Methane (CH₄) contributes ~20% to post-industrial climate warming due to its greenhouse gas effects. Among all methane sources, wetlands are the single largest and climate-sensitive natural source. Estimating wetland methane emissions involves reconciling top-down inversion and bottom-up process-based models. However, these two model types are dependent and exhibit large disparities. To better understand wetland methane emissions and refine the process-based and inversion models, we need independent high-resolution and long-term wetland methane flux data. Here, we develop a high-spatial-resolution (1 km × 1 km) monthly wetland CH₄ flux dataset for the Southeastern (SE) United States (US) from 1982 to 2010 using a data-driven random forest (RF) approach. We utilize CH₄ flux measurements from four FLUXNET-CH4 wetland sites to develop the RF regression model along with 11 environmental variables. Wetland CH₄ fluxes estimated using the model fit well with the measured CH₄ fluxes (R² = 0.91) from four representative FLUXNET-CH4 wetland sites across the SE US. Leveraging the developed RF model and wetland distribution data, we map the spatial distribution of CH₄ emissions in the study region. Our mapping reveals large spatial variability in CH₄ emissions, ranging from 0 to 266.0 nmolCH₄ m^-2 s^-1, with the coastal wetland areas, the Mississippi Delta, and the Everglades being the predominant sources of CH₄. Our dataset demonstrates good agreement with the remote sensing-derived wetland CH₄ fluxes from the Carbon Monitoring System Methane Flux for North America product, confirming the credibility of our wetland CH₄ flux estimations. Variable importance analysis highlights that air temperature and the Palmer Drought Severity Index are key environmental predictors. This first-ever high-spatial-resolution (1 km × 1 km) and long-term (1982-2010) monthly gridded regional wetland CH₄ flux product over the SE US provide a benchmark and an added constraint for future wetland CH₄ flux modeling and upscaling studies in the study region.

wetland,methane emissions,machine learning,FLUXNET