Estimation of forest biomass using spaceborne SAR data is an active field of investigation. C-band backscatter is known to provide little information on biomass because of the weak sensitivity to forest biophysical properties. However, if a large multi-temporal dataset is used, it should be possible to improve the estimation. This assumption has been tested with ENVISAT ASAR Wide Swath (WS) data over a 400.000 km 2 large region in Central Siberia. Forest biomass, expressed as growing stock volume, has been retrieved using an approach based on a Water-Cloud like model and a multi-temporal combination of estimates. To avoid any dependence of the modelling procedure upon in situ training data, a novel training approach based on the information content of the MODIS Vegetation Continuous Fields tree canopy cover product has been developed. Biomass has been estimated at 1 km resolution to be comparable both to the inventory data and the coarse-scale biomass simulations by the LPJ Dynamic Vegetation Model. The patterns of biomass estimated from ASAR WS and from the inventory compare well. This result is far beyond initial expectations. Compared to LPJ-based simulated biomass, the WS-based biomass presents a much higher degree of detail making comparison difficult. ENVISAT ASAR WS data appear as a potential candidate for estimation of large-scale forest biomass for a wide range of applications.