[gdal-dev] Multidimensional raster support in GDAL

[Edzer Pebesma]

I changed the subject, which was "WCS driver"; although it is still
relevant for the WCS driver, it's really about GDAL's abstraction.

On 10/24/2017 12:41 PM, Ari Jolma wrote:
> Even Rouault kirjoitti 24.10.2017 klo 13:28:
>>
>>  
>>
>> > The netcdf files I looked at through the GDAL driver had time and/or
>>
>> > depth offered as different gdal layers, and offered different variables
>>
>> > (i.e. different coverages) as subdatasets.
>>
>>  
>>
>> Yes, although in the WCS case, a coverage can have several bands, and
>> thus using the band mechanism to reflect the N>2 dimensions can become
>> confusing.
>>
> 
> Also in NetCDF the time dimension may be as bands - at least I have
> marine research data which opens first as having two subdatasets (mean
> and max wave height) and then as bands (the monthly values).

Yes, bands is what I meant; sorry for confusing this.

>>  
>>
>> The only clean solution to that would be to have a true
>> multidimensional raster support at the GDAL API level, but that's a
>> rather involved effort.
>>
> 
> Above the division to subdatasets and bands seems arbitrary and perhaps
> not user controllable. I agree it would be useful to have a better data
> model in this respect.

In my earlier email to this list [1] I suggested to allow querying bands
about their time, which would help deal with the case where a large set
of bands refers e.g. to both depth layers and time steps, in some order.

I too think that multidimensional raster support would be useful.
Besides by drastically redesigning data structures, could we get there
incrementally?

[1] https://lists.osgeo.org/pipermail/gdal-dev/2017-October/047381.html

> 
> Ari
> 
>>  
>>
>> >
>>
>> > Which other GDAL drivers give direct access to time-varying raster
>>
>> > stacks in single files?
>>
>>  
>>
>> I guess HDF5, as it is really close to netCDF.
>>
>>  
>>
>> GRIB could potentially be used to have time-varying raster, although,
>> from what I saw, generally you put instead different variables for the
>> same forecast time in the same file. But technically you couldx just
>> concatenate any amount of GRIB messages into a single file and this
>> will be seen as a multi-band raster by GDAL.
>>
>>  
>>
>> Even
>>
>>  
>>
>> -- 
>>
>> Spatialys - Geospatial professional services
>>
>> http://www.spatialys.com
>>
>>
>>
>> _______________________________________________
>> gdal-dev mailing list
>> gdal-dev at lists.osgeo.org
>> https://lists.osgeo.org/mailman/listinfo/gdal-dev
> 
> 
> 
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/gdal-dev
> 

-- 
Edzer Pebesma
Institute for Geoinformatics
Heisenbergstrasse 2, 48151 Muenster, Germany
Phone: +49 251 8333081