The extraction of tree species information of forest stands using satellite images

Two different estimation methods to estimate the tree species proportions of forest stands were compared using ERS-1/2 and Landsat TM images as test data. Overall, the estimation method based on the pixelwise classification performed better than neural network method because the differences between the results acquired with training and test sets were smaller. The best results were got using dataset computed from ERS intensity (texture information) and coherence images. The results which were got using TM-image taken during summer 1994 were little bit worse