The forests of the Russian Taiga can be described as an enormous biomass and carbon reservoir. Therefore, they are of utmost importance for the global carbon cycle. Large-area forest inventories in these mostly remote regions are associated with logistical problems and high financial efforts. Remotely-sensed data from satellite platforms may have the capability to provide such huge amounts of information. This study presents an application-oriented approach to derive aboveground growing stock volume (GSV) maps using the annual large-area L-band backscatter mosaics provided by the Japan Aerospace Exploration Agency (JAXA). Furthermore, a multi-temporal map has been created to improve GSV estimation accuracy. Based on information from Russian forest inventory data, the maps were generated using the machine learning algorithm, RandomForest. The results showed the high potential of this method for an operational, large-scale and high-resolution biomass estimation over boreal forests. An RMSE from 55.2 to 63.3 m3/ha could be obtained for the annual maps. Using the multi-temporal approach, the error could be slightly reduced to 54.4 m3/ha.