From Space to Species: Monitoring Forest Tree Species with Sentinel-2 Satellites

Forests are vital for global biodiversity, climate regulation, and ecosystem functioning, yet monitoring their dynamics at the species level remains challenging. This dissertation develops and evaluates remote sensing approaches for deriving tree species information from Sentinel-2 satellite imagery, with a focus on spatially and temporally continuous forest monitoring across Switzerland. A central contribution lies in advancing the transition from static mapping to dynamic monitoring, shifting the focus from categorical species distributions toward continuous, process-oriented indicators of forest function and change. This is made possible by an innovative Sentinel-2 processing framework that ensures spectrally consistent, temporally continuous, and geometrically stable time series, suited for mountainous landscapes. Bridging the gap between satellite observations and in situ measurements, the research advances species-level forest monitoring. The thesis consists of five articles covering different aspects of forest monitoring. First, it demonstrates that Sentinel-2 vegetation index time series can capture intraspecific variation related to structure, pigment, and water content, with variation dominated by spatial patterns in evergreens and temporal dynamics in deciduous species. Second, it shows that phenology metrics derived from Sentinel-2 reliably reflect interannual and regional patterns of leaf unfolding and peak greenness in Fagus sylvatica and Picea abies, complementing but not replacing invaluable in situ observations. Third, it evaluates the relationship between vegetation indices and stem-derived tree water deficit, finding that remote sensing detects sustained drought stress, particularly with phenology-aware processing and temporal aggregation. Fourth, it develops a framework to map early leaf discoloration in Swiss European beech forests, revealing spatiotemporally heterogeneous but drought-driven patterns between 2017 and 2023, with hotspots concentrated in northwestern Switzerland. Fifth, it explores intra- and interannual dynamics of functional diversity derived from vegetation indices, showing that Sentinel-2 index time series vary across a season and lead to differences in assessing functional divergence and richness. Taken together, these studies demonstrate that Sentinel-2 time series can provide ecologically meaningful, scalable, and transferable indicators of tree species function and health. By linking in situ with continuous, large-scale satellite observations, this work advances operational pathways for nationwide forest monitoring, enabling early detection of stress events, improved tracking of functional diversity, and better-informed forest management under climate change. The findings highlight both opportunities and limitations of Sentinel-2 time series, while pointing toward future integration of high-quality in situ data, multisensor approaches, and artificial intelligence to build dynamic, adaptive monitoring systems for forest monitoring. Ultimately, this thesis offers stepping stones toward an adaptive, science-based monitoring of forest tree species to inform research, policy, and practice.