[fitswcs] Improving FITS header astrometric information

[Doug Mink]

I would like to jump start the FITSWCS discussion with a new topic.
As the WCSTools package gets more widely deployed and more
capabilities are required, I have found that I have to decide
just what makes up a sky position.  Inside the package, a position,
such as that of the reference pixel, is given in equatorial
coordinates by a right ascension, a declination, an epoch, an
equinox, and a coordinate system (RADECSYS).  Seldom, in my
experience, are all five of these parameters defined in a FITS
header, so the software is forced to make assumptions concerning
those parameters which are not present or are given in a variant
manner.  My current methods are mostly described--as usual for
one-person projects, the software is slightly ahead of the
documentation--in

http://tdc-www.harvard.edu/software/wcstools/wcstools.wcs.html

Of course I use the EPOCH keyword for the equinox and system
if EQUINOX and RADECSYS are not present, but I am interested in
suggestions on how to interpret such possibilities as

EPOCH = 'J1991.25 '

which I assume indicates FK5 coordinates at epoch 1991.25, equinox
2000.0.  Such character values for EPOCH do really appear in real
data; hence the apparent need for a standard set of subroutines
to interpret them.  Currently, the use of an epoch (usually for
an image the date it was exposed), makes J2000 <-> B1950 conversions
more accurate (and reversible).  When reference catalogs with
proper motions, such as Tycho, ACT, or future USNO catalogs are
used to set the WCS, the epoch will be more important, so I would
like to make sure that everyone understands just what they are
putting in FITS headers.

-Doug Mink
 Telescope Data Center
 Harvard-Smithsonian Center for Astrophysics