Please join us in the NCAR ML MSR on February 5, 2013 from 10am - 11am.
Future tracer transport in CAM-SE
Finite Volume based Semi-Lagrangian schemes in HOMME
Recently, a highly scalable spectral-element-based version of the High-Order Method Modeling Environment (HOMME) was integrated in the Community Atmosphere Model (CAM-SE). In next generation climate modeling efforts there is the need for on the order of 100 prognostic tracers. Thus, scalable, monotonic, and conservative multi-tracer efficient advection schemes are of high interest in HOMME.
To address this, we integrate conservative semi-lagrangian multi-tracer transport schemes into HOMME. The algorithms are specifically designed for multiple processors and the cubed-sphere grid with the element-wise domain decomposition as employed by HOMME. Our data structures and design allow us to provide finite volume based numerical schemes with only one commu- nication of tracer values for each time step. Our semi-lagrangian schemes conserve mass up to machine precision with monotonic option also on multiple processor systems. In particular, we discuss and compare the approach of the Conservative Semi-LAgrangian Multi-tracer transport scheme (CSLAM) algorithm and a new scheme, called SPELT (SPectral Element Lagrangian Transport), specifically designed for arbitrary quadrilateral grids. Both schemes run on multiple processor platforms, are highly scalable and multi-tracer efficient.
The ultimate goal is to replace the spectral element tracer transport scheme in CAM-SE while still using the velocity provided by the spectral element discretization. We compare the effec- tiveness of our hybrid approach on the baroclinic wave test example. This strategy promise to be very suitable and computationally cheaper for future climate modeling with CAM-SE.