Hysteretic temperature sensitivity in wetland CH4 emission modeling

Wetlands account for nearly one-third of total global methane (CH₄) emissions. Recent studies found that the change in substrate availability and microbial activities can cause positive hysteretic temperature sensitivity of wetland CH₄ emissions, which could serve as a critical metric for model evaluation in terms of the timing of peak emissions and their underlying mechanisms. Here, we quantified the CH₄ hysteresis across 25 eddy-covariance sites and compared it with those from 42 models, including 13 biogeochemical models, 22 atmospheric inversion models, and 7 machine learning models. We found 21 out of 25 sites exhibited positive hysteresis of wetland CH₄ emissions in temperate and arctic climate zones. The machine learning models showed a widespread negligible hysteresis pattern. While biogeochemistry and atmospheric inversion models displayed a prevalent positive hysteresis, the atmospheric inversions guided by observed CH₄ concentrations estimated a larger positive hysteresis than that of biogeochemistry models. Overall, most of CH₄ models underestimated temperature hysteresis to different extents compared with the observations at 25 eddy covariance sites. Our research highlights the necessity of constraining the hysteretic temperature sensitivity of wetland CH₄ emissions in process-based biogeochemistry and machine learning models, which will provide more reliable prior information for atmospheric transport and inversion modeling.