Description
Currently the ADF is capable of comparing model simulations to certain reanalysis and observational datasets, which are accessible by all ADF users, at least when running on Casper. However, right now all of those observational datasets are under people's personal directories, for example here:
/glade/work/nusbaume/SE_projects/model_diagnostics/ADF_obs
I suspect that this strategy will be difficult to maintain long-term, especially as other non-ADF diagnostics are brought into CUPiD that need their own specialized datasets.
Given this, should there be work to identify a common location where all of the CUPiD-relevant observational datasets are stored? Along those lines are several things that should probably be discussed at some point:
- The observations directory should be globally readable, but who should have write access?
- How should these files be organized? Should it be a flat directory structure, or should there be subdirectories?
- What, if any, metadata should we require "official" observational data files to have?
- Should the files be backed up somewhere, or at least the scripts that were used to generate the files?
- Should the data located in this directory be accessible outside the NCAR machines? This would probably only matter if someone was trying to run CUPiD on a machine that didn't have access to the glade filesystem.
I assume it will take multiple group discussions to figure all of this out, but I just wanted to open this issue now, as it may impact how at least the ADF is integrated into CUPiD.