Artificial intelligence and machine learning are opening new opportunities across atmospheric science, enabling information to be extracted from increasingly complex observations, improving the representation and understanding of atmospheric processes, and providing new approaches to weather and climate modelling and prediction. At the same time, the growing use of AI raises fundamental scientific questions about how data-driven approaches can be integrated with physical knowledge, numerical models, and observations while maintaining physical consistency, interpretability, robustness, and generalizability.
This session aims to bring together researchers developing and applying AI and machine-learning approaches across the atmospheric sciences. We welcome contributions spanning scales from clouds, aerosols, precipitation, radiation, and boundary-layer processes to regional and global weather and climate. Topics may include, but are not limited to, AI-assisted analysis and retrievals from atmospheric observations; process discovery and understanding; data-driven and hybrid parameterizations; emulation of computationally expensive model components; differentiable and hybrid physics–AI modelling; foundation models and other data-driven approaches for weather and climate prediction; AI-enhanced data assimilation; and methods for improving physical consistency, interpretability, uncertainty quantification, and robustness.
By connecting methodological developments with atmospheric applications, the session seeks to foster exchange between the AI, observational, process-modelling, weather, and climate communities. We particularly encourage studies that go beyond predictive skill alone to investigate how AI can advance physical understanding, improve atmospheric models, and contribute to more reliable prediction across spatial and temporal scales.