AI for the study of Environmental Risks (AI4ER) CDT Showcase 2026

Join us for our 2026 Annual Showcase at the University of Cambridge. We are excited to welcome speakers from Frontier Development Lab, Goldman Sachs, Moody's and AI4ER at this year's showcase.

The UKRI Centre for Doctoral Training (CDT) in the Application of Artificial Intelligence to the study of Environmental Risks (AI4ER) trains researchers to develop and apply leading edge computational approaches to address the critical environmental challenges that our planet is currently facing.

Explore the latest advancements in Artificial Intelligence and its application in understanding environmental risks. Connect with experts, researchers, and enthusiasts passionate about leveraging AI for a sustainable future. Don't miss this opportunity to be part of a cutting-edge discussion on environmental challenges and innovative solutions.

See you there!

Info: Climate emulation is fundamentally an out-of-distribution (OOD) prediction problem. While ML parametrisations offer skilful, computationally efficient alternatives to explicit simulation of sub-grid processes, how they perform under distribution shift remains poorly understood. Using a novel evaluation framework grounded in emergent constraints, we characterise robustness across state-of-the-art ML super-parametrisation emulators and find they systematically degrade under shift. We identify compositional generalisation — the ability to form novel combinations from previously observed elementary components — as a principled route toward more robust ML parametrisations, demonstrating that physically motivated decompositions improve OOD performance with only modest in-distribution skill degradation.

Info: Artificial intelligence is redefining how we observe, simulate, and anticipate weather and climate risks, but its impact depends on applying the right tools to the right problems. Recent AI weather forecast models have delivered striking gains in short‑term prediction, particularly for large‑scale dynamics and tropical cyclone tracks, transforming operational forecasting and event response. In parallel, a new generation of AI climate emulators is emerging, designed not for days, but for decades, prioritizing stability, physical consistency, and the large‑scale correlations that underpin multi‑peril risk. Together, these advances mark a shift in how AI supports environmental risk analytics. The future lies not in replacing physics‑based models, but in combining AI and physical understanding to build faster, more integrated, and more decision‑relevant climate risk frameworks.

Know what’s Happening Next — before everyone else does.

Get App

Host or Publisher AI4ER CDT