-
Notifications
You must be signed in to change notification settings - Fork 64
CCPP Framework Meeting Minutes 2019 04 18
Dom Heinzeller edited this page Apr 19, 2019
·
3 revisions
EGU 2019
- Library of psychrometric conversions: https://github.com/psychrometrics/psychrolib
- Model post-processing verification system: https://github.com/metoppv/improver
- Potential new "customers"/users of the CCPP: KIM, AIR Worldwide's new wavelet model (https://www.air-worldwide.com), maybe even ICON for RRTMGP (piggyback on their GPU port of it)
CCPP users, support:
- How do we handle increasingly large groups of users of the CCPP, how do we handle a growing number of collaborators
- DTC and NCAR in general has many years of experience in this area, but it's new for NOAA
- CCPP is not something that can be used by itself, i.e. using it already means doing some development
- Key question is to educate users how to find standard names, ...
- How do we manage going through all these updates and changes when the user base is constantly increasing
- transparency on the plans we have
- not trickle the changes every month, but bundle them (for example when releasing code)
- who has to fix code that breaks because of changes coming in from someone else (e.g. fractional landmask -> RUC LSM)
- Regression tests! Once CCPP is in the master and we have regression tests, then these won't pass -> developer is responsible
Discussion of issues around CCPP principles:
- How to handle "bunch of boxes" instead of a "horizontal loop extent"
- Is vertical coordinate a fundamental property?
- vertical_coord as new keyword in metadata
- vertical layer (midpoint)
- vertical level (interfaces)
- surface data?
- CF convention: puts vertical information in standard name, this has less flexibility as using attributes (at_2m, downward_flux)
- how about other declarations that we are using now (_for_diagnostics, _on_physics_timestep)
- NUOPC's field dictionary - discussions still ongoing, prototype available
- we wouldn't have to support, but would be able to contribute
- as of now, using yaml format for their dictionary
- plans for backwards compatibility (can add attributes as needed, but should work without these)
- national ESPC call on Mondays at 1pm (...)
- Grant to join these weekly telecons?
- goal of this group is to increase interoperability of models
CCPP prebuild version of multi-suite static build - standard for suite name and suite definition filename: <suite name="XYZ" ... /> --> suite_XYZ.xml
Alex's request to "augment" our parser to handle Fortran 2003 standards (for typedefs) - New metadata parser can handle this - Call f_typedecl when encountering a type declaration --> then use whatever we need from that object