Complex Filtering of CMIP6 Model Data

[savannahferretti]

Does anyone have experience searching and loading CMIP6 data from ESGF that has multiple variables and time frequencies?

I am trying to get CMIP6 Historical run data for models with: 1) daily precipitation (pr), and 2) either 3-hourly or 6-hourly instantaneous surface pressure (ps), air temperature (ta), and specific humidity (hus). I want to then output an Xarray.Dataset with a "source_id" dimension, where for each model, the variables present are pr, ta, hus, and ps at the respective time frequencies.

Is this doable? I haven't found a complicated example like this in any of the Pangeo CMIP6 gallery notebooks.

[jbusecke]

Do you want to work with only one time frequency? Or perform analysis on multiple time frequencies and compare results.
In general I recommend using the pangeo cloud CMIP6 data, which is now actually maintained by LEAP (https://github.com/leap-stc/cmip6-leap-feedstock).

I want to then output an Xarray.Dataset with a "source_id" dimension, where for each model, the variables present are pr, ta, hus, and ps at the respective time frequencies.

This generally does not work well, since each of the models can have different grids, thus different shapes in x, y dimensions, but you can use @TomNicholas xarray-datatree package for this.

I made a quick example (fully reproducible on the hub 😁).
Happy to keep iterating on this if you can expand a bit more about what the goal is.

[TomNicholas]

I also think datatree is the right tool for your problem @savannahferretti

@jbusecke, is there are bug in your notebook / datatree? What's going in in the output of the last cell there? I presume that was intended to print the whole tree?

[jbusecke]

Oh yikes something went wrong there. @savannahferretti if you could expand on your use case a bit more Ill take another look at the notebook and try to incorporate your use case a bit more.