Spanning the Hydrostatic Limit with the CAM-SE Atmospheric Dynamical-Core

Monday, May 12, 2014 - 07:00
Add to Calendar

Investing in very high resolution simulations will enable regional climate predictions, increased solver accuracy, inclusion of mesoscale effects, and the conversion of subgrid-scale parameterizations into resolved dynamics. A necessary condition for achieving such resolutions is an atmospheric dynamical-core that is capable of solving the nonhydrostatic equations of motion while maintaining excellent parallel scalability. CAM-SE is the most scalable atmospheric dynamical-core in CESM achieving high throughput by way of near optimal strong-scaling up to 100,000 cores, and it is well positioned to satisfy these goals. Our group has been working diligently to integrate nonhydrostatic capabilities within CAM without sacrificing its performance. We have developed a spectral-element nonhydrostatic dynamical core, based on the model of Laprise, which is formulated to be as similar as possible to the existing CAM-SE model, to facilitate code reuse and its rapid integration into CESM. We have also developed a second, more experimental model which we are using to explore the advantages of the Discontinuous-Galerkin method as well as alternative vertical coordinate systems. Our group has accelerated the performance of the CAM-SE model in general through improved parallel communication routines, including asynchronous MPI exchange and communication hiding, and we have developed a HEVI (horizontally-explicit vertically-implicit) time-integration scheme for dealing with fast vertical acoustic waves in the nonhydrostatic system. Looking forward, continued increases in resolution will require high-accuracy representations of vertical processes, the use of variable-resolution grids to focus computational resources, and integration of CAM-SE with scale-aware parameterization packages. In this talk we will demonstrate the progress we have achieved to date in achieving these goals and outline our plans for the continued expansion and improvement of CAM-SE.

Funding Program: