Open oceanaguiar opened 10 months ago
The development of a workflow to run ORCA36 stems from the porting of GOSI9 at NEMO 4.0.4 for ORCA025 to NEMO4.2.1, then 4.2.2, which then becomes the GOSI10 prototype. The steps :
[x] Porting of GOSI10 workflow for ORCA025 from cylc7 to cylc8: GMED ticket (visible to internal Met Office users only): https://code.metoffice.gov.uk/trac/gmed/ticket/666
[x] Adapt the GOSI10 workflow for ORCA025 to include options to run on both XC40 (Met Office) and on ECMWF ATOS supercomputer (the workflow includes already the changes from the previous step) https://code.metoffice.gov.uk/trac/roses-u/ticket/579
[x] Add optional archiving to the ecfs system when running on ECMWF https://code.metoffice.gov.uk/trac/roses-u/ticket/584
[x] Atmospheric forcings for eORCA36 : https://code.metoffice.gov.uk/trac/roses-u/ticket/593
[x] Prepare the ancil files (done, but needs to be documented)
[x] Adapt the workflow to use ORCA36 (ATOS only) https://code.metoffice.gov.uk/trac/roses-u/ticket/592
The ORCA36 model was set up to run a 1 year experiment . The model completed the 1 year experiment on ECMWF ATOS (1 node=240GB) using the following main settings:
Example output ( Surface velocity field averaged from a 24 hourly output on 15-08-1980)
after 1 year run ( Surface velocity field averaged from a 24 hourly output on 31-Dec-1980)
Main differences from the DEMO:
ORCA36 (This implementation) | ORCA36 DEMO | |
---|---|---|
NEMO nodes | 78 | 100 |
XIOS nodes | 150 | 200 |
time to run 1 day | 60 min | 30 min |
ln_isf (*) | false | true |
ln_isfcav (*) | false | true |
atm forcing | CORE2 | IFS |
ln_ncar | true | false |
ln_ecmwf | false | true |
initial conditions | WOA13 | WOA13 |
(*) The input data to build the domain with ln_isfcav have not been provided this file is also not provided