Predicting forest structural complexity in Europe through an integration of radar, optical data and machine learning

Forests stand as vital components of the Earth's biosphere, comprising a significant fraction of the world's terrestrial biomes. The management of forest ecosystems is pivotal in addressing environmental challenges, including the development of climate mitigation strategies. The three-dimensional architecture of forest ecosystems, defined by canopy height, height heterogeneity, and horizontal canopy distribution, is known to be a major driver of ecosystem processes. Thus, quantifying structural heterogeneity of forest ecosystems is fundamental for predicting their resilience and ability to moderate environmental fluctuations. Historically, comprehensive data on forest structure at a macro scales have been scarce. However, advancements in spaceborne Light Detection and Ranging (LiDAR), particularly through the Global Ecosystem Dynamics Investigation (GEDI) mission, have revolutionized our capacity to monitor forest structure. In this study, we integrated various earth observation datasets, including Synthetic Aperture Radar (SAR), along with optical imagery, within a machine learning framework to predict structural complexity. We constructed a forest structural complexity dataset encompassing Europe, including eight structural metrics that characterize the three-dimensional nature of forests. The metrics encapsulate the variability, dispersion and asymmetry in vertical stratification, the dispersion and volume of the canopy in the horizontal plane. Our findings elucidate the multifaceted nature of the structural complexity forest ecosystems. Furthermore we provide a prognostic framework for monitoring changes in this key ecosystem property. By providing a comprehensive picture of forest structural complexity across Europe, our study offers tangible support for the development of effective forest management strategies and climate change mitigation plans.