Numerical Fluid Dynamics in the SciDAC Multiscale project aims to improve and advance the treatment of atmospheric dynamics in current and next generation general circulation models (GCMs).
- Development of CAM-SE’s variable resolution capability, including improving numerical formulations (such as dissipation schemes) for the regionally-refined model (RRM). Numerical elements include assessing resolution-aware schemes for dissipation and evaluating. Ongoing efforts include assessing high-resolution regional climate predictions, for interactions between important features (e.g., tropical cyclones) and physics parameterizations.
- New mesh generation and remapping tools for spectral element unstructured grids. This includes techniques for variable resolution grid generation, such as tools for variable-resolution topography generation and low-connectivity templates. Advanced remapping between structured and unstructured grids are being developed to couple multi-resolution mesh components of climate models. Idealized model assessment tools are being developed to verify RRM and AMR techniques.
- Motivated by the need for more accurate and efficient multiscale simulations when there are large scale separations, for example resolving extreme weather events within planetary wave anomalies, an implicit time stepping capability is being developed within the CAM dynamical core. This capability will provide an option for stable time integration using large time step sizes that scale within global and local spatial refinement. Once complete, coupled dynamics and and cloud physics parameterizations can occur with measured accuracy and stability.
Multiscale also has multiple efforts focused on the “grey zone” of atmospheric modeling – below 10km global resolution – where gravity waves and other non-hydrostatic effects appear. The goal in this case is to explore interactions with physics parameterizations at extremely high resolution in space and time, to better characterize and improve multiscale parameterizations.
- Development of an Adaptive Mesh Refinement (AMR) dynamical core, which merges a global non-hydrostatic model with space-time refinement capabilities provided by CHOMBO. This will be used to explore convergence of parameterizations in dynamic features near 1 km resolution.
DAM is a fully compressible large-eddy model coupled to cloud microphysics and interactive shortwave and longwave radiation. DAM is outfitted with a variety of Eulerian and Lagrangian tracers to probe the turbulent statistics of moist convection, providing a benchmark for convective parameterizations, such as SPM. Related work with LBNL CCSE is using AMR to explore the interaction of reversible moist processes on advective time scales.