Review of recipe_GCM_and_RCM_ALP-3.yml

[nhsavage]

as a first task in QA of the work here, I will start with the recipe_GCM_and_RCM_ALP-3.yml recipe file. I note that some aspects of this are planned to change with the refactoring to a single recipe but I wanted to start somewhere

[nhsavage]

Some questions:

• Why is project commented out ?
• Where models are commented out - would it be better to have a list of which not used and why?
• Do you still plan to use this model ? TODO Need to find IPSL driven WRF331F run
• Models not used to to missing FX files - are these needed now - I thought the regrid meant they weren't? Or am I missing something
• All regrids seem to use linear - is that ok for precip?
• ALP3 - copied in lots of places in the ALP-3 model domain sections, can' t region_name: be defined at the top in an anchor and used here as well as later?

[nhsavage]

@thomascrocker

[thomascrocker]

Some questions:

* Why is project commented out ?

The project tab at the top is a metadata tag used in recipes for provenance generation, see https://docs.esmvaltool.org/projects/esmvalcore/en/latest/recipe/overview.html#recipe-section-documentation but.. to work properly the project needs to be defined in a config-references.yml file. I hadn't gotten around to doing that when I started the recipe, and am not massively bothered by recording provenance info right now. However I've just seen that in the latest release "EUCP" is a project included in the file now, see: https://github.com/ESMValGroup/ESMValTool/blob/6c57a55267e8c10c48cce96360f5ffe3ac7554c6/esmvaltool/config-references.yml#L684

* Where models are commented out - would it be better to have a list of which not used and why?

Probably, this is something I need to discuss with Carol ultimately the models used is kind of arbitrary I think initially I just grabbed a list of common CMIP5 and CMIP6 models from a paper or two and then commented out the ones I couldn't get working, with reasons where I could work that out.

* Do you still plan to use this model ? TODO Need to find IPSL driven WRF331F run

Given the time left to work on this probably not... When this was originally written I was hoping to replicate analysis from the Boe paper mentioned in the comments at the top. The IPSL model mentioned in the comment and the paper is not present on ESGF.. but it's not been my priority to get hold of it. I do have the majority of the other models that that paper used.

* Models not used to to missing FX files - are these needed now - I thought the regrid meant they weren't? Or am I missing something

It could be, ultimately there must have been some operation I was trying to do that ESMValTool was complaining it wouldn't do because of missing FX files. Either land sea masking, or possibly area averaging (they may have been missing coord bounds so it wanted areacella files to help.) I may try uncommenting these to add them back into the models we are using.

* All regrids seem to use linear - is that ok for precip?

I chatted with Erasmo about this a couple of times, he seemed to think that since we were only looking at seasonal mean values of total rainfall, and not extremes that linear regridding was OK and conservative regridding wasn't necessary.

* ALP3 - copied in lots of places in the ALP-3 model domain sections, can' t region_name: be defined at the top in an anchor and used here as well as later?

It could.. Ultimately though the variety of CPM models to be used changes depending on the domain chosen, so I figured seperate files for each domain would be needed (since yaml has no IF style logic). Ultimately though the new pan european recipe makes this kind of obsolete (and I think is now working OK)