Assessing Grid Refinement Strategies in the Chombo Adaptive Mesh Refinement Model

Complex multiscale atmospheric phenomena such as tropical cyclones challenge conventional climate models and their relatively coarse uniform resolutions. Adaptive mesh refinement (AMR) techniques offer the flexibility to permit sufficiently high-resolution grid patches over features of interests without compromising accuracy or being computationally prohibitive. One such approach, being developed by the Applied Numerical Algorithms Group at Lawrence Berkeley National Laboratory and collaborators, is a non-hydrostatic finite-volume General Circulation Model (GCM) on a cubed-sphere grid with dynamic adaptive mesh refinement capabilities called Chombo-AMR. The 2D shallow-water equations exhibit many of the complexities of 3D GCM dynamical cores. Thus they serve as an effective method for testing the dynamical core and the refinement strategies of adaptive atmospheric models. We investigate the accuracy and efficiency of refinement strategies with the Chombo-AMR dynamical core in shallow-water mode. In particular, we use a suite of test cases consisting of both simple and complex dynamical flows, including a test case designed to simulate tropical cyclone-like vortices. We use the suite to test the effectiveness of the refinement criteria for both static and dynamic grid configurations as well as the sensitivity of the model results to the criteria.