Abstract. This study provides improved methanol emission estimates on the global scale, in particular for the largest methanol source, the terrestrial biosphere, and for biomass burning. To this purpose, one complete year of spaceborne measurements of tropospheric methanol columns retrieved for the first time by the thermal infrared sensor IASI aboard the MetOp satellite are compared with distributions calculated by the IMAGESv2 global chemistry-transport model. Two model simulations are performed using a priori biogenic methanol emissions either from the new MEGANv2.1 emission model, which is fully described in this work and is based on net ecosystem flux measurements, or from a previous parameterization based on net primary production by Jacob et al. (2005). A significantly better model performance in terms of both amplitude and seasonality is achieved through the use of MEGANv2.1 in most world regions, with respect to IASI data, and to surface-and air-based methanol measurements, even though important discrepancies over several regions are still present. As a second step of this study, we combine the MEGANv2.1 and the IASI column abundances over continents in an inverse modelling scheme based on the adjoint of the IMAGESv2 model to generate an improved global methanol emission source. The global optimized source totals 187 Tg yr −1 with a contribution of 100 Tg yr −1 from plants, only slightly lower than the a priori Correspondence to: T. Stavrakou (jenny@aeronomie.be) MEGANv2.1 value of 105 Tg yr −1 . Large decreases with respect to the MEGANv2.1 biogenic source are inferred over Amazonia (up to 55 %) and Indonesia (up to 58 %), whereas more moderate reductions are recorded in the Eastern US (20-25 %) and Central Africa (25-35 %). On the other hand, the biogenic source is found to strongly increase in the arid and semi-arid regions of Central Asia (up to a factor of 5) and Western US (factor of 2), probably due to a source of methanol specific to these ecosystems which is unaccounted for in the MEGANv2.1 inventory. The most significant error reductions achieved by the optimization concern the derived biogenic emissions over the Amazon and over the Former Soviet Union. The robustness of the derived fluxes to changes in convective updraft fluxes, in methanol removal processes, and in the choice of the biogenic a priori inventory is assessed through sensitivity inversions. Detailed comparisons of the model with a number of aircraft and surface observations of methanol, as well as new methanol measurements in Europe and in the Reunion Island show that the satellite-derived methanol emissions improve significantly the agreement with the independent data, giving thus credence to the IASI dataset.