Opened 2 years ago

Closed 2 years ago

Last modified 18 months ago

#2742 closed help (fixed)

Nesting suite in free-running mode

Reported by: peterh Owned by: willie
Component: Nesting Suite Keywords: start data
Cc: Platform: ARCHER
UM Version: 11.1


I'm testing out the nesting suite on ARCHER, following the instructions

This all worked fine, but now I'd like to switch to a free-running mode. I've ticked 'FREE_RUN' in the 'cycling options'. However, the suite fails at
install_glm_startdata for 20150324T0000Z which is the day after the starting conditions which are 20150323. Can anyone explain what I'm doing wrong please?
Many thanks.

Change History (18)

comment:1 Changed 2 years ago by grenville

Hi Peter

Is this still a problem — if so, please tell us the suite id and machine it's on.


comment:2 Changed 2 years ago by peterh

Hi Grenville,
It's u-bf065 on ARCHER.

comment:3 Changed 2 years ago by grenville


/home/n02/n02/ggpoh/cylc-run/u-bf065/log/job/20150324T0000Z/install_glm_startdata/01/job.out says:

Source /work/y07/y07/umshared/um-training/NS2018/2015032400_glm_t+0 does not exist

because the value of FREE_RUN is set to false in the Optional override file. Edit opt/rose-suite-ncas.conf:FREE_RUN=false, to say true.


comment:4 Changed 2 years ago by peterh

Grenville, I've made that change, cleared out the cylc directory on ARCHER and I'm getting the same error.

From my limited understanding(!), I think I need new initial conditions for the global model for this next cycling period? Only the 20150323 are available in the Nesting suite tutorial example I'm running from (/work/y07/y07/umshared/um-training/NS2018/). Do you know if there are further timesteps on ARCHER somewhere already?

Many thanks.

comment:5 Changed 2 years ago by grenville


I don't see the same error - it looks like your suite had a communication error, see /home/peterh/cylc-run/u-bf065/log/suite/log

is your ssh agent still running on PUMA


comment:6 Changed 2 years ago by grenville

Hmmm - something going on here which I don't understand (other than ssh)


comment:7 Changed 2 years ago by peterh

I tried re-running the ancil generation and then re-running the suite in free-run mode. I'm not sure whether this will help though?

comment:8 Changed 2 years ago by peterh

This gives the same error for 'install_glm_startdata'. I'm suppose this means that I need these extra external global fields at all timesteps that the nesting suite is set to run for?

comment:9 Changed 2 years ago by grenville


I have been confused by this though misunderstanding what FREE RUNNIG means. You're right - you will need more global start files. We don't keep them, but you can request some though us.


comment:10 Changed 2 years ago by peterh

Excellent, OK can I request them here? If so, eventually I'd like to run from say July 15th to September 15th 2016. Thanks again.

comment:11 Changed 2 years ago by willie

  • Keywords start data added
  • Owner changed from um_support to willie
  • Platform set to ARCHER
  • Status changed from new to assigned
  • UM Version set to 11.1

Hi Peter,

Do you need daily start dumps at 0Z for the whole period? MASS is down at the moment. I'll start to retrieve these when it's back up.


comment:12 Changed 2 years ago by willie

That's about 60 files at 25GB each, so you'll need about 1.5 TB storage just for these.


comment:13 Changed 2 years ago by peterh

My work directory is 0.75 TB … what's the best way to do this? Can they be stored on the RDS?

Also, eventually I want to run a nesting suite with different solar insolation. Do you know how I could generate the 0Z dumps for this? Is there a global setup I can run on ARCHER?

Many thanks.

comment:14 Changed 2 years ago by willie

Hi Peter,

If you want to do a complete run through the start dumps then you will need them all to be on /work. So you need to estimate how much space the run will take. You can do this by scaling up a shorter run. Then put in a ticket asking for more quota on ARCHER if necessary. After the run is complete you can move the data onto the RDF, if you have no further use for the compute nodes.

I don't know about the solar insolation. In the model UM Science Settings → Planet constants you can change the solar irradiance at 1AU (currently set to 1361 W/m2), but the other changes that might be required by this are not clear to me.


comment:15 Changed 2 years ago by peterh

Is there a suite and STASH I copy to run to generate the required global dumps? I think I know how to alter the insolation as we have done this with GC3 previously.

I think I will start with a 5 day run to estimate the space requirements in the first instance.


comment:16 Changed 2 years ago by willie

Hi Peter,

You data is now ready on the RDF in /nerc/n02/shared/wmcginty/data/2742/. Please copy it to your own workspace as soon as possible. You will need to copy this data onto /work in order to run the UM.


comment:17 Changed 2 years ago by willie

  • Resolution set to fixed
  • Status changed from assigned to closed

comment:18 Changed 18 months ago by peterh

An update based on a message from Stuart Webster:

"So FREE_RUN = true and CYCLE_INT=24 means that you will indeed be rerunning the global each day. This means that the inner model keeps freely evolving, but the LBCs are never more than 24 hours from reality (where reality here is the global model analysis). This also means that the filenaming never gets beyond 024. Effectively this is similar to running a limited area climate model, forcing it with analyses (but with forecast fields between those analyses to give a higher resolution temporal updating of the LBCs).

So you can either do that, or you can set FREE_RUN to false and set INITIAL and FINAL cycle points to be the same, and then just set the run length (in the model resolution window) to whatever length in hours that you want. Then you will have the global model running as a single long forecast and the inner model running inside that same long forecast for the duration. Doing this will mean that the global model will drift away from reality, but perhaps that is what you're after (effectively then downscaling a global climate-like run)."

Note: See TracTickets for help on using tickets.