Combining TanDEM-X and Sentinel-2 for large-area species-wise prediction of forest biomass and volume

In this study, data from the satellite sensors TanDEM-X and Sentinel-2 were combined with national field inventory to predict forest above-ground biomass (AGB) stem volume (VOL) over a large area in Sweden. The sources evaluated both separately combination. study covers approximately 20,000,000 ha corresponds about 70% of Swedish area. was divided into tiles 2.5 × km2, which processed sequentially. plots inventoried on 7 m 10 circular by National Forest Inventory, plot AGB VOL at year estimated based 10-year period data. modelled using k nearest neighbor (kNN) algorithm, = 5 neighbors. use two different scene extents enabled generation seamless maps. Moreover, kNN algorithm provided per tree species, used for classification dominant species stand-level. overall accuracy 77%. predicted rasters 549 stands distributed RMSE predictions 31.4 t/ha (29.1%) AGB, 59.0 m3/ha (30.2%) VOL. By estimating removing variance due sampling (the stand values sample plots), improved 18.0 (16.6%). approach suitable variables combination sensors, sufficient reference are available. most important predictions, while essential map species.