Skip to content

2024‐07‐18

Jump to bottom
Teagan King edited this page Jul 18, 2024 · 6 revisions

July 18, 2024

Agenda:

  1. Global / hemispheric / regional reductions
  2. Further plans for x4c + existing scripts

Notes

  • Motivation: Our goal is to share standard methods for global / hemispheric / regional data reductions

  • Starting Points:

    • Relevant CUPiD issue (#45)
    • x4c: https://github.com/fzhu2e/x4c
      • Some features:
        • Xarray extension fundamental features: regrid, get_plev, annualize, geo_mean, plot
        • Advanced features: load, get_paths, check_timespan
        • High-level workflows: gen_climo
      • May need generalization
      • Future plans? Would parts fit into CUPiD or GeoCAT?
    • Keith’s Analysis tools:
    • CVDP currently calls NCL to do some of these tasks; transition to python for SOME CVDP capabilities is in the works?
    • Other tools that exist? Anything else in GeoCAT we could leverage?
      • xCDAT: DOE/E3SM - haven’t used, but also does geospatial weighted and temporal averaging

  • Discussion:

    • Future of x4c & existing scripts and if these tools could fill some needs

    • Define needs (eg, globally averaged temp, NH/SH sea ice area) *

    • Geocat has previously mentioned not having ability to support CESM-specific projects

    • x4c: on pop grid, not mom

    • for other models, would need to map variable names & understand various timeseries directory structures

    • Does GeoCAT have access to data from other models? Examples gallery does include other model data. UXarray has also been working with other models and may have better sample data as a starting point.

    • Can X4C be folded into UXarray (Orhan Eroglu)? This is being developed alongside Project Raijin (John Clyne).

    • xCDAT: DOE (Tom, who is involved in UXarray as well). Orhan mentions this is a very narrowly focused tool (climatology analysis). xCDAT uses structured Grids.

    • Could collaborate between x4c & UXarray on regriding unstructured grids. UXarray handles things in a different way than how x4c extends xarray

    • x4c does regriding from unstructured to structured grids-- there could be collaboration opportunities with UXarray here.

    • x4c allows for temporal operations on, eg, ne30 grids-- this is one of the main things that we will need within CUPiD

    • x4c is an extension of xarray. Feng expressed interest in some community support for x4c.

    • Would bits and pieces of x4c go into GeoCAT/CUPiD, or should we focus on growing x4c in a more robust way? Could potentially bring bits in, but based on the design of the packages, GeoCAT wouldn't really leverage x4c. UXarray operates on different types of objects and could potentially collaborate but uses a different approach. Realistically, GeoCAT can't provide much support for x4c growing as a standalone package.

    • ADF currently incorporates regriding and climo generation from x4c, as an example.

    • CUPiD can run notebooks & python scripts & provide environment, but we want to rely on outside code as much as possible and keep CUPiD lightweight.

    • x4c's API is one of it's biggest strengths; also this is quite specific from GeoCAT perspective & want to focus on things being agile and long lasting

    • Would be great to have an NCAR team that packages tools generated in labs...

    • Could x4c be a hub that pulls in GeoCAT and other packages? Importing the existing supported packages to replace some of the internal bits of x4c.

    • Currently x4c only leverages GeoCAT for p-lev. It's strength is in a unified API.

    • Annualize & regriding would be really useful functions for sea ice as well as regional reductions.

    • Is CGD thinking about putting resources behind x4c? Would want to see on conda-forge, documentation, and testing, and maybe institutional support before bringing in to GeoCAT examples gallery. Seems like funding is not available at this point.

    • OGS call for relevant funding may be worth looking into.

Slides

Clone this wiki locally