Machine learning is reshaping the representation of complex physical processes in Earth system models, offering new avenues for parameterisation, emulation, and hybrid modelling. This session focuses on the use of machine learning to emulate computationally expensive or unresolved processes, accelerate physical simulations, enable data-driven discoveries, and improve representation across domains such as convection, turbulence, radiation, hydrology, sea ice, and other Earth system components. Topics include (but are not limited to):

- Subgrid-scale parameterisations via machine learning
- Emulators of physical processes, model components, or whole weather and climate models (including end-to-end learning and foundation models)
- Data-driven discoveries
- Hybrid ML-physics modelling frameworks
- Physics-informed neural networks, neural operators, and differentiable programming
- Reinforcement learning and other approaches for ensuring physical consistency, stability, and optimising model behaviour
- Calibration and parameter optimisation using ML
- Physical behaviour, encoding and analysis of ML models
- Verification and explainability (XAI) of data-driven models (including AI forecasting)
- Representation of extremes and downscaling
- Coupling of ML models with physical models
- Cross-domain applications (atmosphere, ocean, cryosphere, land)
- Impacts of architecture choices and design on physical behaviour
- Links between ML methods and fundamental physics or applied mathematics
- Novel AI methods and applications