Trees outside forests

Our research combines Machine Learning, Remote Sensing and Geography to discover unprecedented knowledge at the level of individual trees.

We use PlanetScope, Skysat, Maxar and Gaofen satellite images to produce maps on tree locations, count, cover, density, biomass, and canopy height, from local to continental scale.

Our work has been published in Nature, Nature Climate Change, Nature Sustainability, PNAS Nexus, Science Advances, Nature Plants, Nature Food, Nature Geoscience, Nature Ecology, Nature Communications, etc.

Latest

Join us as fully paid PhD candidate!
Nature Reviews: High-resolution sensors and deep learning models for tree resource monitoring.
The Tree Explorer is regularly updated (all still experimental).
We mapped about 3 million baobabs all over the Sahel, published in Nature Ecology.
New paper at ECCV 2024 including codes and dataset
New paper in Nature Sustainability also covered in "The Hindu".
New paper in Nature Food on mapping greenhouses

Research

Mapping trees outside forests

ERC starting grant (TOFDRY) with a focus on global drylands

We aim to quantify the worlds non-forest trees by using PlanetScope satellite imagery and a deep learning technique which is able to identify objects within imagery at unprecedented accuracy. See example.

Tree mapping in Africa

DFF Sapere Aude Danish Research Leader Grant in collaboration with NASA

We have a special focus on Africa to produce large scale maps at tree-level at national and continental scale. See example.

Tree biomass in European Forests

In collaboration with INRAE, Kayrros, LSCE, and others

We use deep learning, NFI data, aerial and PlanetScope images to map tree biomass in European forests. See example.

The Carbon Sink of Southern China

In collaboration with the Chinese Academy of Sciences

We investigate the impact of afforestation projects on the karst landscapes of Southern China. See example.

Global carbon stock and vegetation monitoring

Together with INRAE, LSCE, and others

We use passive microwaves (L-VOD) at coarse scale for monitoring global carbon sinks and vegetation dynamics, and what biotic and abiotic factors drive them. See example.