Dynamics–Physics Coupling in the New GEM Dynamical Core with Height-Based Vertical Coordinates

A new dynamical core of the Global Environmental Multiscale (GEM) model, hereinafter referred to as GEM-H, has recently been developed at Environment and Climate Change Canada (ECCC). The new dynamical core introduces the option of a height-based terrain-following coordinate (TFC) in the vertical in addition to the existing log-hydrostratic-pressure-type TFC employed by the operational GEM-P core. The GEM-P model exhibits strong numerical instability when subjected to steep orography (terrain slopes larger than 45°). This renders the existing operational model unsuitable for developing very high resolution (sub-kilometer) weather forecasting systems that will be crucial for advancing ECCC’s numerical weather prediction (NWP) capabilities in the near future. Initial tests have demonstrated the potential of the height-based GEM-H core in improving the stability behavior over steep terrain. Furthermore, the scalability limitation of the direct solver employed in GEM-P is driving extensive research at ECCC aimed at developing optimized three-dimensional iterative solvers for future generations of massively parallel supercomputers. The height-based GEM-H dynamical core is expected to be more amenable to such iterative solvers as the metric terms attributable to the vertical coordinate transformation appear explicitly in the discretized elliptic boundary value problem.

In the absence of subgrid-scale physics forcings, predictions by the new height-based dynamical core is found to be statistically equivalent. Efforts were then made to couple the ECCC’s Physics Package with the new dynamical core without any additional calibration. During the development stage, a number of dynamics-physics coupling approaches were considered. The operational GEM-P model utilizes the ‘split method’ for dynamics-physics coupling where the dynamical equations are resolved in the absence of any physical forcing and at the end of the dynamics sub-step the physics contributions are incorporated as adjustments to some of the prognostic variables in the so called ‘split mode’. Another approach for dynamics-physics coupling works by directly incorporating the impact of physics forcings as tendencies into the discretized dynamical equations, and hence is referred to as the ‘tendency method’. Researchers in the past have demonstrated possible erroneous behavior of the split method, particularly for large time steps that are permissible by the semi-Lagrangian semi-implicit (or iterative implicit) approach. Efforts to couple ECCC’s Physics Package with the GEM-H dynamical core reaffirmed some of the troubling aspects of the split method.

Overall, the tendency method for dynamics-physics coupling is found to be more acceptable, and it also leads to a very good agreement between the two dynamical cores. A split-tendency hybrid approach for GEM-H has also been developed that results in objective forecast scores that are equivalent to GEM-P with the split method. In addition to providing further details pertaining to the issue of dynamics-physics coupling, results comparing the two GEM dynamical cores with respect to the different dynamics-physics coupling approaches will be presented at the conference.