Opened 4 years ago

Closed 4 years ago

#2038 closed help (fixed)

SST and sea ice ancillary files for the UM Nesting Suite

Reported by: dsergeev Owned by: jeff
Component: UM Model Keywords: sea-ice, sst, ancils, ostia
Cc: Platform: MONSooN
UM Version: 10.2



I am trying to run the UM Nesting Suite with "custom" sst and sea-ice ancillary files for the LAM run. The ancillaries are created from OSTIA data.

The reconfiguration stage finishes without errors, but the forecast job fails on the first timestep (so output at tstep=0 time profile is empty) with the following error:

Error   from routine: eg_sl_helmholtz
Error   message: Convergence failure in BiCGstab, suspected NaNs

I suspect there is something wrong in land/sea points, but then why is the reconfiguration giving me no errors?

What would you recommend me to do?

Attachments (1)

mask_ostia_metum_mismatch.jpg (329.3 KB) - added by dsergeev 4 years ago.
Mismatch in OSTIA and UM masks after

Download all attachments as: .zip

Change History (5)

comment:1 Changed 4 years ago by jeff


The problem could be you have some points which the UM thinks are sea points but are actually land points in the ancillary file, hence you get missing data values in the SST.

How are you creating the ancillary files? What I would normally do is extrapolate the sea points over the missing data land points before creating the ancillary.


Changed 4 years ago by dsergeev

Mismatch in OSTIA and UM masks after

comment:2 Changed 4 years ago by dsergeev

Hi Jeff,

Sorry for a slow reply.
I checked the UM land-sea mask and OSTIA data mask, and indeed they are not matching (see the figure attached).

I'm using Iris Python package to regrid OSTIA data to the ancillary file grid.

comment:3 Changed 4 years ago by jeff


Does the OSTIA SST data have missing data values for land points? If so then all the blue points in your figure will have incorrect SST values. What I would normally do is use xconv to extrapolate the SST values over the missing data values before regridding to the ancillary file grid. I would guess it's possible to also do this using Iris as well, but I don't know how to do it.


comment:4 Changed 4 years ago by dsergeev

  • Resolution set to fixed
  • Status changed from new to closed
Note: See TracTickets for help on using tickets.