November 25th, 2026 | 2:00 PM CEST
Live on Microsoft Teams.
On November 25th, the ESA Φ-Lab Collaborative Innovation Network will host a new Φ-talk. Details are below.
Meet the speaker
Ioannis Papoutsis is an Assistant Professor of Remote Sensing and Artificial Intelligence at the National Technical University of Athens and an Adjunct Researcher at the National Observatory of Athens and the Archimedes/Athena Research Center. His research focuses on artificial intelligence for Earth observation, with an emphasis on foundation models, multimodal representation learning, and deep learning for forecasting environmental extremes. He leads research activities at the Remote Sensing Lab, developing AI methods for large-scale satellite data analytics, natural hazard monitoring, and climate change impact assessment. He coordinates several European research projects, including PROTEUS, ThinkingEarth, MeDiTwin, and SeasFire. He has previously served as Operations Manager of the Greek ESA Collaborative Ground Segment and as Copernicus Emergency Management Service Manager for Risk and Recovery. He is a member of the ELLIS Society.
Talk abstract
Wildfire danger is shaped by processes acting on different time scales: day-to-day weather, the memory of fuel and soil moisture, and large-scale ocean-atmosphere patterns that can precondition a fire season far away. Operational systems still rely mainly on weather-driven indices such as the Fire Weather Index, which capture only part of this picture.
This talk shows how machine learning applied to Earth observation and climate data can capture these interactions. A few days ahead, models that incorporate satellite observations of vegetation and moisture outperform weather-only indices. Weeks to months ahead, predictability shifts to slow land-surface states and remote climate signals. Along the way, I discuss the methods this requires: positive-unlabeled learning and self-supervised pretraining to cope with scarce and incomplete labels, Bayesian deep learning for uncertainty quantification, and Earth system deep learning to capture interactions across space and time.
Register Here!