Weather prediction and climate modelling extensively use numerical models of the Earth system. Both the atmosphere and ocean components of such models consist of a fluid dynamics solver (dynamical core) that solves a system of partial differential equations numerically. The dynamical core is coupled to physical parameterizations that represent processes that occur below the grid scale (physics). Enabled also by substantial improvements of the underlying numerical algorithms, these models can deliver accurate and efficient simulations.

Researchers are constantly working to further improve the accuracy, efficiency, and scalability of the dynamical core, the physics, and their coupling. The rapid development of computing systems towards massive use of graphics processing units and extreme parallelism requires adaptation of algorithms to further increase model efficiency via strong or weak parallel scaling. Recent years have also seen a rapid increase in hybrid approaches that combine physics-based modelling with data driven techniques, adopting techniques from scientific machine learning for Earth system modelling.

This session invites presentations on the development, testing, and application of novel numerical techniques for Earth system models in a broad sense. The scope includes modifications to the governing equations, horizontal and vertical discretizations, structure preserving methods, time stepping schemes (including parallel in time schemes), advection schemes, adaptive multi-scale models, physics-dynamics coupling, regional and global models, classical and stochastic physical parameterizations, as well as hybrid schemes combining numerical methods and machine learning.