An Adaptive Mesh Refinement (AMR) Framework for Future Weather and Climate Models

The paper discusses a novel mesh refinement technique that dynamically adapts its grid resolution to atmospheric features of interest, such as strong vortices or rainfall patterns. The model can span a wide spectrum of grid resolutions that range from hundreds of kilometers to convection-allowing scales of a few kilometers. The paper reviews the current state-of-the-art of AMR and variable-resolution modeling, and introduces a high-order, finite-volume, multi-block AMR framework for solving the 2D shallow water and 3D nonhydrostatic equation sets on the sphere. The framework is built upon the AMR library Chombo which has been designed at the Lawrence Berkeley National Laboratory. Chombo supports the 'cubed-sphere' grid geometry which serves as the base computational grid for the atmospheric AMR applications.

Idealized 2D and 3D test cases are used to illustrate the variable- and high-resolution characteristics of the dynamically adaptive atmospheric model. Both the 2D and 3D configurations include simplified physical parameterization suites. The moisture interactions provide nonlinear forcing effects which challenge the AMR technique and the scale dependencies in the moist atmospheric model. The dynamical core test cases include a 2D moist barotropic wave and tropical cyclone, as well as selected examples from the Dynamical Core Model Intercomparison Project (DCMIP) and a 3D colliding-modons test case. The results suggest that AMR dynamical cores have the potential to serve as the basis for future-generation weather and climate models. They allow the flow-dependent generation of high-resolution domains while limiting the overall computational workload.