Xancil AMIP SSTs

,
1

I am trying to make an SST ancillary. I make this by reading in the standard AMIP SST ancillary ('/work/y07/shared/umshared/CMIP6_ANCIL/data/ancils/n96e/timeseries_1870-2016/SstSeaIce/sst_amip_n96e.anc) and adding 4K to it where there is less than 10% ice. I produce a netcdf file and then use xancil to convert the output netcdf file to an ancillary file. When I do this I get warnings from the UM of the form:

WARNING:controlled time interp for SST not available: Mismatch in SST and SEA-ICE ancillary data times in FIXED HEADER
position= 27 SEA-ICE=-32768
position= 27 SST=16
WARNING:controlled time interp for SST not available: Mismatch in SST and SEA-ICE ancillary data times in FIXED HEADER
position= 34 SEA-ICE=-32768
position= 34 SST=346
WARNING:controlled time interp for SST not available: Mismatch in SST and SEA-ICE ancillary data times in FIXED HEADER
position= 41 SEA-ICE=-32768
position= 41 SST=0

Assuming that this error means something and so I should remove it. I used xconv to look at my sst ancil and see that the ancillary file is a monthly mean field (I can’t copy paste the xconv field info).

See /home/n02/n02/tetts/cylc-run/tc003_140277/share/data/etc/ancil/qrclim.sst

The sea ice ancil at /home/n02/n02/tetts/cylc-run/tc003_140277/share/data/etc/ancil/qrclim.seaice does not show up as a monthly mean. Nor does the standard AMIP ancillary at /home/n02/n02/tetts/cylc-run/tc002_140277/share/data/etc/ancil/qrclim.sst.

As well as not showing as monthly means, the standard ancillaries have a version number of 0.5 while my created SST ancillary has a version number of 13.8. See (on archer2) /work/n02/shared/tetts/OptClim_UKESM/processing/sst_p4.job for my current xancil job.

Any help appreciated including if the warning message matters – if the interpolation is doing strange things then I can imagine the SST ancillary will be a bit strange when the model gets it…

Simon

2

Hi Simon,

I haven’t seen that error before but when I have made SST and SI ancils before, I have found this page useful https://code.metoffice.gov.uk/trac/ancil/wiki/CMIP6/ForcingData/ModelSstSeaIce.

Cheers,

James

3

Hi James,

do you have code? I might go back to the people I am working with at the Met Office/NCAS and see if they have a 360 day AMIP+4K SST ancillary. They do have a Gregorian calendar version but my setup is 360 day!

Simon

4

Hi Simon,

I’ve not managed to get those scripts (e.g. pptoanc) to work due to issues with the ants library (although I believe it may be available for Met Office people). I’ve tended to extract the stash items mentioned from the run of interest, perform modifications on the netcdf files and then use xancil to generate an ancil file.

Cheers,

James

5

Hi James,

that is what I do though my problem is xancil is making it a monthly mean… When the original ancillary was not a monthly mean… It might be the error is actually that the original data should be monthly means as the AMIP-II method is used to update them.

I will try running the raw sea ice through the same process. That might fix my problem.

Simon

6

My problem was fixed by regenerating the sea-ice ancillary using xancil. xconv thinks both files are monthly means, which is what I think the AMIP method expects. No more warning messages, and my perturbed parameter +4K run did not blow up. Lack of blowup may be getting lucky rather than anything else.

Simon