been removed offshore with the thaw stream or redeposited within the slump. With addition of seawater, CO 2 release from cliff debris increased, yet not significantly (ANOVA, p < 0.05), alongside with TOC losses (2.2 ± 0.2%), whereas for mud debris CO 2 production was similar for the set-ups with and without seawater added. Although the higher relative TOC losses from mud debris imply that it is more prone to degradation than the relatively young and OC-enriched cliff debris, the addition of seawater resulted in an increase of CO 2 production and corresponding TOC loss only in cliff debris. This might suggest that within mud debris the most labile compounds were already removed, and not available for further degradation in seawater. During transport within the slump system, this removal may have occurred via melting of massive ice that promoted leaching and liberation of more labile OC compounds and mineralization to CO 2 or CH 4 onshore (Vonk and Gustafsson, 2013;Abbott et al., 2014;Mann et al., 2015;Tanski et al., 2017). The relatively low mineral SA and OC loading of mud debris compared to cliff debris also suggest that the finer sediment fraction hosting OC may have been removed quickly from the mud debris with the thaw stream to the sea (Jong et al., 2020). Incubated mud debris and cliff debris showed no significant changes in TOC contents, TOC/TN-ratios as well as in δ 13 C-TOC and 14 C-TOC signatures (Supplementary Figure 6, Table 1). This is not surprising given the minor TOC losses (max. 2.5% of initial OC as CO 2 -C) during these short-term incubations of two months, and considering measurement uncertainties and natural heterogeneity of organic matter in individual samples. On a molecular level, however, organic matter degradation seems to be more pronounced (Figure 4, Supplementary Figure 7, and Table 1). The CPI for n-alkanes decreased in all setups, indicating degradation of this organic matter fraction. In mud debris the CPI decreased from 4.6 ± 0.2 to 2.8 ± 0.3 and 2.4 ± 0.2 without and with seawater added, respectively ( Figure 4E). This is similar, although less distinct, in cliff debris, where the CPI decreased from 5.5 to 3.5 ± 0.1 and 4.1 ± 0.5. Although the marked decrease of the HMW/LMW-ratio of n-alkanes mainly indicates the reworking of plant wax lipids (Figure 4F), the increase of LMW-even to HMW-odd n-alkanes (Supplementary Figure 7) support the overall degradation of organic matter (Sánchez-García et al., 2014). Yet, LMW and HMW n-alkane concentrations increased in all set-ups indicating in situ production processes during the course of the incubation (Table 1 and Supplementary Figure 8). Therefore, the used degradation proxies cannot unambiguously reflect the OC degradation process as production probably distorts the signal.
The concentration increase of LMW and HMW n-alkanes was highest for set-ups without seawater added. In the incubated mud debris, n-alkane concentrations increased for LMW n-alkanes from 0.4 to 6.0 ± 1.0 (5.0 ± 0.9 µg gdw −1 when mixed with seawater) and for HMW n-alkanes from 7.6 to 18.5 ± 2.0 (15.5 ± 3.4 µg gdw −1 with seawater). In cliff debris, this pattern was similar with LMW n-alkane concentrations increasing from 0.5 to 8.0 ± 2.7 µg gdw −1 (6.6 ± 1.2 µg gdw −1 with seawater). For the HMW n-alkane concentrations, there were virtually no differences between set-ups with or without seawater added. Concentrations increased from 7.2 to 34.4 ± 4.2 FIGURE 4 | Cumulative CO 2 production normalized to gram dry weight (gdw) and %TOC loss as CO 2 -C at the end of the two month-incubation plotted against TOC content (A,B), cumulative CO 2 production normalized to TOC content (C) and surface area (D) plotted against initial TOC content and surface area and cumulative CO 2 production normalized to gdw plotted against CPI n-alkanes (E) and HMW/LMW n-alkanes ratio (F). Open symbols on the x-axis indicate samples from mud lobe and cliff that were not incubated to show the natural variability in the mud lobe and cliff. The mud debris (MUD) and cliff debris (CD) samples used for the incubation (T = 0) are indicated by the yellow symbols, incubated samples by the black symbols and incubated samples with seawater (SW) added by the blue symbols. and 35.0 ± 5.1 µg gdw −1 with seawater added, respectively. While HMW n-alkane increase by up to four times during the incubation, LMW n-alkane concentrations increased up to 14 times compared to the initial concentrations (Table 1,  Supplementary Figure 9). This substantial increase of n-alkanes under aerobic conditions, particularly LMW n-alkanes, may derive from heterotrophic microbial decomposers (Li et al., 2018;Chen et al., 2019). The predominance of even C-homologues in the LMW n-alkane fraction suggests a bacterial source and autochthonous production processes, which has been observed in estuarine and marine sediments (Grimalt and Albaigés, 1987). Under anoxic conditions, these LMW n-alkane C-homologues can also derive from reduction of fatty acids (Debyser et al., 1977;Dastillung and Corbet, 1978). Yet, as aerobic conditions were applied during the experiment this process was probably restricted. Although no light was applied during the incubations, increases of LMW n-alkanes may have been also caused by algal growth as certain algae communities can thrive under extremely low radiance conditions (Suzuki et al., 1997). The lower even C-homologues in the HMW n-alkane fraction may point to the reduction of fatty acids or wax alcohols during degradation (Kvenvolden, 1970;Welte and Waples, 1973). Finally, contamination of lab equipment or by septa could also have been a potential source of n-alkanes. Yet, without further evidence the cause of the n-alkane increase remains speculative and biomarker from the post-incubation results should therefore be dealt with caution. However, the relatively low standard deviation of replicate analyses seems to rule out the random effect to be expected from septum leaching or contamination (Supplementary Figure 9).

DOC Release and Acidification
Over the course of the incubation, a minor portion of the sediment OC pool (< 0.2%) was transferred into the DOC pool of the seawater. After the incubation DOC concentrations in seawater increased from initial 1.7 ± 0.03 mg L −1 to 7.2 ± 0.1 mg L −1 and 27.8 ± 1.0 mg L −1 with addition of mud debris and cliff debris, respectively. The DOC increase in seawater mixed with cliff debris was ∼four times higher mainly due to the higher initial TOC content (Supplementary Figure 10 and Supplementary  Table 2). This higher DOC release probably resulted in the higher CO 2 release as OC has to be in the form of DOC to pass cell membranes before metabolization by microbes (Battin et al., 2008;Vonk and Gustafsson, 2013). Thus, when normalized to initial TOC content, DOC yields for mud debris and cliff debris after 2 month of incubation are similar with 3.1 ± 0.1 and 3.6 ± 0.1 mg g −1 OC, respectively. This indicates that DOC leaching and adsorption processes with mineral surfaces are of similar magnitude for both debris types during the two months of incubation (Neff and Asner, 2001;Littlefair et al., 2017). Higher DOC yields (6.6 ± 1.7 and 7.4 ± 1.4 mg g −1 OC, respectively) have been observed in seawater mixed with mineral and organic-rich permafrost in previous studies, where permafrost had similar initial TOC contents (1.7 ± 0.2 and 7.8 ± 0.7 wt%, respectively) to mud debris and cliff debris, yet the incubation period was twice as long (Supplementary Figure 10; Tanski et al., 2019). This suggests that leaching capacity and DOC yield may increase over time or that DOC components prone to leaching were already leached from mud debris and cliff debris during transport onshore. A dynamic balance of hydrolysis from particulate OC to DOC and subsequent mineralization into CO 2 was probably established within days (Kristensen et al., 1995). This is indicated by stabilization of CO 2 production rates after the initial CO 2 release during the first weeks stemming from rapid mineralization of the most labile compounds. Leaching of labile compounds from plant debris or mineral-bound OC is more likely for cliff debris since labile compounds were probably still available as transit time onshore until accumulation at the cliff toe was short (Dou et al., 2008;Blair and Aller, 2012;Vonk et al., 2015). In combination with the initial liberation of labile DOC compounds, priming and co-metabolization might have caused the initial increase of CO 2 production in cliff debris mixed with seawater (Bianchi, 2011). Input of additional nutrients with seawater might have promoted decomposition of organic matter (Mack et al., 2004). Furthermore, input of nutrients from the adjacent cliffs topsoil would have likely supported microbial metabolic efficiency in the cliff debris  and thus CO 2 production. The increase of DOC concentrations in unfiltered seawater controls from 1.7 ± 0.0 to 2.6 ± 1.0 mg L −1 (Supplementary Table 2) indicate potential production during the incubations and may support the increase of LMW n-alkanes observed. This could have further increased the DOC content in the seawater, which was mixed with mud debris and cliff debris and influences the CO 2 release from the seawater.
In combination with increasing DOC concentrations, a substantial acidification of the seawater mixed with mud debris and cliff debris from pH 8.2 to 7.0 ± 0.1 and 7.0 ± 0.0, respectively, was observed ( Supplementary Figure 10 and Supplementary Table 2). This could have been caused by a low pH of soil organic matter within eroded debris or dissolution of CO 2 produced from the debris in seawater. An even stronger pH decrease from 8.4 ± 0.1 to 5.2 ± 0.2, was observed during a longer incubation experiment with organic-rich permafrost mixed with seawater, which was likely caused by the low initial pH (6.1 ± 0.1) of the permafrost and a high DOC yield (Supplementary Figure 10; Tanski et al., 2019). Depending on the ratio of sediment to seawater, acidification could occur during erosion, which was shown for coastal waters in the Siberian Arctic that receive large amounts of permafrost OC. This, in turn, could result in elevated pCO 2 and outgassing to the atmosphere (Semiletov et al., 2013(Semiletov et al., , 2016.

Incubations Cannot Account for Dynamic in situ Environmental Conditions
The measured CO 2 release rates and OC losses from eroded organic matter were obtained through experimental settings with low constant temperatures, maximum oxygen availability and dark conditions, which are classically used to study permafrost thaw through gradual active layer deepening (e.g., Lee et al., 2012;Knoblauch et al., 2013;Schädel et al., 2014). The CO 2 release observed under laboratory conditions would very likely differ from in situ observations. Yet, for coastal erosion processes onshore and in the nearshore coastal waters, environmental conditions are far from stable. The soil-to-water ratio under natural conditions is much lower resulting in a lower acidification as observed in the incubation vials. With a pH typical for ambient seawater (∼8.2), CO 2 production rates would probably remain similar but carbon storage in seawater would shift towards bicarbonate and carbonate (Wolf-Gladrow et al., 2007). The most important environmental parameters affecting the CO 2 production under natural conditions are most likely temperature and oxygen availability. Incubation experiments carried out at 15 or 16 • C (compared to 4 • C) showed that CO 2 production increased substantially compared to lower temperatures <4 • C (Dutta et al., 2006;Lee et al., 2012;Knoblauch et al., 2013;Schädel et al., 2016;Tanski et al., 2019). For our study area Qikiqtaruk -Herschel Island temperatures in the coastal waters depend on wind direction and coastal orientation and range between ∼5 and 14 • C during the warm season (Klein et al., 2019). This fits with observations from Cunliffe et al. (2019), which reported seawater surface temperatures between ∼6 to 10 • C close to our sampling sites. During the warm season, air temperatures onshore are on average ∼8 • C and soil temperatures in the active layer between 4.2 and 6.6 • C with maximum values of 9.8 • C (Burn, 2012;Wolter et al., 2016). The higher ambient temperatures would result in stronger OM mineralization rates onshore and in the nearshore under natural conditions compared to our incubation results. Using a Q10 value of 2 from Schädel et al. (2016), an average air temperature of ∼8 • C, and a conservative seawater temperature of ∼6 • C, the CO 2 production rates measured at 4 • C would be ∼32% higher onshore at 8 • C and ∼15% higher in the nearshore at 6 • C. With warmer in situ seawater temperatures the solubility of CO 2 also decreases which further promotes the release of CO 2 to the atmosphere from coastal waters.
Another important driving factor for OC mineralization is the availability of oxygen. In contrast to the steady aerobic conditions during our incubation, which support efficient OC degradation and CO 2 production, less efficient anaerobic CO 2 and CH 4 production may apply onshore, where oxygen cannot penetrate deeper layers of the mud debris and cliff debris or under water-logged conditions. Under anaerobic conditions, OC mineralization is about 3 times slower than under aerobic conditions (Schädel et al., 2016;Knoblauch et al., 2018). This would result in an approximately three times lower theoretical OC loss of in mud debris and cliff debris (0.4 and 1.3%, respectively), where anaerobic conditions apply under natural condition, compared to the observed OC loss of 1.1 and 3.9%, respectively, under aerobic conditions in the experiment. In contrast to less efficient anaerobic mineralization of OM, drainage and drying of eroded material onshore could further increase oxygen penetration into eroded debris and support aerobic carbon mineralization. Inter or intra-annual refreezing cycles could further enhance macro-pore spaces and infiltration capacity of sediments (Ding et al., 2019). This could supply oxygen to pore spaces or further mechanical destruction of organo-mineral bonds, which may cause short-lived CO 2 pulses (Blair and Aller, 2012;Höfle et al., 2013;Walz et al., 2017). Additionally, drying and rewetting cycles (caused by rain, waves or high tides) could also further amplify CO 2 production by microbial mineralization of OC and weakening the physical protection of the soil matrix (Fierer and Schimel, 2003).
During transport onshore or in the nearshore water column, photo-oxidation can promote DOC degradation (by up to 40%), which is coupled to CO 2 release (Cory et al., 2013). The input of fresh nutrients, organic matter or plant litter from erosion itself, surface runoff or seawater could trigger priming effects and co-metabolism of recalcitrant organic matter combined from active layer and permafrost, thus further promoting CO 2 release (Abbott et al., 2014;Wild et al., 2014Wild et al., , 2016Walz et al., 2017;Knoblauch et al., 2018). Light conditions could further promote those priming effects in the ocean between produced algal exudates and eroded OC (Bianchi, 2011). The majority of these factors would likely lead to higher OC loss and CO 2 production under natural coastal erosion conditions compared to the experimental incubations set-ups employed in this study. Once deposited in nearshore sediments, OM degradation pathways depend on sedimentation rates and oxygen penetration depth. In our study region sedimentation rates range from 0.1 mm yr −1 offshore to 2 mm yr −1 in the vicinity of the Mackenzie River but can reach up to 0.33 cm yr −1 at shelf locations which promote deposition (Harper and Penland, 1982;Grotheer et al., 2020). In combination with an oxygen penetration depth of a few mm to cm in marine sediments (Glud, 2008;Arndt et al., 2013), we would assume that aerobic conditions similar as in our experiment would apply for eroded and deposited sediments within the first years of deposition and during resuspension, winnowing or longshore transport. In subsequent years, anaerobic conditions may gradually take over in the initially deposited sediments, limiting efficient OC degradation and promoting OC burial on geological time scales (Hilton et al., 2015;Bröder et al., 2018;Grotheer et al., 2020). Furthermore, due to competition with marine sulfate reduction, most likely no CH 4 would be produced in the marine surface sediments .

CO 2 Release Dynamics Along the Yukon Coast
The erosion type and coastal morphology may strongly influence the GHG release potential along vulnerable permafrost coastlines. Our main study area, Qikiqtaruk -Herschel Island, is dominated by high cliffs (mean elevation ∼25 m) and retrogressive thaw slumps, which together affect ∼7% of the coast (Figures 1, 2, Lantuit and Pollard, 2008;Ramage et al., 2017). Thermoabrasion of the cliff tops and headwall retreat in thaw slumps may result in an initial CO 2 pulse onshore with up to ∼4% of OC mineralized into CO 2 within one season . Longer transport pathways of exposed organic matter along the sloping cliff faces or within thaw slump systems (Figures 1A,B) then result in a delayed release into the sea with further OC mineralization (∼5% OC loss as CO 2 measured during one warm season) at low rates (mimicked by mud debris incubations in this study) taking place in the subsequent warm seasons assuming the material discharges into the sea in the following years or even later. This adds to the initial permafrost CO 2 pulse onshore before release into the nearshore zone (Figure 5). Within thaw slump systems, eroded sediments can reside on long time scales onshore (years to decades) before release into the sea (Cray and Pollard, 2015). This long transit time can result in a substantial OC loss (up to ∼50%) with portions mineralized into CO 2 or CH 4 Tanski et al., 2017Tanski et al., , 2019. Similarly, eroded debris can further accumulate on the beach for longer time spans (potentially exceeding one warm season) before being washed into the sea during storm surges or high tides (Obu et al., 2015;Cunliffe et al., 2019). This transition stage could further promote reworking of OC and direct CO 2 release to the atmosphere onshore. Photooxidation processes may have further increased degradation and FIGURE 5 | Infographic summarizing potential organic carbon loss (%OC as CO 2 ) during coastal erosion processes from different erosion types in an Arctic warm season and entire year upon erosion in the sea. The magnitude of absolute carbon dioxide (mg CO 2 gdw) release is indicated by the size of yellow bubbles. Release of mud debris (gray flow) from thaw slumps or other thaw features into the nearshore is likely delayed with longer residence times onshore, resulting in a steady CO 2 release at low rates complementing CO 2 pulses during initial permafrost thaw and erosion. Cliff debris (brown flow) along low bluffs is likely to be transported more directly into the nearshore, resulting only in a short-term CO 2 pulse and overall low OC loss during transit (Figure 1). After erosion into the nearshore CO 2 is produced, which can be amplified by seawater, in particular for cliff debris and in situ permafrost where transport time is short. Carbon losses were modeled for the Arctic open-water season (∼four months) and the entire year based on short term (∼two months) incubation experiments at 4 • C under dark and aerobic conditions. Carbon loss rates from permafrost erosion are taken from Tanski et al. (2019) and displayed as reference for in situ permafrost (∼four month incubation at same conditions). The asterisk (*) indicates that once eroded into the sea OC degradation takes place during the entire year and is not limited to the warm seasons. Burial estimates are taken from Couture et al. (2018) and Grotheer et al. (2020).
CO 2 production during transit from land to sea (Cory et al., 2013;Ward and Cory, 2016).
In contrast, gully channels that intersect high bluffs, and thaw streams that drain slump systems (Figures 1A,B) may transport eroded permafrost OC more quickly and on a direct pathway towards the sea, potentially diminishing CO 2 release onshore. This direct pathway likely also applies to coastal stretches with low cliffs, which can be found especially on the southeast side of Qikiqtaruk -Herschel Island, which we tried to mimic with the cliff debris incubations. Along these cliffs (Figures 2, 5), approximately 3% of OC can be mineralized into CO 2 within one warm season during transit onshore. The debris along low cliffs is likely removed quickly within one season with only larger collapsing permafrost blocks remaining for a longer period onshore on the beach (season to years). Correspondingly, the initial CO 2 pulse is likely complemented by continuing CO 2 production at lower rates. For this study we assumed aerated conditions during transport. Yet, when eroded material gets redeposited in mud pools or lobes onshore, CH 4 production may replace CO 2 production if water-logged and oxygen-depleted conditions apply (Knoblauch et al., 2013;Schädel et al., 2016).
Upon release into the nearshore waters, OC is likely further mineralized within erosion plumes or deposited sediments. Seawater may further promote OC mineralization (Figure 5), especially if transport is direct (in the case of collapsed permafrost or cliff debris) and if the water column is oxygen-saturated.
Seawater enhances the CO 2 production from cliff debris with corresponding OC losses increasing from 1.6 ± 0.3 to 2.2 ± 0.2% after incubation with seawater mimicking nearshore conditions ( Table 2). Enhanced OC loss was also observed for in situ permafrost mixed with seawater with TOC losses increasing from 1.1 ± 0.0 to 1.5 ± 0.2% in organic-rich permafrost and from 3.9 ± 0.2 to 6.2 ± 1.2% in mineral permafrost . If transport to the nearshore is delayed, OC transferred to the sea probably continues to degrade at the same relatively low rates as onshore with OC losses of ∼5% within a warm season of four months (Figure 5). Yet, for both scenarios organic matter degradation in nearshore sediments likely takes place during the entire year and is not limited only to the warm season since nearshore sediments do not freeze up during the cold season. Once eroded into the nearshore, the total OC loss as CO 2 -C from mud debris and cliff debris could be ∼7-11% after one year (Figure 5). Depending on the depositional environment, OC can be either buried on the shelf or transported towards the continental slope on up to millennial time scales (Naidu et al., 2000;Goñi et al., 2013;Bröder et al., 2018). Locally, large portions of eroded OC (∼12 to 40%) are stored in nearshore sediments and local depressions on the shelf such as Herschel Basin (Couture et al., 2018;Grotheer et al., 2020). On the seafloor, effective aerobic degradation likely only takes place in the upper few millimeters of the sediment column and at the water-sediment boundary, where microbial communities allow for higher respiration rates (Rasmussen and Jorgensen, 1992;Arndt et al., 2013;Brüchert et al., 2018). Bioturbation, currents and ice scouring may facilitate resuspension and therefore oxygen resupply and CO 2 production in deeper sediment columns eventually (Héquette et al., 1995;Arndt et al., 2013). Under anaerobic conditions microbial communities that can produce CH 4 can establish after years and substantial amounts of CH 4 being produced and released either directly as CH 4 or oxidized as CO 2 from the seafloor Knoblauch et al., 2018). Although further GHG production of eroded material is very likely in the nearshore, it is unclear if or from which water depth produced CO 2 evades to the atmosphere. Outgassing of degradation-derived CO 2 or CH 4 from oversaturated water layers, however, can occur locally and is promoted by resuspension, storms or upwelling events (Mathis et al., 2012;Vonk et al., 2012;Semiletov et al., 2016;Pipko et al., 2017).
Along the adjacent Yukon mainland coast approximately one third of the entire coastline is characterized by actively eroding cliffs (of variable elevation), with the remainder being gently sloping tundra (∼17%) and inundated coastlines (∼11%) as well as beach and gravel spits . A delayed release similar to the high cliffs on Qikiqtaruk -Herschel Island may apply to the actively eroding cliffs in the formerly glaciated part of the Yukon coast towards the Mackenzie Delta. Here, cliffs are high (mean elevation of up to ∼30m) and thaw slumps are ubiquitous (up to ∼16% of the coastline) promoting substantial OC turnover onshore before release into the ocean (Couture and Pollard, 2015;Ramage et al., 2017). In the formerly unglaciated part towards the Alaskan border ( Figures 1C,D), cliffs are much lower (∼6.5 m mean elevation) and residence times of eroded OC onshore likely to be very short (weeks to years) due to the particularly high erosion rates in this region (∼1.4 m yr −1 ), which results in a shorter time span for OC mineralization during erosion (Couture and Pollard, 2015;Irrgang et al., 2018). Along coastal parts where tundra is sloping towards the beach and thereby restricting exposure of permafrost, OC loss might be even lower and the cycling of released OC may primarily take place in the seawater column or upon deposition in nearshore sediments. In turn, inundation and potential permafrost degradation could promote OC mineralization onshore and release substantial amounts of CO 2 before transport into the ocean . At present, how much OC is converted into CO 2 along the different coastal types is highly speculative. Yet, it is likely that the response of OC to erosion and the magnitude of CO 2 release varies between coastal and erosion types. Based on the annual TOC flux from land to sea of 36 Gg yr −1 for the Yukon coast (Couture et al., 2018), we would estimate that up to 1.8 Gg CO 2 yr −1 could be released as CO 2 during the erosion process onshore assuming an OC loss of up to ∼5% during the warm season and up to 3.9 Gg CO 2 yr −1 from nearshore sediments assuming an OC loss of ∼11% for an entire year since eroded sediments on the seafloor do not freeze anymore (Figure 5). Local dynamic environmental conditions (e.g., higher air temperatures and sunlight exposure) and processes (e.g., drying and rewetting of debris in the coastal zone) may increase OC degradation.

CONCLUSION
Coastal morphology and erosion type play a substantial role in determining the greenhouse gas release potential during erosion along permafrost coasts in the Arctic. Mud debris and cliff debris from an outflow of a thaw slump system and a low collapsing bluff, respectively, representing different coastal erosion scenarios released substantial amounts of CO 2 (1.4 ± 0.1 to 4.0 ± 0.9 mg CO 2 gdw −1 ) over a short period (∼two months) corresponding to an initial OC loss ∼1.6-2.6%. The relatively fresh and OC-rich cliff debris resulted in a ∼four times higher CO 2 release than for mud debris. Yet, when normalized to OC content, CO 2 release was ∼three times higher from mud debris indicating higher biodegradability of the older and more mineral OC. CO 2 release was particularly high with addition of seawater, indicating potential priming effects in the coastal realm. Upon release into the ocean, only a small portion of OC initially enters the seawater DOC pool. Our results suggest that high bluffs or coasts with presence of thermoerosional features are characterized by a delayed release that consists of an initial CO 2 pulse from permafrost erosion followed by a subsequent CO 2 release at lower rates in mud debris, promoting OC degradation during protracted transit from land to sea in the warm season. In contrast, erosion of cliff debris along low bluffs that may contain proportionally more OC results only in a short-term CO 2 pulse with OC quickly removed by waves. In the nearshore zone, CO 2 production from eroded sediments continues provided oxygen is available, likely taking place the entire year. After the duration of an Arctic warms season (∼four month) OC losses of up ∼5% were modeled and after one year, up to ∼11% of OC could be decomposed and released as CO 2 once eroded sediments are deposited in the nearshore. The magnitude of CO 2 release may be underestimated when using traditional incubation experiments since coastal environmental conditions are difficult to account for yet likely enhance degradation. More sophisticated incubation set-ups that mimic changing environmental and site-specific variables, or continuous GHG measurement systems onshore (e.g., eddy covariance towers) and in the nearshore (e.g., pop-up buoys with in situ pCO 2 sensors), accounting for seasonal fluctuations and uptake processes from vegetation and phytoplankton, would be needed to derive the actual GHG release signal along eroding permafrost coasts in different Arctic regions.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
GT, HL, DW, JV, and LB created the study design and incubation set-up. GT, DW, and TS carried out the gas concentration measurements. CK and GT recalculated gas concentration and estimated TOC losses. CB calibrated and applied the dynamic carbon decomposition model. GT and LB carried out the extraction of lipid biomarkers. TT measured lipid biomarkers. LB, JV, GT, and TT processed and interpreted the data. HL, MF, and JS supported the bulk geochemical and stable carbon isotope analysis and data interpretation. BK measured and evaluated DOC concentrations. LB, NH, and TE carried out radiocarbon analysis and data analysis. GT wrote the manuscript with contributions from all co-authors. All authors contributed to the article and approved the submitted version.