I am trying to a rerun a MetO Cray suite, u-br700, on Monsoon to output extra diagnostics. The suite I am using is u-dm032. I am initialising it from u-br700 start dumps for 2065 with recon off and following the approach of Bit comparability for re-run cycles for bit comparability. Specifically:
Recon disabled
astart set to the desired dump file
BITCOMP_NRUN = true
l_nrun_as_crun = true
I appreciate bit comparability may be impossible due to running on a different machine but the output from the first 3 months of running is very different to that of u-br700.
Could you have a look at my suite and see if I am doing anything wrong with the initialisation?
The results should bit-compare between Meto-cray and Monsoon.
The differences between u-br700 and your suite can be seen here.
While most of these are either technical or moving of ancillary files from ‘temporary’ to the central locations, there are some differences in the ocean branch revised used, as well as some nemo_cice settings. I am not user how much these will affect model evolution.
More importantly, I think dm032 uses UKCA emission files from a CH4-emission run, and while the ‘emissions’ themselves should be ignored in concentration-driven runs, the flux adjustment values could still be added to methane tracer?
Many thanks for your help. I will look at the ocean files. Re the methane emissions, I’m a bit confused as I think u-br700 and u-dm032 are effectively emissions driven runs as they use Gerd’s emission driven branch even though emission driven logical isn’t selected (as I’m not sure it existed at 11.2). Is that right?
Ah, yes. I had missed the CH4 emission files in br700 due to the change in order of the files !
So, as far as I can see the only possibly relevant changes are on the ocean side. How much of a change are you seeing between output?
Looking at the ozone mmr on pressure levels (51001), I’m seeing fluctuations of ~ 10 ppb in the lower and mid stratosphere which isn’t huge but it’s making the signal I’m getting when I compare runs with the control and perturbed methane emissions harder to interpret.
Sorry, one further thing, could I just clarify if there is any technical difference for coupled suites between:
Having recon=FALSE and pointing astart to the desired dump file (in namelist:nlcfiles)
and
Having recon=TRUE, not reinitialising any prognostics, setting ainitial to the desired dump file (in namelist:recon_technical) and then having astart=‘$ROSE_DATA/$RUNID.astart’
Likewise, for AMIP suites,
specifying recon=FALSE and pointing astart to the desired dump file (in app/um/rose-app.conf namelist:nlcfiles)
and
having recon=TRUE and setting ainitial in rose-suite.conf to the desired dump file (again not not reinitialising any prognostics)
I have tended to do option 2 for no particular reason but wanted to check!
Option 2 is not guaranteed to produce the start file used in option 1. The reconfiguration can introduce bit-level differences - you can check by using mule-cumf to compare files.
Thanks, Grenville, and sorry, I should have asked earlier, but which files should I be comparing? Would it be the dump file (which is the input to the recon task) and the astart file which is output by that recon task and is stored in cylc-run?