Page Comparison

Abstract

Our group is developing a number of ideas exploring several technologies designed to improve the accuracy, throughput, and usability of future generations of the ACME model. These include our active project to develop a nonhydrostatic atmospheric modeldycore, together with exploratory research in in immersed boundary topography, deep learning architectures for physical paramterization, and an end-to-end graphical user interface for simplifying the ACME workflowprojects in immersed boundaries, deep learning, and computer graphics. A nonhydrostatic atmospheric model is required to perform accurate simulations at resolutions beyond 10km per grid cell, achievable today in variable-resolution meshes. An immersed boundary representation of topography has the potential to solve problems occurring in highly mountainous regions, particularly at high resolution. Deep learning physical parameterizations potentially allow ACME to use physical paramaterizations that are both faster, more detailed, and more physically accurate than currently possible. They also have the potential to address the multi-scale parameterization issue. And our graphical user interface is designed to simplify the use of ACME for many common tasks, while reducing the learning curve for new users.