Recent research indicates that the global soil methane (CH₄) sink—primarily aerobic upland soils consuming atmospheric methane via methanotrophic bacteria—is larger than many prior bottom-up estimates suggested, at roughly 39 Tg CH₄ per year (2003–2018).

Soils are the main known biological sink for atmospheric CH₄ (after the dominant chemical sink from OH radicals). Prior bottom-up estimates often fell in ranges such as ~11–49 Tg yr⁻¹ or centered near ~30–32 Tg yr⁻¹; the higher figure better matches atmospheric constraints.

Cold grasslands and deserts contribute nearly 30% of the total SMU (previously under-appreciated), while disturbed agricultural soils show the lowest uptake.

Future global SMU is projected to respond mainly to temperature and rising atmospheric CH₄ concentrations; local fluxes are more sensitive to soil moisture.

Related work (e.g., incorporating soil organic carbon effects in forest soils) has also revised upward specific components of the sink.

Overall, the evidence supports a stronger natural soil CH₄ sink than many earlier bottom-up inventories assumed, with implications for refining the global methane budget and climate feedback.

Methane-oxidizing microbes (methanotrophs) in soils form the primary biological sink for atmospheric methane (CH₄), and multiple studies indicate this sink is larger or more dynamic than earlier bottom-up estimates often suggested.

Upland (aerobic) soils remove roughly 30–40 Tg CH₄ per year (with recent data-driven upscaling around ~39 Tg yr⁻¹ for periods such as 2003–2018—about 30% higher than many prior bottom-up figures and more consistent with top-down atmospheric constraints).

This represents ~5% of the total atmospheric CH₄ sink (dominated by chemical oxidation via OH radicals). High-affinity methanotrophs can consume dilute atmospheric methane even in dry mineral soils.

These microbes occur across diverse environments, from Arctic/permafrost soils and tundra to forests, grasslands, deserts, and agricultural lands. In drier Arctic sites, methanotrophs (including atmospheric methane consumers) can outcompete methane-producing methanogens, potentially turning some soils into net sinks.

Deserts and cold grasslands contribute substantially (nearly 30% of the global soil sink in some estimates) and have been under-appreciated.

Activity depends on soil moisture (drier conditions often enhance uptake by improving gas diffusion and oxygen availability), temperature, organic carbon, nitrogen, and land use.

Some analyses show the sink capacity has weakened over recent decades in certain biomes (e.g., grasslands/croplands shifting toward sources), while others project increases under warmer/drier scenarios or rising atmospheric CH₄. Hybrid modeling (process-based, machine learning and inversions) is refining both magnitude and trends.

Methane has a 100-year global warming potential of about 27–30 times that of CO₂ (IPCC values; short-term potency is higher). Enhancing or protecting the soil sink (e.g., via reduced disturbance in agriculture or managing moisture) offers a natural mitigation pathway, though it is modest relative to anthropogenic emissions.

Soil microbial diversity is still being mapped (via metagenomes, marker genes like pmoA, and flux measurements), revealing more about these “methane-munching” communities and their potential under climate change.

_____________________________________________________________________________________

Revising the Magnitude and Trends of the Global Methane Soil Sink with Process-Based, Machine-Learning, and Atmospheric Inversion Modeling Approaches

The title refers to research revising estimates of the global soil methane (CH₄) sink (primarily microbial oxidation in aerobic upland soils) by integrating process-based models, knowledge-guided machine learning (KGML), and atmospheric inversion approaches.

This aligns closely with ongoing work by groups including the AI for Natural Methane (AI4NM) working group (e.g., Youmi Oh, Chris Smith, Licheng Liu, and collaborators from NOAA/CIRES, universities, and labs).

Related abstracts and presentations (AGU, NOAA GML meetings) describe hybrid frameworks that combine these methods to better quantify the magnitude, spatial patterns, and long-term trends of the soil CH₄ sink.

Key context and findings from related studies

Background uncertainty:

The soil sink is the second- largest (and only major biological) sink in the global CH₄ budget after atmospheric OH oxidation. Prior bottom-up estimates centered around ~30 Tg CH₄ yr⁻¹ (range often 11–49 Tg or wider, up to >100 Tg in some cases), with large uncertainties from incomplete microbial process representation, parameterizations, sparse observations, and mismatches with top-down (inversion) constraints. Trends over recent decades have also been uncertain.

Hybrid/KGML approach:

Process-based models (e.g., reaction-transport or biogeochemical schemes capturing diffusion, temperature/moisture sensitivity, and methanotrophy) provide physical/scientific foundations and synthetic training data.

Machine-learning components (often hierarchical submodules for thermal, hydrological, and biogeochemical processes) improve flexibility and fit to multi-source observations (FLUXNET-CH₄, chamber fluxes, soil temperature/moisture).

Knowledge-guided constraints (e.g., embedding known responses to CH₄ substrate, temperature, and moisture in loss functions) enhance interpretability and out-of-sample performance compared to pure process-based or black-box ML models. Atmospheric inversions supply independent top-down constraints for validation or joint optimization.

Revised magnitude and trends:

Results from these efforts (and closely related upscaling) point to a higher global soil sink than many prior bottom-up estimates—consistent with other recent data-driven work reporting ~39 Tg CH₄ yr⁻¹ (2003–2018, ~30% higher than typical earlier bottom-up figures and better aligned with top-down assessments).

Preliminary process-based ranges in related presentations showed wide uncertainty (e.g., 30–90 Tg yr⁻¹ depending on parameters and microbial assumptions).

Warmer/drier conditions are suggested to enhance uptake in some projections; cold grasslands, deserts, and other biomes receive greater attention as contributors. Long-term products (e.g., daily/4-km resolution from the 1980s–2020s) are under development.

Related complementary findings:

Independent machine-learning upscaling of field data (1993–2022) has indicated a generally reduced sink capacity of upland soils over three decades in some analyses (e.g., shifts in grasslands/croplands toward sources, reduced forest uptake), driven largely by precipitation and temperature changes—highlighting the need for multi-method reconciliation of both magnitude and trends.

These hybrid approaches aim to reduce biases in the global CH₄ budget, improve understanding of microbial/environmental controls, and support better future projections under climate change.

Exact published results matching the precise title may be in press, recently presented, or emerging from the AI4NM/KGML efforts; the core theme is actively advancing the field beyond traditional single-method estimates.

Published: Journal of Geophysical Research: Biogeosciences (2026)

DOI: 10.1029/2025jg009668

Provided: American Geophysical Union

Authors: Youmi Oh, Licheng Liu, Jaehyun Lee, Lori Bruhwiler, Xin Lan, Sylvia Michel, Sourish Basu, John B. Miller, Qing Zhu, Sparkle Malone, Gavin McNicol, Qianlai Zhuang

Abstract

Methane (CH₄) oxidation by microbes is the largest biological sink of global methane, yet its magnitude and long-term variability remain uncertain. Here, we combined process-based (PB), machine-learning (ML), and atmospheric inversion approaches to evaluate the global methane soil sinks and its implication for the atmospheric CH₄ budget in this study. Both PB and ML approaches estimated annual global methane soil sink to be 40–45 Tg CH₄ yr⁻¹, 30%–50% larger than conventional estimates. Although the two approaches agreed on total magnitude, they differ in their representation of variability. PB models simulate stronger spatial heterogeneity, seasonality, and long-term increases in CH₄ uptake because environmental sensitivities are explicitly represented through mechanistic equations. In contrast, ML models reproduce site-level observations more closely but exhibit muted spatial and temporal variability due to their limited environmental sensitivities from sparse and discrete observations used for model training. Atmospheric inversions further indicate that incorporating the larger soil sink improves agreement with observed atmospheric CH₄ and its stable carbon isotope changes and requires larger microbial CH₄ emissions. Together, these results suggest that the global methane soil sink may have been underestimated and demonstrate the value of integrating PB and ML modeling, and atmospheric constraints to improve understanding of global methane cycling.