Forest biomass disturbance monitoring with Earth Observation embeddings

During this webinar, Damien Robert (University of Zurich) will present the latest findings from the Forest Biomass Disturbance Monitoring use case.

Forests are changing faster than maps can follow, reshaped by storms, fires, pests, and harvesting. While a growing number of methods can detect forest disturbances from satellite imagery, near real‑time (NRT) forest disturbance detection and attribution remain challenging. Existing approaches are often designed for either detection or attribution, rely on a single satellite modality, or are framed as retrospective annual analysis.

Building on tools developed in the Embed2Scale project, the webinar will show how Earth‑Observation (EO) embeddings enable multimodal NRT monitoring of forest biomass disturbances from Sentinel‑1/2 time series. By working with embeddings workflows rather than raw imagery, the approach can leverage relatively few labelled examples while lowering the expertise and compute barrier to applying modern deep learning methods.

Aided by a new machine‑learning‑ready benchmark dataset, the presentation will both quantitatively and qualitatively compare established methods with embedding‑based models on real‑world case studies, and invite your feedback and expertise to shape the future of forest disturbance monitoring.

Register here