Running nudged AMIP UM13.9 past 2014

Unified Model
1

Hello,

I am running a nudged AMIP vn13.9 suite (u-du013 copy of u-ds836). It has run through to 2014, but I now want to run it to 2015 to get a dump file for 20150101 to initialise a free-running CMIP SSP2-45 simulation from 2015 to 2080.

I was wondering what ancil files need to be changed to get the suite to run past 2014 to 2015, or to be pointed in the direction of instructions if they already exist.

Thanks!

2

Hi Isabelle,

As far as I can see most of the ancillaries (even those labelled as 2014) have buffered data into 2015. The CMIP6 historical period ended in Dec-2014 so all the data was set up to enable models to run till then.
So, you can just try increasing EXPT_RUNLEN by an year and doing a ‘cylc vr’ for the suite.

Mohit

3

Hi Mohit,

I initially tried running from 2014 for 13 months with the current suite set up, but it failed in 2015 with a BICGstab error which I assumed was a result of some ancillaries because it happened at 2015 after not having any issues for the 14 years prior.

I’ve just tried that again and will see if it fails with the same error.

4

Ah ok, then most likely cause is the namelist:clmfcg entries which will need an additional year for all GHG species.

5

Actually, the run3/work/../atmos_main has failed due to input NetCDF files not being 365-day compliant. I believe these will be the easyaerosol ones, so you can access the 365-day version from /common/share/monsoon_ancils/atmos/GC5/n96e/easyaerosol/cmip6_stratos/

6

Hi Mohit,

That was from a run where I tried making the same adjustments suggested for the future simulations. I have since reverted the suite back. Could you tell me which ancils need to change from the default nudged AMIP vn13.9 suite?

Thanks for your help!

Cheers,

Izzy

7

I think you can try running for 12 months from Jan-2014 without changing anything as the ancillaries should have data for January 2015 at least.

If you try to run for Jan 2015 then you might go beyond the data in the easyaerosol files as well as the clmchfcg namelist I mentioned above.

8

Hi Mohit,

Running for 13 months (to get the 2015 dump file) results in the BICGstab error which is why I tried to change the ancillary files. Running for 12 months doesn’t seem to output a 20150101 dump file.

9

Hi Hannah,
The 20150101 file will be output in share/data/History_Data -just not archived due to a postproc vs Cylc-8 quirk.
The fix for this has been added to the base UKESM suites, so you can either obtain the file from share/data, or implement the fix to allow the file to be archived.

https://code.metoffice.gov.uk/trac/roses-u/changeset/343015/d/r/7/9/9/trunk

10

Hi Mohit,

Thanks - I’ll try that! One more question about this dump_file. To use it as AINITIAL in my free-running future simulation (in suite u-ea318 copy of u-dr682 ), I can’t find where I should change the start dump file. There doesn’t seem to be an AINITIAL in the rose-suite.conf file.

Thanks again.

Cheers,

Izzy

11

In this case the file is specified in app/um/rose-app.conf: [namelist:recon_technical]
ainitial=

12

Thanks! It’s now giving this error while running:
???
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 30
? Error from routine: RCF_RESET_DATA_SOURCE
? Error message: Section 0 Item 250 : Required field is not in input dump!
? Error from processor: 0
? Error number: 15369
???

I am using a strattrop vn13.9 start dump file for ainitial

13

The field 0-250 is CO2 Ocean flux. I think this is required when running a coupled model (even in a non-interactive mode). You could try to specify 0-250 as an additional field for reconfiguration to initialise (see exercise in UKCA tutorials ) and using the fixed value from the radiation namelist, but I suspect you will hit other mismatch in fields while initialising the coupled run from an atmos-only dump.

14

Is there a coupled dump at vn13.9 that can be used to initialise from 2015? I’m not sure where to find start dumps available for different model configurations.

15

I cannot say off hand - especially as this is UKESM1.1 and I do not think there have been many ‘science’ runs made with this configuration centrally. You might have to contact someone in the UKESM team, or find out if anyone has carried out full historical or ‘future’ runs.