#3165 closed help (completed)

grid-T files not dehaloed

Reported by: m.couldrey Owned by: um_support
Component: Rose/Cylc Keywords: dehalo
Cc: Platform: JASMIN
UM Version:

Description

Hi CMS

I recently ran the dehalo suite (u-bo201) on my suite u-bm978. The output was stored in
/gws/nopw/j04/rdf_migrate_vol2/mpc18/NEXCS_OUTPUT/u-bm978
I ran the suite and it seemingly succeeded. But it seems that it hasn't dehaloed the grid-T files.

When I look in e.g.
/gws/nopw/j04/rdf_migrate_vol2/mpc18/NEXCS_OUTPUT/u-bm978/18520101T0000Z
I can see that there are monthly grid-U, V, and T files.
nemo_bm978o_1m_18520101-18520201_grid-T.nc
nemo_bm978o_1m_18520101-18520201_grid-U.nc
nemo_bm978o_1m_18520101-18520201_grid-V.nc

When I look in the corresponding output directory:
/gws/nopw/j04/rdf_migrate_vol2/mpc18/NEXCS_OUTPUT/u-bm978-cdds/18520101T0000Z
there are dehaloed grid-U, V files but not grid-T.
nemo_bm978o_1m_18520101-18520201_grid-U.nc
nemo_bm978o_1m_18520101-18520201_grid-V.nc

Not sure what's happened here, any help would be very much appreciated! Thanks!

Change History (7)

comment:1 Changed 10 months ago by jjabram

Hi Matt,

I'm just looking into this now for you. I've had a quick look at all the files and it looks like it is processing some grid-T files, but not others. Leave it with me.

Joseph

comment:2 Changed 10 months ago by m.couldrey

Thanks for for looking into it! Yes it seems like it picks up the 1d files but not the 1m ones. Whereas when I ran the suite on u-bi909, it seemed to process both the 1d and 1m files no problem.

comment:3 Changed 10 months ago by jjabram

Hi Matt,

It seems to be struggling to recognise the dimensions of the '1m grid-T' files. Can you think of any reason why the dimensions would be different or perhaps why the dimensions would be reported in a different way to the other files?

I'll keep having a look myself.

comment:4 Changed 10 months ago by jjabram

Doing ncdump to look at the dimensions.

A grid-T file that successfully delhaloed:
netcdf nemo_bm978o_1d_18500101-18500111_grid-T {
dimensions:

axis_nbounds = 2 ;
x = 362 ;
y = 332 ;
nvertex = 4 ;

A grid-U file that successfully delhaloed:
netcdf nemo_bm978o_1m_18500101-18500201_grid-U {
dimensions:

axis_nbounds = 2 ;
x = 362 ;
y = 332 ;
nvertex = 4 ;

A grid-T file that did not delhalo:
netcdf nemo_bm978o_1m_18500101-18500201_grid-T {
dimensions:

axis_nbounds = 2 ;
x_grid_T = 362 ;
y_grid_T = 332 ;
nvertex_grid_T = 4 ;
deptht = 75 ;
x_grid_U = 362 ;
y_grid_U = 332 ;
nvertex_grid_U = 4 ;
x_grid_V = 362 ;
y_grid_V = 332 ;
nvertex_grid_V = 4 ;
deptht300 = 35 ;

It looks like the dehalo script wasn't able to identify the x and y co-ordinates correctly of the 'lm grid-T' files

comment:5 Changed 10 months ago by m.couldrey

Wow, good spot. That's really peculiar. I'm really surprised that the naming conventions of the files is different in this output compared with previous suites like u-bh150 and u-bi909. This certainly isn't something I've done knowingly. I'm looking into why this is, I can't think of a good reason why the output is labelled up differently like this. The field_def.xml and iodef.xml files that manage the formatting of NEMO's output are basically the same across these suites, so some difference is coming in elsewhere. I'll keep looking.

Is the best way forward to add support for these odd axis names? Or to change the axis names in the output for this suite?

comment:6 Changed 10 months ago by m.couldrey

Ok, this won't help solve the problem but I realise why this has happened now. I saved some extra diagnostics to the grid-U.nc and grid-V.nc files in this run (but not the other runs). For some reason that I don't understand, those extra diagnostics have also been output to the grid-T file, which I did not expect to happen. I think that the NEMO I-O system then adds "grid_U/V/T" to every field in the grid-T.nc files to clarify which variables are on which grids. Normally it doesn't do this, I guess because every field in a grid-T.nc file is on the T-grid. Annoying, because this issue will affect another experiment that is currently running.

If it helps, one solution could be to simply delete the offending grid-U and grid-V variables in the grid-T.nc files, then rename the axes names to simply x and y, nav_lat, nav_lon. This would all be possible using NCO tools.

comment:7 Changed 10 months ago by jjabram

  • Resolution set to completed
  • Status changed from new to closed

Hi Matt,

I hope you've got this all back under control again. I'm going to close the ticket now after our conversation last week. If you need anymore help please make another ticket or come and see me.

Cheers,

Joe

Note: See TracTickets for help on using tickets.