Remotely sensed 3D ecosystem structure to explain biodiversity distribution

Abstract

Ecosystem structure is a core Essential Biodiversity Variable (EBV), yet scalable, interpretable indicators have been slow to enter biodiversity models. NASA's Global Ecosystem Dynamics Investigation (GEDI) mission provides freely available waveform LiDAR measurements of forest vertical structure at ∼25 m footprints worldwide, creating a practical pathway to bring structure into biodiversity studies. In this study, we explore and evaluate how remotely sensed 3D ecosystem-structure indicators explain Mediterranean plant distributions and community patterns, complementing macroclimate. Over Mediterranean-type communities in northern Iran, we assembled families of structural indicators (e.g., Plant Area Volume Density, PAVD; Foliage Height Diversity, FHD; Plant Area Index, PAI), along with a set of new 3D voxel-based indicators that integrate the horizontal configuration of multiple footprints with their vertical waveforms, benchmarked them against climate variables using an ensemble of advanced machine learning algorithms for species distribution modeling (SDMs). GEDI-derived structure improved models beyond climate baseline; voxel-based 3D indicators were frequently among the top contributors across species, alongside PAVD and FHD, whereas relative-height summaries were comparatively uninformative in this weakly stratified system. In stacked distribution maps, GEDI delineated discrete community patches, while climate identified broader, more continuous climatic surfaces; the combined configuration performed best, pairing climatic capacity with realized, patch-level structure. Framed as EBV-style indicators, GEDI-based metrics—especially voxel 3D dispersion together with density and vertical heterogeneity—offer interpretable, scalable inputs for biodiversity assessment and conservation planning in heterogeneous, patch-forming ecosystems. We recommend deploying families of structure indicators alongside macroclimate, with spatially aware validation, to improve prediction and ecological interpretability.