[gdal-dev] spatial_ref and GeoTransform attributes for HDF5 datasets

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

[gdal-dev] spatial_ref and GeoTransform attributes for HDF5 datasets

Agram, Piyush S (334D)

Hi,

    I’m looking for a simple convention to represent SRS and Geotransform information for HDF5 arrays. A fairly thorough reading of available HDF5 geospatial data representation specs hasn’t resulted in a satisfactory solution. I know that HDF-EOS5 is a published spec but the unintuitive naming of swaths and grids is not a satisfactory solution for what we are trying to accomplish. We would like a spec that allows us to

 

  1. Support complex datasets (possible since GDAL 2.3). HDF-EOS5 does not support this.
  2. Generate HDF5 files with different grids in the same file – one can think of some grids as low resolution support grids which can be interpolated on the fly to match the primary high resolution grids. This might be an issue for CF-compliant representation? We might want support to include different grids for different layers depending on support needed by interpolators.

 

I’ve also been looking through the netcdf driver in GDAL since this is closely related to HDF5 and noticed that there is a non-CF compliant representation that is still supported. I’m happy to work on implementing a similar solution for HDF5 – namely each dataset gets its own spatial_ref (str) and GeoTransform (array of floats/doubles) attributes.

 

I’m reaching out to see if similar/ alternate solutions have been proposed in the past or if something similar is in the works. If not, would a PR for a feature like this be of interest?

 

Thanks

Piyush


_______________________________________________
gdal-dev mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/gdal-dev
Reply | Threaded
Open this post in threaded view
|

Re: spatial_ref and GeoTransform attributes for HDF5 datasets

Even Rouault-2
On mercredi 20 juin 2018 04:05:18 CEST Agram, Piyush S (334D) wrote:
> Hi,
>     I’m looking for a simple convention to represent SRS and Geotransform
> information for HDF5 arrays. A fairly thorough reading of available HDF5
> geospatial data representation specs hasn’t resulted in a satisfactory
> solution. I know that HDF-EOS5 is a published spec but the unintuitive
> naming of swaths and grids is not a satisfactory solution for what we are
> trying to accomplish. We would like a spec that allows us to
 
>
>   1.  Support complex datasets (possible since GDAL 2.3). HDF-EOS5 does not
> support this.
 2.  Generate HDF5 files with different grids in the same

> file – one can think of some grids as low resolution support grids which
> can be interpolated on the fly to match the primary high resolution grids.
> This might be an issue for CF-compliant representation? We might want
> support to include different grids for different layers depending on
> support needed by interpolators.
> I’ve also been looking through the netcdf driver in GDAL since this is
> closely related to HDF5 and noticed that there is a non-CF compliant
> representation that is still supported. I’m happy to work on implementing a
> similar solution for HDF5 – namely each dataset gets its own spatial_ref
> (str) and GeoTransform (array of floats/doubles) attributes.

Piyush,

I don't know whih the best forums to discuss HDF5 related specifications are,
and how much data producers care about standardization in HDF5 world, but
ideally we would want this to be largely adopted to avoid having yet-another-
way-of-encoding-georeferencing in HDF related products.
So a potential idea could be to have some website / github page with a
proposed spec, make some noise about it to try to get feedback and come to
some concensus around it. Perhaps the scope could also encompass geolocation
arrays / GCPs which seem to be quite common in HDF productions.

Even

--
Spatialys - Geospatial professional services
http://www.spatialys.com
_______________________________________________
gdal-dev mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/gdal-dev