Hi everyone,
I have spent a lot of time since the CMIP6 Hackathon working on cmip6_preprocessing - a package to facilitate complex analysis workflows on the vast amount of CMIP6 data available in the cloud.
@rabernat and I have recently put together a (citeable!) notebook demonstrating the capabilities of cmip6_preprocessing and xgcm together, and I am extremely excited about these two tools working together.
Here I would however like to mention an aspect of this workflow that does currently require many hoops to jump through: Most CMIP6 models do not provide a full grid description (metrics in xgcm terminology) as part of the output. For many models the cell area of the ocean tracercells areacello and the vertical thickness thkcello are given, but other metrics like horizontal distances and areas for other staggered grid points are usually not available. We have ways to approximate them, but run into problems in particular around the Arctic (for examples see the end of the notebook).
I think this is a problem that the pangeo community is uniquely qualified to tackle.
Lets collect the ocean grid files for many (perhaps all) CMIP6 models and put them into the cloud!
I would like to start this off by a role call:
Who here has access or knows someone who has access to ocean grid files of any of the CMIP6 models? If you could reply with the particular model in this thread, I can put together a mailing list and see if we can get the permissions from each of the modelling centers to use that data and then see how we can upload it to the cloud (perhaps using pangeo-forge? @jhamman).
3 Likes
@Thomas_Moore do I recall right that you have the ACCESS files?
Hey @jbusecke - ACCESS is a broad umbrella for the Australian modelling community. Iām one step removed from their generation but can both confirm with others and document as best as possible the grid info for the ACCESS submission to CMIP6. I assume that is your goal here?
Yes that would be awesome! Ultimately I need the actual file(s) with values, but for now I just wanted to see if there are a bunch of folks on here that could help getting to these files (and eventually permissions).
1 Like
I can help with GFDL models. I also work with some folk at the UK MetOffice, so can enquire about getting the Hadley Centre model data.
2 Likes
Excellent! Thanks @gmacgilchrist.
I just wanted to ping this issue. I think that getting the raw static grid files from as many models could help cloud based CMIP workflows tremendously.
Here is a relevant issue in cmip6_preprocessing, which would very much profit from this information.
Does any of @Thomas_Moore @gmacgilchrist @raphaeldussin have one of these files (matching a CMIP6 model) on an ftp server? If that is the case this might be another nice application for pangeo-forge to get these into cloud storage? cc @cisaacstern @rabernat
1 Like
Happy to work on recipe development if you locate the right source file, @jbusecke. Your issue for smap-sss was a great guide, another issue like that is most welcome 
1 Like
@jbusecke thanks for all your great work - these grid-metric files should be complete, transparent, and have provenance for all models. Iām not (yet) working with CMIP6 and Iām also not a āprimary authoritative sourceā for this information in the ACCESS community. But I will have a chance to bring this up next week with key folks and will do so.
That is super helpful! Thanks
1 Like
This is continuing a conversation started over at cmip6_pp: Calculating unprovided grid metrics from CMIP output? Ā· Issue #237 Ā· jbusecke/cmip6_preprocessing Ā· GitHub
@jbusecke - Thanks for starting to collect CMIP6 ocean grids on the pangeo-forge! Iām curious what the status of this effort is and how I might plug in to help? From poking around the cmip6_static_grids-feedstock page on the pangeo-forge and its associated github, it looks like there are a bunch of grids available or soon-to-be available. Iāve not used the pangeo-forge before, but this effort does align with my current project, so I can spend some time helping before the next semester begins.
2 Likes