Negative Tracer Values#2

Unified Model
1

I got a failure in running an atmospheric simulation:

(ModelOptimisation) tetts@ln03:/work/n02/shared/tetts/OptClim_UKESM> more /home/n02/n02/tetts/cylc-run/opt_dfols46/dn00h/work/20110901T0000Z/atmos_main/pe_output/dn00h.fort6.pe262

???
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 2
? Error from routine: UKCA_MAIN
? Error message: ERROR in UKCA_TRACERS_COPY_TO_UM: negative tracer values
? Error from processor: 262
? Error number: 42
???

The model run started at timestep 25920 (2011-09-01 00:00) and last time step in the output is 28317 (2011-10-04 07:00)

I had a look at the 2022 request ( "negative tracer values" ) which suggested setting BITCOMP_NRUN=False.

But I already had BITCOMP_NRUN=false set. I assume that case does not matter, but nevertheless made the case change, reinstalled the suite and triggered the failed atmos_main. Job is in the Q and I’ll update this topic when it runs, either to report failure or success.

The simulation is a perturbed physics case running with a +4K SST which I imagine can push the chemistry into odd places…

Simon

2

And that attempt failed. So I then followed the instructions for perturbing the dump at: https://code.metoffice.gov.uk/trac/moci/wiki/tips_CRgeneral#Restartingifthemodelblowsup

I had to do some digging to find out where pertrub_theta.py is. It is at $UMDIR/scripts/perturb_theta.py. Could the documentation at https://cms.ncas.ac.uk/tools-and-utilities/ be expanded to document all the utilities in UMDIR? (or perhaps a pointer to $UMDIR/scripts & UMDIR/bin).

Simon

3

And indeed perturbing the cycle start dump got the run to work.