The Potential Hidden Age of Dissolved Organic Carbon Exported by Peatland Streams

Radiocarbon (14C) is a key tracer for detecting the mobilization of previously stored terrestrial organic carbon (C) into aquatic systems. Old C (>1,000 years BP) may be “masked” by postbomb C (fixed from the atmosphere post‐1950 CE), potentially rendering bulk aquatic dissolved organic C (DOC) 14C measurements insensitive to old C. We collected DOC with a modern 14C signature from a temperate Scottish peatland stream and decomposed it to produce CO2 under simulated natural conditions over 140 days. We measured the 14C of both DOC and CO2 at seven time points and found that while DOC remained close to modern in age, the resultant CO2 progressively increased in age up to 2,356 ± 767 years BP. The results of this experiment demonstrate that the bulk DO14C pool can hide the presence of old C within peatland stream DOC export, demonstrating that bulk DO14C measurements can be an insensitive indicator of peatland disturbance. Our experiment also demonstrates that this old C component is biologically and photochemically available for conversion to the greenhouse gas CO2, and as such, bulk DO14C measurements do not reflect the 14C signature of the labile organic C pool exported by inland water systems more broadly. Moreover, our experiment suggests that old C may be an important component of CO2 emissions to the atmosphere from peatland aquatic systems, with implications for tracing and modeling interactions between the hydrological and terrestrial C cycles.


Introduction
Radiocarbon ( 14 C) analysis of dissolved organic carbon (DOC) has become a key tool over the past~30 years for detecting the mobilization of C into streams as a signal of environmental change . Since the Last Glacial Maximum~20,000 years ago, up to 620 Pg of C has accumulated in peatland soils around the globe (Page et al., 2011). These soils are vulnerable to environmental disturbance, which can impact terrestrial C loss through both greenhouse gas emissions (CO 2 and CH 4 ) and aquatic export of DOC, particulate organic C(PIC), and dissolved and particulate inorganic C (Evans et al., 2014;Hemes et al., 2018). As an environmental indicator of peatland C loss, 14 C is particularly effective for detecting the lateral export of DOC containing old C (defined here as >1,000 years BP) released as a result of land clearance, drainage, and/or climate change (Figure 1). In pristine peatland catchments or those affected by only limited disturbance, 87-90% of DO 14 C values are modern (Figure 1). This matches what is seen at the global scale, with most studies finding predominantly modern DO 14 C exported by rivers (Marwick et al., 2015). Conversely, 70% of DO 14 C values from extensively disturbed peatland catchments show a clear aged signal (Figure 1).
In peatland catchments, stream DOC is the main form of lateral C export (often 50-80%; , incorporating C from multiple terrestrial sources and soil depths within its source area (Aiken et al., 2014;Evans et al., 2014;Leach et al., 2016). In most peatland aquatic DO 14 C studies to date, predominantly in the UK, North America, Scandinavia, Malaysia, and Indonesia, DO 14 C concentrations are generally modern (i.e., with a radiocarbon content of >100 percent modern C [pmC]; Figure 1), despite inputs from deep organic soil layers where the oldest C is stored, potentially 3,000-6,000 years BP or older (Dargie et al., 2017;Evans et al., 2014;Garnett & Hardie, 2009;Garnett et al., 2011;Hulatt, Kaartokallio, Oinonen, et al., 2014;Leith et al., 2014;Moore et al., 2013;Tipping et al., 2010). The notable exceptions to this are the disturbed sites shown in Figure 1. Attempts to reconcile the multiple sources contributing to bulk aquatic DO 14 C samples have involved matching the measured 14 C content to the predicted soil organic matter 14 C profiles in their catchment area (Figure 2; Dean, van der Velde, et al., 2018;Evans et al., 2014;Moore et al., 2013;Raymond et al., 2007). Soil 14 C profiles are affected by a peak in 14 C concentrations caused by atmospheric nuclear bomb testing (Dutta, 2016). This results in a substantial pool of highly 14 C-enriched C in upper soil profiles that can skew the 14 C signature of the bulk DOC pool, essentially masking any old C (Raymond & Bauer, 2001). For example, a "modern" DOC sample with a 14 C concentration of 102 pmC could contain 2% old C, yet a slightly older DOC sample with a 14 C concentration of 97 pmC can potentially comprise 14% old C; at 90 pmC, old C could contribute as much as 38% to the total DOC pool ( Figure 2).
During transport and storage in inland water systems, DOC can decompose to produce CO 2 , contributing to the supersaturation of many global freshwaters with CO 2 (Catalán et al., 2016;Cole et al., 2007;McCallister & del Giorgio, 2012). On their own, CO 2 emissions from inland waters are equivalent to~19% of current anthropogenic emissions (Holgerson & Raymond, 2016;Raymond et al., 2013). In the USA, it is estimated that 28% of the CO 2 emitted from inland waters is from the photodecomposition and microbial decomposition of DOC during aquatic transport (Hotchkiss et al., 2015); in UK peatlands, 12-18% of exported DOC may be lost in this way (Dawson et al., 2001). DOC decomposition occurs via both microbial respiration (Logue et al., 2016;Raymond & Bauer, 2001), and photo-oxidation (Köhler et al., 2002;Pickard et al., 2017). Microbial respiration is often considered to be the primary process for the conversion of DOC to CO 2 , and 14 C values for dissolved organic carbon exported from peatland catchments that have been heavily disturbed (e.g., by drainage; Moore et al., 2013), semidisturbed (e.g., by long-term impacts from minor human activity; Billett et al., 2007), or relatively undisturbed by human activity (Aiken et al., 2014;Dean, van der Velde, et al., 2018;Campeau et al., 2017;Hulatt, Kaartokallio, Oinonen, et al., 2014;Hulatt, Kaartokallio, Asmala, et al., 2014;Ledesma et al., 2015;Leith et al., 2014;Marwick et al., 2015;Müller et al., 2015;Stimson et al., 2017aStimson et al., , 2017b. Modern DO 14 C values plot above the dotted line; the number of modern values for each catchment type is shown in blue, and the number of values older than modern is shown in red (total n = 209; see supporting information for data sources). many investigations into aquatic DOC decomposition focus on this process (Dean, van Hal, et al., 2018;Vonk et al., 2015). However, photo-oxidation can stimulate or inhibit the microbial respiration of DOC through partial oxidation and can also directly oxidize DOC to CO 2 , so both photo-oxidation and microbial respiration are important for the decomposition of inland water DOC (Cory & Kling, 2018). Organic C (OC) carried by inland waters can also be buried and stored in sediments in the form of POC (Wohl et al., 2017). POC is often defined as OC particles larger than 0.7 μm, as opposed to DOC, which is defined here as <0.7 μm. POC can form through the biological and physical flocculation of smaller OC and mineral molecules and can contain significant amounts of heterotrophic and autotrophic biomass from respiration of OC (i.e., generating free CO 2 in the water) and photosynthesis (i.e., fixing free CO 2 ), respectively (Hunter et al., 2016).
Despite the importance of microbial processes for regulating the fate of C mobilized from temperate peatlands, there is a clear disconnect between the 14 C age of stream DOC and CO 2 (Billett et al., 2007). DOC is predominantly modern (Figure 1), while the majority (62%) of 14 CO 2 values are older (average 14 CO 2 concentration = 96.5 ± 6.5 pmC; ±1σ; n = 92; Billett et al., 2006Billett et al., , 2007Billett et al., , 2012Garnett et al., 2012Leith et al., 2014;Tittel et al., 2013). Figure 2 shows that bulk DO 14 C samples have the potential to contain a wide range of C ages from modern through to several thousand years old. If components of a certain age in the DOC pool are preferentially decomposed to CO 2 , then this could explain the disparity between in situ DO 14 C and 14 CO 2 measurements (Galy & Eglinton, 2011;Loh et al., 2006;McCallister & del Giorgio, 2012). Aquatic CO 2 may also be generated from source material in the soil zone that is of a different age to the DOC pool, or even 14 C-dead (0 pmC) geological sources, further complicating bulk 14 CO 2 signals (Billett et al., 2007;Dean, van der Velde, et al., 2018). This suggests that DO 14 C measurements are not a good proxy for the age of CO 2 emitted from inland water systems.
Here we present an incubation experiment isolating the CO 2 released from the combined photodecomposition and microbial decomposition of peatland stream DOC. We tested the hypothesis that there is an old C component hidden within modern peatland stream DOC pools and that this old C is available for decomposition through natural in-stream processing to CO 2 . We collected a large bulk 14 C-modern DOC sample from a temperate peatland stream in the UK with the potential to receive old C from the deep peat soils within its catchment. We incubated this sample under a simulated day-night cycle for 140 days to represent the interaction between photodecomposition and microbial decomposition of the stream DOC pool. We measured the 14 C of DOC and CO 2 concurrently during the incubation and POC formed by microbial growth at the end of the experiment.

Field Sampling
The DOC-rich water for incubation was collected on 17 March 2015 from Black Burn, a first-order stream draining a 3.4-km 2 catchment of Auchencorth Moss peatland (55°48′N, 03°15′W) in central Scotland C ages to a soil profile in an exponential distribution (Dean, van der Velde, et al., 2018;Hua et al., 2013;Levin et al., 2013;Reimer et al., 2013). (b) The potential contribution of C of different ages to example peatland fluvial DO 14 C values (in percent modern C [pmC]) based on inputs from the soil 14 C profile in (a). (Dinsmore et al., 2010). This is an ombrotrophic peatland (85% of the catchment is peat), drained by a narrow open stream channel 0.7 m wide on average. Mean air temperature was 3.1°C, total rainfall was 102 mm, and mean discharge in the Black Burn was 41 L/s for the month prior to sampling, with no unusual hydrological or climatic events for this time of year (Dinsmore et al., 2013;Pickard et al., 2017).
The aquatic C flux is dominated by DOC, with an annual yield of 19.3 ± 4.6 g C·m −2 ·year −1 . CO 2 evasion is the second biggest component of aquatic C flux in the study catchment (10.0 g C·m −2 ·year −1 ; range = 2.33-24.0 g C·m −2 ·year −1 ). Other greenhouse gas fluxes are negligible compared with DOC export and CO 2 evasion (Dinsmore et al., 2013). Peat between 0.5 and 5 m thick overlies an extensive layer of glacial-derived boulder clay, which means that the underlying bedrock (predominantly Upper Carboniferous/Lower Devonian sandstones with occasional thin bands of limestone, mudstone, and coal) is likely less important to stream CO 2 fluxes compared with deep peat layers (Billett et al., 2007). Peat soils in the riparian zone of the Black Burn have provided 14 C ages of up to 1,400 years BP (83.93 pmC) at 70-cm depth (dates for deeper layers are not available but are likely older; Leith et al., 2014).
We prefiltered 60 L of stream water through a coarse nylon mesh (~225 μm) and homogenized it across two prerinsed 30-L barrels. The samples were collected at 3.7°C (water temperature) and kept at that temperature and in the dark until the start of the incubation. Field parameters (pH, electrical conductivity [EC], and temperature) were measured at the site using a Hach PH101C pH probe and a CDC401 conductivity probe connected to an HQd portable meter; these measurements were repeated for the sample waters at the end of the incubation. Dissolved CO 2 (Hope et al., 1995) and DOC concentrations, and initial (t 0 ) measurements of DO 14 C and 14 CO 2 , were also collected in the field. The DO 14 C sample was collected in a 1-L acid-washed bottle and filtered to 0.7 μm on preashed (~500°C for~3 hr) Whatman glass fiber filters (GF/F) in the laboratory. The 14 CO 2 sample was collected using the super-headspace method (see supporting information), collecting 3.5-ml CO 2 on a molecular sieve cartridge .

Incubations
The 60-L DOC-rich water sample was refrigerated at the University of Stirling overnight after sampling and thoroughly mixed and filtered the following day through preashed 0.7-μm GF/F filters at low pressure. This At day 138, three replicate samples were collected for DO 14 C and 14 CO 2 ; at day 140, three replicate samples were collected for PO 14 filter size isolates the DOC size fraction but allows sufficient bacteria (both heterotrophic and autotrophic) to pass through the filter for incubation (Dean, van Hal, et al., 2018). Fourteen liters of filtrate was then transferred into each of the three replicate incubation chambers. The filled chambers were refrigerated until the incubations commenced on 19 March. Three water aliquots were collected during filtration to confirm the homogeneity of the incubation water (Table S1).
Incubation chambers were 17-L white high-density polyethylene (HDPE) barrels (Ampulla UK), UN Class 1 approved for storage of pharmaceuticals and food products ( Figure 3). The chambers were cleaned with dilute Decon90 and thoroughly rinsed with distilled water prior to the experiment. We conducted blank test incubations using MilliQ®-grade deionized water in the same-class HDPE barrels to account for potential leaching from the incubation vessels. After 1 month, blank water samples contained 0.006 mg C/L, equivalent to <0.05% of the sample DOC content (Table S2). After 101 days, the blank water contained 0.01 mg C/L, approximately 0.1% of the sample DOC content (blank samples were processed in the same way as the DO 14 C samples, but insufficient material was obtained to measure either δ 13 C or 14 C content; see section 2.3). This shows that potential leaching from the HDPE incubation chambers used in this experiment was negligible and would not have impacted the 14 C results or DOC concentrations.
When filled with sample water, chambers had a 3-L headspace that was continuously flushed at a rate of approximately 150 cm 3 /min (replacing the headspace in~20 min) with outdoor ambient air scrubbed of CO 2 using an "indicating" soda lime trap (26-cm 3 volume) and Mg (ClO 4 ) 2 moisture trap via Tygon® medical-grade E-3606 tubing and mounted CPC couplings (Colder Products Co., USA) and then sealed with plastic paint (Plasti Dip, USA; Figure 3). The soda lime trap was replaced approximately halfway through the experiment, remaining fully functional for the duration of the experiment. Unfortunately, we were not able to monitor CO 2 concentrations in the system except during sampling.
A 140-mm-diameter circular hole was drilled into the lid of each chamber and covered with a 150-mmdiameter quartz glass window that allowed all wavelengths of light to pass through while preventing gas diffusion and then sealed with silicone sealant (Figure 3). The quartz glass window allowed both primary production and photo-oxidation to take place in the incubation chambers (see below). The chambers were equally irradiated with a T5 UV-B lamp (Arcadia Reptile, UK) designed to emit the full spectrum of natural light. The irradiance value measured on the sample surface was 14.5 W/m 2 (3.19 W/m 2 between 300 and 400 nm)-this is approximately twice the typical irradiance of the site where the incubation sample was collected, but the aim of the experimental design was to allow both autotrophic and heterotrophic DOC decomposition rather than specifically focus on either decomposition process. A night/day regime was used with a 12-hr equal split (8 a.m. to 8 p.m. light).
The incubated water was recirculated continuously using an external 12-V water pump at a rate of 15 ml/s, turning over the full water volume in 16 hr via mounted CPC couplings and E-3606 tubing, in order to replicate in-stream physical processes and limit stratification and heterogeneous mixing of solutes ( Figure 3). The incubation system was designed to limit the ingress of ambient CO 2 by sealing the chambers and flushing the headspace with CO 2 -free outside air (flushing was not possible during CO 2 sampling, which took between 40 and 170 min). Any ingress of ambient air that did occur during CO 2 sampling was corrected for using an established methodology for field-based chamber sampling (Gaudinski et al., 2000;Walker et al., 2016). Briefly, two δ 13 C isotopic end-members, ambient lab air (Bilzon et al., 2002) and the oldest CO 2 sample collected, were used to estimate the relative contributions of each end-member to a given sample. This is then compared with our estimate of ambient air ingress into each sample based on measurements of air ingress rates into the chambers (see Supplementary Methods; Figure S2). The mixing of the incubated sample water and the regular supply of oxygen to the headspace meant that any CH 4 production was minimized. The incubations were started on 19 March 2015 and stopped on 6 August 2015. The experiment ran for 140 days to simulate longer potential residence times for DOC during aquatic transfer from the terrestrial to marine zone. While the half-life of OC within inland water systems is 2.5 ± 4.5 years (Catalán et al., 2016), average residence times in peatland streams could be much lower than this. But with drain blocking becoming an increasingly common peatland restoration technique (Evans et al., 2018), and peatlands providing a substantial proportion of water supply to reservoirs , residence times in the order of 140 days or longer are possible for peatland DOC within inland water systems.

Sample Collection and Analysis
Incubation aliquots for DO 14 C and 14 CO 2 were collected from one chamber only (C2) on days 4 (t 1 ), 8 (t 2 ), 14 (t 3 ), 32 (t 4 ), 76 (t 5 ), and 138 (t 6 ; Figure 3). These were spread based on expected higher rates of DOC decomposition over the initial incubation period . At the final time point, 14 C samples were collected from all three chambers. Incubation water aliquots were collected from each chamber via an outlet on the water pump line; headspace 14 CO 2 samples were collected via the upper mounted CPC couplings ( Figure 3). DO 14 C samples were collected in 500-ml acid-washed clear HDPE bottles after prerinsing with sample water, filtered to 0.7 μm on preashed GF/F filters and refrigerated until analysis (less than 90 days in all cases, which should not have impacted the stable or radiogenic isotopic signatures of the DOC sample; Gulliver et al., 2010). 14 CO 2 samples were collected by first purging the headspace of CO 2 by cycling it through a soda lime CO 2 trap until the concentration was <60 ppm. The headspace was then allowed to equilibrate with the water by diffusion for 1 to 2 hr (overnight in one case; see Supplementary Methods). The headspace was then cycled through a molecular sieve trap to capture the CO 2 for 14 C analysis (1.5-to 4.1-ml CO 2 ; Table S1). Where insufficient sample was captured on the molecular sieve in one cycle, the headspace was then allowed to build up again for another 1 to 2 hr and sampled again.
Incubation aliquots for DOC concentrations (DOC and total C) and optical properties (absorbance and fluorescence) were collected in 30-ml sterile HDPE tubes (first rinsed with sample) during preincubation filtration and on days 1,4,7,8,14,32,34,48,62,76,97,124,138, and 140 ( Figure 3). Dissolved organic matter (DOM-contains the DOC pool) quality parameters from absorbance and fluorescence measurements provide insights into the structure and molecular weight of the DOC pool (Fellman et al., 2010;Helms et al., 2008). Absorbance aliquots (50 mL) from the three replicate chambers were collected in sterile containers and then filtered through precombusted 0.7-μm Whatman GF/F filters under low pressure. Absorbance measurements were carried out immediately after sample collection using a dual-beam spectrophotometer (Agilent Cary 100; see Supplementary Methods). From the absorbance values, we calculated the ratio of the absorption spectral slopes between 275-295 and 350-400 nm (slope ratio, S R ), which is inversely correlated with DOM molecular weight (Fichot & Benner, 2012;Helms et al., 2008). Aliquots for the measurement of DOM fluorescence were only collected from chamber C2. DOM fluorescence was measured with a spectrofluorometer (Instant Screener®, Laser Diagnostic Instruments AS; see Supplementary Methods; Lawaetz & Stedmon, 2009;Murphy et al., 2010;Patel-Sorrentino, 2002) and used to calculate the humification index (HIX; the extent of humification in the DOM-i.e., how decomposed it is) by dividing the area under the emission spectra 435-480 nm by the peak area 300-345 nm + 435-480 nm, at excitation 254 nm (Ohno, 2002). Lower HIX values indicate more decomposition (increased humification) of the DOM pool. DOC quality and concentration aliquots collected before day 32 were not filtered prior to analysis. From day 34 onward, minor flocculation was first observed, so these samples were then filtered through 0.7-μm preashed GF/F filters, with negligible difference seen between days 32 and 34 (±1.0 mg C/L). DOC concentrations were measured at the University of Stirling on a Shimadzu Total Organic Carbon Analyzer (TOC-V series) with a lower detection limit of 0.5 μg/L and an accuracy of 1.5%.
POC concentration and 14 C samples were collected on day 140 only, by first stirring to homogenize the sample thoroughly, then filtering through 0.7-μm preashed GF/F filters, and measuring the captured material. One liter of sample water was filtered to obtain the PO 14 C samples, and 200 ml of water was filtered for triplicate particulate organic matter concentrations; the filters were then dried at 70°C. POC concentration samples were combusted in a Carlo-Erba EA 1108 elemental analyzer, using sulfanilamide as the standard. 14 C samples were prepared for analysis at the Natural Environment Research Council Radiocarbon Facility (East Kilbride, UK). DO 14 C samples were processed to solids using rotary evaporation and then acidfumigated (Dean, van der Velde, et al., 2018). DO 14 C and PO 14 C samples were combusted, and the CO 2 produced was cryogenically recovered. Molecular sieve cartridges were heated, and the desorbed sample CO 2 was cryogenically recovered . For all 14 C samples, an aliquot of the recovered CO 2 was analyzed for δ 13 C using isotope-ratio mass spectrometry (Thermo Fisher Delta V), with results reported relative to the Vienna Pee Dee Belemnite standard. A further aliquot of CO 2 was graphitized using Fe-Zn reduction and the 14 C content determined by accelerator mass spectrometry (AMS) at the Scottish Universities Environmental Research Centre (Xu et al., 2004). Following convention, all 14 C results were normalized using the δ 13 C values and presented as pmC and conventional radiocarbon ages (years BP,

Journal of Geophysical Research: Biogeosciences
where 0 year BP is 1950 CE). The 14 C background associated with these methods was quantified by analysis of 14 C-dead materials (blanks); standards of known 14 C content processed alongside the samples gave values within 1σ of the consensus.

Field Results
DOC and dissolved CO 2 concentrations at the field site were 18.4 ± 0.6 ( Figure 4) and 2.3 ± 0.1 mg C/L, respectively, while pH was 5.9 and EC was 60.3 μS/cm. These values are usual for this time of year at the study site (Dinsmore et al., 2013), with average values in the 3 months prior to sampling being 5.1 for pH and 59.1 μS/cm for EC (Pickard et al., 2017). DO 14 C was 103.79 ± 0.48 pmC (i.e., modern; Figure 5), while 14 CO 2 was 88.14 ± 0.38 pmC (1,014 ± 35 years BP; Figure 5). The 14 C results agree with previous work at the site and similar temperate peatlands where DOC was modern, while CO 2 was consistently older (Billett et al., 2007;Billett & Garnett, 2010;Leith et al., 2014).

DOC Decomposition and 14 C Dynamics
DO 14 C content decreased from 102.45 ± 0.47 pmC (modern) at t 1 , very similar to the field sample, to 99.78 ± 0.46 pmC (17 ± 17 years BP) over 138 days of incubation ( Figure 5). δ 13 C-DOC changed only slightly, from −28.7‰ to −28.2‰ during incubation (Table 1). The final DO 14 C concentration was consistent across all three chambers with a range of 99.16 ± 0.45 to 99.83 ± 0.44 pmC. While the reduction in DO 14 C content over the 138 days of incubation was relatively small (Table 1 and Figure 5), it is statistically significant (p < 0.05 for the three t 6 samples; Student's t test, R version 3.4.3) and implies a small shift toward older DOC.
DOC concentrations in chambers C2 (where the primary 14 C measurements were collected) and C3 behaved similarly, remaining relatively stable for the first 8 to 14 days at a maximum of 18.6 mg C/L, before  S R increased across all chambers, indicating a decline in molecular weight that suggests larger DOM molecules were continually broken down over the course of the incubation (Figure 4b). The increase in S R was reflected in a 28.7-35.7% reduction in the absorption coefficient at 254 nm, a 31.8-40.2% reduction at 300 nm, and a 30.9-44.9% reduction at 440 nm, suggesting an overall decline in aromaticity and molecular weight (Table S1). HIX values, calculated from the DOM fluorescence measurements from chamber C2, decreased from >0.99 (measurement saturated) to 0.95, indicating a reduction in humic substances in the DOC pool (Figure 4b).

14 CO 2 dynamics
During the incubation, headspace 14 CO 2 signatures fell from 89.92 ± 0.90 pmC (853 ± 100 years BP) at t 1 to 74.58 ± 3.53 (2,356 ± 767 years BP) at the final time point ( Figure 5). The t 1 14 CO 2 age was slightly younger than the field 14 CO 2 sample, indicating that, initially, younger DOC was decomposing to produce CO 2 . The subsequent increase in 14 CO 2 age was supported by replicate 14 C measurements at the final time point in chambers 1 and 3 of 80.47 ± 0.84 and 83.60 ± 1.45 pmC (1,746 ± 124 and 1,439 ± 280 years BP, respectively; Table 1 and Figure 6). The pH and EC of the incubated water remained stable at 5.8-5.9 and 59.8-63.7, respectively, indicating that dissolved inorganic C was predominantly present as dissolved CO 2 , which would be in isotopic equilibrium with the sampled headspace CO 2 .
The δ 13 C values can also indicate CO 2 source, alongside 14 C. The field CO 2 sample had a δ 13 C signature of −18.6 ‰, which was higher than the CO 2 produced directly from the decomposition of the DOC pool (−27.1 ‰; Table 1). The field δ 13 CO 2 value was matched by an old 14 C age, suggesting it was sourced in part by the decomposition of deep peat layers (Billett et al., 2007;Dean, van der Velde, et al., 2018) and potentially 13 CO 2 fractionation during evasion from the stream (Doctor et al., 2008). The lower δ 13 CO 2 value from the incubation matches that of DOC (−28.4‰ to −28.8‰; Table 1), supporting the case for the old 14 CO 2 ages measured during the incubation being directly produced from the decomposition of the DOC pool. CO 2 isotopic values corrected for atmospheric contamination, and the uncertainty from this correction (95% confidence interval) is included in the uncertainties here (see Table S3 for uncorrected values).

The Production of CO 2 and POC From Decomposition of the DOC Pool
POC was formed in the incubation chambers during the experiment, and this also had an old 14 C signature of 1,824 ± 37 to 2,094 ± 37 years BP (Table 1 and Figure 6). This POC formed from large-scale microbial growth, driven by heterotrophic and autotrophic production, causing flocculation of living and dead microbial biomass and OC, and the formation of microbial biofilms ( Figure S2; Battin et al., 2016;Logue et al., 2016). POC was not present at the beginning of the incubation as we filtered the sample water to isolate the DOC pool.
Isotope mass balance equations can be useful for partitioning C sources (Billett et al., 2007). During the incubation, old C was predominantly partitioned into the CO 2 and POC pools ( Figure 6). To satisfy isotope mass balance, the DOC age would be expected to remain constant or become younger (more 14 C enriched) as old C was lost to CO 2 and POC. For example, we used equation (1) to estimate the 14 C content of the DOC pool that was lost during the course of the incubation: where Δ represent the 14 C content (pmC) and M the mass of the DOC lost during the incubation ( lost ) and at the start ( t0 ) and end ( t6 ) of the incubation. Because the age of DOC increased over time (i.e., 14 C depleted; Figure 5), equation (1) predicts the 14 C content of the CO 2 produced from the decomposition of the DOC pool to be 14 C enriched (117.6 pmC). However, this value contrasts with our direct observations of the predominantly old C in the CO 2 and POC pools ( Figure 6).

Carbon (Re)Cycling in the Incubation Chambers
The observed increase in DOC age in conjunction with the production of old CO 2 and POC pools during the experiment could be explained by the rapid loss of a modern component of the DOC pool early in the incubation and the reintroduction of old C into the DOC pool by autotrophic fixation. If there was rapid respiration of a labile modern component of the DOC pool (e.g., plant exudates and leachates from freshly decayed organic matter), then the resultant CO 2 could have been flushed from the headspace and thus missed in the first days of the incubation (the chambers were not closed systems). This is hinted at by the t 1 CO 2 , which was generated directly from the decomposition of the DOC pool, being younger than the field CO 2 , which was an integrated CO 2 pool generated from the decomposition of OC throughout the catchment peat soil profiles as well as within the stream itself ( Figure 5). As this labile modern DOC pool rapidly decomposed over the early phase of the incubation and was flushed from the headspace, the old C signature would have become more pronounced in the respired CO 2 , which can be seen in Figure 5. An additional mechanism, the growth of autotrophic microorganisms, would have fixed free (dissolved) CO 2 into the microbial-DOC pool through photosynthesis. This mechanism can at least partially explain the partitioning of C ages in the incubation systems. The free CO 2 in the incubation system may have been younger, initially, as described above, but rapidly aged between t 1 and t 2 (4 days; Figure 5), yielding old CO 2 for subsequent autotrophic fixation and incorporation into autotrophic biomass and POC formation. Microbial biomass and detritus less than 0.7 μm in size are also considered part of the DOC pool and are included in the DO 14 C and DOC concentration measurements for this experiment, so the reintroduction of old C in the DOC pool could also explain the increase in the bulk DO 14 C age (Fellman et al., 2010).

The Potential Hidden Age of Peatland Stream DOC
Our study supports the hypothesis that peatland stream DOC with a modern age can contain a fraction of old C that is available for decomposition during aquatic transport. Old C was identified in the CO 2 produced from the decomposition of the predominantly modern DOC pool within 20 days ( Figure 5), well within the half-life of OC within inland water systems (2.5 ± 4.5 years; Catalán et al., 2016). The increase in DOC concentrations observed in chamber C1 may have been due to increased autotrophic microbial growth in this chamber compared with the others (some microbial biomass is incorporated into DOC; see section 3.4). However, shifts in DOM structure across all three incubation chambers show that the terrestrial DOC pool was continually being decomposed for the duration of the experiment, despite DOC dynamics being influenced by autotrophic and heterotrophic microbial growth, (Figure 4 and Table S1). The observed 10.1029/2018JG004650 Journal of Geophysical Research: Biogeosciences decline in HIX values (Figure 4) is likely due to the preferential decomposition of terrestrial-derived (humic) DOC, rather than the decomposition of heterotrophic and autotrophic microbial biomass that also contributes to the overall DOC pool (Fellman et al., 2010). The terrestrial-derived component of the DOC pool is likely old (Galy & Eglinton, 2011;Leith et al., 2014), and the HIX dynamics suggest that this old terrestrial OC was continually decomposed over the incubation, fueling CO 2 production as reflected in the old 14 CO 2 signature we observe ( Figure 5). DOM structural shifts were consistent across all the incubation chambers (Figure 4b), suggesting that although the kinetics of DOC decomposition in chamber C1 may have been different from the others, the overall decomposition of the DOC pool and 14 C results are comparable across all chambers ( Figure 6). However, fluxes of CO 2 produced in the different chambers were not measured to corroborate this.
There was an increase in the difference between the incubation DO 14 C and 14 CO 2 signatures as the experiment progressed. This difference rapidly increased over the first 14 days and then plateaued ( Figure 5). This is consistent with the rapid respiration of a labile modern component of the DOC pool and increased decomposition of old C once this labile modern pool was lost. Previous work also demonstrated the likely presence of a labile young C component with bulk riverine DOC loads (Galy & Eglinton, 2011;Loh et al., 2006;Raymond & Bauer, 2001). We were not able to quantify the amount of modern labile C that was rapidly lost in the initial stages of the incubation. But this fraction cannot have been the majority of the modern DOC component because the bulk DO 14 C signature remained close to modern for the duration of the experiment. This indicates that there was a substantial recalcitrant modern DOC component. In this experiment, we clearly demonstrate that there is also potential for an important component of labile old C to be hidden within bulk "modern" riverine DOC loads. This old C signal may have been exaggerated in our experimental system due to the rapid respiration and flushing from the system of a labile modern DOC component and the later reintroduction of C into autotrophic biomass (see section 3.4). However, that the presence of old C is observed at all, and at such clear levels in the CO 2 and POC measurements, unequivocally supports our hypothesis that labile old C can be concealed within bulk modern DOC pools.
Previous work has shown that heterotrophic respiration of temperate lake DOC produced old CO 2 (1,000 to 3,000 years BP) despite the bulk 14 C signature of the incubated DOC being significantly younger (modern, with 14 C contents of 101.4 to 111.0 pmC; McCallister & del Giorgio, 2012). The degradation of ancient C (>11,300 years BP) has also been demonstrated in Arctic yedoma permafrost (Drake et al., 2015;Mann et al., 2015;Spencer et al., 2015;Vonk et al., 2013). In contrast, earlier work suggests that modern C is preferentially respired from bulk riverine DOC loads, with old C remaining recalcitrant (Galy & Eglinton, 2011;Raymond & Bauer, 2001). However, concurrent in situ measurements of aquatic DO 14 C and 14 CO 2 signatures are rare, limited to four studies, of which we are aware. Aquatic CO 2 is generally older than DOC in UK peatlands, including the study site where we collected our DOC sample for incubation (Billett et al., 2007Leith et al., 2014). This was thought to be driven by a mixed signal comprising 14 C-dead CO 2 from geological sources (~5-30%) and modern CO 2 derived from soil respiration in the upper peat layers (~70-95%). DOC was older than CO 2 in a western Canadian Arctic peatland underlain by continuous permafrost (Dean, van der Velde, et al., 2018). There, the shallow permafrost prevented contributions of 14 C-dead geogenic sources, with the CO 2 and DOC sourced from both old and young shallow peat layers. The results of the incubation study presented here, where external C inputs to the 14 CO 2 signal are restricted, suggest that respiration of old C within bulk DOC pool also contributes to in situ riverine 14 CO 2 signals. However, this is highly system dependent, with different DOC age components available for decomposition to CO 2 during aquatic transport and storage.

Implications and Future Work
Bulk DO 14 C measurements are not sensitive enough to detect old C in peatland inland water systems, as demonstrated by the production of old 14 CO 2 from the decomposition of a bulk modern DO 14 C sample in the experiment presented here-with the exception of severe cases of landscape disturbance (Figure 1). The findings of this study also suggest that bulk DO 14 C measurements do not represent the 14 C signature of the labile OC pool exported by inland water systems more broadly (Fellman et al., 2014). In situ aquatic 14 CO 2 measurements in UK peatlands were previously thought to show a mixture of modern soil and 14 available for decomposition after its release by variable mixing with water flowing through the peat soil layers (Dean, van der Velde, et al., 2018;van der Velde et al., 2012). This shift from recalcitrant to labile may be driven by the change from oxygen-poor to oxygen-rich conditions, differing microbial populations in the aquatic zone from the soil zone (Dean, van Hal, et al., 2018), and/or combined photodecomposition and microbial decomposition. All of these processes were operating in the experiment presented here (Figure 3). However, we did not separate them out by experimental design, seeking only to simulate natural in-stream conditions for a single site. Further studies should be carried out to distinguish the importance of these individual processes to DOC lability during transport and storage in inland water systems and across a range of ecosystems.
These findings may extend beyond aquatic respiration and may be relevant to the interpretation of 14 C studies on soil respiration and porewater DO 14 C. Further, other aquatic C species may also be affected by components of differing C age and lability. Incubations such as the one presented here can aid this research strand, although they require multiple 14 C dates and are time-consuming. For bulk POC and DOC samples, ramped pyrolysis can reveal the 14 C signature of C combusted at different temperatures (i.e., indicating lability; e.g., Zhang et al., 2017), although this has only recently been developed for bulk aquatic DOC samples (Hemingway et al., 2017). Matching ramped pyrolysis DO 14 C signatures and integrated soil OC age distributions (Sierra et al., 2017) with bulk DO 14 C samples via age distribution models (e.g., Figure 2) would improve our estimates of the likely proportions of different aged C within bulk DO 14 C samples (Evans et al., 2014). This could be assisted by compound-specific 14 C methods (e.g., Feng et al., 2013Feng et al., , 2017, although this can also require multiple 14 C dates for a single sample depending on the target organic compounds, and it is possible that C of multiple ages can be present within the isolated pools. Direct in situ measurements of 14 CO 2 and 14 CH 4 signatures are needed to improve our understanding of the contribution of old C to inland water C emissions across a range of environmental settings. Methodological advances now allow for the collection of 14 CO 2 and 14 CH 4 samples relatively quickly and easily in even remote field locations (Billett et al., 2006;Dean et al., 2017;. For aquatic CO 2 and CH 4 14 C samples, age distribution analysis (e.g., Dean, van der Velde, et al., 2018) or multisource isotopic mass balance (e.g., Elder et al., 2018) can help determine relative contributions of different aged C.

Conclusions
In the incubation experiment presented here, old C up to 2,350 years BP was clearly identified in the CO 2 produced from the decomposition of a bulk "modern" DOC pool. Bulk 14 C dating of peatland stream DOC may therefore be an insensitive environmental indicator for the export of old riverine C. Consideration should be given to collecting a suite of 14 C measurements (e.g., PO 14 C, 14 CO 2 , and 14 CH 4 ) to complement DO 14 C analyses and obtain an accurate representation of the age of mobile OC within fluvial systems. Our study also suggests that old C may be an important contributor to CO 2 emissions from inland waters, which are a large source of C to the atmosphere. Understanding the proportion of old C released to the atmosphere within this flux is of great importance for modeling ecosystem productivity and C cycling and for predicting the impacts of future climate change at the global scale (Cole et al., 2007). This is especially true for peatlands, which are considered a key C sink in the global C cycle (Billett et al., 2015). An increased proportion of old C in riverine DOC export means a potential increase in old C lost to the atmosphere and delivered to the oceans (Battin et al., 2009;Bogard & Butman, 2018). This suggests that there may be more old C involved in the global C cycle than previously thought Fellman et al., 2014;Guillemette et al., 2017).