Future Forest

Deep Learning for forest disturbance detection using Sentinel-2 time series

Forest disturbance is a viral topic in times of climate change, which leaves stressed trees vulnerable to calamities such as bark beetle infestations and drought damage.

To this end, Deep Learning methods are applied pixelwise on Sentinel-2 time series to detect and anticipate forest disturbance.

The Transformer models, a state-of-the-art Deep Learning architecture, are pre-trained on published forest disturbance data and fine-tuned on forest dieback and logging that was manually delineated on aerial imagery. The goal of the work package is a near real-time forest disturbance product all over Germany for forest practitioners and scientists.

More information

Forest disturbance detection is just one objective in the FF.ai project! Find out more on https://future-forest.eu/