| Speaker: | Christiane Jablonowski, Atmospheric, Oceanic & Space Sciences, U of M | Title: | Adaptive Grids in Climate and Weather Modeling |
| Date: | Monday, March 31, 2003 |
| Time: | 3:10 - 4:00pm |
| Place: | 3096 East Hall |
| Abstract. |
Climate and weather models are amongst the many fluid
dynamics applications that are characterized by multi-scale
interactions. But although today's atmospheric General Circulation Models
(GCMs), and in particular weather prediction codes, are already capable
of uniformly resolving horizontal scales of order 20km, the motions of
interest span many more scales than those captured in a fixed resolution
model run.
Dynamically adaptive grid approaches have long been used in aerospace engineering, astrophysics, aeronautical and other computational fluid dynamics problems. In atmospheric science they were first applied a decade ago and mostly restricted to regional models in Cartesian coordinates. This research project is now aimed at adaptive methods for PDEs on the sphere. The talk will give a design overview of an adaptive GCM that is based on NASA's next generation global climate model. This conservative model in flux form uses a local finite volume discretization that is third order accurate when applying the piecewise parabolic method (PPM). The main building block of the adaptations is a block-structured data structure that allows high performance computations with minimal changes to the pre-existing PDE solvers. Blocks are refined or coarsened by a factor of two and neighboring blocks can only differ by one refinement level. This guarantees accurate influx and outflow conditions at the interface boundaries. Questions concerning the optimal refinement criteria that are capable of capturing the atmospheric `features of interest' will be discussed. Furthermore, the challenging aspects of grid adaptations on parallel computers will be addressed. |