A vertical vegetation structure model of the Earth
Abstract
Knowing the global distribution of vegetation structure is an essential prerequisite to better quantify, model, and manage carbon, water, energy, and biodiversity in the Earth system. Despite its recognized importance, vertical structure measurements remain sparse at global scales, limiting our understanding of the links between structure and ecosystem functions and resilience. Here we present global vertical vegetation structure data for the year 2020 at 10 m spatial resolution. A deep learning approach was developed to combine satellite images from Sentinel-2 with 178 million sparse height profile measurements from the GEDI spaceborne LiDAR mission. While prior work has focused on modeling canopy top height or other scalar descriptors of structure, we provide a detailed global representation of the full vertical vegetation structure. Not only does this improve canopy top height estimates by up to 37% compared to commonly used top height proxies, but more importantly, the vertical structure estimates reveal conservation-relevant differences in canopy organization, lower-canopy structure, and canopy closure that are otherwise compressed or missed by top height and scalar complexity products. For example, we find that our full profile data improve the identification of natural forests by 42% relative to using canopy top height alone, likely being a result of the improved sensitivity to fine-scale structural signatures associated with anthropogenic disturbance, including structural changes caused by forest degradation or selective logging. By revealing subtle signatures associated with forest management and potential degradation, the presented vertical structure model provides a foundation for implementing global conservation initiatives and advancing early warning systems in ecosystem monitoring.
Related
- CLIVeg: Contrastive Image-Vegetation Alignment for Botany-Aware Earth Observation Representations
- Fixed-ground multisensor monitoring resolves longitudinal vegetation condition and fuel-state structure in Southern California chaparral and oak scenes
- Improved permafrost soil representation in a dynamic global vegetation model enhances simulations of carbon dynamics and vegetation structure in boreal forests of Northeast China
- Explainable GeoAI for urban wildfire structure loss: vegetation and built environment in the 2025 Palisades Fire
- Extreme flooding and vegetation fragmentation: post-event changes in landscape structure in southern Brazil
BibTeX
@misc{zhang2026vertical,
title = {{A vertical vegetation structure model of the Earth}},
author = {Hui Zhang and Nico Lang and Mikolaj Mazurczyk and Stefan Oehmcke and Ke Huang and Martin Brandt and Rasmus Fensholt and Ankit Kariryaa and Christian Igel},
howpublished = {EarthArXiv},
year = {2026},
month = oct,
doi = {10.31223/x5m22f},
url = {https://eartharxiv.org/repository/view/15380/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - A vertical vegetation structure model of the Earth
AU - Hui Zhang
AU - Nico Lang
AU - Mikolaj Mazurczyk
AU - Stefan Oehmcke
AU - Ke Huang
AU - Martin Brandt
AU - Rasmus Fensholt
AU - Ankit Kariryaa
AU - Christian Igel
PY - 2026
DA - 2026/10/06
PB - EarthArXiv
DO - 10.31223/x5m22f
UR - https://eartharxiv.org/repository/view/15380/
AB - Knowing the global distribution of vegetation structure is an essential prerequisite to better quantify, model, and manage carbon, water, energy, and biodiversity in the Earth system. Despite its recognized importance, vertical structure measurements remain sparse at global scales, limiting our understanding of the links between structure and ecosystem functions and resilience. Here we present global vertical vegetation structure data for the year 2020 at 10 m spatial resolution. A deep learning approach was developed to combine satellite images from Sentinel-2 with 178 million sparse height profile measurements from the GEDI spaceborne LiDAR mission. While prior work has focused on modeling canopy top height or other scalar descriptors of structure, we provide a detailed global representation of the full vertical vegetation structure. Not only does this improve canopy top height estimates by up to 37% compared to commonly used top height proxies, but more importantly, the vertical structure estimates reveal conservation-relevant differences in canopy organization, lower-canopy structure, and canopy closure that are otherwise compressed or missed by top height and scalar complexity products. For example, we find that our full profile data improve the identification of natural forests by 42% relative to using canopy top height alone, likely being a result of the improved sensitivity to fine-scale structural signatures associated with anthropogenic disturbance, including structural changes caused by forest degradation or selective logging. By revealing subtle signatures associated with forest management and potential degradation, the presented vertical structure model provides a foundation for implementing global conservation initiatives and advancing early warning systems in ecosystem monitoring.
ER -