This dataset provides an upscaled estimate of Boreal-Arctic wetland CH4 emissions at a weekly time scale from 2002 to 2021 at 0.5 by 0.5-degree spatial resolution. Ground truth data on wetland CH4 emissions from eddy covariance towers (139 site years) and chambers (168 site years) were used to train and validate a causality-guided machine learning model. The trained model was then used to estimate CH4 emissions at grid cells that have wetlands and located above 44 degrees north. The data are provided in netCDF format.