Improving Consistency, Accuracy and Stability of the Global Environmental Multiscale Atmospheric Model

Research and development activities are conducted at the Meteorological Research Division (MRD) of Environment and Climate Change Canada on a continuous basis to address different issues related to numerical consistency, accuracy and stability of the dynamical core of the Global Environmental Multiscale (GEM) atmospheric model. Several modifications have recently been implemented to improve general stability of the model with particular emphasis on the high-resolution forecasting aspect. At sub-kilometer resolutions, the model exhibits strong instability over complex orography where it may encounter mountains with steep slopes. Off-centered averaging of the source terms in the semi-Lagrangian scheme as well as the explicit horizontal numerical diffusion scheme – available within the GEM model to control high wave number noise – are inadequate to address this strong orography-induced instability. Increasing the level of off-centering for the equations attributable to the non-hydrostatic aspects of the atmospheric flow, on the other hand, is found to significantly improve model stability over steep mountains. Furthermore, at sub-kilometer resolutions the three-dimensional elliptic problem resulting from vertical discretization imposes severe restrictions on the vertical resolution and the time-step size in order to maintain vertical separability that permits the use of a direct solver. Readjusting the contributions of the nonhydrostatic pressure perturbation is found to circumvent this separability issue for the direct solver.

The elimination of inconsistencies in the temporal averaging and interpolation methods employed in the semi-Lagrangian advection and the trajectory calculations within GEM has been another important recent development. Elimination of these inconsistencies is found to have a considerable positive impact on the accuracy of the GEM model in both hydrostatic and non-hydrostatic scenarios. Controlling spurious numerical resonance due to stationary orographic forcing is also an active research interest at the MRD. The present approach to suppress orographic resonance is based on off-centering that leads to damping of the meteorological fields over a wide range of length scales, and as a consequence, results in less accurate forecast. Work is therefore currently in progress to eliminate or reduce GEM’s dependence on off-centering to control this resonance by treating the orography-related terms in the model equations in a spatially-averaged Eulerian manner.

Results pertaining to the aforementioned enhancements implemented in the GEM model to improve its numerical consistency, accuracy and stability will be presented at the conference.