Item

Abstract

Conclusions of discussions:

Following the presentations on the first day of the workshop a round-table discussion were
held on the second day with the participants (explecitely) mentioned in the list above.

It was concluded that the operational semi—Lagrangian models developed at CMC and ECMWF
seemed to work satisfactory in weather forecast mode. Control Eulerian forecasts with a short
time-step and semi—Lagrangian forecasts with relatively long time—steps had been found to give
almost equivalent accuracy. So far they had not been tested in climate mode, i.e. with lower
resolution in very long term integrations. The semi—Lagrangian version of the CCM2 had been
developed at NCAR for applications in climate research, but so far it had been tested only in
relatively short term integrations.

It was decided to initiate coordinated experiments with the semi-Lagrangian schemes currently
available at the three centres in order to test the suitability for the use in climate simulations.
Each center should perform two five year integrations, an Eulerian control integration and a
semi-Lagrangian integration, with similar resolution and sea—surface temperature (SST). CMC
would run its model at resolution T63/L20, NCAR would run the CCM2 at T42/L18 and
ECMWF would run at T63/Ll9. Annual—cycle climatological SST's and sea-ice distributions
(AMIP) should be used and mass conservation would be (approximately) enforced by fixing
the horizontal mean of In ps (pg beeing the surface pressure). The results obtained from the
ECMWF experiments should be analysed at MP1.

The cell-inte grated semi—Lagrangian scheme proposed by Machenhauer were found intriguing
and it was recommended that this work should be continued as intended with tests of the
scheme at first in a shallow water model. A coorporation'might be established with Rene
Laprice and his student Andre Plante at the University of Quebec in Montreal who
independently had developed and tested cell—integrated semi—Lagrangian schemes for horizontal
advection of a passive scalar.