AM3 access3-configs suite failing when set to use non-default aerosol options

I’m trying to use AM3 from the access3-configs github to test the cost of different aerosol options (in order to guide decisions made at the Centre’s high-res AM3 project), but setting anything other than the default option is resulting either in the run failing with an error, or not completing any suite tasks beyond fcm_make_um.

  1. If I run access3-configs out of the box (where the default is fully prognostic aerosol, prognostic in suite conf => Model Configuration => Aerosol Configuration), I get a successful standard run.

  2. If I change the aerosol option to aeroclim in suite conf => Model Configuration => Aerosol Configuration, the run fails with the following error:

???!!!???!!!???!!!???!!!???!!!       ERROR        ???!!!???!!!???!!!???!!!???!!!
?  Error code: 54921
?  Error from routine: PREPARE_FIELDS_FOR_RADAER_MOD:PREPARE_FIELDS_FOR_RADAER
?  Error message: Outfield contains negative values at:
?        item: 921 section: 54
?  Error from processor: 1226
?  Error number: 31
?  Warning from routine: ANCIL_CHECK_MOD::REPORT_ANCIL_ERRORS
????????????????????????????????????????????????????????????????????????????????
  1. If I change the aerosol option to aerosolclim_chem, the suite doesn’t run at all. It looks like it gets through the fcm_make_um step, but although there are no errors recorded in the fcm_make_um log files no further suite tasks run.

The three cylc directories are in /scratch/fy29/bw6466/cylc-run

@bethanwhite , as you likely know, ACCESS-AM3 has not been released yet. We have simply given access to a few people to the configuration we use for development and the codes with no guarantee of support. For this reason, I am tagging this request with outofscope for our own support triage. But it is something we have an interest in solving during the development of ACCESS-AM3 configurations.

In the development of ACCESS-AM3, we haven’t yet changed the aerosol scheme (or any option in the UM or JULES, I believe), so I am not surprised it does not work straight out of the box.

We cannot look into this issue right away as the final push for ESM1.6 is of higher priority than furthering the development of AM3. I have, however, copied your issue in a GitHub issue for AM3 configurations so we will remember to look into it. And we will report back here when we have made some progress.

That said, this topic will stay open, and other people in the community might have expertise with the aerosol schemes and might be able to give some insights here.

It looks to be a problem with the input files. It is possible that different aerosol schemes need different input data, but I don’t know more than that. If the schemes existed in previous UM versions, maybe it would be useful to look at CM2 suites using these aerosol schemes and see if the input data is compatible.

2 Likes

Thanks @clairecarouge!

Hi Bethan, Claire,
Turning off the prognostic aerosol will require (scientifically at least, but ideally also technically) prescribing some substitute quantities, usually cloud droplet number and some aerosol optical depth parameters, to the radiation scheme. This is not something I’m familiar with, though I believe it is technically possible.

1 Like

Thanks Matt, that is actually very helpful information.