Understanding forests

Forests fulfil many essential functions in the Earth system at different scales, from storing carbon to hosting diverse ecosystems. The vast amounts of remote sensing data captured since several decades together with advances in machine learning methods have enabled to make achieve great advances in forest monitoring. However, improvements in the accuracy and interpretability of such methods need to be made so that they can benefit scientists and decision makers.We thus aim at building machine learning methods that are rooted in expert knowledge from forestry,  and take decisions that can be understood by the user. Furthermore, we aim at not only extracting information from remote sensing data, but also achieving better understanding of forest processes themselves.At the local scale, we work on mapping and understanding alpine treeline dynamics in the Swiss Alps based on historical aerial images. We also work at much larger scale, by studying the identification of drivers of deforestation globally.


  • Mapping forest in the Swiss Alps treeline ecotone with explainable deep learning, Remote Sensing of Environment, 2022 (paper, code)