I agree, it appears that GeoTiff does not support the curvilinear case,
you would have to regrid. One could output netcdf/CF, i think.
On 6/15/2011 10:09 PM, Tom Kunicki wrote:
Having worked with the importing WCS output via GeoTIFFs for use with
NetCDF-Java i'm not sure it's possible to define a curvilinear model
coordinates using the current GeoTIFF tag set.
The pertinent portion of the GeoTIFF specification is here:
http://www.remotesensing.org/geotiff/spec/geotiff2.6.html#2.6.1
There are two transformation types:
1) Define a tie point as x,y[,z] along with pixel x,y,[z] deltas.
*or*
2) Define a transformation matrix such that m(x,y,z,1) = M_t *
i(x,y,z,1) with M_t being a 4x4 transformation matrix and m being
model coord and i being image coord.
I've validated the GeoTIFF IOSP against all the sample GeoTIFFs
available at the remotesensing.org <http://remotesensing.org> site and
have yet to see an implementation of
type 2. I would take this to mean that even if one were to output a
transformation of type 2 it would be unlikely that client software
would support it (since there doesn't seem to exist validation data
for it). That being said I don't think a type 2 definition could be
used to define curvilinear mode coordinates.
I guess the point is that WCS support for curvilinear grids would need
an output format other than GeoTIFF. The only way to specify a
curvilinear grid with GeoTIFF would be to provide model data as a
regular grid with a (known) projection mapping to the curvilinear
coordinates. I don't think this provides the desired flexibility
(projections are limited to known ESG
projection-algorithm/parameter-sets or projections via EPSG codes).
NetCDF-CDF output from WCS would be needed. I apologize if i've
stated the obvious...
Tom
On Jun 15, 2011, at 10:00 PM, john caron wrote:
some terms as far as i understand:
regular grids: 1d x, y coordinates evenly spaced.
only ones WCS 1.0 allows
irregularly spaced grids: 1d x,y, but dont have to be even, eg
gaussian latitudes.
curvilinear grids: 2d x,y but each section is a
quadilateral and connected to its neighbor (implicit topology)
CF allows both irregular and curvilinear grids. one could extend WCS
to return netcdf/CF irregular and curvilinear grids, without regridding.
unstructured grids: arbitrary meshes, explicit topology, no CF
standard yet.
John, why do you want WCS? To get geotiffs, or ??
On 6/15/2011 7:25 PM, Ben Domenico wrote:
Hi John et al.,
I am currently at the OGC Technical Committee meetings so your note
is quite timely. My concept of how this sort of capability might be
worked into the standards -- from the netCDF point of view anyway --
is via the CF conventions. We are currently working on a CF
extension for the OGC netCDF core standard. At the same time we are
working on a CF-netCDF encoding standard for WCS 2.0. But my
understanding is that irregular/curvilinear grids are not formally
adopted yet as part of the CF conventions. When these grids
officially become part of CF, we can then do a revised extension to
CF-netCDF and to WCS.
Please correct me if I'm wrong about the status of
irregular/curvilinear grids in CF. If you have suggestions for
better ways to accomplish this, let me know. But for now this is
the general outline of the plan for getting this into the formal
standards world.
For what it's worth.
-- Ben
On Wed, Jun 15, 2011 at 6:23 PM, John Maurer <jmaurer@xxxxxxxxxx
<mailto:jmaurer@xxxxxxxxxx>> wrote:
Thanks for the explanation, Ethan. If TDS could handle the
regridding itself, then it could work. I assume this is what's
happening on the backend via TDS/ncWMS for GetMap requests since
the WMS spec doesn't handle irregular/curvilinear grids either
(right?). I was hoping a future enhancement could someday do the
same for TDS WCS. Something for the suggestion box.
Cheers,
John
On Wed, Jun 15, 2011 at 2:17 PM, Ethan Davis
<edavis@xxxxxxxxxxxxxxxx <mailto:edavis@xxxxxxxxxxxxxxxx>> wrote:
Hi John,
TDS currently only supports WCS 1.0.0. The WCS 1.0.0
specification
itself does not support requesting or returning
irregular/curvilinear
grids. I believe the same is true for WCS 1.1.0.
I don't belive this is necessarily the case for WCS 2.0.
However, the
2.0 specification is written in the new core/extensions
paradigm and I
don't believe the extensions necessary to request or return
irregular/curvilinear grids are yet written.
The TDS WCS implementation does not currently support
regridding data.
So, currently one gets the data from the WCS in the same
projection in
which the data is stored. The combination means the TDS WCS
can only
serve regularly gridded data.
Ethan
On 6/15/2011 5:31 PM, John Maurer wrote:
> Hi All,
> We now use the WCS service through TDS to serve GeoTIFF
files to users.
> Since it only works for regular grids, I was wondering if
TDS could be
> enhanced in the future to support irregular/curvilinear
grids via WCS?
> TDS/ncWMS already handles curvilinear grids via WMS, so I
was hoping the
> same sort of intervening magic could be applied for TDS'
WCS service?
> Thoughts?
> Thanks for listening!,
> John Maurer
> Pacific Islands Ocean Observing System (PacIOOS)
> University of Hawaii at Manoa
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx <mailto:thredds@xxxxxxxxxxxxxxxx>
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx <mailto:thredds@xxxxxxxxxxxxxxxx>
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit:http://www.unidata.ucar.edu/mailing_lists/
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx <mailto:thredds@xxxxxxxxxxxxxxxx>
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
