Access to global model forecast data on ARCHER

Unified Model
,
1

Good morning NCAS-CMS,

Hope you had a good weekend!

I’m attempting to re-run a nested forecasting suite focused over East Africa. The run is under the name “u-cu605” and is a copy of James Warner’s (UKMO) “u-ct683”. Originally, the forecast was performed on MONSOON, however I was hoping to test the simulation on ARCHER.

After spending some time porting the simulation onto ARCHER2 (/home/jtalib/roses/u-cu605/site/ncas-cray-xc30/suite-adds.rc), I hit a stumbling block when calling the appropriate start data for the simulation. The “install_glm_startdata” process uses MOOSE to extract the appropriate rerun data, in this case, /opfc/atm/global/rerun/202007.file/20200705T0000Z_glm_t+0.

How I can access forecast start data for the global model on ARCHER? I don’t mind changing the start date, as I’m only testing different model configurations (six hour run-time). Once I’ve got a successful set of configurations, I plan to initialize the forecasts with ERA5 data? I will also use ERA5 for the boundary conditions. Do you have any documentation on creating ERA5 ancillary files for start data and boundary conditions?

To be honest, it may actually be easier to copy a pre-existing nesting suite which uses ERA5 to initialize and set boundary conditions. To begin with, I thought I would copy James’ forecast run as this focuses on the same domain.

Also note - I’m currently using “ncas-training” to test simulation set-up. Is this the best project code to be using at the moment. I am currently writing up an HPC application form.

Finally, is it also possible for someone to check my edits in “/home/jtalib/roses/u-cu605/site/ncas-cray-xc30/suite-adds.rc”? Whilst I’ve got the forecast working so far, it may be the case that certain options have been excluded due to me getting confused with labeling.

Thanks in advance for your help and wisdom,

Kind regards,
Josh.

2

Hi Josh

We don’t keep start data on ARCHER2 - we suggest that users get data from MASS either on Monsoon or through the MASS client on JASMIN (Setting up your JASMIN account for access to MASS - JASMIN help docs)

See Met Office Science Repository Service in general and Met Office Science Repository Service in particular for how to get ECMWF data.

It’s too tall an order to sanity check your suite-adds.rc - easier to run the model and see where/how it fails , but this doesn’t look right:
UMDIR = /work/n02/n02/jostal

probably should be UMDIR = /work/y07/y07/umshared

Grenville

3

Josh,

My mistake:

UMDIR = /work/y07/shared/umshared

4

Can I just check: So I need to extract from MO MASS to JASMIN. Then transfer start and LBC files from JASMIN to ARCHER? Or can I get the UM to read the files from JASMIN?

This is requiring much more data than I envisaged!

Josh.

5

You’ll need to copy files from JASMIN to ARCHER2 – ARCHER2 compute nodes can only see /work.

6

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.