Arctic greenhouse gas fluxes are key for climate feedback but the Arctic greenhouse gas balance is poorly constrained due to a limited understanding of the spatial variation in these fluxes. This study combines extensive chamber-based flux measurements and remote sensing data to develop a machine-learning model to predict greenhouse gas fluxes across a tundra landscape in Finland. The analysis revealed that the system was a net greenhouse gas sink and showed widespread CH4 uptake in upland vegetation types, almost surpassing the high wetland CH4 emissions at the landscape scale.