top of page

Trees outside forests in global drylands (TOFDRY)

ERC starting grant

TOFDRY aims 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.

Contact: Florian Reiner



We form global mosaics at 3 m resolution from millions of PlanetScope scenes having a size of several hundred terabyte.


Deep Learning

We develop deep learning models being able to identify trees and groups of trees - at continental scale and with a precision at tree level.


Tree cover maps at unprecedented accuracy

Tree cover maps at 1-m resolution will be linked with various variables, like soil fertility, agroforestry, carbon stocks, climate, etc.

bottom of page