NDBC long time series NetCDF files

[emiliom]

Opening this issue to follow up on progress and discussions on Craig's NDBC NetCDF files.

To start it off, here's an update from @crisien from a week ago (reminder: agate THREDDS is Craig's test, unadvertised THREDDS server):

I have been working on the NDBC NetCDF files and the agate THREDDS server, see http://agate.coas.oregonstate.edu:8080/thredds/catalog/ndbc/catalog.html

[emiliom]

One minor comment: You converted temperatures to Kelvin. There's really no need. Kelvin is the CF "canonical" unit, but canonical units are not required at all. What's required is that the unit used can be converted to the canonical unit using a linear transformation, and that the units string be specified using udunits. Celsius is equally compliant, and IMHO, far more user-friendly since it won't require unit conversion for most users. That's my 1.5cent, anyway.

[crisien]

OK I will revert back to Celsius.

[emiliom]

Thanks. FYI, I pushed my changes to the repository just now:

• reorganized the repo to make it cleaner and easier to find things
• added the NDBC 46088 file you sent me via email
• added a CDL (ncdump -h) and compliance-checker output for the NDBC 46050 file you pushed to the repo and the 46088 file you sent me via email. These are in the new ndbcnc directory

The compliance-checker score for one of your new, multi-variable NDBC nc files is great, 162/164. But I haven't looked at any details of the report.

I've been thinking about and exploring the structure of your current nc files. There might be issues with the depth implementations, but I'm looking into it. I'll follow up once I feel more confident about what I say; you did implement exactly what we agreed to at the meeting.