AI4PEX
@ai4pex.bsky.social
We are a consortium of 19 institutions from all over Europe involved in the Horizon Europe funded project: Artificial Intelligence and Machine Learning for Enhanced Representation of Processes and Extremes in Earth System Models. https://ai4pex.org
💡 New paper! How does 🔥 wildfire risk in Southern Europe change by 2100? Using ML + climate + future land-use & population scenarios, wildfire risk increases strongly under high-emission SSP3-7.0. Future land use and population changes significantly shape the projected risk. 👉 buff.ly/rFeObPM
📢 New preprint! Introducing DROP 💧 a physically informed deep learning framework for river discharge prediction. By combining AI with river routing, DROP delivers: ✅ Strong spatial generalization ✅ Better flood peak prediction ✅ Interpretable, spatially explicit diagnostics 👉 buff.ly/VxZ2uQf
💡 New Paper! Why do observed and simulated winter precipitation trends often disagree? Our study shows that models capture the warming-driven response well, but often underestimate changes in atmospheric circulation, especially in the Mediterranean. 👉 buff.ly/6B2t2Rl
💡 New Paper! Can AI improve how Earth System Models represent clouds? ☁️ We present a hybrid #ML radiation scheme that learns radiative impact of subgrid-scale clouds from 5 km storm-resolving simulations. ✅ 4–10× lower errors than conventional radiation schemes. 👉 buff.ly/nxo8Wdm
⏰ Reminder: #AI4PEX Science Talks take place this Thursday! 🌍 Hadi Shokati AI for soil erosion, flood segmentation & rainfall erosivity forecasting 🧠 René Geist SoftJAX & SoftTorch: differentiable programming for scientific ML 📅 Thursday, 25 June 2026, 13:00 CEST 👉 Register now! buff.ly/ISWKn3W
💡 New Paper! ⛵ Can sailboats improve estimates of the ocean CO₂ sink? Citizen-science measurements help reveal a stronger Southern Ocean sink, but current sampling is still too sparse to fully constrain long-term trends. 👉 buff.ly/FfqGxiZ #OceanCarbon #CitizenScience #ClimateScience
🎓 Next up in the #AI4PEX Science Talks! 🌍 Hadi Shokati AI for soil erosion, flood segmentation & rainfall erosivity forecasting 🧠 René Geist SoftJAX & SoftTorch: differentiable programming for scientific ML 📅 Thursday, 25. June 2026 at 13:00 CEST 👉 Register now! buff.ly/il9q9XX
🎥 Looking back at the ELLIS Winter School 2026 in Athens! A week full of AI, climate science, hands-on projects, inspiring discussions & new collaborations. 🌍🤖 Watch the full highlight video: buff.ly/UWRH9cF Special thanks to Olyvon for producing this amazing video.
💡 New Paper! 🌡️ How much of recent summer warming is circulation-driven? Our new study shows that atmospheric circulation trends explain up to 50% of Europe’s summer warming (1979–2023) — with strong but contrasting effects across the northern mid-latitudes. Dynamics matter. 👉https://buff.ly/4jCxbSQ
💡 New Paper on global plant functional diversity 🌍🌿 Using remote sensing, biodiversity observations & trait databases, this study maps key plant traits worldwide at 1-km resolution — offering new insights into ecosystem functioning, biodiversity & resilience. 🛰️🌱 👉 buff.ly/9pEsi7p
🌍 AI4PEX 2nd General Assembly in Brest, France! The week combines scientific exchange, collaboration & hands-on on CNNs, AI agents, UQ & Generative approaches. A highlight: dedicated discussions for ECRs on careers, funding & supervision — followed by a beautiful coastal hike 🌊 Thanks for hosting us
💡 New Paper on physics-aware AI for #EarthObservation 🌍🤖 By combining physical models with symbolic regression & sparse ML, this work moves toward more interpretable and trustworthy emulators for atmospheric radiative transfer. 👉https://doi.org/10.1109/TGRS.2026.3676858 #ExplainableAI #AI4Earth
🎤 #EGU26 Spotlight AI must go beyond prediction 🌍 Gustau Camps-Valls on AI for climate resilience — from explainable AI to early warning systems & causal models. 🔗 buff.ly/Vr5jo6Z #AI4PEX
🎤 #EGU26 Spotlight Climate extremes don’t always happen alone 🌡️ Sonia Seneviratne on compound events — rising risks, near-permanent extremes, and limits to adaptation. How multiple climate extremes interact & amplify impacts. 🔗 buff.ly/h6FuBR9 #AI4PEX
🎤 #EGU26 Spotlight How do soil moisture & temperature drive extremes? 🌍 Feini Huang combines causal AI + ML to disentangle land–atmosphere feedbacks — and benchmark climate models. 🔗 buff.ly/56imoYB #AI4PEX
🎤 #EGU26 Spotlight Are AI climate models really more efficient? 🤖🌍 Tom Beucler compares AI, hybrid & GPU-based models — and finds the advantage isn’t always clear-cut. 🔗 buff.ly/2LSZLP5 #AI4PEX
🎤 #EGU26 Spotlight Some regions are “behind” or “ahead” in warming 🌡️ Dominik Schumacher shows: areas lagging today may see faster rises in extreme heat soon. 🔗 buff.ly/WLiZV8W #AI4PEX
🎤 #EGU26 Spotlight Fires, beetles… or management? 🌱 Thomas Pugh shows that harvest decisions may shape Europe’s forests more than disturbances. 🔗 buff.ly/sf0ejeT #AI4PEX
🌍 AI is transforming climate modelling! Catch us at #EGU26 presenting work on AI climate models, hybrid approaches & scaling ⚡️ Don’t miss: Beucler, Grundner, Ivanov, Camps-Valls + more! #AI4PEX #AI4Earth #ClimateAI
3 days, 9 models, 1 goal: benchmarking mortality in #EarthSystemModels 🌍 From data & model overviews to hands-on coding sessions, we worked toward a shared framework, common metrics, and first results for a joint paper. Great collaboration across teams & time zones! 🚀 #AI4PEX #ClimateScience
✨ Huge thanks to all instructors & participants for making the ELLIS Winter School on AI for Earth System, Hazards & Climate Extremes so engaging and interactive! 👏 Special thanks to OroraTech, ESA Academy & ICEYE — your support enabled 10 participants to join, helping make the school more diverse.
Final day at the ELLIS Winter School on “AI for Earth System, Hazards & Climate Extremes.” 🎉 Group challenge presentations, keynote reflections and well-deserved awards marked the conclusion of an intense week of collaboration and learning. Thanks to everyone who made it happen!
Day 4 of the ELLIS Winter School on “AI for Earth System, Hazards & Climate Extremes” kept the momentum going. 🔬 Challenge teams refined their solutions, supported by expert insights and keynote talks on AI for climate and Earth systems and stakeholder engagement.
We’re in Athens for the ELLIS Winter School 2026 🌍 This week, researchers and students are coming together to explore how AI can help understand and respond to climate extremes, from floods and wildfires to droughts. Hands-on learning in action. ✋
Momentum continued on day four of the ELLIS Winter School 🚀 From Zero2Hero sessions to lively challenge sessions, participants sharpened their skills and exchanged ideas across disciplines. Evening at the National Observatory of Athens — no stargazing due to clouds, but a great tour instead 🌥️🔭
🇬🇷 Day 2 of the ELLIS Winter School took us to Lavrion Campus for keynotes on AI for Earth at scale, foundation models, and ocean extremes — plus a great career panel highlighting the diverse and often non-linear paths into climate & AI. Science + sea air + career insights 🌊
Day 1 of the ELLIS Winter School in Athens 🇬🇷 Keynotes on early warning of climate risks and a planetary data layer for an observable Earth, plus a powerful talk from Zisoula Ntasiou (Hellenic Fire Service) on how science and AI can support firefighters making decisions in the field.
