Weather forecast and climate models
often use spectral or pseudo-spectral methods
for solving the PDEs that describe
the atmospheric motions. Compared to
grid-point based methods, such methods lose
their advantage when very fine grids are used.
In addition, grid-point methods
can much better exploit massive parallelism.
Because massive parallelism will play an important role in future
computational modeling, there is a renewed interest in
grid-point methods for atmospheric hydrodynamical
calculations. This project aims at participating in this
development.
Our long term goal is
a comprehensive comparison between newly developed and
existing methods, taking into
account
- space grid generation and discretization,
- use of extremely fine grids,
- efficient, long term time integration and
- massive parallelism.
Such a comparison should attempt to answer the question:
which numerical grid-point based method is most promising for
future atmospheric hydrodynamical
calculations.
In 1999 we have examined the spatial discretization
and time integration for the
shallow water equations on the sphere.
These equations may be considered as a simplified
set of atmospheric motion equations.
Special attention has been paid to an Osher-type
finite-volume discretization on lat-lon
and stereographic grids. In 2000 we will continue
with time integration research and extend the research
to a prototype more realistic than the shallow water equations.
The project is funded by NWO (PhD student Debby Lanser).
In 2000 cooperation with the GMD has started. A Postdoc
(Jason Frank), is shared by CWI and GMD.
People involved:
Last update 00/01/20.