Jared O. Ferguson

Adaptive Mesh Refinement (AMR) techniques have the potential to address many of the challenges current climate models with their relatively coarse grids face in resolving complex multi-scale atmospheric phenomena like tropical cyclones. Models with AMR balance the benefits of high-resolution with increased computational costs by increasing resolution only over features of interests thus limiting the computational burden of requiring such high-resolution grids globally.

AMR allows the creation and removal of additional grid points, effectively increasing or decreasing resolution in local areas when needed. Thus AMR can ensure that salient features and areas of high gradients are sufficiently resolved while reducing computational costs by not over-resolving large areas of smooth gradients and calm weather systems. For climate applications, AMR techniques could allow a single model to track small- scale features such as tropical cyclones while also resolving how those features interact with the global dynamics. Adaptive refinement techniques are a relatively new approach to global atmospheric models and significant challenges such as developing the complex numerical methods and data structures to run efficiently and accurately across parallel systems have only recently started to be overcome.

The research explores the effectiveness of AMR techniques in resolving and tracking such multic-scale features in Atmospheric General Circulation Models (GCM). The research focuses on assessing how AMR techniques affect both the dynamical core component, which uses numerical methods to resolve the fluid motion, and the physical parameterizations, which approximate all unresolved subgrid-scale physical processes (such as clouds and precipitation) at each grid point. Specifcally we are using a new multiscale climate model deigned to implement AMR is being developed by the Applied Numerical Algorithm Group (ANAG) at Lawrence Berkeley National Laboratory (LBNL) with several collaborators. The Chombo-AMR model is a non-hydrostatic 4th-order finite-volume GCM on a cubed-sphere grid with block-structured adaptive mesh refinement capabilities that allow it to refine in both space and time. With the model we investigate the AMR methods in a 2d shallow-water framework that serves as an ideal testbed for 3d model developments. We will then test a variety of refinement criteria for selecting where increase resolution and investigate the valididty of physical parameterizations at the different spatial scales in the 3D version of the model.

Figure: 2D shallow-water AMR tests. (Left) Spiral advection test in which a passive tracer is wrapped up into tight spiral bands. AMR (shown by black boxes)is tagging on the gradient of the tracer. (Right) Idealized vortices moving about the sphere and merging with other vortices. AMR is tagging on relative vorticity.