A follow-up to the first RPN AMIP integration is reported upon. The new simulations utilise higher spatial resolutions both in the vertical and in the horizontal. Furthermore, several modifications to the physical parametrizations have been implemented in order to correct for deficiencies identified in the first run. We compare a control (using a repeating but specified seasonal cycle of sea surface temperature and sea ice coverage) with the standard AMIP-type simulation (i.e. which uses varying monthly mean values of these fields for the 1979-1988 decade). We also compare our results with NMC analyses for the simulated decade.
The RPN AMIP integrations are done with a version of the Canadian global spectral forecast model, a state-of-the-art forecast engine. The model uses a semi-implicit, semi-Lagrangian time integration scheme which allows for very efficient use of existing computer resources. These runs are used by RPN to better understand and correct for our model's systematic errors.