Environmental Science and Engineering Seminar
Abstract: The talk will review recent trends in Earth System Modeling and connects them to emerging data science opportunities. In particular, the talk will survey the latest variable-resolution and Adaptive Mesh Refinement (AMR) modeling approaches, which include the development of statically-nested and adaptively-moving meshes in the fluid dynamics component (the `dynamical core') of weather and climate models. These variable-resolution techniques provide refined grid spacings in areas of interest, and even allow scales down to the kilometer regime in the selected areas. Such resolutions are often computationally too expensive for the global domain, but are needed to reliably represent many weather extremes, such as tropical cyclones. Severe weather events will be used as a physical motivator throughout the talk. The modeling concepts are illustrated with examples from NCAR's Community Earth System Model (CESM) and the Chombo-AMR dynamical core. The latter has been jointly developed at Lawrence Berkeley National Laboratory and the University of Michigan.
Furthermore, it is discussed how the design of future-generation weather and climate models can be informed by modern statistical and data science techniques, such as stochastic physical parameterizations or machine-learning approaches. The talk concludes with a brief overview of NCAR's ‘Simpler Model' hierarchy, that provides a wide spectrum of research opportunities for the university community.