Machine learning methods show great promise in emulating physical, dynamical, and chemical processes critical to weather and climate modeling. These methods comprise applications of AI techniques such as deep neural networks, random forests, and conditional generative adversarial networks to address problems including turbulent dynamics, radiative transfer, and aerosol-cloud interactions.

Despite these recent breakthroughs, additional advances are necessary to warrant operational use of these methods in assessing their stability, trustworthiness, and interpretability. This session thus solicits relevant progress in this area, including but not limited to tests of physical credibility, interpretability, stochastic parameterization, and especially prognostic stability in increasingly realistic operational settings. We seek applications from across the weather and climate modeling community exploring the use of a wide variety of AI techniques in operational online computational settings.