Particulate dominance of organic carbon mobilization from thaw slumps on the Peel Plateau, NT: Quantification and implications for stream systems and permafrost carbon release

Climate change is increasing the frequency and intensity of thermokarst, and accelerating the delivery of terrestrial organic material from previously sequestered sources to aquatic systems, where it is subject to further biochemical alteration. Rapid climate change in the glacially conditioned ice-rich and ice-marginal terrain of the Peel Plateau, western Canada, is accelerating thaw-driven mass wasting in the form of retrogressive thaw slumps, which are rapidly increasing in area, volume and thickness of permafrost thawed. Despite major perturbation of downstream sedimentary and geochemical fluxes, few studies have examined changes in flux and composition of particulate organic carbon (POC) in streams and rivers as a result of permafrost thaw. Here we show that the orders of magnitude increase in total organic carbon, nitrogen, and phosphorus mobilized to streams from thaw slumps on the Peel Plateau is almost entirely due to POC and associated particulate nitrogen and phosphorus release. Slump-mobilized POC is compositionally distinct from its dissolved counterpart and appears to contain relatively greater amounts of degraded organic matter, as inferred from base-extracted fluorescence of particulate organic matter. Thus, slump-mobilized POC is potentially more recalcitrant than POC present in non-slump affected stream networks. Furthermore a substantial portion of POC mobilized from thaw slumps will be constrained within primary sediment stores in valley bottoms, where net accumulation is currently exceeding net erosion, resulting in century to millennial scale sequestration of thermokarst-mobilized POC. This study highlights the pressing need for better knowledge of sedimentary cascades, mobilization, and storage reservoirs in slump-affected streams, and baseline assessments of the biodegradability of POC and cycling of particulate nutrients within a sedimentary cascade framework. Explicit incorporation of POC dynamics into our understanding of land-water carbon mobilization in the face of permafrost thaw is critical for understanding implications of thermokarst for regional carbon cycling and fluvial ecosystems.


Introduction
Abrupt climate-driven permafrost thaw (i.e. thermokarst) is prevalent across the circumpolar north (Olefeldt et al 2016, Kokelj et al 2017a and is accelerating in ice-rich glaciated landscapes (Segal et al 2016, Lewkowicz andWay 2019). Degradation of ice-rich ground liberates sequestered materials, saturates thawing soils and sediments, and enables mixing of various soil layers that may substantially differ in composition (Vonk et al 2015, Kokelj et al 2015, Lacelle et al 2019. Thermokarst on slopes can rapidly translocate large volumes (10 2 to 10 6 m 3 ) of previously frozen materials downslope (Kokelj et al 2015, van der Sluijs et al 2018 to new storage reservoirs in valley bottoms where material is subject to fluvial erosion and entrainment (Kokelj et al 2013), thereby exposing material to light, dissolution, and microbial degradation (Vonk et al 2015).
The growing need to include thermokarst processes (Kokelj and Jorgenson 2013) into Earth system or land surface models (Turetsky et al 2020) has advanced research on quantifying carbon storage in thermokarst landscapes (Olefeldt et al 2016, Fuchs et al 2018, 2019 and measuring carbon fluxes and composition from thermokarst features (Vonk et al 2015, Littlefair et al 2017, Tanski et al 2017, Ramage et al 2018. However, studies on organic carbon mobilization from thermokarst features have predominantly focused on dissolved organic carbon (DOC) (e.g. Mann et al 2015, Abbott et al 2015, Littlefair et al 2017, with effects varying from orders of magnitude increases in concentrations (Manning et al 2015) to little response or a decrease in concentrations (Littlefair et al 2017) due to a combination of landscape state factors that control organic matter accumulation in permafrost and its release (Ewing et al 2015, Mu et al 2017, Tank et al 2020. Most work examining thermokarst-mobilized particulate organic carbon (POC) flux has focused on coastal erosion (Vonk et al 2012, Tanski et al 2017, Ramage et al 2018, and studies detecting signatures of permafrost thaw in major northern rivers (Guo and Macdonald 2006, Guo et al 2007, Wild et al 2019, Bröder et al 2020, the latter of which have noted POC is more aged than dissolved or colloidal phases, thus reflecting a greater degree of mobilization from deeper permafrost deposits. Although work on Melville Island has shown that active layer detachments can increase the delivery of aged POC to stream networks Lafreniere 2014, Beel et al 2020), studies on POC mobilization in fluvial systems is extremely limited.
Organic matter is a vital source of energy and nutrients to the microbial base of aquatic food webs (Manning et al 2015). The composition of organic matter, such as its C: N: P stoichiometry, molecular structure, and degree of mineral association, can impact the relative importance of carbon transfer through food webs, relative to mineralization to CO 2 (Sardans et al 2012, Manning et al 2015, Welti et al 2017. Thermokarst may alter the relative abundance of organic matter from different sources (Wauthy et al 2018) and phases (particulate vs. dissolved) in recipient systems. If sources of organic carbon differ substantially in their composition, then rapid transfer of terrestrial material via thermokarst processes may have the potential to abruptly alter energy transfer to and through aquatic systems. 'Autochthonous' organic matter, derived from living material in the stream channel (e.g. algae) is generally considered to be more labile and of greater nutritional quality for upper trophic levels than terrestrial organic matter, partly due to the lower carbon-tonutrient ratios, higher algal polyunsaturated fattyacid content, and the complex amorphous structure of terrestrial organic matter (Guo et al 2016, Brett et al 2017. However, the origin and nature of terrestrial deposits can vary. Lipid degradation proxies have suggested that permafrost-sourced particulate organic matter (POM) in the Kolyma river is more labile than POM in a headwater stream with POM dominated by in-stream production and recent vegetative sources (Bröder et al 2020). Pautler et al (2010) also found elevated microbial activity in soil organic matter redistributed by active layer detachments on Melville Island, suggesting the release of labile soil organic matter from thawing permafrost. Many studies have found permafrost-derived DOC to be more biodegradable than relatively more modern DOC present in adjacent aquatic systems (Abbott et al 2014, Spencer et al 2015, Littlefair and Tank 2018. In some landscapes, this elevated lability has been attributed to the preservation and/or accumulation of biolabile compounds in permafrost (Ewing et al 2015, Strauss et al 2017, while in others it may be due in part to biochemical processes occurring when multiple soil profiles mix prior to delivery to stream systems (Littlefair and Tank 2018). However, the elevated lability of permafrost-derived organic matter is not consistent throughout the Arctic (Burd et al 2020, Wickland et al 2018 highlighting the need for contextualizing the fate of organic matter mobilized by permafrost thaw in relation to landscape state factors that may control the diagenetic state of terrestrial organic matter and how it is released during thaw (Tank et al 2020). But assessments on thermokarst-mobilized organic matter composition and lability are dominantly made for the dissolved phase. Across boreal aquatic ecosystems, POC can have significantly elevated mineralization rates relative to DOC (Attermeyer et al 2018), which suggest assessments based on DOC could be inaccurate for total organic carbon pools if POM and associated POC are the dominant form of organic matter and carbon mobilized to recipient aquatic systems.
One of the most dramatic forms of thermokarst is the retrogressive thaw slump (henceforth 'thaw slump'). Thaw slumps occur at high densities in glacially conditioned permafrost landscapes including ice-marginal moraine systems, and glaciofluvial, glaciolacustrine, and glaciomarine deposits (Rudy et al 2017, Kokelj et al 2017a, Ward Jones et al 2019. Thaw slumps develop as ice-rich permafrost thaws and collapses, and saturated materials flow over sloping terrain, transporting material downslope to streams, lakes, or coastal environments. These dynamic thermokarst systems predominantly affect low order streams where they overwhelm transport capacity and can transform long-term sediment flux (Kokelj et al 2017a, Kokelj et al under review). On the Peel Plateau in the western Canadian Arctic, a warming and wetting climate has intensified thaw slump activity (Kokelj et al 2015, Segal et al 2016). The region has one of the highest density of large thaw slumps in northwestern Canada with most features occurring in headwater streams of east-flowing tributaries of the Peel River (Kokelj et al 2017a). As thaw slumps erode hillslopes, they can increase sediment loads in streams by orders of magnitude (Kokelj et al 2013), and expose materials normally unavailable for transport to stream systems, such as glacial tills maintained in permafrost since the Pleistocene (Lacelle et al , 2019. The degree to which thaw slumping changes the nature and quantity of organic matter mobilized to streams may depend on the characteristics of surficial materials and thickness of permafrost thawed, in addition to ground-ice content, topography, and intensity of climate conditions that drive slope thermokarst processes (Kokelj et al 2015, 2017a, Lacelle et al 2019.
Here, we target three key goals. First, we quantify the relative magnitude of POC vs. DOC mobilization to stream water columns on the Peel Plateau. Second, we assess how thaw slumping affects the source and composition of organic matter present in recipient streams. Third, we examine how changes in the quantity and composition of mobilized organic matter varies in relation to thaw slump morphology and landscape variation. We undertake this work across a series of thaw slump features that vary in their morphology and landscape position as it relates to the nature of materials mobilized by thaw. Assessing POC mobilization and its overall composition is critical to understand and predict the impacts of abrupt permafrost thaw on aquatic ecosystems and carbon cycling, and to advance our understanding of the biogeochemical effects of permafrost thaw within aquatic networks.

Study region
The Peel Plateau is a 24 000 km 2 glacially conditioned landscape characterized by ice-rich, fluvially incised terrain that is representative of other permafrost preserved glaciated landscapes across northwestern Canada, Alaska, and Siberia (Kokelj et al 2017b) (figure 1). The headwaters of east flowing Peel River tributaries originate in mountainous unglaciated terrain where surficial materials are dominated by colluvium, and transition to glacial till with decreasing topographic relief towards the east (Duk-Rodkin and Hughes 1992, Kokelj et al 2017b). The transition from tundra to shrub to stunted boreal forest in lowlands is associated with the regional topographic gradient that generally occurs West to East (O'Neill et al 2015) and may affect the organic carbon content in active layer and Holocene-age permafrost (Dyke 2005).
The region's ice-rich, fluvially-incised landscape is characterized by a high density of thaw slumps (Segal et al 2016) compared to many glaciated permafrost landscapes to the east (Kokelj et al 2017b).
In the late-Pleistocene (ca. 18, 000 to 15, 000 cal. yr. BP; Lacelle et al 2013), the region was briefly covered by the Laurentide Ice Sheet (LIS). During this time, glacial tills, glaciofluvial and glaciolacustrine materials were deposited in a dynamic ice-marginal setting  and preserved as permafrost was established following glacial retreat (Lacelle et al 2004. Subsequent early Holocene warming was associated with an increase in active layer thickness to~1.5 fold that of present day conditions (ca. 9000 cal. yrs. BP, Burn et al 1997), and an acceleration of mass-wasting (e.g. thaw slumping) and development of colluvial deposits. These processes enabled soil formation, geochemical modification and the incorporation of organics to the depths of maximum thaw, followed by a cooling period and upwards aggradation of the permafrost table (Lacelle et al 2019). These modified deposits have been preserved in a 'relictthaw layer' situated immediately below present-day active layer soils, distinguished from the unmodified tills below by a thaw unconformity (Lacelle et al 2019). Thus, the relict-thaw layer can be comprised of a paleo-active layer and/or sediments deposited at the base of a hillslope from past slumping activity (colluvial deposits, Lacelle et al 2019).
Considering this history, we refer to three main categories of terrestrial material eroded by thaw slumps: (1) the currrent active layer, where soil has developed from the relict-thaw layer; (2) Holoceneage permafrost containing previously thawed diamicton (relict-active layer) and colluvium (slumped sediments); and (3) Pleistocene-age permafrost containing tills that have remained preserved within permafrost. Headwall height, scar zone area, and debris tongue length (figures 1(a)-(b)) vary by orders of magnitude (table 1, Kokelj et al under review), potentially affecting the relative contributions of different sources exposed in headwalls and their delivery to streams.

Site selection
Eight thaw slump sites were selected for sampling within the Stony Creek and Vittrekwa River watersheds on the Peel Plateau (figure 1, table 1). Sites were selected based on the following criteria in order of importance: (1) connection of channelized runoff to a stream network; (2) provision of a range of variation in slump morphology; and (3) availability of data from previous studies to build on a baseline body of knowledge (Kokelj et al 2013, Littlefair et al 2017. This selection was constrained by accessibility of sites due to the remote location of the region. The Stony Creek and Vittrekwa watershed were targeted due to the ability to access sites from the Dempster Highway.
Sites were visited two to three times from June to August 2015. Samples were obtained from: channelized runoff within each thaw slump (IN); a downstream location where all channelized runoff entered the valley-bottom stream (DN); and an unimpacted , and soluble reactive phosphorus (SRP). Dissolved organic nitrogen (DON; TDN-DIN) and organic phosphorus (DOP; TDP-SRP) were calculated from measured nutrient species. Any DOP concentrations below the limit of quantitation for TDP (4.9 µg l −1 ) were replaced with half the limit's value (supplementary S1.2). Suspended particulate material was collected on pre-weighed and pre-combusted (450 • C, 5 h) glass fibre filters (Whatman, GF/F) within 24 h of collection and stored frozen for analysis of POC & POδ 13 C, particulate nitrogen (PN) & Pδ 15 N, particulate organic phosphorus (POP), and PO 14 C. PO 14 C samples were collected from three of our eight sites (FM2, FM3, SD, capturing slump morphology range; table 1) on pre-combusted (500 • C, 5 h) quartz fibre filters (Whatman, QM/A, 2.2 µm). In 2016, sites accessible by foot (5 of 8) were re-visited to obtain samples to measure optical characteristics of base-extracted particulate organic matter (BEPOM) (Osburn et al 2012, Brym et al 2014. For further details, see supplementary S1.

Quantification of organic carbon delivery (Goal 1)
We calculated instantaneous yields of POC, DOC, and total organic carbon (TOC) upstream and downstream of thaw slumps using paired measurements of constituent concentration and stream discharge, and watershed areas delineated in ArcGIS (supplementary S4). We determined slump effects on TOC yields with linear mixed effects models using the R package lme4 (Bates et al 2015;supplementary S5). We included random slope and random intercept terms, accounting for repeated measures by allowing the slope (i.e. slump effect) and intercept to vary per slump site.

Source contributions to stream organic carbon (Goal 2)
Stream periphyton and headwall samples were collected at six slump sites in 2017 for POδ 13 C end member values (supplementary S6). Headwall sources consisted of an upper active layer (O-horizon), lower active layer (resembling A and/or B soil horizons), and Holocene and Pleistocene permafrost. Headwall samples were additionally analyzed for Pδ 15 N, and %POC, %PN, and %POP. FM2 and FM3 headwalls were also sampled for PO 14 C to match 2015 stream PO 14 C. Contributions of sources to upstream, within, and downstream POC were quantified using a dual carbon ( 13 C, 14 C) mixing model (MixSIAR, Stock et al 2018; supplementary S7).

Assessment of compositional changes in particulate organic material via geochemical analyses (Goal 2)
We used linear mixed effects models (section 2.4) to examine slumping effects on: POC:PN and POC:POP stoichiometry (Sardans et al 2012, Manning et al 2015; POδ 13 C and Pδ 15 N, used as metrics of organic matter source (Finlay and Kendall 2008) and stage of decay ( 13 C enrichment via anaerobic decomposition; Gundelwein et al 2007); and percent POC (%POC; as POC:TSS), used as a metric of organic carbon content of sediments. We further examined DOC:DON and DOC:DOP ratios to compare the dissolved and particulate response. We used principal components analysis (PCA) to (a) build upon our linear mixed models and assess broad patterns in geochemical parameters across sites, and (b) compare the geochemical composition of streamwater particulate organic matter (POM) to headwall sources (section 2.5). Variables with highly skewed and non-normal distributions were transformed prior to analysis.

Assessments of compositional changes via optical analyses of base-extracted particulate organic matter (BEPOM) (Goal 2)
Absorbance and fluorescence spectra of BEPOM were analysed following Osburn et al (2012) and Brym et al (2014). We calculated spectral slopes, slope ratios (S R ), and total absorbance from 250-450 nm per unit of suspended sediment (a tot/TSS , Helms et al 2008, Brym et al 2014) as proxies for molecular weight and relative quantity of extractable chromophoric organic matter per unit of sediment. A parallel factor analysis (PARAFAC) model was fit to excitation-emission matrices (drEEM toolbox in MATLAB; Murphy et al 2013) to resolve fluorescent components associated with fluorescent organic matter pools with different bulk molecular structure. Relative contribution of components to sample fluorescence was determined by normalizing maximum fluorescence intensities (F max ) of each component to the sum of component intensities for a given sample (%C = F max ⁄( ∑ F max )). Excitation-emission matrices were corrected and smoothed prior to any calculations and modelling. Further details are in supplementary S2. To aid component interpretations, we calculated commonly-used fluorescence indices (table S2), and assessed relationships between these indices and our PARAFAC components using PCA, transforming any highly skewed or non-normal variables (supplementary S3). Follow-up analyses on variables of interest were conducted using repeated measures ANOVA (supplementary S8).

Landscape and environmental controls on composition and delivery (Goal 3)
We used a linear model (lm() in R) to examine the relative importance of morphological, meteorological, and landscape effects on variation in upstream vs. within and upstream vs. downstream comparisons (i.e. slump effects) across slump sites. Slump effects were represented by extracting the random slopes per slump site from the aforementioned linear mixed effects models of TOC yield (section 2.4) and %POC (section 2.6, table 2). We ran two separate models for upstream vs. within and upstream vs. downstream random slopes. Explanatory variables consisted of: debris tongue length and maximum headwall height Maximum slope across the scar zone to immediate valley bottom was also used as a landscape variable for the TOC yield model. Through our study design we attempted to control meteorogical variation by sampling all sites close in time, but meteorological variables were included to account for the fact that sites were sampled on a series of sequential days. For further details on explanatory variables see supplementary S9.

The effects of slumping on TOC and nutrient delivery to streams (Goal 1)
POC accounted for the majority of organic carbon present within the channelized runoff of slumps. Within-slump POC concentrations reached greater than 9000 mg l −1 , while the maximum DOC concentration recorded was 33.6 mg l −1 . Within-slump sediment concentrations at some sites (FM2, SE) accounted for 10%-30% of the water volume, further decreasing per-volume DOC estimates and increasing the importance of POC fluxes and yields. Instantaneous total organic carbon (TOC) yields increased, on average, by an order of magnitude downstream of thaw slumps (p = 0.005, table 2, figures 2(a) and (d). Increases in POC yields accounted for 85%-100% of TOC increases, except for two out of three instances at SD, where POC accounted for 31% of TOC increases, and in one instance, TOC yield decreased downstream by~6 mg s −1 km −2 due to decreases in DOC yield (figures 2(a) and (e)). In 43% of our observations, DOC yields decreased downstream of slumps. With the exception of site SD, the high concentrations of particles mobilized from within-slump shifted streams from being DOC to POC dominated (figures 2(a), 3(a) and (d).
Increases in suspended sediment delivery and associated PN and POP to streams also led to orders of magnitude increases in instantaneous total nitrogen (as TDN + PN) and phosphorus (as TDP + POP) yields (figures 2(b)-(c), figures 3(b)-(c)). Greater DON and DOP concentrations within slumps relative to upstream did not result in elevated downstream concentrations relative to upstream (figures 3(e)-(f)). Across stream locations, POC:POP ratios were lower than DOC:DOP ratios (figure 3(l)). POC:PN ratios were also lower than DOC:DON and DOC:DIN ratios at all downstream locations with the exception of one sample downstream of SC (figures 3(j)-(k)).
High DIN concentrations within slump runoff did result in elevated concentrations downstream, largely driven by increases in NH 4 + (figure 3(h)). In contrast, no consistent difference was noted in SRP concentrations (figure 3(i)). TDN:TDP ratios were often more than double the Redfield ratio of 16:1, a guideline for nutrient limitation in marine and freshwater systems (figure 3(g)). With the exception of site SE, ratios were elevated within-slump, relative to upstream sites, sometimes leading to elevated ratios downstream relative to upstream (figure 3(g)).

The effect of slumping on sources of stream POC and DOC (Goal 2)
Terrestrial sources accounted for upwards of 80% of POC upstream, within, and downstream of sites incorporated in the dual-carbon isotope model (SD, FM2, and FM3, figure 4, table S5). Across all eight sites, upstream POδ 13 C was variable and sometimes more 13 C depleted than active layer sources (figure 5(b)), indicating differing contributions from 13 C depleted in situ sources (i.e. periphyton, POδ 13 C = −33.39 ± 4.19, mean ± SD) across sites. Slumping increased POC mobilization from both Pleistocene-age and Holocene-age permafrost, but POC from Pleistocene-age permafrost appeared to contribute more on average to withinslump and downstream POC ( figure 4(b)). The small permafrost contributions to upstream POC are due to incorporation of FM3-UP in the model, which began to experience streambank erosion and has more 14 Cdepleted POC than other upstream sites ( figure 4(a)). DOδ 13 C  and DO∆ 14 C (Littlefair et al 2017) suggest DOC is also predominantly terrestrial, with upstream DOC sourced from O-horizon soils ( figure 4(a)). Similar to upstream POδ 13 C, upstream DOδ 13 C values were more 13 C depleted than within-slump DOδ 13 C, suggesting relatively greater contributions from 13 C depleted periphyton to upstream DOC. However within-slump DOC is relatively more 14 C-enriched than within-slump POC (i.e. younger; figure 4(a)).
A principal components analysis of DOM and POM composition showed two major gradients (figure 5(a)). Only %POC, POC:POP, POC:PN, and POδ 13 C variables had arrow lengths greater than the circle of equilibrium for the PCA analysis. Therefore these variables contributed the most to dispersion of sites in the reduced space and we focus on them for description of the gradients. PC1 (29%) clearly separated within-slump and downstream sites from upstream locations along a gradient that was indicative of variation in POM diagenetic state and periphytic contributions to POM, as indicated by %POC and POδ 13 C. Along PC1, within-slump and downstream sites were more mineral-and 13 Cenriched relative to upstream sites, suggesting slump sediments had a greater proportion of degraded terrestrial material relative to fresher terrestrial material or in situ production ( 13 C depleted periphyton, figure 5(b)) characteristic of POC upstream (figure 4). No slump effect was apparent along PC2 (24%). Instead, PC2 reflected variation in POC: PN: POP stoichiometry. Runoff and upstream POM showed greater variability than downstream POM along this gradient, driven by nutrient-poor POM from active layer-dominated runoff (shallow slump SD) and an upstream site draining a catchment near the mountains (SE). This variation disappears for downstream samples, where POM is more universally nutrient enriched.
Incorporating headwall end-member sources into a PCA of POM geochemistry resulted in similar gradients of organic matter diagenesis/in situ contribution (along %POC and POδ 13 C; PC1) and nutrient composition (POC:POP and POC:PN; figure 5(b), PC2). Organic matter from Holocene and Pleistocene permafrost was compositionally indistinguishable between horizons and from most within-slump samples. The upper active layer (O-horizon) had a greater organic carbon content (%POC) than permafrost and lower active layer samples, was more closely associated with upstream and within-slump runoff samples from SD (the shallowest slump), and was generally more nutrient depleted than permafrost sources. Both upper and lower active layer sources had greater variability in POC:PN and POC:POP than permafrost sources, which clustered at the more nutrient enriched end of PC2. Decreasing C:N ratios from upper active layer to permafrost has also been noted by Lacelle et al (2019) and is attributed to carbon mineralization, and thus degradation of organic matter since production.
A PCA of BEPOM-derived optical characteristics revealed two clear gradients (figure 5(c)). The first was driven by contributions of C1, C3, and a tot/TSS . Increasing a tot/TSS indicates greater chromophoric organic matter per unit sediment (i.e. Stubbins et al 2014), while C1 and C3 contrast two pools of terrestrial humic-like materials (see below). Along this gradient, within and downstream locations had greater values of %C1, but lower values of %C3 and a tot/TSS than upstream locations. We found a significant difference in %C1, the variable most strongly associated with this gradient, between stream locations (repeated measures ANOVA, F 2,4 = 8.34, p = 0.011) due to differences between upstream %C1 (29 ± 3 se) vs. within (43 ± 3 se) and downstream (40 ± 3 se) locations (least-squared means; Lenth 2016). The second gradient in PCA space separated more biogeochemically-processed material (indicated by HIX, peak A:B and C:B; table S2) with greater molecular weight (S R , Helms et al 2008) from 'fresh' or protein-like organic material (%C5). Across all sites, within-slump and downstream locations tended to be most closely associated with the more processed end of this gradient ( figure 5(c)). The %C1-%C3 gradient described above reinforces that slump sediments are more biogeochemically processed. C1 strongly resembles C1 of Stubbins et al (2014), which was identified as, 'an aggregation of highly diverse, relatively high molecular weight' terrigenous molecules that were 'carbon-rich and nitrogen poor' . In contrast, C3 resembles Stubbins-C3 and -C4 which both had greater nitrogen content, less conjugation or lower molecular weight, and greater homogeneity, suggesting these components have undergone less reworking since production. Thus, these results are consistent with inferences made from POM bulk geochemistry that withinslump and downstream material are relatively more degraded since production than upstream material (section 3.3).
Contributions of permafrost-origin materials to within-slump and downstream POM increased with headwall height (figure 4(a), table S5). Holocene-and Pleistocene-age permafrost contributions to downstream POC increased from 18.5 ± 18.5% and 14.4 ± 15%, respectively (mean ± se) at SD (~2 m max. height, table 1) to 31.5 ± 34.4% and 50 ± 35.3% at FM3 (~10.3 m max. height) and 31.4 ± 32.1% and 49.6 ± 29.1% at FM2 (~25 m max. height). In particular, POC from Pleistocene-age permafrost doubled in contribution from SD to FM2. Within-slump POM molecular structure, inferred from BEPOM, also varied with headwall height, and this variation was similarly lost downstream. Within-slump BEPOM fluorescence ranged from being indistinguishable from upstream sites at the shallowest slump (SD), to being substantially enriched in %C1 at the largest features (FM2; table 1; figure 5(c)). This variation did not persist downstream as POM downstream of FM2 and SD became more similar to other slump sites.

Discussion
Slumping caused orders of magnitude increases in TOC and nutrient delivery to streams via the thaw-driven mobilization of sediments. This mobilization was dominantly in the form of POC, PN, and POP, shifting streams from dissolved to particulate dominated systems. Slumping increased permafrost organic carbon in stream networks, however within-slump POC values were more 14 C-depleted (older) than values previously recorded for DO 14 C (figure 4(a), Littlefair et al 2017) indicating POC mobilizes a greater degree of aged organic matter from deeper permafrost deposits. Optical properties of base-extracted POM indicated slump-mobilized POM was more processed since production than material present upstream.
Slumping increased TOC yields to a greater degree at sites with greater maximum headwall height and downstream stream power. Increasing thickness of permafrost thawed, associated with headwall height, may increase hydrologic input from melting of massive ice (Rudy et al 2017) which can increase sediment transport at baseflow (Kokelj et al 2013). Across the Peel Plateau, headwall heights and the depth of permafrost thawed generally increase as slumping intensifies and connects to streams, with slumpstream connectivity typically occurring at maximum headwall heights greater than~2 m (Kokelj et al under review). Thus, as slumping intensifies, stream TOC yields may increase due to increased delivery of POC. However this increase will also depend on the stream's transportation potential (i.e. stream power), which we discuss further in section 4.2. Landscape position also played an important role as TOC yields increased to a greater degree westwards on the Peel Plateau, with closer proximity to late-glacial limits, where elevation and relative relief is greater, tundra vegetation is sparser, and fine-grained tills, with a greater abundance of glaciofluvial/glaciolacustrine deposits, are veneered over bedrock (Duk-Rodkin and Hughes 1992). Since maximum scar-valley slope was not a significant predictor of TOC yields, the combination of sparse vegetation and fine-grained, ice-rich deposits veneering bedrock appear to make the terrain near Table 3. Effects of morphological, landscape, and stream power variables on variation in log(TOC yield) and (%POC −1 ) across slump sites. Only explanatory variables retained in final models are shown. Estimates are standardized. Starting variables for all models consisted of maximum headwall height, debris tongue length, total rainfall over the past 96 h, longitude of the slump site, and downstream stream power (except for the upstream: within model). Maximum slope was also used for the TOC yield model. No explanatory variables found to be significant a stream power was not entered as a variable in the starting model b retained to improve model residuals the western margin of the glacial limit particularly susceptible to thaw-driven erosion. Headwall height and landscape position also affected the organic carbon content of within-slump particles and divergence from upstream particles, with headwall height also increasing the proportion of more decayed organic matter of within-slump POM (section 3.5). However, we found no significant predictors of differences in composition between upstream and downstream %POC, and site variation in upstream-towithin BEPOM differences decreased downstream of thaw slumps, despite the fact that Holoceneand Pleistocene-age permafrost contributions to downstream POC were greater at sites with greater headwall heights. We discuss this further in section 4.3. Finally, though our study focused on actively eroding slumps, slump stabilization, which can occur if collapsed debris accumulates at the base of the headwall and insulates exposed ice (Kokelj et al 2015), was evident at FM4 where most of the headwall was not actively eroding during our sampling period. This may be why TOC concentrations did not increase substantially and within-slump POC, PN, and POP concentrations are more similar to SD, the smallest slump in this study. Notably, POC yields still increased downstream of slump FM4, likely due to stream erosion of the~1 km debris tongue.

Significance of thermokarst-derived POC and implications for permafrost-carbon release and fate
Particulate dominance of organic carbon mobilization is likely common in hillslope thermokarst systems throughout the circumpolar Arctic, based on previous records of high POC and sediment con- On the Peel Plateau, slump-mobilized POC appears to be at a greater stage of decay than relatively more modern material present upstream, as inferred from depleted organic carbon content, and relative increase in a fluorescent component (C1) that has been described as a diverse array of relatively high molecular weight molecules (Stubbins et al 2014). This suggests reduced bioavailability of stream POC. This is in agreement with Lacelle et al (2019) where 13 C NMR analysis of slump headwalls has shown that active layers contain young and more biodegradable forms of organic carbon (O-alkyl-C and narrow Alkyl-C), while undisturbed permafrost and relict-thaw layers, if not comprised of colluvium from past slumping , contain older and relatively more recalcitrant aromatic compounds. Calibrated dates of Pleistocene-age permafrost in this study are > 40 000 yrs B.P. (table S4), in agreement with previous work that suggests a significant amount of soil organic carbon within Pleistocene-age permafrost likely originates from regional vegetation growing prior to the advance of the LIS (Lacelle et al 2019). However, shale bedrock, including black shales with > 2 wt% TOC (Allen et al 2015), was likely also eroded and entrained in glacial till (Norris 1985, Calmels et al 2007. Thus petrogenic organic carbon, which is 'radiocarbon dead' , cannot be excluded as another important component of material preserved within Pleistocene-age deposits.
Our findings of relatively more decayed slumpmobilized POC broadens previous findings that slump-derived DOC on the Peel Plateau has reduced aromaticity and molecular weight, and elevated biolability compared to that from unimpacted streams (Littlefair et al 2017, Littlefair and. This suggests that DOM and POM pools are compositionally distinct and slumping affects the two pools in contrasting ways. Compositional differences could be due in part to different source contributions to the POC vs. DOC pools. ∆ 14 C values suggest that organic carbon from Pleistocene-age permafrost contributes proportionately less to DOC than POC, which is further supported by ground ice measurements on the Peel Plateau that show that DOC concentrations are greater in Holoceneage permafrost (15.36 ± 14.16 mg l −1 ; Zolkos and Tank 2019) than Pleistocene-age permafrost (2.58 ± 2.19 mg l −1 ; mean ± SD, n = 8). Furthermore, DOM stoichiometry is starkly different from POM stoichiometry, the latter of which closely resembles Holocene and Pleistocene headwall organic matter stoichiometry. This difference could be due to ongoing processing of DOM in permafrost porewaters (Ewing et al 2015) and partial sorption/desorption that preferentially enables aliphatic moieties to remain dissolved (Littlefair and Tank 2018). These findings are corroborated by work on Herschel Island, an ice marginal feature along the Yukon north coast, where the degradation of permafrost material exposed in slump headwalls was largely driven by degradation in the dissolved phase (Tanski et al 2019). Given that tills on Herschel Island contain a greater proportion of marine sediments, and thus likely differ somewhat in their composition to those from the Peel Plateau (Kokelj et al 2002, Lane et al 2012, Tanski et al 2017, this result suggests that our findings may be relevant across the broader glacial margin region of the western Canadian Arctic. In terms of the broader Arctic, our finding of more decayed slump-mobilized POC does contrast with recent findings that permafrost-influenced POM in the Kolyma River is less degraded than that derived from a headwater stream primarily influenced by instream production and recent vegetation, as assessed by lipid degradation proxies (Bröder et al 2020). This difference may be due to differences in the sources and genesis of permafrost between the two regions. While permafrost on the Peel Plateauwas formed following the deposition of glacial tills within which petrogenic organic carbon is likely an important component (see below), permafrost thaw in the Kolyma River region unearths deposits in the Yedoma domain, formed from the simultaneous accumulation of ground ice and sediment/organic matter deposition (syngenetic) in non-glaciated regions typically with low topographic relief (Schirrmeister et al 2013, Strauss et al 2017. Contrasts in geological origins of these two permafrost environments may also influence the nature of material mobilization. The remobilization of sediments and organic carbon to fluvial networks in the Yedoma domain tends to be documented along larger rivers and coastlines, while direct impacts to fluvial networks on the Peel Plateaupredominantly occur in headwater streams where fluvial incision has engendered high topographic relief (Tank et al 2020, Kokelj et al under review). Thus we need more studies examining/including thermokarst-POC mobilization in different landscapes to understand its release and fate in relation to landscape state factors (Tank et al 2020).
Petrogenic organic carbon, likely an important component of Pleistocene-origin POC in this study, is generally considered resistant to remineralization (Blattmann et al 2018). However, the balance between oxidation of petrogenic organic carbon, a source of

Implications of sediment transport limitation for permafrost-carbon fate
Our study focuses on suspended particles, yet the majority of terrestrial material eroded via slumping accumulates within valley-bottom debris tongues (Kokelj et al under review, Kokelj et al 2015, van der Sluijs et al 2018. Fluvial erosion of these deposits are limited by the stream's transport capacity as illustrated by the significant role stream power plays in explaining suspended TOC yield variations in this study. At FM2 and FM3,~30% of the terrestrial volume excavated by slumping is contained within slump debris tongues (Kokelj et al under review, van der Sluijs et al 2018). When accounting for the fact that the upper 10 m of permafrost can contain greater than 50% ground ice content (Malone et al 2013, Lacelle et al 2015, the debris tongue likely contains greater than 80% of the soils and sediments evacuated by slumping, since melted ground ice is lost via runoff and evaporation. Debris tongues at FM3 and the immediately downstream FM2 have estimated volumes of 19.5 and 1.5 10 5 m 3 (van der Sluijs et al 2018, Kokelj et al under review). If we assume a single density of sediments (2.65 g cm −3 ; Armanini et al 2018) and estimate 2% of sediment is POC (using suspended sediment %POC data from FM2-DN and FM3-DN in this study), this 15.95 km 2 catchment contains~1.1 × 10 11 g slump-origin POC in debris tongue deposits. We estimate that POC may remain within these primary sediment stores (Murton and Ballantyne 2017) for hundreds (~1.4 × 10 2 ) to thousands (~1.1 × 10 3 ) of years. This estimate is based on scaling of average downstream suspended POC fluxes (FM2-DN, this study) through the active thaw season (April 30th to September 30th inclusive, 154 d, O'Neill et al 2015), with an error range of −75% to + 100%. Our error range is roughly based on differences noted between scaling average SD fluxes vs calculations with turbidity calibrated by TSS flux and %POC data (supplementary S10) and uncertainty in the duration of active stream erosion throughout the year. This storage estimate does not take stream beload transport into account, and we recommend future studies include measurements of bedload transport to increase accuracy of storage estimates.
The intensification of thaw slump activity and disturbance enlargement is causing a rapid, net accumulation of materials in valley bottom debris tongue deposits (van der Sluijs et al 2018, Kokelj et al under review). Though not located in the valley bottom, preserved organic matter within colluvial deposits of headwalls suggests that the accumulation of thawed material may inhibit organic matter decomposition (Lacelle et al 2019, Zolkos andTank 2020). POC that is transported away from the thaw site by the stream will also fluctuate between deposition and transport, dependent on particle density and size. When settled out, mineralization of organic carbon can be reduced 50% (Richardson et al 2013) or more, particularly if contained in anoxic sediments (Peter et al 2016), though DOC can diffuse from sediments to be mineralized in oxic waters (Peter et al 2016). Across the western Canadian Arctic, the volume of terrestrial material excavated from landscapes by slumping is accelerating ( . This suggests that hillslope thermokarst models, where the majority of carbon is laterally exported and a substantial proportion is assumed to be mineralized (e.g. one-third, Turetsky et al 2020), may overestimate carbon mineralization. Improved prediction of the fate of permafrost carbon released by slope thermokarst requires that sediment cascade frameworks, and assessments of organic carbon reactivity within that framework, be integrated into modelling efforts.

Ecological implications for stream systems
Increasing sediment mobilization due to thermokarst activity can directly decrease foodweb production by decreasing stream benthic invertebrate abundance (Chin et al 2016), and by shifting food webs to become more reliant on decomposition of terrestrial organic carbon than in-stream primary production, as a result of the negative effects of sediments on autochonous production (Levenstein et al 2018). Thus, the composition of terrestrial inputs may play an important role in changes in stream ecological function and the basal energy resources that support higher trophic level production.
Differential organic matter composition is a wellestablished driver of variation of microbial metabolism (Sinsabaugh and Follstad Shah 2012, Ward and Cory 2015, Roiha et al 2016, Panneer Selvam et al 2017, Jain et al 2019. Thus, the divergent composition of permafrost-origin DOM and POM in this study suggests that DOM-based assessments of microbial production and mineralization rates are unlikely to be accurate for POM. In our study, the low POC:POP ratios relative to DOC:DOP, are closer to mean C:P values of aquatic microbial biomass obtained from the literature (166 with approximately three-fold variation; Sinsabaugh et al 2013), suggesting that microbial biomass production could be greater on POM than estimates based solely on DOM, if organisms are P-limited and if the lower C:P ratios in POM is not solely due to living microbial biomass Follstad Shah 2010, Franklin et al 2011). TDN:TDP ratios were typically more than double Redfield ratios and, in some cases, increased downstream due to high TDN:TDP within slump ratios, suggesting limited phosphorus supply in the dissolved phase and that POM could be an important source of phosphorus to microbes. Yet the diagenetic state of POM in this study suggests low lability, thus low C:N and C:P ratios may be due to incorporation of N and P in humic complexes (Grandy andNeff 2008, Sinsabaugh andFollstad Shah 2011). In addition, some particulate nitrogen may be inorganic nitrogen sorped to clay structures (Schubert and Calvert 2001). Complexation in combination with non-competitive sorption of enzymes to mineral and organic colloids in particle-rich environments, may decrease metabolic efficiency (Schimel andWeintraub 2003, Grandy andNeff 2008) and microbial growth rates (Moorhead and Sinsabaugh 2006). This highlights the need to include particleassociated microbial communities and enzymes in assessments of rates of microbial production and mineralization (Sinsabaugh and Shah 2012) to understand how the abrupt shift in the quantity and composition of organic matter delivered to streams may affect resource flow through stream food webs (Sinsabaugh and Shah 2013), and thus stream ecological function and it is role in regional or global biogeochemical cycles.
Despite differences in within-slump POM, consistent convergence of downstream POM composition suggests that slump-POM material is well mixed prior to entry into modern stream networks. Headwall ablation, accumulation of a saturated slurry of thawed materials, and downslope sediment transport by surface flow and deeper seated flow contributes significantly to physical mixing of slumped materials (Kokelj et al 2015, van der Sluijs 2018) that were previously distinguishable in stratigraphy. Furthermore, biochemical processing within the slump scar zone (Abbott et al 2015, Tanski et al 2017 may be enhanced by the warm and wet conditions that distinguish this environment (Kokelj et al 2009, van der Sluijs et al 2018. This suggests that compositional differences between slumps may be less important than the quantity of POM for driving differences in downstream microbial processing between slump sites and/or that the initial unimpacted stream system state is important in understanding the effect of slumping on change to a stream system since upstream locations in this study had greater variation than downstream locations. Further longterm studies of changing biogeochemical function as slumps undergo variations in headwall erosion rates would help assess this. Of course, this interpretation is limited by the sensitivity of our analyses. Further work detailing specific organic compounds (e.g. via solid-state NMR, Mccallister et al 2018; or Pyrolysis-Gas Chromatography-Mass Spectrometry, Py-GC/MS; Ma et al 2018) is required. In addition, the shift of streams from DOC to POC dominated systems highlights the need to use tools that can be applied to both dissolved and particulate phases of organic matter and associated microbial activity.
Finally, slump-associated particle release along fluvial networks creates potential for ecological and biogeochemical effects to be carried far from the site of thaw. The headwaters of the Peel Plateau drain into the Peel River, which has an annual sediment flux greater than most major Russian-Arctic Rivers (Holmes et al 2002). Currently, the contribution of thermokarst-derived sediments and POC to this flux is an important unknown. Using turbidity data and relationships between turbidity and TSS, POC, PN, and POP (supplementary S10) we estimate that POC, PN, and POP yields downstream of the smallest slump in this study (SD) are approximately 3.0 × 10 5 , 3.9 × 10 4 , and 1.0 × 10 4 kg km -2 from mid-June to mid-August (supplementary S10, figure  S7). These estimated yields are orders of magnitude greater than the combined dissolved and particulate carbon and nitrogen yields of any of the major Arctic rivers, including the Mackenzie River (Mcclelland et al 2016). Thaw slumping along these headwaters is thus likely to increase downstream nutrient and organic carbon subsidies and act as a hotspot for nutrient and organic matter transfer from land to fluvial networks. Though the majority of material excavated from thaw slumps is being temporarily contained within debris tongue deposits, these primary sediment stores are readily available for stream transport, and will cascade through a sequence of storage reservoirs through downstream systems over the coming millennia (Kokelj et al under review). This also highlights, as previously noted (Kokelj et al 2015, Beel et al 2018, that hydrometeorological change will play an important role in propagation of effects to downstream systems.

Conclusions
In permafrost regions where slope thermokarst is a dominant mechanism of thaw-driven change, we show that the vast majority of organic carbon mobilization to stream networks occurs via POC. Particulate organic matter mobilized from thaw slumps, and associated POC, is compositionally distinct from the dissolved phase, in part due to the range of distinct material types that have been preserved by permafrost. This will likely translate into differences in carbon mineralization rates and roles in nutrient cycling between the two phases. Estimated yields from slumps suggest that these features create a hotspot for particulate C, N, and P release to fluvial networks. While the majority of material mobilized by thaw slumping is deposited in valley-bottom stores, these massive repositories of readily-mobilized material can switch streams from dissolved-to particulatedominated systems and will create a step change in the ecological and biogeochemical functioning of downstream systems that is irreversable over century to millenial timescales. The dominance of sedimentassociated C, N, and P release from thaw slumps highlights the pressing need for better knowledge of sedimentary cascades, mobilization, and storage reservoirs in slump-affected streams, and baseline assessments of the microbial processing of POM and cycling of particulate nutrients within a sedimentary cascade framework.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10.21963/13160. Firth, Elizabeth Jerome, Billy Wilson, Keith Colin, Rosemin Nathoo, and Erin MacDonald. Cara Littlefair provided invaluable advice on field logistics, and was involved in initial site selection of HA, HB, HD, and SD. Shawne Kokelj provided meteorological data. Alberto Reyes provided advice on headwall source sampling. Christopher Osburn advised laboratory BEPOM analyses and he and Ashley Dubnick provided advice on PARAFAC modelling. Christine Ridenour assisted with BEPOM analyses. Casey Beel provided advice on interpretation of turbidity sensor data. Research was financially supported by the Aurora Research Fellowship, Garfield Weston Foundation, Northern Scientific Training Program, UAlberta North, Natural Sciences and Engineering Research Council of Canada, Environment Canada Science Youth Horizons Program, Arctic Institute of North America, and Polar Continental Shelf Program. Northwest Territories Geological Survey (NTGS) contribution #0128. Finally, thank you to two anonoymous reviewers whose comments improved this manuscript.