The dataset includes results from a global inversion of GOSAT methane observations documented in Zhang et al. (2021). The inversion is performed with the GEOS-Chem model for the period of 2010-2018. We optimize mean and trend of non-wetland emissions, monthly wetland emissions for 14 subcontinental regions, and annual hemispheric OH concentrations (methane loss rates). Sector attribution of posterior non-wetland emissions in a 4ox5o grid cell is done on the basis of the relative contribution of a sector to prior emissions in the grid cell. Detailed description of the methodology can be found in Zhang et al. (2021). Reference Zhang et al., Attribution of the accelerating increase in atmospheric methane during 2010–2018 by inverse analysis of GOSAT observations, Atmos. Chem. Phys., 21, 3643–3666, https://doi.org/10.5194/acp-21-3643-2021, 2021. Date files The dataset includes 4 files: 1. GlobalInv_v1.nc is the main data file, which includes (1) prior estimates (prior_value), (2) posterior scaling factor (post_sf), (3) prior error covariance matrix (prior_errcov), (4) posterior error covariance matrix (post_errcov), and (5) averaging kernel matrix (post_ak) for the state vector. All the values are expressed relative to prior_values, except for prior_value which is expressed in absolute magnitude. 2. State_vec.csv is a table which indexes the state vector elements, for non-wetland emission (PertA) (1-1009), non-wetland emission trends (PertT) (1010-2018), wetland emissions (PertW) (2019-3530), and hemispheric OH (PertN for Northern Hemisphere and PertS for Southern Hemisphere) (3531-3548). 3. Clusters_nonwetland.nc contains a 72x46 matrix (corresponding to GEOS-Chem 4x5 degree grid) that maps grid cells to non-wetland emission grid-cell ID. 4. Clusters_wetland.nc contains a 72x46 matrix (corresponding to GEOS-Chem 4x5 degree grid) that maps grid cells to wetland emission region ID. Derived dataset We also derive a sectoral monthly gridded methane flux dataset based on the inversion results archived here. This gridded flux dataset is deposited in a separate repository (https://doi.org/10.57760/sciencedb.02328) (in Aug 2022). The gridded dataset represents the best estimate from this inversion but includes no information about error metrics (e.g., posterior error covariance).