Overall, there is a growing need for small-scale climate and weather information in the user community. As discussed before, nested grids have been widely used for local weather predictions and regional climate models to downscale a global large-scale simulation (Skamarock et al. 2005, Caya and Laprise 1999). The main motivation for using nested grids over high-resolution models is not science-driven. It is based on the realization that high-resolution global models are limited by the available computing resources. Nested grids thereby offer an economical approach to the multi-scale problem. However, different numerical schemes, physics parameterizations and dynamical formulations are most often used in the global GCM and nested domain. Therefore, special care must be taken to decrease or even damp the numerical and physical inconsistencies across the fine-coarse grid boundaries. Common deficiencies are artificial rainfall at fine-coarse grid boundaries and reflections of weather systems that enter or leave the refined domain.
Recently, a new approach to regional climate modeling and multi-scale interactions has emerged which builds upon the Adaptive Mesh Refinement (AMR) technique in the global domain (St-Cyr et al. 2008, Jablonowski et al. 2009). We envision for future model generations that high resolution regions will be built into GCMs that consistently treat the boundaries at fine-coarse grid interfaces. The high-resolution domains are kept at a minimum and can be individually tailored towards user-selected features or regions of interest. AMR thereby allows for static and dynamic mesh adaptations that can flexibly focus their computational mesh on certain geographical areas or atmospheric events. Examples are mountain ranges, tropical storms or other extreme events, not adequately captured by coarse GCM resolutions. The computational resources are minimized while allowing for multi-scale interactions in the refined region. The interaction is automatic two-way interactive and provides feedback mechanisms in the coarse domain. Both non-conforming (stepwise) and smooth conforming adaptive mesh transitions are feasible. The latter is realized through triangular meshes. The numerical schemes for AMR can be designed to guarantee physical consistencies like mass conservation, and to minimize wave reflections and refractions at fine-coarse grid interfaces.