NEMO4 Modifications

NEMO4 largely follows the same structure as CanESM with the following differences. These notes are based on the NEMO 4.0.3 codebase.

  • NEMO4 allows for domains with different shapes. The mppgather and mppscatter routines used to collect fields onto a global array had to be updated to support this.

  • The eORCA1 grid has a very similar structure to the ORCA1 grid with the most significant changes being the landmask with a smaller, more realistic Caspian sea, some small changes in the coastlines. If the under ice-sheet cavities are excluded, then the grid has the same 362x292 dimensions as the ORCA1 grid

  • In an attempt to more cleanly incorporate potential upstream changes to the coupling interface, a layer was written between sbccpl.F90 and the OASIS coupler calls. The signature for CanCPL routines was also updated to match the abstract interface description based on the equivalent OASIS routines.

  • XIOS can be run in ‘detached’ mode where a separate set of processors is devoted to I/O. CanCPL initialization and the definition of the colored communicators is so far a blocking call. XIOS may need to be modified to call define_group to enable support. OASIS seems to deal with this by allowing the coupler to return two communicators one for the actual NEMO processes and another for the XIOS processes. An extensive dive into how data are exchanged between XIOS and NEMO has not yet been done.

  • One oddity that seems to be within the NEMO code is that if the land processor exclusion is enabled by setting jpni, jpnj, and jpnij, some of areas of the ocean are incorrectly blocked out.

  • Some consideration could be given for the atmosphere to calculate fluxes among the different ice thickness categories, requiring a 3D array of fluxes (jpcat,jpi,jpj). Some of this machinery appears to exist, but likely has not been tested recently.