Peatland pools are tightly coupled to the contemporary carbon cycle

Peatlands are globally important stores of soil carbon (C) formed over millennial timescales but are at risk of destabilization by human and climate disturbance. Pools are ubiquitous features of many peatlands and can contain very high concentrations of C mobilized in dissolved and particulate organic form and as the greenhouses gases carbon dioxide (CO2) and methane (CH4). The radiocarbon content (14C) of these aquatic C forms tells us whether pool C is generated by contemporary primary production or from destabilized C released from deep peat layers where it was previously stored for millennia. We present novel 14C and stable C (δ13C) isotope data from 97 aquatic samples across six peatland pool locations in the United Kingdom with a focus on dissolved and particulate organic C and dissolved CO2. Our observations cover two distinct pool types: natural peatland pools and those formed by ditch blocking efforts to rewet peatlands (restoration pools). The pools were dominated by contemporary C, with the majority of C (~50%–75%) in all forms being younger than 300 years old. Both pool types readily transform and decompose organic C in the water column and emit CO2 to the atmosphere, though mixing with the atmosphere and subsequent CO2 emissions was more evident in natural pools. Our results show little evidence of destabilization of deep, old C in natural or restoration pools, despite the presence of substantial millennial‐aged C in the surrounding peat. One possible exception is CH4 ebullition (bubbling), with our observations showing that millennial‐aged C can be emitted from peatland pools via this pathway. Our results suggest that restoration pools formed by ditch blocking are effective at preventing the release of deep, old C from rewetted peatlands via aquatic export.


| INTRODUC TI ON
Peatlands are important long-term terrestrial carbon (C) stores.They cover only ~3% of the Earth's land surface yet contain 600-650 Pg C fixed into peat soils from the atmosphere by primary production and maintained for millennia under water-logged conditions (Page et al., 2022;Xu et al., 2018).The magnitude of this C store is comparable to that currently held in the atmosphere (875 Pg C; Friedlingstein et al., 2022).Peatlands release some C laterally to standing and flowing inland waters (i.e., streams and lakes; Dean et al., 2020), as well as emitting carbon dioxide (CO 2 ) and methane (CH 4 ) to the atmosphere (Evans et al., 2021), but under pristine conditions are net C sinks.However, peatlands are at severe risk of destabilization which could release the C they store to the atmosphere and aquatic systems (Lupascu et al., 2020).Climate change and changes in land use may cause the release of additional C that would otherwise have remained in peat soils.This additional C release comes from perturbation of contemporary C cycling and/or the release of millennial aged C that was previously held in deeper peat layers (Moore et al., 2013;Tanentzap et al., 2021).
Pools, or ponds, are small shallow standing water bodies that are commonly found in peatlands worldwide (Arsenault et al., 2022;Hassan et al., 2023).Ranging from <1 to 1000 m 2 in size and up to ~1 m in depth, pools can be hotspots of terrestrial C decomposition and CO 2 and CH 4 emissions, and provide a direct and rapid pathway for the release of C to the atmosphere (Downing, 2010;Holgerson & Raymond, 2016;Peacock et al., 2021;Rosentreter et al., 2021;Turner et al., 2016).Peatland pools are sensitive to both climate and land use change and could potentially serve as sentinels for shifts in patterns of peatland C cycling (Arsenault et al., 2023;Hassan et al., 2023).
Radiocarbon ( 14 C) is a powerful tool for detecting the age of C released from soils (Estop-Aragonés et al., 2020;Schwab et al., 2020;Stuart et al., 2023;Waldron et al., 2019).Inland waters integrate 14 C signals of the air, plants, and soil that mobilized C has moved through prior to its release.Recent advances have enabled sampling of all major C forms found in inland waters: dissolved and particulate organic C (DOC and POC, respectively), dissolved CO 2 and CH 4 , and ebullition (bubble release from aquatic sediments; Dean et al., 2020;Prėskienis et al., 2021).For example, inland waters have shown clear signs of old soil C release from enhanced seasonal thaw of permafrost soils in Arctic catchments (e.g., Schwab et al., 2020) and artificially drained tropical peatlands (e.g., Moore et al., 2013).Often, 14 C markers associated with disturbance can be hidden within the larger signal of contemporary C turnover (Dean et al., 2019).In these cases, age distribution modelling can provide estimates of the likely contributions of different soil C layers to the mixture of C sources found in a single inland water 14 C sample (Dean, van der Velde, et al., 2018;Evans et al., 2014Evans et al., , 2022;;Raymond et al., 2007;Tanentzap et al., 2021).
Many peatlands worldwide have been drained historically for agriculture or peat extraction, and these activities continue in some regions today (Evans et al., 2021).Heavily drained peatlands can release C that is ~1000-4000 years old into drainage waters (Hulatt et al., 2014;Moore et al., 2013;Waldron et al., 2019).In many regions, including Europe, Canada, and Indonesia, peatlands are being restored by blocking drainage ditches in an attempt to raise the water table, thereby reducing C loss to the atmosphere and aquatic systems (Evans et al., 2021;Parry et al., 2014;Putra et al., 2021;Waldron et al., 2019).Ditch blocking often results in the establishment of pools behind small dams (Chapman et al., 2022;Holden et al., 2018).These "restoration" pools may retain C that would otherwise be directly lost from peatlands via runoff into drainage ditches (Gaffney et al., 2021;Parry et al., 2014).It is therefore important to know whether pools formed through restoration efforts are receiving and emitting old C destabilized by legacy drainage, in addition to the contemporary C turnover expected under raised water table conditions (Evans et al., 2021;Parry et al., 2014;Waldron et al., 2019).Given that pools are important sites of organic C decomposition and emissions of CO 2 and CH 4 within peatland landscapes, climate change and further anthropogenic landscape disturbance could readily impact the substantial fluxes of C released into and from peatland pools (Arsenault et al., 2023;Chapman et al., 2022;Turner et al., 2016).Understanding the role of peatland pools is crucial to quantifying current and future C cycling in peatlands, yet little is known about the source of pool C. Do pools only receive and decompose contemporary peatland C, or do they have the potential to process and emit additional old C from deeper peat layers?
Here, we present a set of 14 C and stable C isotope (δ 13 C) observations of DOC, POC, and dissolved CO 2 collected simultaneously from a Flow Country blanket peatland in the north of Scotland across three sampling campaigns in 2014 and 2015.These observations are augmented by snapshot sampling in six pool locations on blanket peatlands across the United Kingdom that focused on CO 2 and DOC but also included rare ebullition 14 C samples.Organic C has accumulated in the peat soils of our study locations and in the United Kingdom's blanket peatlands in general over the past 8000-10,000 years (Billett et al., 2010;Ratcliffe et al., 2018).We sought to determine the source and age of C in peatland pools to establish where this C comes from, to what degree these pools receive C from the surrounding peat and release it to the atmosphere, and whether C sources differ between natural and restoration pools.Specifically, we aimed to answer the following research questions: 1. What is the age (source) of C found in pools across a range of temperate blanket peatlands?2. How does the age (source) of pool C vary between two distinct types of peatland pools (natural and restoration)?
3. What is the relationship between pool C isotopes and key peatland pool biogeochemical parameters?
To our knowledge, this is the first comprehensive study of the isotopic composition of peatland pool C. We hypothesized that pools contain mainly contemporary C but will display clear signs of old C release if peatland destabilization is occurring.

| Study locations
We carried out isotopic measurements in three distinct geographical regions of the United Kingdom: northern and southwest Scotland and Northern Ireland (Table 1).All the pool complexes were on ombrotrophic plateaux; all sampled pools were situated in the peat matrix, and thus, geologic C inputs to these pools are unlikely.The Cross Lochs, Loch Lier, and Munsary locations are in the Flow Country of northern Scotland, the largest blanket peatland complex in Europe (~4000 km 2 ).The Silver Flowe is a blanket peatland located in a glacial valley in Galloway, southwest Scotland.Slieveanorra is an upland raised bog in County Antrim, Northern Ireland.Garron Plateau is also in County Antrim and is one of the largest areas of nearnatural upland blanket bog in Northern Ireland.Vegetation in the pools included both submerged and emergent vegetation, and can be broadly classified into aquatic Sphagnum (e.g., S. cuspidatum and S. denticulatum), Eriophorum (e.g., E. angustifollium), Menyanthes (e.g., M. trifoliata), and algae including cyanobacteria (Turner et al., 2016).
All sites experience similar climate; natural pools, sampled across all locations, shared similar hydrological and geomorphological characteristics, and pool waters shared similar pH, dissolved oxygen (DO), DOC, and CO 2 contents (Turner et al., 2016).
Natural pools were sampled at all locations.Restoration pools created by drain blocking were only sampled at the Cross Lochs and Loch Lier locations (Table 1).For the studied pools across all locations, natural pools tended to be larger than restoration pools but had similar depths (Figure S1; Table S1).Natural pools tended to have longer water residence times than restoration pools, and this was consistent across pools at the Cross Lochs location where data were available (Holden et al., 2018)

| Pool sampling
The greatest sampling intensity was in the natural and restoration pool systems at Cross Lochs (Dean et al., 2023).Three natural and three restoration pools at this location were sampled in May 2014, November 2014, and September 2015.Additional pools at Cross Lochs were sampled a single time (Table S1; Dean et al., 2023).At Loch Lier, two natural and two restoration pools were sampled in November 2014 during the same campaign as the Cross Lochs sites.In September 2015, all locations were sampled except Loch Lier due to logistical constraints (Table S1; Figure S1c).Pools were sampled under the same conditions in each campaign as far as possible; timing of the sampling campaigns had no clear impact on the observed patterns of C isotopic composition (Figures S1 and S3).Ice cover, which was not present on any pools during sampling, is generally limited to surface freezing for a few days or weeks during the coldest parts of winter.
Pool water sampling for isotope analysis focused on DOC and dissolved CO 2 .In many cases, POC levels and dissolved CH 4 concentrations were too low for sufficient material to be collected for 14 C analysis.
Isotope sample collection followed Dean et al. (2020).Water samples were collected in 1 L glass bottles previously furnaced for 3 h at 450°C, and were kept in a cool box immediately following sampling then placed in a refrigerator at 4°C at the end of the day and stored there for the duration of the field campaign (up to 3 days).The water samples were then transported to the NERC Radiocarbon Facility in East Kilbride in a cool box (up to 1 day).Upon arrival at the facility, samples were placed in a refrigerator at 4°C.The water samples were filtered within the next 3 days using pre-ashed glass microfiber filters (GF/F 0.7 μm).
Where enough material was present, POC samples were collected by combusting these GF/F filters (Dean et al., 2020).Previous work has TA B L E 1 Study site characteristics.See Table S1 for full sample details.Data from Turner et al. (2016).demonstrated that this treatment of samples has no significant impact on DOC isotopic composition (Gulliver et al., 2010).After collection on the filters, POC samples were refrigerated (4°C) until combustion.CO 2 samples were collected by degassing 3 L of sample water into a 1 L CO 2free headspace for 3 min, and the headspace then injected onto a molecular sieve cartridge in the field (Garnett et al., 2016).We also include a single dissolved CH 4 sample collected from Loch Lier in November 2014 as part of a separate study (Dean et al., 2017).
Bubble traps were deployed to collect ebullition (bubble) samples from six natural and five restoration pools at Cross Lochs.Two to six traps were deployed in each of the larger natural pools, and one trap was deployed in the smaller restoration pools due to space constraints.The bubble traps comprised stoppered inverted glass funnels of inside diameter 246 mm (area = 0.0475 m 2 ).The funnels were kept on the pool surface by a circular "lifebuoy" type float made from pipe insulation.Funnels were filled with pool water which is drawn into the funnel by a large syringe via a three-way gas-tight valve.When deployed, rising bubbles displace the water in the funnel, allowing ebullition bubbles to accumulate.The traps were deployed on May 21, 2014 and checked fortnightly until collection on November 5, 2014.
Ebullition samples were extracted from the funnel traps using a syringe purged of ambient air and injected into an N 2 -flushed gas bag.
Where multiple traps were deployed in a single pool, the bubbles from each trap were combined into a single sample.CO 2 was separated from CH 4 in the gas bags by filtering the gas sample through molecular sieve cartridges in the laboratory during sample processing (Garnett et al., 2011).Six samples yielded sufficient material for 14 C-CH 4 analysis (four from natural pools and two from restoration pools), and only one sample yielded enough for 14 C-CO 2 (natural pool; Table S2).
Sediment grab samples were collected from three pools at Cross Lochs, two natural pools and one restoration pool (Table S1).The top ~20 cm of pool floor sediment was collected using a simple corer and refrigerated at 4°C prior to analysis.Each pool sediment sample was cold acid washed (1 M HCl) overnight to remove carbonates, homogenized by stirring with a glass rod, freeze-dried, and then combusted to CO 2 for isotope analysis using the sealed quartz tube method (Boutton et al., 1983); glassware was cleaned by combusting at 450°C for at least 3 h.

| Isotopic analyses
The samples were processed at the United Kingdom's Natural Environment Research Council Radiocarbon Facility in East Kilbride, Scotland.The filtered DOC samples were rotary evaporated to solids.
DOC solids and POC filters were acid fumigated to remove any inorganic C, then combusted to CO 2 .POC samples were combusted using the standard Boutton et al. (1983) sealed quartz tube method which involves complete thermal oxidation of the organic matter; the POC and the filter were combusted together.Dissolved CO 2 samples were recovered from the molecular sieve cartridges by heating to 425°C (Garnett et al., 2019).The ebullition samples were first filtered through a molecular sieve cartridge to remove CO 2 , and the CH 4 was then oxidized using a platinum bead catalyst at 950°C to CO 2 .In one case, there was enough CO 2 in the ebullition sample for 14 C analysis as well.For all 14 C samples, the CO 2 produced was graphitized using Fe-Zn reduction.
Where there was >1 mL CO 2 , an aliquot was extracted for δ 13 C analysis using a dual-inlet isotope ratio mass spectrometer (Thermo Fisher Delta V) and reported relative to the Vienna Pee Dee Belemnite (VPDB) standard. 14C content was determined using accelerator mass spectrometry at the Scottish Universities Environmental Research Centre, East Kilbride.
Where sample size was <0.5 mL CO 2 , samples were analyzed at the Keck Carbon Cycle Accelerator Mass Spectrometer facility at the University of California Irvine.Following convention, 14 C content was normalized to a δ 13 C value of −25‰ and presented as fraction modern carbon (F 14 C); conventional radiocarbon ages were provided for samples with F 14 C content <1 in years before present (yBP) where "present" is defined as 1950 CE (Stuiver & Polach, 1977).Radiocarbon-dead and known-age standards were processed alongside the samples to quantify the 14 C background and verify the radiocarbon measurements, respectively.The 14 C-dead standards (bituminous and anthracite coal) quantified the extraneous C associated with the processing and were used to background correct the sample results using the method described by Donahue et al. (1990).
Barley mash from the Third International Radiocarbon Intercomparison (Gulliksen & Scott, 1995) and an internal humin standard (96H humin; Xu et al., 2004) were used as radiocarbon reference standards to verify the accuracy and precision of the sample results.

| Statistical analyses
The aquatic samples we collected contain mixtures of C of different ages that contribute to a single "'bulk" F 14 C value for each sample (Dean et al., 2019).We used an age distribution model to estimate the different age contributions to a given mixture for each 14 C sample (Dean, van der Velde, et al., 2018;Evans et al., 2014Evans et al., , 2022;;Raymond et al., 2007;Tanentzap et al., 2021).This approach assumes that C sources in a given mixture are dominated by C fixed into peat soils in the year prior to sampling, with contributions of exponentially less C from each preceding year up to a maximum peat age of 10,000 years.
This model conceptually follows empirical observations of both C accumulation and loss in peatland soils (Ratcliffe et al., 2018) and that discharge contributions to pools in blanket peatlands are mainly generated from near-surface flow paths (Holden et al., 2018).We estimate the age contribution to each 14 C mixture (sample) based on reconstructed atmospheric 14 CO 2 which was assumed to be fixed into plant material and formed the organic matter in peat soils each year over the past 10,000 years by solving Equation (1) for λ (the rate parameter, termed "k" in e.g., Evans et al., 2022): where F 14 C aq is sample 14 C content in F 14 C, F 14 C air is atmospheric and associated data files used to analyze and present the data in this study can be found in the associated Github repository (https:// github.com/ jfdea n1/ UKpea tland pool14C).The full atmospheric 14 CO 2 reconstruction is available in the Supporting Information.Analytical uncertainty was propagated through the model by giving an upper and lower bound for each sample (Table S1).We report the median value of the modelled age distribution for each sample: The age distribution modelling and other statistical analyses were carried out in R version 4.1.1(R Core Team, 2021).To explore differences between the isotopic composition of different C forms and pool locations, we used nonparametric Kruskal-Wallis tests with the kruskal.testfunction in R, supplemented by post hoc analyses consisting of Conover-Iman tests using the conover.testfunction and unpaired two-sample Wilcoxon tests using the wilcox.testfunction.We undertook linear regression analyses using the lm function to quantify the relationship between the isotopic composition of DOC, POC, and dissolved CO 2 , as well as between DOC and dissolved organic matter (DOM) absorbance parameters where available (Chapman et al., 2022).We used Cook's distance to identify potential outliers in the linear models using the cooks.distancefunction.To explore potential correlations between the isotopic composition of DOC, POC, and dissolved CO 2 and additional biogeochemical parameters available at the Cross Lochs location (collected within ±3 days of the isotope samples; Chapman et al., 2022), we used Spearman's rank correlation in the rcorr function.The details of where each analysis is applied are provided in the figure and table captions in the main text and Supporting Information.
We also utilized Keeling and Miller-Tans plots to estimate possible sources for the observed dissolved CO 2 (Campeau, Wallin, et al., 2017).
Both approaches assume a mixing of isotopically distinct C sources and conservation of mass (adapted from Campeau, Wallin, et al., 2017): where I obs is the observed sample isotopic composition and C obs is the paired sample CO 2 concentration, which results from a mixture of isotopic composition and CO 2 concentration in a background ( b ) and a single source ( s ).The composition of the source can be derived from the y-intercept of the simple linear equation in a Keeling plot or the slope of the linear equation in a Miller-Tans plot, assuming linearity in the mixing between two C sources without additional isotopic fractionation.

| Age and source of carbon in peatland pools
Across six peatland pool locations, we collected 94 aquatic and three sediment 14 C samples (Figure 1).Of the aquatic samples, (2) Median Age = −1 ln(2).
(3)  S1).Only the oldest solutions to the age distribution model are shown (Figure S4).CH 4 , methane; CO 2 , carbon dioxide; DOC, dissolved organic carbon; POC, particulate organic carbon.by primary production since ca.1955 CE when nuclear testing elevated the 14 C content of the atmosphere (Figure S4; Evans et al., 2014).DOC contained more 14 C-enriched samples than any other C form (median F 14 C content = 1.0663, range = 0.9690-1.1019),indicating a generally younger age dominated by bomb-peak C (Figure 1A).Dissolved CO 2 was the next most 14 C-enriched (median = 1.0294, range = 0.9420-1.0550),followed by CH 4 ebullition (1.0279, 0.8732-1.0758),and then POC (1.0050, 0.9370-1.0426),with pool sediments containing the lowest 14 C content (0.7608, 0.7371-0.9274).Dissolved CO 2 and CH 4 ebullition were statistically alike, as were CH 4 ebullition, POC, and pool sediment.The oldest 14 C age of 1090 ± 25 yBP was from ebullition CH 4 ; although it should be noted that two out of three pool sediment values were noticeably older than the other C forms.
We estimated the potential contribution of different aged peat layers to the C mixtures that make up the F 14 C observations (Figure 1B; Dean et al., 2019;Tanentzap et al., 2021).DOC likely contains high contributions (~50%) of 0-to 100-year-old C sources, while the other C forms are dominated by contributions of 0-to 300-year-old C (>50%).Organic C as old as 5000 years is common in the upper ~2 m of peat at the study sites (Ratcliffe et al., 2018), but there is no evidence of C this old in the pool waters.Millennial aged C (>1000 years) is likely only present in notable quantities in pool sediments (~50%; Figure 1B) and potentially some POC and ebullition CH 4 (~10%).For each sample, we also estimated its median age in yBSD from the age distributions (Table S1; Dean, van der Velde, et al., 2018).Due to elevated atmospheric F 14 C levels during the bomb peak, some samples can have median ages either side of this peak (Figure S4; Evans et al., 2014).For example, a dissolved CO 2 sample from pool P07 collected in November 2014 could have a median age of 3 or 133 yBSD (Table S1).We do not attempt to determine if one solution is correct or not, and instead plot all solutions together with the atmospheric F 14 C-CO 2 record to demonstrate that 49 of 94 (52%) of our F 14 C observations (excluding pool floor sediment) could have both a very young (<10 yBSD) and decadal to centennial (>56 yBSD) median age.These potential young median ages indicate that C in some pools could be almost entirely dominated by primary production in the decade prior to sampling (Figure S4).In Figure 1B, we only show the mean potential contributions of different aged peat layers from the old median ages.
There were some statistically significant differences in F 14 C between the different peatland pool locations (Figure 2).There is also dissimilarity between the number and type of C forms sampled across the sites, which was confounded by logistical challenges in obtaining enough aquatic C for 14 C analysis (see Section 2.2; Table S1; Dean et al., 2017).However, we found no notable patterns in the F 14 C observations due to the timing of the sampling campaigns (Figure S1).Despite the variability in sample number and timing, the F 14 C observations across 27 pools at six different peatland locations shared several statistical similarities; the majority of samples contained F 14 C > 1, which indicates a predominance of contemporary C fixed post-1950 CE (Figure 2).
Isotopic differences between natural and restoration pools were much more pronounced in the δ 13 C data (Figure 4).During CO 2 formation, most likely via respiration in these systems, δ 13 C is heavily fractionated so the boxplots in Figure 4  ).However, we found no notable patterns in our δ 13 C observations due to the timing of the sampling campaigns (Figure S3).
For dissolved CO 2 , F 14 C and δ 13 C were significantly related in both natural and restoration pools (p < .05; Figure 5C).F 14 C-CO 2 trended from enriched (F 14 C > 1) toward 1 as δ 13 C-CO 2 trended from −20‰ to −10‰.For DOC and POC, F 14 C and δ 13 C were not related in either pool type (Figure 5A,B).These results suggest that as the main sources of C in dissolved CO 2 trended toward either modern (~2013 CE) or ~1950 CE in age (depending on which side of the 14 C bomb peak the samples fall on), the concurrent δ 13 C content of dissolved CO 2 became progressively less negative (δ 13 C-CO 2 produced from organic matter respiration is generally in the −30‰ to −25‰ range; Dean et al., 2020).

| Relationship between carbon isotopes and other peatland pool parameters
Across all pools where F 14 C was concurrently measured in DOC and dissolved CO 2 , there was a statistically significant linear relationship between the two C forms (p < .05; Figure 6A).The statistical signifi-  S1).Only the oldest solutions to the age distribution model are shown (Figure S4).CO 2 , carbon dioxide; DOC, dissolved organic carbon; POC, particulate organic carbon; yBP, years before present; yBSD, years before sampling date.
type and for δ 13 C in natural pools (Figure 6B).An unclear relationship was also observed for DOC and POC (Figure 6E,F).F 14 C in DOC and POC was significantly related only in natural pools, while δ 13 C was significantly related across all observations but not when separated by pool type.For dissolved CO 2 and POC, only δ 13 C was significantly related and only across all observations, not when separated by pool type (Figure 6C,D).These regressions are likely impacted by the limited sample n when separating the observations by pool type.
For the Cross Lochs pools, we have additional observations of pool biogeochemical parameters from a parallel study (Chapman et al., 2022).We explored possible correlations between our isotopic observations and these additional data were available from concurrent measurement campaigns (campaigns ±3-4 days of each other).
These additional parameters were concentrations of DOC, POC, and dissolved CO 2 and CH 4 , pH, electrical conductivity, DO, water temperature, and the water table depth in the adjacent peat.F 14 C-DOC and F 14 C-CO 2 were significantly (p < .05)correlated with pool DOC concentrations, dissolved CO 2 concentrations, and pH; no other pool biogeochemical parameters were significantly correlated with F 14 C.
For δ 13 C, again the same pool parameters were significantly correlated: DOC and dissolved CO 2 concentrations, and pH (Table S3).Most of these correlations are impacted by low sample n, skewness in the data and grouping by pool type, which is also evident in Figures 5 and 6 and is primarily driven by differences in δ 13 C (Figure 4) rather than F 14 C (Figure 3).We therefore do not wish to overemphasize the importance of these relationships but present them for completeness.
We additionally looked for relationships between the quality (structure) of DOM in the pools and the DOC isotope observationsunfortunately, due to timing of the parallel sampling campaigns, only data for Cross Lochs in November 2014 and September 2015 were available (Chapman et al., 2022).DOC F 14 C and δ 13 C content were significantly related to SUVA, a proxy for DOM aromaticity (Figure S5; Helms et al., 2008).Neither DOC isotope was significantly related to DOM E4:E6 ratio, a proxy for molecular weight (Helms et al., 2008).
The Keeling and Miller-Tans analyses returned statistically significant linear relationships, despite only being available for a small number of Cross Lochs pools (n = 17; Figure S6).For F 14 C, the Keeling regression was statistically significant when considering all samples and for natural pools; the Miller-Tans regression was significant for all samples and when separated by pool type.For δ 13 C, the Keeling regression was only significant for all samples, not by pool type; the Miller-Tans regression was significant for all samples and when separated by pool type.Combined, these analyses estimated a CO 2 source for all samples with an isotopic composition of −18.6‰ and 1.0427-1.0486F 14 C.For natural pools, the source isotopic composition estimate was −15.8‰ and 1.0593-1.0767F 14 C; for restoration pools, the estimate was −18.4‰ and 1.0415 F 14 C. production (Figure 1; Figure S4).The majority of our samples indicate contemporary C mixtures dominated by C younger than 300 yBSD (Figure 1B), and these findings were generally consistent across the broad spatial coverage of the snapshot sampling campaign (Figure 2)

| Peatland pools dominated by contemporary carbon but old carbon is present in some forms
and pool type (Figure 3).This finding is surprising given that the top 0.5-1 m of peat surrounding our study pools contains substantial stores of C that is 1000-3000 years old (Ratcliffe et al., 2018), and the pool sediments contained C up to ~2500 years old (Figure 7).Pool location did not have a clear pattern of influence on our F 14 C observations (Figure 2), meaning the processes discussed hereafter were unlikely to be driven by differences between sampling locations.
That DOC is the youngest C form in these peatland inland water bodies follows previous 14 C studies on peatland streams, and rivers globally, where DOC is generally found to be modern (Billett et al., 2007(Billett et al., , 2012;;Campeau, Bishop, et al., 2017;Dean et al., 2019;Evans et al., 2007Evans et al., , 2022;;Marwick et al., 2015;Tipping et al., 2010), except where there is clear disturbance within their catchments (Butman et al., 2015;Hulatt et al., 2014;Moore et al., 2013).DOC-F 14 C was positively correlated with DOC concentration in the Cross Lochs pools, indicating that increasing DOC concentrations may be driven by contemporary C inputs (Table S3).
Globally, riverine dissolved inorganic C, of which dissolved CO 2 is a constituent, also tends to be modern but can show clear signals of geogenic C release from weathering, including in some United Kingdom peatlands (Billett et al., 2007;Marwick et al., 2015).Our dissolved CO 2 observations are strongly indicative of contemporary primary production, and unlikely to contain inputs from geologic C (Figure 1).Keeling and Miller-Tans plots for our dissolved CO 2 observations from Cross Lochs where CO 2 concentrations were available further indicate a source with a F 14 C content ranging from 1.0415 to 1.0767 and δ 13 C content ranging from −15.8‰ to −18.6‰, neither of which indicate substantial contributions from older peat or geologic C sources (Figure 6).These findings follow previous work on the age of peatland stream dissolved CO 2 , which tends to be dominated by contemporary primary production (Campeau, Bishop, et al., 2017;Leith et al., 2014).CO 2 -F 14 C was positively correlated with DOC and CO 2 concentrations, again indicating contemporary C inputs dominate in these pools (Table S3).
POC was older than DOC and CO 2 , statistically indistinct from pool sediments, and was estimated to contain 1%-20% C older than 1000 yBSD (Figures 1, 3 and 7).This pattern is reflected in the few studies that have looked at 14 C of DOC and POC concurrently in temperate peatland inland waters (Billett et al., 2012), with millennial-aged POC only evident in heavily disturbed catchments (Hulatt et al., 2014).Work in permafrost peatlands also indicates that POC tends to be older than other inland water C forms (Dean et al., 2020).
Previous work synthesizing global riverine 14 C-POC further suggests the age of POC is correlated with, but generally older than, DOC (Marwick et al., 2015).
Our ebullition data were statistically similar to pool sediments and POC, both of which tended to be older, and dissolved CO 2 which tended to be modern (Figure 1A).We present seven ebullition observations (one CO 2 and six CH 4 ).Five of our ebullition observations had F 14 C values >1, indicating a predominance of post-bomb C; two had older 14 C ages indicating mixtures dominated by centennial to millennial C; these observations were estimated to contain 0%-35% C older than 1000 yBSD (Figure 1B; Table S1).To our knowledge, these are the first 14 C ebullition data for temperate peatland inland waters.In permafrost peatland inland waters, ebullition can be an important pathway for the substantial release of millennial CO 2 and CH 4 (Bouchard et al., 2015;Walter Anthony et al., 2016).Our POC and ebullition observations suggest that millennial C is present in the water column of some peatland pools and can be emitted to the atmosphere.
Although data are limited (n = 3), pool floor sediments had the oldest 14 C ages in our pools, with ~50% of the C in these samples estimated to be older than 1000 years (Figure 1B).This indicates that there is potential for millennial-aged C input into natural peatland pools possibly via erosion.While older C is unlikely to be contributing to the DOC and CO 2 , POC and ebullition CH 4 displayed statistical similarities with these limited sediment data (Figure 1B; Table S1) meaning older C in pool sediments could be contributing to POC or ebullition CH 4 .

| Limited variability of F 14 C across pool location and type
No clear pattern emerged when considering all F 14 C observations by location (Figure 2).The four locations sampled in the wider snapshot campaign that covered only natural pools were all statistically alike (Munsary, Silver Flowe, Garron Plateau and Slieveanorra, sampled in September 2015).There were inconsistent statistical similarities and differences between these four sites and Loch Lier (only sampled once in November 2014 and included natural and restoration pools) and Cross Lochs (sampled in May and November 2014, and September 2015)-these comparisons were hampered by differences in field campaign timing and C form and pool type sampled.For DOC, POC, and dissolved CO 2 , which were consistently sampled across both natural and restoration pools, only DOC-F 14 C was significantly different between the two pool types (Figure 3A).This difference is driven by a likely predominance of younger C in restoration pool DOC relative to the natural pools (Figure 3B).

| δ 13 C content distinctly different between natural and restoration pools, suggesting differences in CO 2 degassing and DOC and POC transformations
Differences between natural and restoration pools were much more pronounced in the δ 13 C observations than F 14 C, with δ 13 C consistently higher (less negative) and with a greater range in natural pools compared to restoration pools (Figure 4).
For CO 2 , the higher δ 13 C in natural pools indicates increased mixing with atmospheric CO 2 and/or higher CO 2 co-production with CH 4 relative to restoration pools (Campeau, Wallin, et al., 2017).F 14 C and δ 13 C were significantly related in our CO 2 samples (Figure 5c).This relationship is indicative of pool CO 2 mixing with atmospheric CO 2 (atmospheric CO 2 F 14 C = 1.02-1.03,δ 13 C = −8.5‰).Assuming the CO 2 isotopic composition we observed is a mixture between atmospheric CO 2 and another source, Keeling and Miller-Tans analyses indicated dissolved CO 2 is primarily derived from a relatively young (F 14 C = 1.0415-1.077)biogenic (−15.8‰ to −18.6‰) source, likely respired modern plant material (Figure 7; Figure S6).These linear models assume that mixing between atmospheric CO 2 and pool CO 2 remains fixed across observations, which is not the case here; furthermore, δ 13 C fractionation is known to occur during CO 2 degassing from water.Hence, these source isotopic compositions should be considered estimates.
It is possible that the relationship between CO 2 F 14 C and δ 13 C could instead indicate CO 2 production during methanogenesis if the F 14 C content of CH 4 in the pools is consistently lower than that of CO 2 .We do not have enough concurrent observations of F 14 C-CH 4 to answer this question definitively.A single dissolved CH 4 sample in this study was contemporary (Figure 1A), as were previous F 14 C observations of restoration pool dissolved CH 4 (Dean et al., 2017), while our ebullition CH 4 observations ranged from contemporary to millennial in age (Figure 1).There was a significant correlation between CO 2 concentration and δ 13 C-CO 2 , suggesting lower concentrations of CO 2 are associated with higher (less negative) δ 13 C-CO 2 (Table S3).This relationship further supports mixing between pool CO 2 and the atmosphere: δ 13 C-CO 2 increases with decreasing CO 2 concentrations which indicates mixing with the atmosphere and subsequent degassing.We would expect δ 13 C-CO 2 to become less negative as CO 2 concentrations increase if this enrichment of δ 13 C-CO 2 was due to co-production during methanogenesis (Campeau, Wallin, et al., 2017).Concentrations of CO 2 were consistently found to be much lower in natural pools than restoration pools in previous work at the Flow Country sites, which further indicates CO 2 degassing is occurring more readily in natural pools (Chapman et al., 2022).The natural pools tended to be larger in surface area and thus more exposed to wind-driven turbulence that would be higher at the center of wider pools where there is less shelter from the wind, and this turbulence drives atmospheric exchange (Figure S1; Holgerson & Raymond, 2016).Furthermore, Chapman et al. (2022) found a significant linear relationship between CO 2 and CH 4 concentrations in both natural and restoration pools, but the slope of the regression was very shallow suggesting that CO 2 co-production with CH 4 is not the main driver of higher CO 2 concentrations in the pools.
We therefore attribute the difference in δ 13 C-CO 2 between the two pool types to increased exchange with the atmosphere in the natural pools.This demonstrates that natural peatland pools are active sites of CO 2 emission to the atmosphere (Holgerson & Raymond, 2016).
CO 2 dynamics in peatland pools may also be driven by DOC decomposition (Chapman et al., 2022;Pickard et al., 2017).DOC and CO 2 were isotopically coupled when considering all observations together (Figure 6).This isotopic coupling could indicate production of CO 2 from DOC decomposition.Significant correlations were found between both F 14 C-CO 2 and δ 13 C-CO 2 and DOC concentrations, and between both F 14 C-DOC and δ 13 C-DOC and CO 2 concentrations, further indicating coupling between DOC and CO 2 dynamics in these pools (Table S3).Previous work at the Flow Country sites found that concentrations of DOC and CO 2 were more strongly related in restoration pools than natural pools (Chapman et al., 2022).
When separating our observations by pool type, we only saw a statistically significant relationship between DOC and CO 2 for δ 13 C in restoration pools, not in natural pools and not for F 14 C in either pool type (Figure 6A,B).We also repeated this analysis to remove potential outliers identified using Cook's distance (Figure S7).Repeating these analyses without the six identified potential outliers yielded significant relationships between DOC and CO 2 for F 14 C overall and when separated by pool type, but did not substantially alter the relationships for δ 13 C (Figure 6B; Figure S7).These relationships further indicate that DOC and CO 2 are likely from a similarly aged peat C source, though the δ 13 C relationship is complicated by pool CO 2 mixing with the atmosphere (Figure 7).
The wider range of δ 13 C-DOC values in natural relative to restoration pools (Figure 4) may also indicate increased DOC decomposition occurring within the natural pools.Decomposition can cause fractionation of δ 13 C, either microbially or through photo-oxidation, although the impact of decomposition on δ 13 C-DOC in inland waters is not well characterized (Abbott et al., 2016;Dean, van Hal, et al., 2018;Marwick et al., 2015).Our limited concurrent isotope-DOM structural observations in the Cross Lochs pools indicate that increased DOM decomposition (lower DOM aromaticity as indicated by lower SUVA values) is related to higher (less negative) δ 13 C-DOC values (Figure S5b).Both relationships seen in Figure S5 could be artifacts of the isotopic differences in DOC between the natural and restoration pools.Regardless, the lower SUVA (aromaticity) values in the natural pools suggest a more decomposed type of DOC resulting in enhanced production of CO 2 (Chapman et al., 2022;Dean et al., 2019;Worrall & Moody, 2014).The natural pools at our sites were generally larger in surface area and had longer water residence times than restoration pools (Figure S1).Larger surface area facilitates increased likelihood of photo-oxidation due to less shading by surrounding vegetation, while longer residence times increases opportunities for DOC decomposition within the pools (Catalán et al., 2016;Chapman et al., 2022).
The lower (more negative) δ 13 C-POC values in the restoration pools indicate recent POC production from fresh vegetation (Arsenault et al., 2019;Chapman et al., 2022;Dean et al., 2020;Marwick et al., 2015).However, this suggestion is not supported by the F 14 C-POC data which show that POC is not significantly younger in restoration pools as would be expected if fresh vegetation was the primary source of POC (Figure 3).Furthermore, our POC observations were statistically alike with the pool sediments which contains substantial C older than 1000 years (Figure 1A).Decomposition of POC during residence in the pools, or in the surrounding peat prior to its release into the pool, could explain the differences in δ 13 C-POC between the two pool types (Attermeyer et al., 2018).POC decomposition could also contribute to CO 2 production as shown by a significant relationship between δ 13 C in POC and CO 2 (Figure 6d), and a significant correlation between δ 13 C-POC and CO 2 concentrations (Table S3).Chapman et al. (2022) showed that POC deposition rates were three times higher in natural pools compared to restoration pools, and that DOC may be flocculating to form POC in natural pools.The generally longer water residence times in the restoration pools, and the accompanying decrease in the energy of water flowing through pools, could increase the opportunity for DOC flocculation and POC deposition on pool floors (Figure S1; Chapman et al., 2022).
These transformations are likely to impact the isotopic composition of both POC and DOC.DOC and POC F 14 C were only significantly coupled in natural pools (Figure 6E), while δ 13 C-POC and DOC concentrations were negatively correlated (Table S3).This could indicate some exchange between DOC and POC, more so in natural pools, such as DOC flocculation to POC (Attermeyer et al., 2018;Dean et al., 2019).
For each of DOC and POC individually, F 14 C and δ 13 C were not related (Figure 5A,B), suggesting that the differences in DOC and POC decomposition as indicated by δ 13 C fractionation between the two pool types is not driven by the age (F 14 C content) of DOC and POC.
The limited significant relationships between POC and CO 2 isotopic composition suggest that POC degradation is not a main driver of CO 2 production.

| Study limitations and future work
Our  S3).No distinct hydrological events were recorded at Cross Lochs prior to or during our field campaigns, with water tables staying within 0-20 cm of the peat surface for the duration of 2014-2015 (Holden et al., 2018).Drain blocking at our study sites occurred more than 10 years ago, meaning that the system has had time to respond to this intervention and potentially even reach an equilibrium statethis could in part explain the similarity of C ages in the two pool types.We thus consider our observations to represent relatively stable conditions at our study sites.
Our observations may not represent the full potential range of environmental conditions at the study sites.The campaigns likely missed periods with the highest temperatures and longest sunlight hours, but did capture the biologically active early summer conditions (May) and periods of relatively low water tables in late summer (September).Warmer and lighter conditions could be a time of increased organic C transformation, such as photo-oxidation, and/or increased inputs from primary production.Our campaigns also do not represent periods of especially large variations in local water tables (Holden et al., 2018).Substantial water table variations can be important for whether shallow younger C or deeper older C is mobilized into inland water systems (Barnes et al., 2018;Chapman et al., 2022).
Our campaigns also do not capture diurnal fluctuations, known to be important for the magnitude of inland water greenhouse gas emissions (Attermeyer et al., 2021;Gómez-Gener et al., 2021) ing and deeply incised peatlands (Moore et al., 2013;Waldron et al., 2019).
14C-CO 2 content at time (t), and T is the age of the sample in years before sampling date (yBSD).Atmospheric 14 CO 2 was derived fromReimer et al. (2020) for 10,000 years ago to 1950 CE, and fromHua et al. (2022) from 1950 to 2015 CE.These data were then interpolated using a cubic spline (splinefun in R) to create a continuous data series.

1
Distribution of radiocarbon observations.(A) All radiocarbon observations from this study in fraction modern (F 14 C).Separate boxplots are ordered by C form: DOC (n = 34), POC (n = 22), dissolved CO 2 (n = 30), ebullition (bubble) CO 2 (n = 1), dissolved CH 4 (n = 1), ebullition CH 4 (n = 6), and pool sediments (n = 3); thick horizontal lines of the boxes represent medians, limits of the boxes represent upper and lower quartiles, whiskers extend to 1.5 times the interquartile range, dots represent all data points.Lowercase letters indicate statistically significant differences (p < .05) between C forms using a Kruskal-Wallis test and Conover-Iman post hoc (excluded where sample n < 2); the oldest radiocarbon ages observed for each C species are indicated in years before present (yBP) where "present" = 1950 CE; any data plotting above the dashed horizontal line is considered radiocarbon "modern" (i.e., F 14 C > 1, or younger than 1950 CE).(B) The potential contribution of soil age layers (classes) to the observed mixtures (samples), based on the age distribution model(Tanentzap et al., 2021) and ordered by C form.The relative soil age classes are in years before sampling date (yBSD); F 14 C analytical uncertainty (±1σ) was propagated through the age distribution model but is not shown here (Table

Garron
Plateau had the highest median F 14 C content (1), Slievearnorra (1.0378, range = 1.0305-1.1019),Cross Lochs (1.0218, range = 0.8732-1.0962),and Loch Lier (1.0187, range = 0.9370-1.0807).Garron Plateau and Loch Lier were statistically distinct, while statistical similarities were present between these locations and all others (see lowercase letters denoting statistical significance in Figure2).This is partly complicated by the differing number of samples collected at each site.Cross Lochs was the focus of three sampling campaigns across 2014 and 2015, while the other sites were sampled in single snapshot campaigns.

F
All F 14 C observations across the different peatland pool locations (excluding sediments).Separate boxplots are ordered by location: Cross Lochs (n = 59), Loch Leir (n = 13), Munsary (n = 4), Silver Flowe (n = 6), Garron Plateau (n = 6), and Slieveanorra (n = 6); box features follow Figure 1.Lowercase letters indicate statistically significant differences (p < .05) between locations using a Kruskal-Wallis test and Conover-Iman post hoc; the oldest radiocarbon ages observed for each location are indicated in years before present (yBP); any data plotting above the dashed horizontal line are considered radiocarbon "modern." are ordered by C form.Across all pools, DOC had a mean δ 13 C value of −27.5‰ (range = −29.1‰ to −23.4‰), POC −26.9‰ (−29.9‰ to −24.3‰), and CO 2 −14.4‰ (−20.5‰ to −9.4‰), with DOC and POC values statistically alike.In all C forms, δ 13 C was significantly higher (less negative; p < .05) in natural pools.Restoration pool DOC had a noticeably smaller range of δ 13 C values compared to natural pools and compared to the range of δ 13 C values seen in POC and CO 2 .Observations of δ 13 C come from the same samples as 14 C, so face similar problems with consistency in sample collection (see Section 3.1 cance of this relationship extended to δ 13 C across all observations and for restoration pools, but broke down when considering F 14 C by pool F I G U R E 3 Distribution of radiocarbon observations in natural and restoration pool types.(A) F 14 C observations in the two different pool types sampled in this study (excluding where sample n < 10): natural pool DOC (n = 23), POC (n = 11), and dissolved CO 2 (n = 19); restoration pool DOC (n = 11), POC (n = 11), and dissolved CO 2 (n = 11).Box features follow Figure 1; the dashed horizontal lines represent the mean F 14 C value for each pool type (natural = 1.0268, restoration = 1.0347).Lowercase letters indicate statistically significant differences (p < .05)using a series of unpaired two-sample Wilcoxon test by pool type only, and then by C form and pool type (e.g., natural DOC vs. restoration DOC).The oldest radiocarbon ages observed for each C form are indicated in yBP; any data plotting above the dashed horizontal line is considered radiocarbon "modern."(B) The potential contribution of soil age layers (classes) to the observed mixtures (samples), based on the age distribution model (Tanentzap et al., 2021) and ordered by C form and pool type.The relative soil age classes are in yBSD; F 14 C analytical uncertainty (±1σ) was propagated through the age distribution model but is not shown here (Table

F
I G U R E 4 Stable C isotope observations (δ 13 C in ‰ Vienna Pee Dee Belemnite) in the two pool types measured in this study (excluding where sample n < 10): natural pool DOC (n = 23), POC (n = 11), and dissolved CO 2 (n = 18); restoration pool DOC (n = 11), POC (n = 10), and dissolved CO 2 (n = 11).Box features follow Figure 1; the dashed horizontal lines represent the mean δ 13 C value for each C form (DOC = −27.56‰,POC = −26.9‰,CO 2 = −14.4‰).Lowercase letters indicate statistically significant differences (p < .05)using a Kruskal-Wallis test (C form only) and an unpaired two-sample Wilcoxon test (C form and pool type, e.g., natural DOC vs. restoration DOC).CO 2 , carbon dioxide; DOC, dissolved organic carbon; POC, particulate organic carbon.F I G U R E 5 F 14 C versus δ 13 C from all sites where paired data were available: DOC (A; n = 34), POC (B; n = 21), and dissolved CO 2 (C; n = 29).Black lines and text represent natural pools, gray lines and text represent restoration pools.R 2 -values and p-values are from linear regression; * indicates statistical significance (p < .05);lines are only shown for statistically significant relationships.CO 2 , carbon dioxide; DOC, dissolved organic carbon; POC, particulate organic carbon.
, nor extreme climatic events such as erosive storm events or long-term drought due to climate change, each of which could significantly alter hydrologic flow paths and C supply to peatland pools.Therefore, future work could explore the impact of these potential drivers of C supply to the isotopic composition of C mobilized into and emitted from peatland pools.Interesting future work could also consider the evolution of the age of C exported by peatland ditches before, during, and after blocking to determine if, and for how long, the age of aquatic C export changes following restoration efforts.Furthermore, despite the inherent difficulty in collecting reliable ebullition data, further study is warranted as it may be a | 13 of 16 DEAN et al. potentially important emission pathway of contemporary and old peat C.5 | CON CLUS IONSWe show that the dominant age (source) of C in pools across a range of peatland locations in the United Kingdom is contemporary primary production.Our limited evidence for the release of old CH 4 via ebullition and the presence of old POC in pool water columns indicates that millennial C can be mobilized in some peatland pools and that these pools are capable of releasing old, previously stored peat C to the atmosphere.Overall, our results indicate that both natural and restoration pools are dominated by contemporary aquatic C. Our analyses identified few significant relationships between peatland pool C isotopes and other pool biogeochemical parameters, although positive correlations between F 14 C in DOC and CO 2 and DOC and CO 2 concentrations further indicate contemporary C inputs are dominating.We do show that these pools are dynamic, transforming C from one form to another, and releasing some of this C back to the atmosphere, with the generally larger natural pools being more active sites of C transformation and emission.The limited significant relationships between DOC and POC transformations (indicated by δ 13 C) and F 14 C suggest the age of the C being transformed is not an important control on the degree to which C decomposition and emission is occurring in these pools.With pools being active sites for the transformation and emission of peatland C, the lack of consistently significant correlations between pool 14 C ages and C concentrations also indicates that they remain vulnerable to changes in C inputs from both old and young peat layers driven by peatland disturbance.Addressing our hypothesis, that peatland pools contain mainly contemporary C but display clear signs of old C release if peatland destabilization is occurring, we observed an overwhelming dominance of contemporary C in most aquatic C forms.Ditch blocking to rewet our study peatlands was effective in preventing the release of old C via aquatic export, as demonstrated by the limited evidence for mobilization of old C. The question remains as to how peatland pool C cycling will respond to substantial influxes of old C from peatland disturbance, and whether restoration pools are as effective in preventing loss of old C in heavily erod- . Ditch construction to drain the peatlands at Cross Lochs and Loch Lier occurred in the 1970s (exact dates unknown) and targeted areas where peat was naturally shallower and better drained.Ditch blocking for restoration purposes through rewetting by raising the water table took place in 1998 at Loch Lier and 2002 at Cross Lochs, approximately 16 and 12 years prior to our sampling campaigns, respectively.
study provides a unique first insight into peatland pool C age characteristics but remains a "snapshot."From a practical perspective, access could not be afforded by landowners when other priori-