of Low biodegradability of particulate organic carbon mobilized from thaw slumps on the Peel Plateau, NT, and possible chemosynthesis and sorption effects

Abstract. Upon thaw, permafrost carbon entering streams may be mineralized into CO2 or re-sequestered into sediments. The balance between these processes is an important uncertainty in the permafrost-carbon-climate feedback. Warming and wetting in the western Canadian Arctic are accelerating thaw-driven mass wasting by permafrost thaw slumps, increasing particulate organic carbon (POC) delivered to headwater streams by orders of magnitude. Using aerobic incubations of POC from streams affected by thaw slumps we find that slump-mobilized POC undergoes minimal (~4 %) oxidation over a 1-month period and may be predominantly destined for sediment deposition. Mobilization of mineral-rich tills in this region may also protect carbon from mineralization via adsorption to minerals and promote inorganic carbon sequestration via chemolithoautotrophic processes. Region-specific assessments of permafrost carbon fates and inquiries beyond organic carbon decomposition are needed to constrain drivers of carbon cycling and climate feedbacks within stream networks affected by permafrost thaw.



S1 Field Processing
Stream water upstream and downstream of slump sites was collected in 10L and 4L LDPE cubitainers respectively (pre-leached with deionized water for ~24hrs). Within-slump samples were collected in acid-leached 250 -500 mL HDPE/polyethylene bottles. In 2016, additional samples were taken at downstream sites in acid-leached 500 mL HDPE cubitainers to add as unfiltered water. All containers were triple sample rinsed.
Cubitainers containing upstream water were thoroughly shaken and a portion was decanted into a clean 4L LDPE cubitainer and stored in the dark at 4ºC until the start of the experiment.

S2 Detailed Experimental Set-up
Prior to the start of the experiment, all sample water was brought to room temperature (~20 -25 ºC) to prevent changing oxygen concentrations and volumes over the course of bottle filling. Experiment bottles were first filled with filtered water by drawing the appropriate filtered water sample from the cubitainer through pre-vacuumed Tygon tubing. The outlet of this tubing was placed at the bottom of the bottle to prevent air bubble formation. Due to the extremely low particulate concentrations in upstream samples, unfiltered upstream bottle treatments were also filled this way, with cubitainers gently shaken to maintain homogeneity of unfiltered water. For slump-affected treatments, unfiltered slump runoff and/or downstream water was pipetted into bottles filled with the respective filtered water. This tended to result in particulate concentrations being more dilute than in situ measurements. Particle suspension was maintained prior to pipetting either by inverting bottles several times prior to pipetting or using a magnetic stir bar. Pipette tips were cut to prevent clogging. Cutting tips did not affect the volume of water drawn up, which we tested by pipetting deionized water volumes with cut tips on a weigh scale. Bottles also contained pre-sterilized and pre-combusted glass beads (MP Biomedicals, Roll & Grow TM Plating Beads, MP115000550) to ensure particle suspension. In 2015, the 120 mL serum bottles were sealed with Balch blue butyl stoppers. In 2016, glass BOD bottles were used so bottles were sealed with glass stoppers wetted around the rim to ensure an air-tight seal and capped with a plastic cap to prevent any evaporation. In 2019, 120 mL serum bottles were used again and capped with grey chlorobutyl isopropene stoppers (Niemann et al., 2015).
Experiment bottles were set up in replicates of 3 -4, with bottles processed immediately and at the end of the experiment for changes in organic carbon concentrations. Bottles processed at the end of the experiment contained oxygen sensor spots (details below) to continuously monitor oxygen concentrations during the experiment. In 2015, we intended to process bottles after 7 and 28 days, following timepoints recommended for BDOC incubations (Vonk et al., 2015) and previously used for BDOC incubations in our study region (Littlefair et al., 2017). However, due to rapid oxygen consumption at some sites, experiment bottles intended for 28 days of incubations had to be removed at 11 days to prevent anoxic conditions in the bottles that would not be reflective of stream water column conditions. Thus, for 2015 experiments we present results after 7 days of incubation. For following experiments, we aimed to conduct incubations for long enough to see a detectable change in carbon measurements, based on the declines in oxygen concentrations, while preventing experiments from going below 2 mg L -1 O2. This resulted in an incubation duration of 8 days in 2016 and 27 days in 2019.
For the fractionation experiment, slump runoff collected in 2016 was stored in the dark at ~4°C until fractionation which was completed within 48 hours of collection. Briefly, 0.5 mL was poured sequentially through a 2mm 3-inch SS mesh sieve, a 63 µm sieve, and a 20 µm sieve and shaken with a cover and pan. Material that passed through the 20 µm sieve was passed through a 0.65-micron PES filter on a filter tower. Material that passed through the filter was discarded and material collected on the filter rinsed into a beaker for use in the fractionation experiment. GF/F filtered water collected immediately downstream of slump SE was used to help rinse particles from the 0.5 mL sample through the fractionation process and make each fraction back up to an appropriate volume to compare with the unfractionated treatment, where 0.5 mL of slump runoff was diluted with filtered downstream water as necessary to approximate downstream concentrations (Table S2). There was error during this process since the sum of TSS across fractions was 61% of the TSS in the unfractionated sample (Table S2), but we still believe the relative proportion of material in each fraction is representative of the environment and provide the unfractionated control as a reference throughout.
Bottles were suspended either by placing them within a pre-drilled PVC pipe (

S3.1 Oxygen
Oxygen concentrations were measured within air-tight bottles with SP-PSt3-PSUP-YOP-D5 oxygen sensor spots (PreSens GmbH, Regensburg, Germany, Warkentin et al., 2007). These sensor spots were attached to the inner wall of glass bottles with silicone glue. Molecular oxygen quenches the luminescence of inert metal porphyrine complex immobilized in an oxygen-permeable matrix. The photoluminescence lifetime of the luminophore within the sensor spot was measured by a fiber-optic oxygen meter (Fibox 3; PreSens GmbH) placed at the center of the spot sensor outside of the glass bottle. Excitation light (505 nm) was supplied by a glass fiber, which also transported the emitted fluorescence signal (>600 nm) back to the oxygen meter. The method does not consume any oxygen. The average of ~5 fluorescence measurements was used to determine the oxygen concentration at a timepoint. The standard deviation of these replicate measurements was always less than 0.35 mg/L. There are three factors that can interfere with oxygen measurements: (1) temperature, (2) pressure, and (3) salinity. Temperature can affect the fluorescence lifetime and the solubility of oxygen in water.
Both effects are compensated for using simultaneous measurements from the temperature probe placed in an identical water experiencing the same environmental conditions as experimental bottles. Pressure does not affect the spot sensor's measurements of mg/L of oxygen in a closed container. For our 2019 experiment, we added 1 mL of 3.6 M ZnCl2 solution to sterilize half of our bottles so we corrected all bottles for a potential "salting-out-effect" oxygen (Lang and Zander, 1986) sterilized with ZnCl2 in 2019 using a salinity value of 4.1 ‰.

S3.2 Organic Carbon
Filters for POC were stored frozen (1-2 months in 2015 and 2016, <2 weeks in 2019) until they were oven dried at the University of Alberta at 60 C for 24 hours and weighed. Filters were then fumigated under heat (60 C) for 24 hours by placing 25 mL of 12M HCl into a desiccator in an oven to remove carbonates and dolomites (Whiteside et al., 2011). Following fumigation, samples were air dried in a second desiccator and were then re-oven dried at 60 C for 24 hours (Whiteside et al., 2011). Dried filters were packed into silver capsules, with a second layer of tin capsules to promote combustion and shipped to the GG Hatch Laboratory concentrations of NH4 + , NO3 -), and NO2were determined. SO4 tubes were pre-leached with deionized water and stored in the dark at 4 ºC until analysis at BASL where samples were analyzed via ion chromatography following a modified method from US EPA 300.1. Major ions (Na, K, Ca, Mg) and trace metals (Fe, Al, Sr, Se, Zn, Si, Ba, Mn, Ni, P, Ti, V, Rb, Li) tubes were pre-leached with acid and stored in the dark at 4 ºC. To prevent binding to the container, 4 drops of 18% trace-metal grade nitric acid was added to each tube. Samples were analyzed by inductively coupled plasma mass spectrometer (Perkin Elmer Elan 6000 Quadrupole ICP-MS) at the University of Alberta Canadian Centre for Isotope Microanalysis following Cooper et al. (2008).

S3.4 XRD (2019 test)
A subset of experimental water was filtered through a pre-combusted and pre-weighed 25 mm GF/F filter from two bottles at the beginning of the experiment to determine the mineralogy of sediments and assess whether changes in mineralogy occurred due to sterilization procedures. Filters were stored frozen (-20°C, < 2 weeks) until dried at 50-60 ºC for 24 h and weighed to record sediment weights. Mineralogy was analyzed at the University of Alberta by X-ray diffraction (XRD; Rigaku Ultimate IV). The radiation source used was a Cobalt tube at 38 kV and 38 mA. Filters were mounted on a zero-background plate and scans were conducted using Bragg-Brentano parafocusing geometry, from 5 to 90° at 0.02° steps with a scan speed of 2.00 degrees per minute (0.6 s/step). Presence of minerals was determined using JADE 9.6 software with the 2019 ICDD Database PDF 4+, and 2018-1 ICSD databases.
Detection limit of the analyses is typically between 1-5%. Organic matter was not removed prior to analysis, but organic matter content of samples was below 5%. XRD analysis did not reveal any changes in mineral presence due to sterilization procedures.

S3.5 Dissolved inorganic carbon (2019 test)
Dissolved inorganic carbon samples were obtained by filtering incubation water through a 0.45 µm PES syringe filter into pre-acid-leached (10% v/v HCl, 24 hours) and pre-combusted 12 mL glass vials. Vials were sealed airtight with butyl-lined screw caps and stored in the dark at 4°C. Samples were taken in quadruplicate. In two of the replicates, we added 0.05-0.1 mL of 3.6 M ZnCl2 to assess whether there was any difference in storage. Samples were measured on an Apollo SciTech DIC analyzer within 1 month of collection. For all samples, the coefficient of variation between machine replicates was less than 0.1% and the standard deviation was less than 5 µM. For all samples without the addition of ZnCl2, the coefficient of variation between tube replicates was less than 5%, and the standard deviation was less than 24 µM. We found ZnCl2 has significantly lower DIC concentrations in samples and noticed precipitation of salts at the bottom of several tubes, thus we do not report DIC concentrations from DIC tubes with ZnCl2 added after collection from incubation bottles.

S3.7 Absorbance measurements
In 2015 and 2016, absorbance was measured at 254 nm and 750 nm on a Genesys 10 UV spectrophotomer using a 1-cm quartz cuvette. In 2019, absorbance was measured at 240-800 nm at 1nm increments in a 1-cm quartz cuvette using an integration time of 0.1 seconds (Horiba Aqualog). Absorbance values were baseline corrected either using absorbance at 750 nm (2015-16) or the mean absorbance from 700 -800 nm (2019). Absorbance values were then converted to both decadal and Napierian absorption coefficients for calculating absorbance indices (S3.9).

S3.8 Fluorescence measurements
BEPOM samples, and 2019 DOM samples were also analysed for fluorescence (Horiba Aqualog) at excitation wavelengths of 230 -800 nm at 5nm increments, with an integration time of 2 seconds. Emission wavelengths spanned 117.27 -826.70 for 2016, and 118.78 -828.18 for 2019, both with an increment of 2.39, an integration time of 2 seconds and Medium CCD Gain. Samples were diluted when optical density was: (a) greater than 0.4 at 240 nm (Osburn et al., 2012) and/or the sum of absorbance at a pair of wavelengths was greater than 1.5 (Kothawala et al., 2013), or (b) if counts on the machine exceeded 50,000 outside the Rayleigh scatter lines (nearing the maximum number of counts the machine can record).
Excitation and emission spectra were corrected using drEEM version 0.6.3 . Briefly, spectra were: (a) blank corrected, using 18.2 MΩ Milli-Q water for DOM samples and neutralized 0.1 M NaOH blanks for BEPOM samples; (b) inner filter effects were corrected using matching absorbance measurements also collected at 5 nm increments; (c) fluorescence data sets were normalized to Raman Units by dividing by the Raman area of pure water integrated of a λem range 383/384 to 420/425 nm at λex 350 nm (RU350) for BEPOM and DOM samples, respectively.

S3.9 Optical Indices
For all DOM samples, we calculate SUVA254 by normalizing the decadal absorbance at 254 nm to DOC concentrations (mg L -1 ) (Weishaar et al., 2003). SUVA254 values were corrected for Fe concentrations estimated to be within bottles (Poulin et al., 2014) either calculated from in situ Fe measurements or measurements made on a subset of bottles. Using absorbance spectra, we calculated spectral slope ratios (Helms et al., 2008) for 2019 DOM samples and all BEPOM samples. From fluorescence matrices we picked the maximum fluorescence of common peaks (Coble, 2007) and normalized them to the maximum samples fluorescence to assess their relative contribution to the fluorescence landscape. We also calculated humification index for all samples, along with the biological index for DOM samples. Indices used are detailed in Table S5.

S3.10 Radiocarbon analyses
Size fractioned material retained for characterization was stored frozen (-20°C) until it was freeze dried at the University of Alberta. A portion of each freeze-dried material was subsampled using a pre-combusted stainless-steel scoopula into a new 15 mL centrifuge tube (Corning ® ) for radiocarbon analysis. Material subsampled for radiocarbon analysis was spread on pre-combusted glass petri-dishes and examined under a dissecting microscope to remove any large debris or material (e.g., twigs, large rocks) that contrasted with the bulk background sediments. Subsamples of sediments were then pre-treated to remove carbonates using heated acid washes (HCl, 1M, 80ºC, 30 min; "A" treatment from Crann et al., 2017). Acid washes were repeated until effervescence stopped occurring in the samples; across all samples, two rounds of acid washes were sufficient. 14 C was then analyzed by Accelerator Mass Spectrometry following pellitization at the University of Ottawa (A.E. Lalonde AMS Laboratory).

S4 Sterile Experiment Details
In 2018, to test the rate of oxygen consumption on sterilized sediments, we placed 0.15 g of sterilized sediment each in 5 replicate 60 mL glass BOD bottle containing an oxygen sensor spot and glass beads to enable particle suspension, to achieve a concentration mimicking typical in-situ total suspended sediment concentrations. Bottles were filled 18.2 MΩ Milli-Q water (TOC <10ppb, sterilized in the Milli-Q system with UV-light). Bottles were sealed with a glass stopper with water placed around the rim to prevent gas exchange, along with a cap to hold the stopper in place. An additional set of 5 replicate bottles were prepared in parallel containing only Milli-Q water. All bottles and tools were sterilized with 95% ethanol which was evaporated in a fumehood. Sediment was obtained from a parallel study  and sterilized using dry heat in an oven set at 200°C for 24h. This sterilization procedure was validated by mixing sterilized sediment with DI water and placing 1 mL of the slurry onto a nutrient-rich agar plate at 37°C for 7 days. No colonies were observed.
In 2019, sterilization of bottles was achieved by autoclaving filtered water collected downstream of site FM3, and unfiltered water collected within slump FM3 runoff using a 30minute sterilization procedure at 18 psi and 121°C to kill the native bacterial population. To ensure no bacterial growth if any microorganisms were introduced as the experiment was set up, we added 1 mL of ~3.6M ZnCl2 solution to ensure even minimal introduction of microbes during experiment set-up would not result in rapid growth of microbes due to the absence of the original microbial community. We recognize that this removes our ability to assess whether non-stream microorganisms were introduced into the unsterilized bottles during incubation set up. We assumed that the biotic effect of microbes introduced into unsterilized bottles during set-up would be minimal since all tools and surfaces were sterilized with 95% ethanol prior to incubation set-up and there is a microbial community present to compete with. To test sterilization, 1 mL of incubation water was pipetted onto a nutrient-rich "plate-count" agar plate, from both sterilized (see above) and unsterilized treatments. This water was removed from a bottle dedicated to BEPOM analyses where quantification of particles was not required.
A sterilized glass spreader, 95% ethanol and a lighter, was then used to spread the drop around the plate evenly. The plates were then incubated under the same conditions as the incubation (20°C, in the dark). Colonies were observed to form on plates with water from unsterilized bottles, but none formed with water from sterilized bottles (Fig. S8). This test was done using bottles at the beginning and end of the experiment and still showed colony formation for unsterilized bottles but none for sterilized ones. used within-slump particles and downstream filtrate (fractionation experiment) c used donwstream particles and filtrate (for transect experiment) note: low Exp:InSitu ratios for upstream locations with low TSS values may be detection limit issues of TSS In Situ TSS measurements at many upstream sites required filtering of upwards of 1-2 L of water. Experimental bottle volumes were 60-120 mL. Table S3. Presence of minerals, as detected by XRD analysis, of sediments in bottles at the beginning of the experiment.   Figure S1. 2015 experiment oxygen concentrations (µM) for bottles incubated for 7 days (a-c) and 11 days (d-f). Dots show measured concentrations and lines show modelled measurements based on first order exponential decay. HA, HB, and HD differentiate slump sites. Codes: filtered (UF) and unfiltered (UU) upstream, slump material in upstream (SU) and downstream (SD) filtrate, and SU settled out (SS). Figure S2. 2016 experiment oxygen concentrations for fractionation experiment (a-d) and transect validation experiment (e-h). Codes in a-d differentiate filtered controls (F) and particle containing bottles (P) and codes for e-h differentiate Milli-Q control (MQ) filtered controls (F) and unfiltered treatments (U). Note that F in a-d and g are from the same samples, but repeatedly shown for easy comparison. Dots show measured concentrations and lines show modelled concentrations based on first order exponential decay.