Coordinate specification in Data Cubes (Was: Re: Complex data, Cube data

[Doug Tody]

Hi Brian -

On Tue, 16 May 2006, Brian Thomas wrote:

> On Tuesday 16 May 2006 11:43 am, Doug Tody wrote:
> > Hi Arnold -
> >
> > On Mon, 15 May 2006, Arnold Rots wrote:
> > > Several months ago I sent some examples of how STC elements would give
> > > a proper metadata description of data cubes and how to ask for such a
> > > cube using STC metadata.
> > > Could that information be merged into what you are doing in this area?
> >
> > Possibly.  The chief issues are that a data cube is a fairly constrained
> > problem (N-D image matrix plus a WCS) which is already adequately by FITS
> > WCS, and FITS and FITS WCS is what is already in use to store, transport,
> > access, and analyze cube data.  Anything we use for cube coordinates has
> > to either be based on FITS or at least have a clear mapping to FITS WCS to
> > provide reasonable support for existing data and software.  The remaining
> > issues of describing cube data can already be dealt with by the generic
> > dataset metadata (Characterization etc.) being developed with SSA.
>
> There must be a different definition of 'data cube' in play here than
> I understand.  An N-D image matrix would be one kind of data cube,
> but there are many others which are not image based at all.

It is certainly true that there are other kinds of 3-D data than uniformly
sampled (pixelated) data cubes (or N-D versions of such cubes), however
for the SIA upgrade this is the case we are specifically referring to.
Radio data cubes for example are of this type, and they are generally
stored in FITS format, with a FITS WCS, and the applications used to
analyze and visualize these cubes are also FITS-based.  IFU data cubes in
the Euro3D format are also FITS-based although the actual format is a FITS
binary table and they are not simple FITS images.  A more careful look
at Euro3D format is needed to determine what is required to support it.
These are the two most widely-used cases I know of for fully sampled 3-D
astronomy data; does anyone know of any others?  (Event data maybe, but
it probably does not qualify as the energy/spectral measurement precision
is generally very limited).

We could consider using a non-FITS model for WCS within the VO, but if we
do so there at least has to be a simple, well-defined mapping to FITS WCS
so that FITS data can be published to the VO, and processed with existing
FITS-based software.  We can't expect the user community to abandon their
archival data formats or existing software, particularly if it is still
in active use, which it is of course.

None of this is relevant to the argument for a parameter-based interface,
although it could be pertinent to the data model.

	- Doug