DeepExtremes

DEEP EXTREMES is part of the ESA AI4SCIENCE activity. The first AI4SCIENCE ITT was launched in 2021 and had a focus on Extreme Events, Multi-Hazards and Compound Events, and contributes to the ESA Extremes and Natural Disasters Science Cluster.

The ESA AI4SCIENCE ITT had 2 main objectives:

1. Advancing Earth System Science: advancing our capacity to combine EO and AI to address a major scientific challenge: The observation, understanding and characterisation of multi-hazards, compound and cascade events and its impacts on society and ecosystems.
2. Advancing Artificial Intelligence for EO: unlocking the full potential of Artificial Intelligence for Earth System Science with focus on two main AI challenges: physics-driven Artificial Intelligence and explainable AI. The DEEP EXTREMES project has a focus on compound heat and drought events at global scale, looking at detection based on long-term climate and land-surface data, combining EO archives and other observation data, with methods tailored to multivariate event detection.

The principle is to start from sampling a subset of large events in Sentinel era and zooming into the events and in unaffected areas around the event with high-dimensional “mini cubes”.

The activity then aims to train complementary deep-learning methods for prediction and understanding dynamics in such events, implement the tested and validated workflow in a cloud environment and developing it further based on community feedback. Science community engagement is planned via workshops and science discussions to further develop the proposed framework.

Published Datasets

Involved team members

Mélanie Weynants
PostDoc

mweynants@...

Vitus Benson
PhD Candidate

vbenson@...

• Interests: Beers&Coding&Sunsets

Claire Robin
PhD Candidate

crobin@...

Markus Reichstein
Director

mreichstein@...

Partners