Development of a Spectral, Structural and Site Specific Feature Data Base for Mapping of Tree Species in the Bavarian Forest National Park

Abstract

The Bavarian Forest National Park, established in 1970, is a unique area of forests with large non-intervention zones, which promote a large scale rewilding process with low human interference. Thus, there is a high interest of the National Park authority to investigate the structure and dynamics of the forest ecosystems within the park. However, conventional forest inventories are time-consuming and not able to fully record the heterogeneity of natural forests. Our goal is to develop advanced techniques for species mapping based on hyperspectral remote sensing in combination with other remote sensing and in-situ measurements that meet the demands of the National Park and emphasize the ecological value of a forest. These advanced techniques need to be adapted to the heterogeneous appearance of the forest comprising larger standing and lying deadwoods as well as lack of larger pure species stands. Previous studies showed that the combination of hyperspectral and LiDAR data can be very effective and accurate for tree species classification. However, the quality of classification is limited by different factors, e.g. high intraspecies spectral variability and interspecies spectral similarity. Objective of this study is to develop a feature data base for improved tree species mapping using image based spectral information, LiDAR based structural information as well as botanical information. This work aims at building a model transferable to an area wide mapping of tree species based on the needs of the Bavarian Forest National Park. It reveals the requirements for tree species mapping and shows which spectral/spatial features and data combinations generate the best results within a Random Forest modelling approach.