wms metadata about sci-wms HOT 4 OPEN

Good questions. In an ideal world sci-wms wouldn't need to identify the model type, only the grid type. Unfortunately it is used on non CF-UGRID compliant ADCIRC and FVCOM models, so a bunch of model specific checks were added in over time.

I usually check for an R grid first (delta x and delta y are constant in the 2D coordinate variables), then a C grid (data variables are dimensioned by (i, j) and have a "coordinates" attribute which references the 2D coordinate variables which are also dimensioned by (i,j), then a U grid (data variables are dimensioned by a single 2D dimension (node, edge, etc.) and have a "coordinates" attribute pointing to the coordinate variables.

This avoids checking for CF specific things like a global attribute "Conventions", but still uses the CF specific "coordinates" attribute heavily.

So to answer your question: there really isn't a great way to identify a model type. There is no widely used convention for identifying a file as "FVCOM 2.4".

from sci-wms.

Does SCIWMS work with files that actually are CF or UGRID compliant?
I'm guessing not, since that would require a CF or UGRID library. I'd love it if we could add UGRID support to Iris and use Iris for reading both CF and UGRID compliant files.

Folks seem to think that Iris is too big and too hard to build. But now that we have Iris as a conda package, it's pretty easy to install: conda install -c rsignell iris. And the beauty of it is that its got long term support and an established community already using it.

from sci-wms.

Rich, Does CF have the concept of

(see GRIB2 specification section 2.2.2 Product Definition Templates)

• Type of generating process
• Background generating process identifier (defined by originating centre)
• Analysis or forecast generating process identifier

In GRIB2, these sit alongside

• Product Definition Template Number
• Parameter Category
• Parameter number

which I know CF abstracts (so no codes are needed)

One of those (or all of those) could be interpreted as "model", and originating centre/center could represent the organization.

At the moment, we're assuming these are unique (for the time being) to COMT which has high interest in who and what ran the dataset. But if they happen to already be in CF conventions, maybe we can use that.

from sci-wms.

The COMT data should all be UGRID or CF compliant. Thus the only thing you should have to look for is the conventions attribute: Conventions: CF-1.X or UGRID-0.9. If they are not CF or UGRID compliant, we likely can make them so via NcML. This is the point of CF and UGRID -- to get away from model-specific code. But the bigger problem is likely that Sci-WMS doesn't know how to handle CF or UGRID if it finds them.

from sci-wms.