Arctic soil methane sink increases with drier conditions and higher ecosystem respiration

The study investigates whether well-drained Arctic upland soils constitute a substantial biological sink for atmospheric methane (CH₄) and what controls its temporal variability. Although Arctic regions are often treated as net CH₄ sources due to extensive wetland emissions, the overall Arctic CH₄ budget is highly uncertain, and uplands dominate the Arctic-boreal landscape. Prior focus on wetland hotspots may bias regional budgets, overlooking atmospheric CH₄ consumption by upland soils. The authors hypothesize that CH₄ uptake in upland tundra is common and controlled by a combination of abiotic factors (especially soil moisture regulating diffusion) and biotic processes linked to ecosystem respiration and labile carbon inputs, leading to pronounced diel and seasonal dynamics.

The paper situates its work within evidence that soils are the only known biological sink for atmospheric CH₄, removing 11–49 Tg CH₄ annually. Arctic CH₄ budgets remain uncertain (8–55 Tg CH₄ yr⁻¹), partly due to sparse measurements and a focus on wetlands. Prior studies observed atmospheric CH₄ uptake in polar deserts, dwarf-shrub tundra, and forest–tundra ecotones, often linked to soil temperature and especially soil moisture which governs gas diffusion via air-filled pore space. Methodologically, high-precision, field-deployable gas analyzers and automated chambers now enable reliable detection of small fluxes with short closure times, potentially revealing overlooked temporal patterns (diel and seasonal). Biogeochemically, methanotrophs may be stimulated by labile carbon inputs (rhizodeposition) and nutrient availability, suggesting plant–soil interactions can modulate CH₄ uptake.

• Sites and design: Primary site at Trail Valley Creek (TVC), western Canadian Arctic, continuous permafrost. Installed 18 automated chambers (6 each on lichen-, shrub-, and tussock-dominated patches). Hourly CH₄ and CO₂ fluxes measured during growing seasons of 2019 (DOY 172–236) and 2021 (DOY 150–243). Additional manual chamber campaigns at Havikpak Creek, Scotty Creek (Northwest Territories, Canada), and Kilpisjärvi (Finnish Lapland) to assess spatial variability across upland land covers.
• Instrumentation: Automated chambers (transparent; half covered to be opaque) with fans and pressure equalization. Los Gatos Research Enhanced Performance GGA analyzer (1 Hz) with external pump; chamber closure 3 min. Auxiliary measurements: PAR (transparent chambers), chamber air temperature, surface soil temperature and moisture (0–12 cm at opaque chambers), meteorology (air temperature, humidity, wind, rainfall, PAR, pressure). Additional continuous profiles (2021) for soil temperature, moisture, and oxygen at 10, 20, 30 cm (one profile per vegetation type).
• Data processing: Fluxes derived mainly via linear fits for CH₄ (93%), exponential when appropriate. Quality control removed periods with pressure issues and poor mixing (low friction velocity and wind at night). After filtering, datasets comprised 44,848 CH₄ and 44,644 CO₂ flux points. Diel daily sums constructed from hourly data with short gap-filling for gaps <12 h.
• Manual chambers: Portable analyzers (Picarro G4301 or LGR Ultraportable), 5-min enclosures, linear/nonlinear fits; daytime measurements across representative upland plots (WFPS <50%).
• Soil, gas, and vegetation measurements: Collar greenness from photographs; soil pore gas CH₄/CO₂/N₂O at depths; δ¹³CH₄ and δ¹³CO₂ at 10 cm; soil physical/chemical properties (pH, bulk density, WFPS, organic matter, C and N, isotopes). Plant-available N via PRS probes across months at TVC; lab extractions for nutrient turnover, DOC and dissolved N.
• Laboratory incubations: CH₄ oxidation assays at 4 °C and 20 °C; moisture manipulation (20% vs 60% water-holding capacity) and labile C addition (glucose) to assess stimulation of oxidation in soils from Kilpisjärvi (and TVC subset). Replication noted; statistics via Dunn’s test and Welch's t-tests.
• Statistical analysis and modeling: Random forest (RF) models (ntree=500) to rank variable importance and explain variance in CH₄ fluxes at multiple temporal aggregations (hourly, daily, weekly), including models per vegetation type and per chamber. Predictors included WFPS, surface/soil temperatures, PAR (chamber and station), wind speed, air pressure, and in 2021 extended set (WFPS and O₂ by depth, VPD, RH, air temperature). Additional models included ecosystem respiration (ER) as predictor to capture biotic controls. Transfer entropy analysis assessed lagged effects (especially temperature). Manual chamber data analyzed for site/landcover differences via non-parametric tests and Welch’s t-tests.

• Upland tundra is a consistent CH₄ sink at TVC: 95% of measured fluxes indicated uptake. Mean uptake magnitudes by vegetation type: lichen −0.020 ± 0.016 mg CH₄ m⁻² h⁻¹ and shrub −0.024 ± 0.027 mg CH₄ m⁻² h⁻¹, equivalent to −0.49 ± 0.33 and −0.59 ± 0.51 mg CH₄ m⁻² d⁻¹, respectively. Tussock showed uptake 67% of the time, but seasonal means were near zero due to episodic emissions during rainy periods.
• Drier soils enhanced uptake: Late-summer uptake peaks coincided with low water-filled pore space (WFPS; lichen <35%, shrub <15%). 2021 (warmer/drier) showed stronger late-summer uptake than 2019.
• Pronounced diel and seasonal dynamics: In June, uptake peaked afternoons (15:00–16:00) aligned with PAR/temperature and ER peaks; in July diel contrasts weakened; in August the pattern reversed with nocturnal maxima (22:00–04:00) 21–50% higher than daytime minima. Restricting measurements to daytime would overestimate daily uptake by 25–37% in early summer and underestimate by 6–19% in late summer.
• Strong link to ecosystem respiration (ER): Correlations between CH₄ uptake and ER were high, especially in low-light late summer (lichen R² ≈ 0.53–0.54; shrub R² ≈ 0.76–0.81). Maximum uptake occurred at relatively low air (<8 °C) and soil temperatures (<4 °C at 10 cm), aligning with ER peaks, suggesting biotic control via labile C inputs (rhizodeposition) to methanotrophs.
• Abiotic drivers via machine learning: RF models identified WFPS as the dominant abiotic predictor of uptake across vegetation types; temperature importance increased at longer aggregation scales. Lagged temperature effects were generally weak (<4 h). In 2021, WFPS and soil oxygen ranked highest, especially for lichen; for tussock, deeper (30 cm) WFPS was important (indicative of deeper CH₄ production). Including ER in RF improved explained variance markedly in late summer for shrub by 26–45%.
• Model performance: RF explained 48–76% of flux variance for individual vegetation types, highest for lichen (70–76%), with more unexplained variance for shrub (up to 51%) in late summer, consistent with additional unmeasured biotic controls.
• Spatial generality: Across four Arctic/Sub-Arctic sites (176 manual observations), all upland land covers acted as CH₄ sinks. Uptake in Finnish Lapland was significantly stronger (mean −0.143 mg CH₄ m⁻² h⁻¹) than in Canadian Arctic sites (mean −0.041 mg CH₄ m⁻² h⁻¹). Daily medians/means: Finland −2.88 to −3.43 mg CH₄ m⁻² d⁻¹ vs Canada −0.94 to −0.98 mg CH₄ m⁻² d⁻¹. Soil gas profiles showed below-ambient CH₄ to 20 cm depth, corroborating in situ consumption.
• Laboratory evidence: Labile carbon additions (glucose) rapidly increased CH₄ oxidation rates (within 1 h) at both 4 °C and 20 °C; drier incubation (20% WHC) favored oxidation relative to wetter (60% WHC) conditions, supporting field inference of moisture and substrate controls.
• Overall magnitude: Across sites and surface types, the study reports atmospheric CH₄ consumption at rates of 0.092 ± 0.011 mg CH₄ m⁻² h⁻¹ (mean ± s.e.) and higher uptake than reported by recent syntheses that often categorize dry tundra as a source.

The results demonstrate that well-drained Arctic upland soils are widespread and significant sinks for atmospheric CH₄, with uptake strongly modulated by soil moisture and closely coupled to ecosystem respiration, indicating biotic controls via plant–microbe interactions and labile carbon supply. The observed diel and seasonal patterns—particularly the reversal to nocturnal uptake maxima in late summer—cannot be explained by temperature alone and imply that rhizosphere-driven substrate dynamics and plant phenology regulate methanotrophic activity. Because many Arctic CH₄ flux measurements have been made infrequently during daytime, such diel biases can lead to substantial over- or underestimation of daily and seasonal CH₄ budgets. Machine-learning analyses confirm WFPS as the primary abiotic constraint on uptake, while inclusion of ER markedly improves model skill in late summer, especially in shrub communities, underscoring the role of biotic processes. Geographically, consistent uptake across sites and higher rates in Finnish Lapland emphasize that upland sinks may be stronger and more variable than previously represented in models and inventories. Collectively, the findings support the hypothesis that soil drying and enhanced substrate availability amplify CH₄ uptake, offering a potential negative feedback to climate change in Arctic uplands, though with important temporal and vegetation-specific nuances.

This study provides high-frequency, ecosystem-scale evidence that Arctic upland soils commonly act as atmospheric CH₄ sinks, with strongest uptake under dry conditions and during periods of elevated ecosystem respiration. It reveals pronounced diel and seasonal dynamics tied to plant–soil biotic processes, highlights soil moisture as the dominant abiotic control, and demonstrates that including ER substantially improves predictive models. Spatially, multiple Arctic/Sub-Arctic upland systems function as CH₄ sinks, with particularly strong uptake in Finnish Lapland. These insights call for revising Arctic CH₄ budgets and for incorporating upland sinks, diel variability, soil moisture, and biotic drivers into process-based and empirical models. Future research should: expand spatial and temporal measurements across upland classes; integrate microbial community and substrate dynamics; resolve night-time processes; and assess how projected drying, shrubification, and nutrient dynamics will modulate the Arctic CH₄ sink at landscape to regional scales.

• Temporal scope largely limited to growing seasons (June–August); shoulder seasons and winter processes not captured.
• Predictive performance remained limited for shrub sites in late summer and for night-time fluxes, indicating unmeasured biotic or microclimatic controls.
• Manual chamber campaigns were sparse in time and uneven across regions, potentially limiting generalizability.
• Tussock areas exhibited episodic emissions during wet periods, introducing high temporal variability not fully resolved by available predictors.
• While ER improved models, causality among ER, labile C supply, and methanotrophy is inferred rather than directly measured in situ (substrate-specific pathways and microbial community composition not resolved).