station file

Discussion about tangent linear and adjoint models, variational data assimilation, and other related issues.

Moderators: arango, robertson

Post Reply
Message
Author
User avatar
jivica
Posts: 169
Joined: Mon May 05, 2003 2:41 pm
Location: The University of Western Australia, Perth, Australia
Contact:

station file

#1 Unread post by jivica »

Dear Hernan,
just wondering if you changed IO for station files within RPG?
I am using almost ;) latest version and saw that there are multiple time series with outer loops outputs (?) inside station file (actually only missing values are written).
Is there cpp option or whatever switch to turn that off so we have smaller station files with only one increasing ocean_time from the last NLM sweep?

Regards,
Ivica

User avatar
arango
Site Admin
Posts: 1347
Joined: Wed Feb 26, 2003 4:41 pm
Location: DMCS, Rutgers University
Contact:

Re: station file

#2 Unread post by arango »

What is RPG? Did you mean RPCG option for RBL4D-Var minimization algorithm?

I have never turned on the STATIONS option with the 4D-Var algorithms. I suppose that we can use the split 4D-Var algorithms that I released yesterday to only run the STATIONS in the background and analysis 4D-Varphases. And not in the increment phase where the inner-loops minimization is taking place.

Look at the test repository for WC13/RBL4DVAR_split and notice that there are two templates for roms.in: roms_da_wc13.tmp and roms_nl_wc13.tmp. We can have different standard input parameters and executables. Spend some time learning how the split 4D-Var scheme works.

User avatar
jivica
Posts: 169
Joined: Mon May 05, 2003 2:41 pm
Location: The University of Western Australia, Perth, Australia
Contact:

Re: station file

#3 Unread post by jivica »

Thanks Hernan for your swift response.

Yes, I am thinking about rbl4dvar. I thought there is easier way to handle/call station file output (If statement last NLM sweep then activate station output).

However, if ROMS is going into SPLIT architecture then I'll spend some time to familiarize.
Saying that I went through WC13 test RBL4DVAR_split. It seems that you are moving a lots of stuff into SLURM/scripting.
This might be tricky as some users are using cluster based on SLURM other Torque/SGE with all bash/sh/zsh/csh etc.
If it is possible, could you please copy/paste your actual SLURM script you send to cluster queue (i know there is script to create it submit_split_rbl4dvar.sh but it is customized to your cluster and file locations).
For example I am doing all my pre/post-processing in the scripts that are outside SLURM (front node which has python/NCO etc) while actual integrations are done within SLURM/nodes.
From your actual script I should be able to modify my pre/post scripts and if needed used dependent SLURM stacking etc.

Thanks!
Ivica

User avatar
arango
Site Admin
Posts: 1347
Joined: Wed Feb 26, 2003 4:41 pm
Location: DMCS, Rutgers University
Contact:

Re: station file

#4 Unread post by arango »

Nope, that bash script supports both SLURM (BATCH=1) and non SLURM (BATCH=0). You just need to change that flag. See the customizable section. I can run that script in my 14 CPUs Mac or in our SLURM supercomputer. The script is very generic, compact, and complex. It is well documented. You need to study and understand it. Also, check the DRYRUN switch. Study every line of the tunable parameters. You will be surprised that everything is there and very trivial to run. The coupled 4D-Var script is similar but more complex because it has stuff for ESMF/NUOPC and the atmosphere model. This stuff is not for amateurs. I am sure that you can handle it.

The BATCH flag can be expanded to include other queuing posabilities.

Post Reply