Opened 14 years ago

Closed 14 years ago

#52 closed help (fixed)

UM and ancillary files

Reported by: umdoc Owned by: cwang
Component: UM Model Keywords:
Cc: Platform:
UM Version:


Dear all,

I need some help regarding UM and ancillary files.
I have made a sea ice ancillary file using xancil0.40, taking into account the rotated pole. As an input I gave xancil sea ice concentration and sea ice depth but xancil calculated the se ice edge (basically giving every grid point the value -1.0737e+09 ).

I've then attempted to use the ancillary file in a LAM run, using data from a global run as boundaries and as ancillary files, with the exception of the land mask and the orography that were retrieved from BADC.
My job is xckgb and can be found in /hpcx/devt/n02/n02-ncas/gnp/xckgb/ on the hpcx.

Now, when I run the UM the job fails in the first time step:

Atm_Step: Timestep 1
RHS zero so GCR( 2 ) not needed

and then stops when Gathering fields, as most likely no fields to gather.

Do you have any idea what I'm not doing that I should be doing?
Or how I could find out what's going wrong?

My own thoughts were that perhaps there were some incompatibilities between my new sea ice ancillary file and the rest of the variables in the restart file. The land mask is consistent with the sea ice and there is no sea ice over land.
Is the sea ice edge a problem? I'm putting all my sea ice data into the sea ice concentration. Is there any documentation on sea ice ancillary files?
Are there any other variables I should be worrying about?

Any help here would be greatly appreciated.




Guðrún Nína Petersen
School of Environmental Sciences
University of East Anglia

Attachments (1)

ice.png (28.7 KB) - added by jeff 14 years ago.

Download all attachments as: .zip

Change History (8)

comment:1 Changed 14 years ago by jeff

  • Owner changed from um_support to jeff
  • Status changed from new to assigned

Not sure what is wrong here, have you run this configuration before with
different ancils, i.e. has a similar job worked in the past?

xcitb is the same job but with sea ice concentration from the global
dump. The time step is shorter in xckgb at the moment but I get the same
message for the regular no. of timesteps, 288 per day.

I did a cumf between the start dumps for these 2 jobs and noticed the differences are in the sea-ice fields and the orography fields. Your could try only changing one of these at a time to narrow down which is causing the problem.

comment:2 Changed 14 years ago by umdoc

The only difference was that I had switched on the configuration of the standard deviation of the orography. Didn't think it would matter much.

Check box: The ancillary standard deviation of orography to be configured

Job xcitb: Entry is set to 'OFF'
Job xckgb: Entry is set to 'ON'

And it didn't, see new run of xckgb, finished at about 10:41 21.05.

It is clearly my sea ice ancillary file that is making a fuzz. How can I find out why/where things start to go wrong?



Changed 14 years ago by jeff

comment:3 Changed 14 years ago by jeff

I've imposed a land/sea mask over your ice fraction field

and you can see there are some areas which look like they should have ice but don't. Maybe this is causing the problems?

comment:4 Changed 14 years ago by jeff

I also noticed that some of your land points do have very small values of ice fraction on them, maybe that could also be a problem? Using a version of the field with the land mask imposed should fix this.

comment:5 Changed 14 years ago by umdoc

I'm not convinced that the reason the model falls over in the first time step is to do with the ice free areas that look like they should have ice. After all, the ice on the left hand side of Greenland, in the figure above, is identical to the one in my control job, xcitb, and that one runs happily for 72 hours. I would rather have expected the model to blow up … but maybe I'm wrong??

I've tried various versions of the sea ice ancillary file, e.g.
xckgc: I've filled the holes with ice and used the land mask in xancil as you suggested. (I find your comment about sea ice in land points curious as I check all my land points in matlab and make sure that there is no sea ice there. ) → It still fails.
xckgb: I've tried running the sea ice but without using the landmask in xancil and the model fails.
xckgj: I've used matlab to make a NetCDF file with sea ice identical to that in xcitb and the job runs for the length of the test run, 6 hours.

I've fixed my NetCDF file so that the precision is right without having to go through Xconv. Still the job fails.

I'm getting to the end of my list of possible test runs as I have no idea what is the problem.


comment:6 Changed 14 years ago by cwang

  • Owner changed from jeff to cwang
  • Status changed from assigned to new

Hi Nina,

First, I have to point it out that reconfigurating from a global dump without reading in acnillary files almost invariably results in model crashes, most likely in the surface processes routines. There will be problems with the vegetation and soli properties not being consistent with the soil moisture and this results in a model crash every time.

Secondly, you changed the ice ancillary which will likely cause inconsistent with the above mentioned fields. Zeros ice fraction over land (the standard ancillary file will have -1.0737e09) might be another problem as it will think those zeros are legitimate points.

Therefore, the first thing you need to do is to reconfigure your global dump using ancillary following files rather than just reconfiguring ice alone: mask,orog, ice, soil and veg properties.

I hope that will help.


comment:7 Changed 14 years ago by cwang

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