Update on the New Dynamical Core of the GEM-Mars GCM with a Yin-Yang Grid Configuration

Abstract:

We present new simulations of the Martian atmosphere using the GEM-Mars GCM with an updated dynamical core. This new version includes a different horizontal discretization, the Yin-Yang grid, where the domain is decomposed into two limited area latitude-longitude sub-domains, perpendicular to each other. The two domains have an overlapping static halo region similar to a limited-area piloting region where an interpolation is done to update the variables. This type of grid eliminates problems with the convergence of meridians at the poles and improves the computational load balance in message-passing interfaces.

The GEM-Mars model (Neary and Daerden, 2018; Daerden et al., 2019) is able to benefit from the improvements to GEM, the terrestrial version of the model used operationally by Environment and Climate Change Canada (ECCC). We have adopted the most recent version (V5.1) of the dynamical core (see Girard et al, 2016 and Qaddouri and Lee, 2011) for descriptions of some of the recent updates) to apply to the Martian atmosphere.

Previous versions of GEM-Mars used a latitude-longitude grid definition with a horizontal resolution of 4°x4° or 90x45 grid points (~237 km at the equator). For initial tests of the Yin-Yang grid, we have chosen the two domains to have 90x31 grid points each which leads to a maximum grid size of ~178 km. The central longitude of the domain covering the poles is chosen to pass through Hellas basin, so that it does not sit on the border of the sub-domains. The original GEM-Mars latitude-longitude grid box areas vary from 1,960 km² near the poles to 56,000 km² at the equator, while the new Yin-Yang grid leads to a more uniform grid with a smaller range between 22,370 and 31,640 km².

Preliminary results from this configuration are promising, with the basic climatological patterns in the atmosphere well reproduced. The simulation with the Yin-Yang grid took slightly longer to run than the latitude-longitude grid as the total number of grid points is increased (91×31×2 for Yin-Yang, 90×45 for the latitude-longitude grid). One aspect to be investigated is a comparison between the two grid formats at higher resolution, using more processors, where the load-balancing may be a more important factor.