A dataset on the structural diversity of European forests

Abstract. Forest structural diversity, defined as the heterogeneity of canopy structural elements in space, is an important axis of functional diversity and is central to understanding the relationship between canopy structure, biodiversity, and ecosystem functioning. Despite the recognised importance of forest structural diversity, the development of specific data products has been hindered by the challenges associated with collecting information on forest structure over large spatial scales. However, the advent of novel spaceborne LiDAR sensors like the Global Ecosystem Dynamics Investigation (GEDI) is now revolutionising the assessment of forest structural diversity by providing high-quality information on forest structural parameters with a quasi-global coverage. Whilst the availability of GEDI data and the computational capacity to handle large datasets have opened up new opportunities for mapping structural diversity, GEDI only collects sparse measurements of vegetation structure. Continuous information of forest structural diversity over large spatial domains may be needed for a variety of applications. The aim of this study was to create wall-to-wall maps of canopy structural diversity in European forests using a predictive modelling framework based on machine learning. We leverage multispectral and Synthetic Aperture Radar (SAR) data to create a series of input features that were related to eight different structural diversity metrics, calculated using GEDI. The models proved to be robust, indicating that active radar and passive optical data can effectively be used to predict structural diversity. Our dataset finds applications in a range of disciplines, including ecology, hydrology, and climate science. As our models can be regularly rerun as new images become available, it can be used to monitor the impacts of climate change and land use management on forest structural diversity.