Methane (CH4) is a powerful greenhouse gas and a central target for near-term climate mitigation. Despite its shorter atmospheric lifetime compared with carbon dioxide (CO2), CH4 exerts a strong warming effect, contributing ~0.5°C of the observed 1.15°C global surface warming since preindustrial periods (IPCC 2023). Consequently, rapid reductions in CH4 emissions are widely recognized as one of the most effective ways to slow warming in the coming decades. This urgency is reflected in global initiatives like the Global Methane Pledge, which aims for substantial emission cuts by 2030. The atmospheric CH4 budget is controlled by a balance between sources and sinks. While oxidation by hydroxyl radicals dominates global CH4 removal, aerobic upland soils constitute the second-largest CH4 sink, accounting for approximately 4% of global CH4 uptake (Kirschke et al. 2013). This sink is unique. Unlike atmospheric chemical sinks, soil CH4 uptake is embedded within terrestrial ecosystems. This means it can, in principle, respond to land management and stewardship (Song et al. 2024). For decades, upland soils have therefore been regarded as an important and manageable ally in mitigating climate change. However, the upland soil methane sink is dynamic, especially in a rapidly changing climate. This is because methanotrophic microbes are highly sensitive to multiple factors: temperature, soil moisture, substrate availability, and land-use disturbance. Shifts in temperature and precipitation regimes, together with intensifying human land use, are altering the spatial and temporal patterns of CH4 uptake in upland ecosystems (Guo et al. 2023; Wang et al. 2025). Yet the size and direction of these changes remain poorly understood. Observational soil–atmosphere CH4 exchange remains spatially uneven, and existing models differ substantially in their capacity to represent the nonlinear and interacting controls on CH4 fluxes. To address these knowledge gaps, Li et al. (2025) make a major step forward. It provides the first observation-driven, three-decade (1993–2022) assessment of the changing role and spatiotemporal dynamics of global upland soils as CH4 sinks and sources. Their study is built on an extensive compilation of in situ CH4 flux measurements from upland ecosystems worldwide, including croplands, forests, grasslands and tundra. This dataset constitutes the most systematic and comprehensive collection of upland soil CH4 observations currently available. Based on this rich data foundation, Li et al. (2025) upscaled site-level measurements to global gridded estimates with advanced machine-learning approaches, most notably the XGBoost algorithm. By integrating climatic variables, soil properties, and geographic information, their framework produces high-resolution maps of CH4 fluxes across space and time. Importantly, the authors did not rely solely on predictive performance. They combined machine learning with structural causal modeling and interpretability tools to identify dominant environmental drivers and to move beyond simple correlations toward causal understanding. The central finding is clear: the global sink capacity of upland soils for atmospheric CH4 has declined substantially over the past three decades. Grassland and cropland soils, which historically functioned as net CH4 sinks, have weakened markedly. In many regions, they have transitioned toward becoming net CH4 sources. Forest soils have also experienced a pronounced reduction in CH4 uptake capacity, while tundra uplands, long considered sources, show more complex but still consequential changes. Crucially, Li et al. (2025) demonstrate that these trends are not localized anomalies but represent broad and global-scale shifts. Climate variables, particularly changes in precipitation and temperature, explain much of the observed spatiotemporal variability, underscoring the sensitivity of upland CH4 fluxes to ongoing climate change. The significance of Li et al.'s work lies not only in documenting change, but in fundamentally revising how upland soils are represented in the global CH4 budget. Many previous assessments implicitly treated upland soil CH4 uptake as relatively stable over time (Kammann et al. 2009; Murguia-Flores et al. 2018). In contrast, Li et al. show that this sink is dynamic, vulnerable, and already eroding under contemporary environmental pressures. From a broader perspective, the findings have serious implications. The observed decline—and in some regions, reversal—of upland soil CH4 uptake indicates a progressive loss of the natural buffering capacity of terrestrial ecosystems. This makes the global CH4 budget increasingly sensitive to additional emissions. With reduced resilience, there is less room for uncertainty or delayed action. The margin for error in CH4 mitigation strategies tightens, reinforcing the urgency of rapid emission cuts in the energy, agriculture, and waste sectors. It also underscores the importance of protecting and, where possible, enhancing soil CH4 sink capacity, rather than assuming that terrestrial ecosystems will continue to provide a stable regulatory service under ongoing global change. While Li et al. (2025) delivers a robust and comprehensive analysis, it also highlights important limitations and directions for future research. First, like most large-scale modeling efforts, the model predicts the net CH4 flux which results from the offset between CH4 oxidation and methanogenesis. Thus, it cannot directly distinguish whether a weakened sink function arises from reduced oxidation or enhanced production. Although the authors draw mechanistic inferences from existing knowledge, the data themselves do not directly reveal this critical linkage, which limits process-based insight. Second, some ecosystems remain underrepresented in global datasets. Arid and semi-arid regions, for example, may contribute nonnegligibly to global CH4 uptake but are sparsely sampled (Lee et al. 2023; Song et al. 2024; Wu et al. 2024). Similarly, emerging evidence suggests that CH4 uptake by tree woody tissues may be an additional, previously overlooked sink in forest ecosystems, particularly in the tropics (Gauci et al. 2024). Incorporating these processes in future studies could further refine global CH4 budget estimates. In addition, biological and anthropogenic disturbances on CH4 emissions deserve greater attention. Recent studies demonstrate that grazing, rodent activity, and land-use legacies can strongly modify soil structure, moisture, and microbial communities, thereby influencing CH4 fluxes (Gan et al. 2025). Extreme events like droughts, floods, and heatwaves may also exert disproportionate impacts that are difficult to capture with long-term averages. Future work should aim to explicitly represent these factors and assess their interactions with climate change. Finally, the framework developed by Li et al. (2025) offers a valuable platform for exploring management options. Identifying region-specific strategies to protect or enhance upland soil CH4 sinks, whether through grazing management, cropland practices, or forest conservation, represents an important next step with clear policy relevance. Overall, Li et al. (2025) provide a timely reassessment of a critical yet often overlooked component of the global CH4 cycle. Their results demonstrate that upland soils are losing their capacity to act as a stable sink for atmospheric CH4, challenging long-held assumptions about their role in climate regulation. By integrating an unprecedented synthesis of observations with advanced analytical approaches, the study establishes a new benchmark for assessing terrestrial CH4 dynamics and highlights the necessity of protecting and, where possible, restoring soil CH4 sink function as part of future climate mitigation strategies. Jingrui Yang: writing – original draft. Xiaoyuan Yan: writing – review and editing. Longlong Xia: conceptualization, writing – original draft, writing – review and editing. This work was supported by the Progress of Strategy Priority Research Program of Chinese Academy of Sciences (XDB0630302), Distinguished Young Scholars Fund (SBK2024010366), and Carbon Peaking and Carbon Neutrality Special Fund for Science and Technology from Jiangsu Science (BM2022002). The authors declare no conflicts of interest. This article is an Invited Commentary on Li et al., https://doi.org/10.1111/gcb.70248. Data sharing is not applicable to this article as no datasets were generated or analyzed for the current article.
Yang et al. (Sun,) studied this question.