HOWTO setup and Run a UM expt

see also the UWERN UM support pages, the CGAM UM help pages and Rowan's stuff
Why not join & mail the UM email list for further help?

Notes on the move to NEWTON

Mainly, what you need to do is to change a few switches etc to get an old turing job to run on Newton. There are example jobs on the dept_system; do a search_filter for the user umui Then select one of the Newton example jobs and one of your old turings jobs. Do Job_difference to work out which switches and things were changed.

Don't have access to the dept_system? Check out the example basis files page.

Also see the Newton pages for information about switching to newton.

You may also need to run a slightly different um setup script. ie for NEWTON:

      /ohome/lrum/newton/setvars_4.5
and for GREEN etc
      /ohome/lrum/setvars_4.5

Notes on jobs on GREEN

Green has no interactive login, so you have to (compile) and submit jobs from WREN. Under
      submodel independent
	               |
		       job submission, resources and re-submission pattern
		                				 
Set run job on this host to green. Set use special command for compilation to bcompile

Now, umsubmit to wren. The job will then be passed onto green.

Alternatively look at the Green HadAM3 example job (xbepb) under the user umui on the umui.

NOTE

: If you're having problems gettinga job to run , it could be that you haven't been added to the Green UM group -> check this with Lois!

Using the UMUI and copying an existing Job

Start the UMUI using
umui
Click on Search - Filter
Type the name of another user in the Owner box. Click Filter.
Copy the experimt by selecting the an experiment (highlights in blue) and click Experiment-Copy. then Click OK.

Enter a name for the Experiment in the requestor, and click OK.
Return to your experiment by using filter, as above, with you username.

Altering parameters

The first parameters to change in your copied expt are the personal details. Click on the folder icon of your new expt to open the expt. Click on one of Jobs contained in the expt. then click File - Open Read/Write.

Click on Model Selection - User Information ... - General Details
Change the Target Machine User-id, the Mail-id to your own. Click Close

Choosing output variables

You'll probably want some output from your model run! This is done via the STASH Submenu. This can be somewhat confusing. Again - it's helpful to look at an existing job, to work out what's going on here.

You can find a full list of Possible output variables and their associated STASH codes here.

Create an SST Ancillary file to force the atmosphere

If you have an SST file (full values, not an anomaly file) in netCDF then you can simply: this will create a sst.sst file - a PP ancillary file suitable for forcing the model. To use this in the model open the UMUI - select you job and change Atmosphere - Ancillary and input ... - Climatologies and potent ... - Sea Surface Temperatuyre. Enter the name & location of the sst.sst file you have just created. Click on CLOSE.

Processing and Submiting the Job

Click on the PROCESS button at the bottom of the Job window. The files will then be saved in ~/umui_jobs/ in a directory with the relevant job name.

To submit a job you first need to create a ~/.umrc file. Copy Jeff's example file from ~jeff/.umrc.example to ~/.umrc

Change the PASSWORD to your password on the relevant machines. Change jwc to your username on the relevant machine. Then protect the file by doing

chmod 600 ~/.umrc
Now, to submit a given job you need to run eg:
~jeff/bin/umsubmit xbaqa
where xbaqa is the jobname You should get a message like:
To start um job xbaqd, logon to machine turing.cfs.ac.uk and issue command: 

       ksh $HOME/umui_runs/xbaqd-175173936/SUBMIT
Now follow the next section:

Running a Job on Newton

ssh to newton.cfs.ac.uk

type groups
if you do not see 'um' then you need to get someone (eg lois) to add you to the um group.

Ensure that you have the line

. ~lrum/newton/setvars_4.5  #  set up UM 4.5 environment variables
in your .profile Now issue the command noted above

You can monitor the progress of your jobs with qstat -a (on Green/Wren use: bjobs -a)

Continuation Runs

When you (or at least when I) set a model running, it only runs for the first Year and then halts. This is to allow you to check everything is running OK. Once done you can now procede with a continuation run. TO do this: The model should now start from the end of the first year and run to completion.

Model Output

When the model finishes it writes a .leave file in ~/um/umui_out/
'grep ATMOS filename.leave' shows the number of timesteps completed. The atmos timestepis half an hour. Hence 4320 timesteps is 90 days.

output files live in $DUMP2HOLD/um/


    restart files and pp output are directed to $DUMP2HOLD/um/

Channels A, B, C and D are used for the atmosphere, E for the slab ocean and F, G and H for the coupled ocean. The channels can be described as follows:
A Daily upper air fields
B Daily surface and near surface fields
C Monthly mean upper air fields
D Monthly mean surface fields
E Monthly mean slab ocean fields
F Monthly mean ocean potential temperature
G & H  Seasonal Mean Ocean Fields
Brief summary of actual fields is:

   pa - daily height and temperature on four pressure levels
   pb - daily PMSL and surface temperature
   pc - monthly mean u,v, omega, height, temperature, diabatic heating
	  on four pressure levels.  Plus PV on four isentropic levels and
	  surface pressure
   pd - monthly mean surface fields: lots, including surface temperature
	  windstress and surface fluxes.
   pe - monthly mean slab ocean fields including temperature and ocean
	  heat flux convergence

Data volumes:  One year of data requires approx:
   pa - 81Mb; pb - 20Mb; pc - 10Mb; pd - 15Mb; pe - 3Mb
(from Rowan's pages)

Use ~lrum/bin/uminfo filename to examine file in $DUMP2HOLD/um/

use xconv to view

CRACERU* files list the outpur files. All files are UM files (not pp)? Some files are TAR'd others are not.

Stash Codes

You can find out what stash codes mean by using /home/dan/bin/stashc

Did it work?

Did your run work? You probably want to read this first. And then look at the .leave file in ~/um/umui_out

Archiving Data

You may want to read The UGAMP documantation of Automatic Archiving

Data is archived to /hold/year/lrdlrh/um_archive/

Index files: /ohome/lrum/index/

Extracting fields from the archive

I suggest using getf

Cancel a Job!!

do
qstat -a
then use the JOB_ID_NUMBER from that to do:
qdel -k JOB_ID_NUMBER
See the csar pages for more detail.
Page updated on