Researchers will further develop a 3-D mesh adaptation model for climate modeling, allowing computational resources to be focused on phenomena of interest such as tropical cyclones or other extreme weather events. The project will also introduce data-driven machine learning paradigms into modeling of clouds and precipitation.
The goal of this project is to demonstrate that climate modeling significantly benefits from emerging high-performance, multi-scale computational techniques, such as 3D Adaptive Mesh Refinement (AMR), and novel representations of unresolved physical processes, such as stochastic approaches and physics-informed, data-driven machine-learning paradigms for clouds and precipitation. These have the potential to transform the field of climate modeling in fundamental ways that are not possible with brute-force scaling of existing approaches or physical parameterizations that are only valid at specific mesh spacings. The paradigm shift is that computational resources are only used where needed for accuracy, which is dynamically determined by the atmospheric flow field. This guarantees a targeted use of computational resources near “phenomena of interest” like tropical cyclones or other extreme weather events, and enables unprecedented scientific investigations with cloud-resolving (km-scale) small grid spacings in selected, mesh-adapted areas.