Global soil methane uptake inferred from global field measurements

Soil methane (CH4) uptake is an important sink in the global methane budget. However, both the magnitude and the dynamics of global soil CH4 uptake remain highly uncertain. Here, by compiling an extensive data set of 6899 site-months of soil CH4 fluxes observed at 682 sites, we use a machine-learning algorithm (Random Forest) to provide a data-driven estimate of global soil CH4 uptake each month from 1981 to 2018. Varying the combination of predictors, we construct 9 different simulations for three landcover types (Forests, Grasslands, Croplands). The dataset uploaded is the mean of 9 simulations after calculating the area-weighted average of CH4 flux of three landcover types.