WCS for long-slit spectroscopy

[Frank Valdes]

> Malcolm J. Currie <mjc at roe.ac.uk> writes in sci.astro.fits:
> 
> Defining keywords is not straightforward with spectroscopic data,
> especially long-slit spectroscopy.  (For those not familiar with the
> problem, long-slit datasets have sky co-ordinates at each slit position
> in addition to the x-y pixels of the two-dimensional detector/data
> array.)  For a single spectrum I could add dummy third and fourth
> dimensions which specify the sky co-ordinates for that spectrum. 
> However, I'm uncertain what is the best approach (i.e. one that many
> FITS readers would be able to interpret) for long-slit data.  Therefore
> I'd like to know please what schemes other people have devised or
> recommend for such data. 

I am not aware of any common solution and certainly not a standard.  The
IRAF WCS is capable of representing this situation.  I am not advocating
this but it is grounds for discussion.  The IRAF WCS includes a concept of
axes mapping.  Basically this is a keyword that maps the existing raster
axes to a higher order set of axes in a higher order WCS.  This allows a
single column or line stored as a 1D image to still support the coupled
RA/DEC systems for celestial coordinates.  So in the long slit case the
higher order WCS would be three dimensional with two axes for the coupled
spatial part and one for the dispersion.  To be simple the dispersion axis
would match the image raster dispersion axis and if the slit was
predominately along RA that raster axis would map to the RA-TAN (or
whatever) axis.  The third axis would then be the subordinate spatial
axis.  The axis mapping keyword (WAXMAP01 in IRAF) would then map away the
missing axis.  Currently there is no raw or reduced data in IRAF that has
this information but logically that is how it would be done.

Rather than axis mapping one could use a dummy axis.  The problem with
dummy axes and data analysis software, speaking for IRAF software mainly,
is that in this case the software would see NAXIS=3 and many applications
will simply say that they are intended only for 2D data and quit.  Of
course the software could have a "kludge" that notes that an axis length
is one and assume the image is really of a lower dimensionality but this
is not always the case.  In IRAF an IMCOPY to reduce the dimensionality
and remove the dummy axis would automatically set up the axis mapping
described above.

Frank Valdes
NOAO/IRAF Group