Hyperspectral vs. multispectral data for tree species classification – first experiences with EnMAP data

Kurzfassung

Tree species classification using Earth observation (EO) data is crucial for biodiversity conservation and it supports sustainable forest management. Overall, accurate classification helps maintain ecosystem health and resilience, ensuring forests' ecological, economic, and social benefits are preserved. Large-scale tree species maps are feasible only through the use of EO data, as they offer comprehensive, consistent, and repeatable coverage over extensive and often inaccessible regions. Furthermore, remote sensing technologies provide detailed spectral information crucial for distinguishing between different species across large landscapes, making them indispensable for accurate and efficient mapping. However, different tree species often have similar spectral signatures, making it difficult to distinguish between them. This is especially true for species that have similar leaf structures and pigments. Next to the spectral information also the spatial resolution and the temporal availability are crucial. Optimal EO data would be fine enough to capture individual trees or small groups of trees and would have a high revisit rate to detect changes in the spectral properties over time due to seasonal variations, phenological stages, and environmental stressors. However, tree species can be even more robustly identified if we classify/identify them in the biophysical feature space and not in the spectral-temporal feature space.