ESA title
December 15, 2021

ESA-ECMWF workshop: turbocharging EO and Earth System data with AI

Posted in

The Machine Learning for Earth System Observation and Prediction (ML4ESOP) workshop, an annual gathering organised by the ESA Φ-lab and ECMWF, has proven once again to be a key focal point for practitioners in the field. With over 1,100 participants from 85 countries, this virtual event gave truly global insight into the trends, strengths and weaknesses of the application of Artificial Intelligence (AI) to Earth System monitoring and predictive modelling.

In addition to more established applications such as image recognition and medical diagnosis, the benefits of Machine Learning/Deep Learning (ML/DL) techniques have started to garner interest in the last few years in the fields of Earth System Observation and Prediction (ESOP). ML and DL can improve our understanding of the Earth’s complex and wide-scale dynamics by drawing on new methodologies and enhanced computing power to process and analyse huge volumes of data automatically.

Held this year from 15 to 18 November, the ML4ESOP workshop is a platform for leading scientists to share expertise, domain experience and best practices. “The workshop centres on the synergies to be gained from the intersection of Earth observation, Earth System Prediction and ML,” explained Pierre Philippe Mathieu, Head of the Φ-lab Explore Office. “Just like last year, the 2021 edition has been an invaluable opportunity to gather together a fast-growing, multi-disciplinary community of practitioners, all of whom are exploring the power of ML to help scientists quantify the current and future state of our planet and the impact of human activity.”

ECMWF Director of Research Andy Brown also praised the event’s aims and achievements: “There are both huge opportunities and huge challenges in observation processing, data assimilation and modelling as part of ESOP, and ML will undoubtedly play a significant part in meeting those challenges. This workshop, convened jointly by ESA and ECMWF, has been a great opportunity to bring together a broad community to explore the state of the art and the road ahead, and I’ve been extremely impressed by the range and depth of material covered.”

In fact a brief glance at the event’s agenda confirms the variety of topics presented. For the first three days contributors gave talks under the four thematic areas of enhancing satellite observation, approaches to hybrid data assimilation, geophysical forecasting, and post-processing and dissemination. Applications discussed included wildfire forecasting, water resource management and food security, while from the computational perspective there were several presentations on neural networks and an intriguing contribution on autonomous robotic teams. The final day was dedicated to working groups split according to the four themes, with each group discussing current limitations and ways forward in ML4ESOP. The findings were then summarised at the closing plenary session.

The workshop was also accompanied by an e-poster side event. Attendees were treated to an expansive assortment of subject matter, with some 30 presentations from academia, research institutes and industry. The ocean and maritime monitoring was a recurring theme, and the curiously titled ‘It’s a Bird, It’s a Plane, It’s a Meteor!’ from NVIDIA and the SETI Institute drew considerable attention.

“It was a pleasure to lead the organisation of the 2021 ML4ESOP workshop as it provided me with the opportunity to connect with a large number of universities and companies worldwide, all of which share an amazing commitment to AI4EO [Artificial Intelligence for Earth Observation],” commented ESA Research Fellow Rochelle Schneider. “The workshop conveyed strong messages on the potential of replicable, scalable and sustainable ML methods, not only for ESOP topics but also for economy, policy and health applications. This has set the stage for the far-reaching benefits that ML/DL will inevitably deliver.”

The workshop content is available here.

To know more: ECMWF, ESA AI4EO, ESA ɸ-lab