Carbon emission from Western Siberian inland waters

The study addresses how much carbon is emitted from inland waters of Western Siberia, a high-latitude region with extensive peatlands (~70 Pg C) and widespread permafrost that is rapidly warming. Thawing permafrost mobilizes organic carbon, leading to degradation and atmospheric CO2 and CH4 release, and transports substantial terrestrial carbon to rivers and lakes where additional outgassing occurs. Existing assessments of high-latitude inland-water carbon emissions are scarce, often based on indirect or low-resolution data, and largely limited to small catchments, resulting in major uncertainties. Western Siberia is especially important because the Ob’ River basin spans all permafrost zones and has large water coverage over flat terrain, yet no direct regional C emission estimate previously existed. The research aims to quantify annual carbon emissions (CO2 + diffusive CH4) from inland waters (rivers, streams, lakes, ponds) across all permafrost zones in Western Siberia and to place these emissions in the context of the regional carbon budget (land C uptake, river export to the Arctic Ocean, and coastal sea uptake).

The paper situates its work within prior global and high-latitude studies showing significant inland water CO2 and CH4 emissions, but notes that high-latitude assessments are limited and uncertain, often relying on indirect calculations from pH and alkalinity or modeled fluxes. Prior global estimates included Siberian systems but used sparse snapshot data that poorly capture annual emissions. Studies indicate that small streams and ponds can contribute disproportionately to emissions, yet are underrepresented in inventories. Previous research in Western Siberia documented high riverine CO2 emissions near the permafrost boundary and high emissions from thermokarst lakes, while global datasets such as GRWL and GLOWABO provide surface areas for larger rivers and lakes. There is also evidence that pond size distributions may not follow Pareto laws, affecting area extrapolation and emission upscaling.

Study area encompasses the Ob’, Pur, and Taz River basins across all Western Siberian permafrost zones. The authors combined direct measurements of C emission rates with mapped and extrapolated surface areas to upscale to the regional annual total.

• Water surface areas: Used GRWL for rivers and GLOWABO for lakes. Clipped datasets to Western Siberia, separated Ob’ main channel from other rivers, and stratified by permafrost zone using the Circum-Arctic Permafrost map. Rivers <90 m wide were excluded from GRWL as in prior work.
• Extrapolation for small systems: Estimated areas of streams (<90 m wide) via Pareto extrapolation with shape parameter 0.93 and a first-order stream minimum width of 0.32 m. Estimated pond areas (<0.01 km²) using Pareto extrapolation with shape parameter 1.19 and a minimum pond area of 0.000115 km². Recognizing that pond areas may not follow Pareto, an alternative approach used pond fractions from satellite inventories at several sites to calculate pond coverage per permafrost zone.
• Ob’ main channel CO2 and C emission: Conducted first direct continuous pCO2 measurements (Vaisala GMP222) in summer 2016 (31 July–11 August) at 10-min intervals (4938 raw readings; 4396 used across permafrost zones). Calculated CO2 evasion using a gas transfer coefficient k median of 4.464 m d−1 (from floating chamber measurements), a pH-dependent chemical enhancement factor, and atmospheric pCO2 of 390 ppm. Adjusted for seasonal variation by doubling summer pCO2 to approximate spring values and averaging for the open-water period. Estimated CH4 as 1.19% of total C emission rate from river data, then integrated over water area and ice-free season length per zone.
• Rivers and streams: For rivers, generated permafrost-zone-specific normal distributions of CO2 emission rates from 58 rivers (n=116) and assigned rates to each river area pixel (n=882,124). Added CH4 at 1.19% of C emission, estimated season length via a latitude–season regression (R²=0.99), multiplied by area and duration, and summed. For streams, used the published median C emission rate for small watersheds (5.67 g C m−2 d−1), multiplied by extrapolated stream area and median open-water season length (180.6 days).
• Lakes and ponds: For permafrost-affected lakes (76 lakes, n=228), created permafrost-zone-specific distributions of C (CO2 + diffusive CH4) emission rates and upscaled to all mapped lakes (n=612,003 observations), applying a latitude–season regression (R²=0.96). For permafrost-free lakes (13 lakes), used the median C emission rate (0.6 g C m−2 d−1) to upscale to mapped lakes (n=361,777). For ponds, applied median C emission rates for permafrost-affected (1.12 g C m−2 d−1) and permafrost-free (0.6 g C m−2 d−1) regions to extrapolated pond areas and season lengths; alternatively, used satellite-derived pond fractions (Muster et al. 2019) to estimate pond areas and emissions.
• Regional context data: Land C uptake (NEE) from NASA SMAP Global Daily 9 km NEE v4 for 2016, integrated over the region. River DOC and DIC export to the Arctic Ocean compiled for Ob’ (2003–2009) and Pur/Taz (2013–2014). Kara Sea net CO2 uptake derived from 1° gridded flux products for 2014, clipped to Kara Sea boundaries and integrated over area and time.
• Uncertainty: Monte Carlo analysis for Ob’ main channel emissions; for component-wise totals, assumed 15% uncertainty in emission rates, water areas, and season length, propagating errors via standard multiplication and summation rules.

• Emission rates: Rivers exhibited ~4× higher annual C emission rates than lakes (rivers: 0.9 ± 0.5 kg C m−2 yr−1; lakes: 0.2 ± 0.1 kg C m−2 yr−1). Ob’ main channel pCO2 averaged 1546 ± 882 µatm.
• Areas: Mapped rivers and lakes (excluding streams and ponds) cover ~5.2% of Western Siberia (~20,000 km² rivers; ~171,000 km² lakes). Including extrapolated streams and ponds increases lotic area by ~1.6× (to 33,390 km²) and lentic area by ~2.4× (to 425,986 km²), raising total water coverage to ~12% of land area.
• Total emissions (Pg C yr−1): • Excluding streams and ponds: 0.050 ± 0.007. • Including streams and ponds (preferred): 0.104 ± 0.013 (lotic 0.032 ± 0.005; lentic 0.071 ± 0.012). • Component breakdown: streams 0.013 ± 0.003; rivers 0.018 ± 0.003 (Ob’ main channel 0.004 ± 0.001; 24% of river emissions); lakes 0.032 ± 0.006; ponds 0.039 ± 0.010. • Alternative pond-area method (satellite fractions) yields 0.076 Pg C yr−1 total (≈31% lower than preferred estimate).
• Permafrost gradient: C yield (emission per land area) increases with permafrost extent, peaking in the discontinuous zone; lake contributions drive this pattern due to higher lake coverage and emission rates in permafrost-rich regions.
• Regional carbon budget context: Inland-water C emissions (0.076–0.104 Pg C yr−1) equal 35–50% of land C uptake (NEE −0.198 ± 0.009 Pg C yr−1 in 2016). Emissions exceed riverine C export to the Arctic Ocean (0.011 Pg C yr−1) by 6.8–9.0×, indicating only ~10% of laterally exported terrestrial C reaches the ocean. Inland-water emissions are 2.4–3.0× higher than Kara Sea CO2 uptake (−0.031 Pg C yr−1 in 2014). Compared to earlier global/model-based estimates, mean river pCO2, CO2 emission rates, and total regional emissions are ~3×, ~6.3×, and ~1.4–4.6× higher, respectively.

Findings demonstrate that Western Siberian inland waters are a major atmospheric C source, substantially offsetting the terrestrial carbon sink and dwarfing downstream export. Emission controls differ among systems: rivers are primarily influenced by lateral terrestrial C inputs, temperature, and water transit times, leading to higher rates in warmer, permafrost-influenced regions but lower in permafrost-free zones; lakes and ponds are driven more by in-lake processes (shallow depths, benthic mineralization of high-quality organic C from recently thawed sediments, reduced algal CO2 fixation in cold permafrost-rich areas). The flat terrain and high water coverage increase water residence times, promoting mineralization and outgassing. The strong permafrost-zone dependence of C yield underscores the sensitivity of cold, permafrost-rich landscapes. However, predicting future outgassing is complex due to dynamic changes in water surface areas (e.g., thaw pond/lake formation and drainage), multiple interacting drivers, and spatial heterogeneity, necessitating coupled land–water, interdisciplinary approaches and improved monitoring.

This study provides the first region-wide, observation-based estimate of carbon emissions from Western Siberian inland waters, showing high total emissions (0.076–0.104 Pg C yr−1) that represent 35–50% of regional land C uptake and far exceed riverine export to the Arctic Ocean. Emissions are driven by both high per-area fluxes (especially in rivers) and extensive water coverage that increases when small streams and ponds are accounted for. The results highlight that excluding inland waters leads to significant overestimation of terrestrial carbon sink strength. Future work should prioritize improved spatial and temporal coverage of emissions, robust inventories of small waterbodies (especially ponds), better characterization of floodplain/wetland inundation dynamics, and mechanistic understanding of drivers across permafrost zones to project responses to warming.

• Temporal coverage: Measurements emphasize open-water seasons and key periods (spring, summer, autumn); higher-resolution year-round data, especially for small systems, are limited.
• Spatial gaps: Despite extensive coverage, additional data are needed in the permafrost-free zone and across the full region to refine upscaling.
• Surface area dynamics: Floodplains and wetlands expand substantially during spring floods; assuming an 85% area increase for 30 days raises annual emissions by ~11%, indicating sensitivity to seasonal hydrology not fully captured.
• Small-system inventories: Areas of the smallest streams and ponds are extrapolated; true distributions are uncertain. Pond areas may not follow Pareto scaling; an alternative satellite-based method reduced total emissions by ~31% (to 0.076 Pg C yr−1), highlighting a key uncertainty.
• Model assumptions: Use of median CH4 fraction (1.19%) for rivers; assumed 15% uncertainty in key variables; permafrost-zone substitutions for missing Ob’ pCO2 in the continuous zone; seasonal pCO2 adjustments introduce uncertainty.
• Interannual variability: Comparisons with land NEE and Kara Sea uptake are from non-overlapping single years; variability may affect exact ratios.