Tree species from space: a new product for Germany based on Sentinel-1 and -2 time series

Abstract

German forests are increasingly threatened by climate change, highlighting the need to understand tree species composition to preserve biodiversity, ecosystem resilience and climate regulation. Assessing tree species distribution is essential for effective forest management and adaptation to changing environmental conditions. The current remote-sensing-based dominant tree species products for Germany are based on reference data from the National Forest Inventory (NFI). The NFI data for Germany is not accessible to the general public, due to considerations pertaining to the safeguarding of individual privacy and the avoidance of unintended disruption to the dataset. Information on specific tree species can also be obtained from alternative sources. We collected a total of over 80 000 samples on dominant tree species across Germany by utilizing a range of databases, including city tree registers, Google Earth Pro, Google Street View, and our own on-site observations in order to generate a reference database. Spatio-temporal composites derived from Sentinel-2 (S2) and Sentinel-1 (S1) satellites, combined with a digital elevation model (DEM), were utilized to generate products showcasing the distribution of 10 specific tree species groups across Germany in 2022. This approach enabled continuous mapping of dominant tree species at a 10-m resolution. The accuracy of different machine learning algorithms was assessed using various data combinations. The combination of S2, S1, and DEM yielded the highest overall F1-Score of 0.89. S2 alone achieved results that were nearly as accurate with an F1-Score of 0.86. The optimal model (S2-S1-DEM) demonstrated an F1-Score range of 0.76 to 0.98 for the four primary tree species (pine, spruce, beech, and oak). For other common tree species, including birch, alder, larch, Douglas fir, and fir, the F1-Score ranges from 0.88 to 0.96. Here, we present a cost-effective and reproducible method for monitoring tree species in response to the rapid changes occurring in German forests.