Pilot-project tree species classification for canton Aargau
For the two very diverse and high structured study areas Bremgarten and Lägern up to 9 different tree species were classified using multinomial logistic regression models based on digital aerial images from Ultracam X and ADS80 and field samples. The obtained overall accuracies are between 75-80% for nine tree species revealing that digital aerial images are very appropriate since they are regularly used to cover entire Switzerland. The focus was laid on operationality, data handling, amount of reference data needed and the detection of the main tree species. Problems occurred for deciduous tree species due to few training samples (maple and ash in both study areas), spectral heterogeneity (beech in Bremgarten) vitality (larch in Bremgarten) and small crowns (maple in both study areas). The tree species classification maps will serve as potential information for up-dating forest stand maps and distribution of tree species.
Contact
| Keywords: |
digital aerial images, ADS80, tree species, large-area, forest parameters |