Process-based land surface models are important tools for estimating global wetland methane (CH4) emissions and projecting their behavior across space and time. So far there are no performance assessments of model responses to drivers at multiple time scales. In this study, we apply wavelet analysis to identify the dominant time scales contributing to model uncertainty in the frequency domain. We evaluate seven wetland models at 23 eddy covariance tower sites. Our study first characterizes site-level patterns of freshwater wetland CH4 fluxes (FCH4) at different time scales. A Monte Carlo approach has been developed to incorporate flux observation error to avoid misidentification of the time scales that dominate model error. Our results suggest that 1) significant model-observation disagreements are mainly at short- to intermediate time scales (< 15 days); 2) most of the models can capture the CH4 variability at long time scales (> 32 days) for the boreal and Arctic tundra wetland sites but have limited performance for temperate and tropical/subtropical sites; 3) model error approximates pink noise patterns, indicating that biases at short time scales (< 5 days) could contribute to persistent systematic biases on longer time scales; and 4) differences in error pattern are related to model structure (e.g. proxy of CH4 production). Our evaluation suggests the need to accurately replicate FCH4 variability in future wetland CH4 model developments.

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.