East Siberian Arctic inland waters emit mostly contemporary carbon

High-latitude permafrost regions store about half of the global soil organic carbon, much of which is vulnerable to release as Arctic temperatures rise. Inland waters act as key pathways for this terrestrial carbon to be transformed and emitted as CO2 and CH4. Carbon entering waters can be contemporary (from seasonal topsoil turnover and vegetation dynamics) or pre-aged (old Holocene and ancient Pleistocene carbon, including Yedoma deposits). Prior work shows both young and old carbon can fuel inland water emissions, but the relative contributions at the landscape scale remain uncertain, especially in regions with substantial Yedoma. This study aims to quantify the age and source contributions of all major inland water carbon forms (DOC, POC, dissolved CO2 and CH4) across a representative northeast Siberian permafrost landscape during summer, testing whether contemporary carbon dominates emissions or whether thaw-released pre-aged carbon is a major driver.

Previous radiocarbon studies in Yedoma-influenced regions indicate that pre-aged carbon is highly susceptible to microbial decomposition to CO2 upon release into streams, and thermokarst lakes can emit very old CH4 via ebullition (up to ~42,900 yBP). Conversely, in many non-Yedoma permafrost regions, atmospheric emissions are dominated by contemporary carbon, and ice-core records suggest large-scale emissions of pre-aged CH4 have been unlikely over the last ~15,000 years. Studies also show seasonal controls: winter can favor older dissolved gases due to isolation from contemporary inputs under ice. Overall, the literature highlights site- and process-specific variability but leaves unresolved the landscape-scale balance of contemporary versus pre-aged carbon fueling inland water emissions.

Study area and period: Oligotrophic tundra in the Kytalyk Nature Reserve, Indigirka lowlands, northeast Siberia (70.83°N, 147.49°E), underlain by continuous permafrost. Landscape includes drained lake basins, fluvial floodplain, numerous ponds, small lakes, and a thermokarst lake eroding into a Yedoma ridge. Sampling occurred during peak summer (25 July–17 August 2016) when active layer depth and thermokarst activity were maximal. Sampling design: Sample types included shallow tundra ponds (P01–P12), a fluvial system (stream and connected pond; S01–S04), small shallow lakes (L06–L12), a thermokarst lake eroding Yedoma (L01–L05), and direct Yedoma meltwater from thawing ice-rich sediments adjacent to the thermokarst lake. For each inland water type, concurrent measurements were made of concentrations and isotopes for dissolved organic carbon (DOC), particulate organic carbon (POC), and dissolved CO2 and CH4. Concentrations for DOC/POC and dissolved gases were typically sampled three times at 3-day intervals; δ13C and 14C were collected once per site group due to cost constraints (with ponds and stream sampled at multiple points; lakes at one location each). DOM absorbance and fluorescence spectroscopy were performed where 14C samples were collected to infer source and degradation pathways. Isotopic measurements: 14C and δ13C were analyzed for DOC, POC, dissolved CO2, and dissolved CH4 using established field collection and laboratory methods, including headspace equilibration and molecular sieve trapping for gases. Radiocarbon results are reported in pmC and conventional 14C ages. Isotope mass balance: A Bayesian mixing model (simmr in R) estimated proportional contributions of five source end-members to observed 14C of each carbon form by site group: (1) modern atmospheric CO2 (2013; 103.13 pmC), (2) topsoil organic matter representing 1950–2012 accumulation (123.64 ± 22.61 pmC), (3) basal peat (250–1180 yBP; 91.95 ± 7.16 pmC), (4) old early Holocene to late Pleistocene soil carbon (55.13 ± 17.19 pmC), and (5) ancient Pleistocene Yedoma carbon (3.09 ± 6.19 pmC). Model fitting used MCMC with propagated source uncertainties; convergence criteria followed published guidance. Flux calculations: Diffusive fluxes of dissolved CO2 and CH4 were computed as F = k × (Cobs − Ceq). Gas transfer velocity (k) was derived from k600 and Schmidt number scaling: for lakes, wind-based parameterization (k600 = 2.07 + 0.215 U10^0.5); for ponds, low k600 = 0.36 cm h−1; for streams, k600 = 13.1 cm h−1. U10 was obtained from the on-site eddy covariance tower. Atmospheric equilibrium concentrations used measured ambient CO2 and CH4. Tundra fluxes: Net ecosystem CO2 exchange and CH4 fluxes were measured by eddy covariance (10 Hz instruments at 4.7 m) and processed with EddyPro, including standard corrections, despiking, and quality control, yielding daily mg C m−2 d−1. Upscaling and statistics: Fluxes over the 23-day period were upscaled to a 15.9 km² area encompassing the sampled waters and surrounding tundra using mapped surface coverage (tundra 89.4%, ponds 4.9%, fluvial 1.2%, small lakes 1.2%, thermokarst lake 3.2%). Uncertainty was propagated to produce ranges. Proportions of contemporary vs pre-aged contributions to gaseous fluxes were derived by applying source proportions from the mixing model. Statistical analyses used linear models, ANOVA with Tukey tests, and Pearson correlations in R; certain correlations excluded Yedoma meltwater due to incomplete data for gases.

• Radiocarbon ages showed a clear gradient: ponds were youngest (modern to ~1–2 kyr BP), fluvial and small lakes similar, while thermokarst lake and Yedoma meltwater contained much older carbon (up to 29,355 ± 2,967 yBP).
• Across all inland waters, >80% of total carbon was contemporary during summer; in thaw-impacted systems (thermokarst lake, Yedoma meltwater) pre-aged sources contributed >50% of carbon.
• Gaseous carbon (CO2 and CH4) was consistently younger than DOC and POC, indicating emissions were primarily fueled by contemporary carbon decomposition.
• δ13C signatures and ε (δ13C-CO2 − δ13C-CH4 = 32.7–45.6; median 42.1 ± 4.8) indicated microbial methane production mainly via bacterial methyl-type fermentation; all sites were supersaturated in CO2.
• Concentrations: DOC was highest (mean 15.8 ± 12.3 mg C L−1), POC and dissolved CO2 were similar (2.6 ± 3.0 vs 2.8 ± 3.2 mg C L−1), CH4 lower (0.1 ± 0.3 mg C L−1). Yedoma meltwater had very high organic carbon (DOC 42.5 mg C L−1; POC 8.5 mg C L−1).
• Higher dissolved CO2 concentrations were associated with younger 14CO2 (p < 0.05, R2 = 0.71; exponential fit).
• DOM spectroscopy indicated predominantly terrestrial origin (HIX 0.6–0.7) with decomposition via both microbial processing and photo-oxidation; younger DOC tended to be more photosensitive, older DOC more microbially processed.
• Diffusive emissions (mg C m−2 d−1): fluvial systems had the highest CO2 and CH4 fluxes (CO2 1,820 ± 1,118; CH4 56 ± 29). Ponds, small lakes, and thermokarst lake had lower and statistically similar diffusive fluxes (CO2 ~290 ± 232 to 394 ± 89; CH4 ~18 ± 6 to 21 ± 7).
• Upscaled to 15.9 km² and 23 days: inland waters emitted 20.4 ± 13.2 Mg C (16.8 ± 10.4 Mg C from ponds/fluvial/small lakes and 3.6 ± 2.8 Mg C from the thermokarst lake). Contemporary C emissions were 17.0 ± 10.9 Mg C; pre-aged emissions 3.5 ± 2.3 Mg C, offsetting 1.9 ± 1.2% and 0.4 ± 0.3% of the tundra sink, respectively.
• The landscape acted as a net carbon sink during the study period (−876.9 ± 136.4 Mg C), dominated by tundra CO2 uptake (−2,784 ± 349 mg C-CO2 m−2 d−1) partially offset by tundra CH4 emissions (43 ± 5 mg C-CH4 m−2 d−1).

The results show that contemporary terrestrial carbon predominantly fuels inland water CO2 and CH4 emissions in this northeast Siberian permafrost landscape during summer. Younger ages of dissolved gases relative to DOC and POC imply that the most labile organic matter pools are young and are rapidly converted to gases, with photo-oxidation likely enhancing the decomposition of contemporary DOC under continuous summer daylight. Although thaw mobilizes pre-aged carbon (evident in thermokarst lake and Yedoma meltwater), its contribution to diffusive gaseous emissions was generally secondary where contemporary inputs are abundant. Instances of very old CH4 reported elsewhere (notably ebullition from deep sediments) likely reflect environments where contemporary labile carbon is limited (e.g., taliks, winter ice cover), allowing microbes to utilize older substrates. The study further suggests that lateral export of DOC and POC—often older and at higher concentrations than dissolved gases—could be an important pathway for pre-aged carbon loss to downstream systems where it may be decomposed or buried. Overall, inland water emissions in this landscape offset only a small portion of the tundra CO2 sink over the study period, but their sensitivity to changes in contemporary carbon cycling (and to thaw-driven changes in hydrology and waterbody coverage) indicates potential for increased emissions with ongoing Arctic change.

Simultaneous radiocarbon measurements across all major carbon forms reveal that summer inland water CO2 and CH4 emissions in an East Siberian permafrost landscape are dominated by contemporary carbon, while thaw-affected sites exhibit significant contributions from pre-aged carbon to all forms. The younger ages of dissolved gases relative to DOC and POC highlight rapid turnover of young, labile organic matter, likely aided by photo-oxidation. At the landscape scale during late summer, inland waters constituted a small offset to the tundra carbon sink. Future work should quantify year-round dynamics (including winter under-ice periods), incorporate ebullitive fluxes, resolve lateral DOC/POC export and fate, and assess how changing inland water coverage and connectivity modulate the balance between contemporary and pre-aged carbon in emissions.

• Temporal scope limited to late summer (25 July–17 August 2016); results may not represent shoulder seasons or winter when older dissolved gases can accumulate under ice.
• Ebullitive CH4 fluxes (often older and potentially large) were not measured, possibly underestimating total and pre-aged CH4 emissions in some systems.
• Radiocarbon sampling density was limited by cost (single locations for lakes; few ponds/stream points), which may miss spatial heterogeneity.
• Isotope mixing model relies on predefined source end-members and assumes mixing without fractionation effects on 14C; uncertainties in end-member characterization propagate to source apportionment.
• Upscaling restricted to a 15.9 km² area and 23-day period; generalization beyond this spatial/temporal window is uncertain.
• Lateral transport of DOC/POC was not included in the comparison with the tundra sink, potentially underrepresenting total carbon loss pathways.
• Some correlations excluded Yedoma meltwater due to incomplete gas data, limiting inference for that end-member.