Re: [Fwd: Re: XYT and XYZT grid with only 1 X and 1 Y point in axis]

> > Hi Roland:
> >
> > The problem is that the z coordinate doesnt have a "positive" 
> > attribute. I was assuming that it needed it for COARDS, but on closer 
> > read it is in fact optional, if the units have dimension of length, 
> > pressure, level, layer, sigma_level or another vaguely worded section 
> > that seems to allow a bunch more things.
> >
> > Anyway, I have changed the code to follow the spec, but I fear that 
> > non-vertical coordinates will get misidentified as vertical. I would 
> > reccommend always using the "positive" attribute to disambiguate.
>
>
> I agree that the spec is terribly vague and using the positive attribute 
> is a good idea.  More below.
>
> > Get the latest 2.2.10 release at
> > http://my.unidata.ucar.edu/content/software/netcdf-java/index.html
>
>
> Yep, I got it.  Works great.
>
> > FYI, I would recommend using a convention for observation data, rather 
> > than using singleton latitude/longitude to create degenerate "grids". 
> > One that I have been working on is at
> >
> > http://my.unidata.ucar.edu/content/software/netcdf-java/formats/UnidataObsConvention.html 
>
>
> Again I agree with you in principle.  Jon Callahan in our group has 
> begun working with this convention and some examples you provided.  
> We're very impressed.  Ferret can work with these data files quite well 
> and the access is fast.  In defense of the folks and NDBC, I think their 
> needs pre-date the Unidata Obs Convention and I think what they've done 
> is 1000x better than other attempts I've seen to cram observations into 
> netCDF in some haphazard way.  At least these files are usable as is.  
> Even so, I will pass your comments along to the folks that create and 
> manage these files.  Perhaps there will be some tricks we can use from 
> the jnc 2.2 code and/or NCML to massage these files into the Obs 
> Convention.

theres nothing "wrong" with these as they stand, but they dont 
generalize to the case of multiple buoys in the same file.

i think in fact, nj22 can already read them into our ObsDataset 
interface (this is at a higher abstraction layer than the convention).

> > Can you send me a sample file with the depth dimsnion, so I can test 
> > in my stationObs reader ?
>
>
> I'll see what I can do about getting you an example from NDBC.  
> Meanwhile the only access I'm aware of for these files is through the 
> DODS server.  E.g.  
> http://dods.ndbc.noaa.gov/cgi-bin/nph-dods/dods/adcp/41012/41012a2002.nc.

ill have a look, thanks

> Roland