[Gdal-dev] RE: GDAL netcdf convention CF support

[Denis Nadeau]

> 
> -----Original Message-----
> From: Frank Warmerdam [mailto:fwarmerdam at gmail.com]
> Sent: Thursday, August 18, 2005 2:24 PM
> To: Norman Barker
> Cc: gdal-dev at lists.maptools.org; galeon at unidata.ucar.edu
> Subject: Re: [Gdal-dev] RE: GDAL netcdf convention CF support
> 
> 
> On 8/18/05, Norman Barker <nbarker at rsinc.com> wrote:
> > [Norman Barker]
> >
> > Hi,
> >
> > this is a really exciting announcement. I have run ncdump on my netcdf 
> file
> > and checked the variables, and gdalinfo picks up the lat, lon, and ozone
> > vars ok. It puts the z-levels in bands within the data file which is ok
> > since that was the way that we had done it with geotiff. It has lost the
> > information attaching the band level to an atmospheric height. Even if 
> this
> > was a comma separated index string it would be useful. It has also lost
> > the Time dimension stamp within the file. Again is this could just be a
> > metadata tag that would be great.
> 
> Norman / Denis,
> 
> I agree that attaching the dimension value(s) unrolled as bands as
> band metatadata would be desirable.


I could write the metadata such as 

GDALInfoBands = "times X isobaric (2 x 19 )"

Gdal order the bands in the following way:

band time isobaric
------ ----- ---------
1 1 1
2 1 2
... .... ....
19 1 19
20 2 1
21 2 2
... ... ...
38 2 19



> I would like to help wrap this into MapServer WCS if possible, if I am
> > missing something with my comments above please let me know. This is 
> great
> > though, is has caused great excitement in our office!
> 
> Are you interested in having GDAL write netCDF files so that
> MapServer can return netCDF files? As far as I know, there is no
> support for creation/writing in the generic netcdf module in GDAL but
> it seems like it might not be that hard to implement.
> 
> <snip>
> 
> Ideally having MapServer return netCDF would be a nice to have, but as we 
> have discussed before about 3D BBOX + parameters (and this is an aim of OGC 
> GALEON), there is little to differentiate between a geotiff and a netCDF 
> file (or any banded file) when implemented this way (2D + bands for the 
> z-level). If we could access netCDF through MapServer in an ND sense it 
> would be different, I hope this is something we can work through in GALEON. 
> It is still possible to make a very workable WCS with 2D + bands though, and 
> I am interested in a way to keep the metadata within the file.
> 
> However most of the work (for example when you have 40TB of data as our 
> domain user does!) is in the conversion, so having gdal and hence MapServer 
> read CF netCDF is a major plus (really good!), and makes using MapServer WCS 
> even easier.
> 
> This is great though, I am compiling up MapServer with this new GDAL as we 
> speak!
> 
> Hats off to Denis and Frank!


Thanks a lot for the Kuddo!! The driver should work for more than 4 
dimensions, but I have not data to check it. 
If you can provide me with 5D or 6D data, I could test the netcdf driver. 
The number of bands grows a lot with 4D data, the RUC.nc test file provided 
by GALEON has about 296 bands....

Denis
P.S. I will check in an new version of netcdfdataset.cpp, I found an error 
in the program this morning.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/gdal-dev/attachments/20050818/1de5112a/attachment.html