Sorry to bother you, but I have a question regarding the sea surface temperature (SST) initial conditions in a standard fully coupled GC31 suite (e.g. u-df570, which is a copy of the official CMIP6 preindustrial control).
As far as I am aware, there are 2 sources for SST initial conditions:
An ocean version (on the ORCA1 curvilinear grid i.e. 362,332) at /work/y07/shared/umshared/hadgem3/initial/ocean/eORCA1v2.2x/EN4_v1.1.1995_2014.monthlymean_eORCA1T_NEMO_L75.nc or TS_1950-1954_1-12_EN.4.1.1.g10_eORCA1_v2.2x_L75_degC.nc
An atmosphere version (on the standard N96 grid i.e. 192, 144), at /work/y07/shared/umshared/ancil/atmos/n96e/orca1/sst/reynolds/1981_2012_360/v3/qrclim.sst
The ocean version contains a monthly climatology both temperature and salinity, at all levels. The atmosphere version only contains a monthly climatology of SST (i.e. single level).
My question is: which file overrides which? Or do neither override each other?
If, for example, I wanted to modify one of these to impose, say, a large area of heating in the central Pacific, would I need to modify both? If I modified just the ocean initial conditions, would the atmosphere not see the change? Likewise if I modified just the atmosphere SST initial conditions, would the ocean not see the change?
In an ocean-coupled run the SST will be obtained as a ‘online’ field from NEMO via the coupler. The ‘atmospheric’ sst and seaice files will not be used (and cannot see them specified in the app/um/rose-app.conf under [namelist:items]).
The eORCA1 file above is used to initialise NEMO, after which the values be calculated independently.
For your purpose it looks like an AMIP configuration will be more suitable, by specifying a bespoke N96 sst file with the forcings proposed to be applied.
Firstly, it is a good to know that the atmospheric SST and sea ice files are not used. I thought this was the case, because (as you correctly say) they are not specified under the namelist:items. Great.
As to your other suggestion: yes, I initially considered during an AMIP run, but my purposes are slightly different to a standard AMIP simulation. I am wanting to do what we call a ‘restorative simulation’, in which the fully coupled model is run but where the initial conditions are ‘restored’ (a.k.a. changed) to something else at the beginning, and then let the model do its thing. To that end, I have 2 further (but very related) questions:
Once I have modified my NEMO initial conditions and started the run, is there a way to basically reinitialise again at regular intervals i.e. impose my anomaly again? Given that the file is a monthly climatology I wouldn’t necessarily want to do this every year, but perhaps every 5 years or every 10 years? In other words, impose my anomaly at the beginning, run for n years, impose the anomaly again, run for n years, and so on? Without having to stop the model each time and restart?
The eORCA1 is a full depth ocean i.e. 75 levels, containing both temperature and salinity. Is there a way to effectively mask out all of the lower levels, so that the only thing the atmosphere sees is the surface? A little bit like an old-fashioned slab ocean experiment. I wondered if I could just convert all the values below the surface to NaNs, but something tells me the model would crash. Alternatively, is there a switch somewhere to do this, or a branch that could be added?
Whether the above 2 ideas are scientifically appropriate is something I will need to think about. For now, I just want to know if they are technically possible and, if yes, how?
1. I cannot see any way of updating the sea surface temperature without maybe ‘pausing’ the run before the coupled task at a given year and then modifying the last ocean restart file before releasing the workflow. However, in general I suspect this will lead to instabilities in the model.
Only the surface i..e topmost level values are passed from Ocean to the Atmosphere model, essentially replacing the fields that would have come from the sst and seaice ancillaries.