Skip to content

Roadmap

Teagan King edited this page Dec 4, 2023 · 3 revisions

Welcome to the CESM3 Diagnostics Roadmap!

Below are our development milestones.

Goal: Minimal diagnostic generation for CESM3 development

  • Produce diagnostics from all working groups
  • Atmosphere (including chemistry)
  • Ocean (including BGC)
  • Sea-ice
  • Land and river
  • Glacier
  • Diagnostic Output
  • Plots and tables
  • Comparison with obs (requires regridding, provided by ESMF?)
  • Comparison with one (or numerous) other CESM cases
  • HTML output page
  • Feature-complete API
  • Well-defined environment (available on casper & derecho)
  • Can run inside CESM cime workflow (easy to have CESM run package as part of case.submit)
  • Can run outside of CESM workflow
  • Users can write & run one-off diagnostics scripts
  • Well-documented examples of how to run (external API)
  • Migrate some ADF/NBscuid functionality to GeoCAT (.viz or .comp)?
  • Mechanisms for feedback from developer community (issue tickets & feature requests)

Goal: Diagnostics for CMIP PI run

  • Port remaining NCL-based code to python (possibly in GeoCAT)
  • Include additional post-processing steps from other projects
  • Time series generation is separate effort from Brian Dobbins and CISL
  • Data compression (part of CISL time series generation efforts?)
  • CMORization is coming from Gary S (?)
  • Remapping (ESMF)
  • Needed for comparing to observations
  • Might help with plotting (CAM-SE grids are unstructured)
  • Climatology generation (could be moved to milestone #1 if needed sooner)
  • Improve extensibility / add specific diagnostics that are needed but not available
  • Include outside packages such as ILAMB, MDTF, CVDP, ESMValTool, DOE extremes packages, Met Plus, MOM6 tools, CMAT, Metrics packages, etc.
  • Packages not needed for PI run would be incorporated later

Goal: Diagnostics for CMIP DECK experiments

  • Generate figures for CESM3 paper based on PI control

Goal: High resolution /ensemble support

  • Would require better parallelization of individual steps
  • Support for initialized prediction/large ensembles