GPU Developments for Applications in Climate and Weather

This talk will contain two primary topics: The first topic will provide an overview on the NVIDIA HPC strategy for numerical models in Earth system science, and present the current state of GPU-acceleration for widely-used models and their HPC performance achievements. Specific GPU progress will be examined for atmosphere models and select examples will provide relevance to science-scale HPC practice that quantifies the benefits of a heterogeneous GPU-based approach vs. CPU-only. The second topic will present NVIDIA developments in AI models from recent collaborations that apply emerging techniques to improve and accelerate research in areas of automatic feature  detection of severe weather events from model data, improved data assimilation strategies, emulations and improvements in physics parameterizations, and other applications. Examples will be provided from a range of NVIDIA collaborations with scientists and organizations in the climate and weather communities.