Abstract Wetlands are responsible for 20%–31% of global methane (CH 4 ) emissions and account for a large source of uncertainty in the global CH 4 budget. Data‐driven upscaling of CH 4 fluxes from eddy covariance measurements can provide new and independent bottom‐up estimates of wetland CH 4 emissions. Here, we develop a six‐predictor random forest upscaling model (UpCH4), trained on 119 site‐years of eddy covariance CH 4 flux data from 43 freshwater wetland sites in the FLUXNET‐CH4 Community Product. Network patterns in site‐level annual means and mean seasonal cycles of CH 4 fluxes were reproduced accurately in tundra, boreal, and temperate regions (Nash‐Sutcliffe Efficiency ∼0.52–0.63 and 0.53). UpCH4 estimated annual global wetland CH 4 emissions of 146 ± 43 TgCH 4 y −1 for 2001–2018 which agrees closely with current bottom‐up land surface models (102–181 TgCH 4 y −1 ) and overlaps with top‐down atmospheric inversion models (155–200 TgCH 4 y −1 ). However, UpCH4 diverged from both types of models in the spatial pattern and seasonal dynamics of tropical wetland emissions. We conclude that upscaling of eddy covariance CH 4 fluxes has the potential to produce realistic extra‐tropical wetland CH 4 emissions estimates which will improve with more flux data. To reduce uncertainty in upscaled estimates, researchers could prioritize new wetland flux sites along humid‐to‐arid tropical climate gradients, from major rainforest basins (Congo, Amazon, and SE Asia), into monsoon (Bangladesh and India) and savannah regions (African Sahel) and be paired with improved knowledge of wetland extent seasonal dynamics in these regions. The monthly wetland methane products gridded at 0.25° from UpCH4 are available via ORNL DAAC ( https://doi.org/10.3334/ORNLDAAC/2253 ).

Wetland CH4 emissions are among the most uncertain components of the global CH4 budget. The complex nature of wetland CH4 processes makes it challenging to identify causal relationships for improving our understanding and predictability of CH4 emissions. In this study, we used the flux measurements of CH4 from eddy covariance towers (30 sites from 4 wetlands types: bog, fen, marsh, and wet tundra) to construct a causality-constrained machine learning (ML) framework to explain the regulative factors and to capture CH4 emissions at sub-seasonal scale. We found that soil temperature is the dominant factor for CH4 emissions in all studied wetland types. Ecosystem respiration (CO2) and gross primary productivity exert controls at bog, fen, and marsh sites with lagged responses of days to weeks. Integrating these asynchronous environmental and biological causal relationships in predictive models significantly improved model performance. More importantly, modeled CH4 emissions differed by up to a factor of 4 under a +1°C warming scenario when causality constraints were considered. These results highlight the significant role of causality in modeling wetland CH4 emissions especially under future warming conditions, while traditional data-driven ML models may reproduce observations for the wrong reasons. Our proposed causality-guided model could benefit predictive modeling, large-scale upscaling, data gap-filling, and surrogate modeling of wetland CH4 emissions within earth system land models.