[PROJ] Grid CDN Update and Crowdfunding Request

[Even Rouault]

Chris,

> I'm very interested in the possibilities hinted at for GeoTIFF for grid
> files - but not very clear what capability that would provide.   It sounds
> like it could provide a way of encoding some aspects of the deformation
> model we use in NZ.

The scope of this campaign was directly dedicated to 'classic' horizontal and 
vertical shift grids. I know we had a discussion a few months ago about 
deformation models.

> The specific capability I am hoping for is:
> 
> - multiple levels of nested grid
> - 1,2, or 3 (more ?) dimensions of data at each grid node (eg de, dn, du,  +
> possibly uncertainties of each, NTv2 only provides dx/dy and uncertainties
> of them IIRC) - values to be floating point or possibly I4 with a defined
> offset and scale factor for each dimension

GDAL can encode offset and scale factors in a TIFF tag (as XML metadata), but 
I'm not sure of the value of that for the need we want to address, rather than 
use Float32 directly
Regarding uncertainties, we don't make currently use of them, so I'd imagine 
that we would not include them in the TIFFs to make them more compact.

> Presumably this would also define an interpolation technique (bilinear in
> each dimension).  NTv2 places what appear to me to be unnecessarily tight
> constraints on relationships of subgrids with parent grids.  How do you
> envisage the selection of grid to use at a specific grid cell would be
> implemented?

I was only thinking to a single-level of subgrids, like the Canadian 
ntv2_0.gsb, which is the most complex grid we have in the proj-datumgrid 
repository. The algorithm is simple: select the (first) subgrid whose extent 
intersects the point to transform.

TIFF files have a concept of SubIFDs that could potentiall be used to 
implement multiple levels but there's no support for this in GDAL and I'm not 
completely sure about the state of this in libtiff itself.

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com