[fitsbits] WCS keywords in tables

[William Pence]

The following question about FITS WCS keyword usage occurred during some 
recent discussions with a group that is in the process of designing the 
format of their FITS data files:

This project intends to produce a large volume of FITS binary tables in 
which several of the columns will each contain a small image (in a 
2-dimensional vector cell) in every row of the table.   Table 8.2 in the 
FITS standard defines the set of WCS keywords that should be used in 
this case (where the images are in column 4 of the table in this example):

WCAX4   =                    2 / number of World Coordinate System axes
1CRPX4  =             86.88896 / reference pixel
2CRPX4  =            261.93919 / reference pixel
1CRVL4  =    284.6423916705557 / right ascension  at reference pixel
2CRVL4  =    39.92821811003438 / declination (degrees) at reference 
pixel
1CTYP4  = 'RA---TAN'           / the coordinate type for the first axis 

2CTYP4  = 'DEC--TAN'           / the coordinate type for the second axis 

1CDLT4  = -0.00110560439808150 / degrees per pixel, increasing eastward 

2CDLT4  = 0.001105604398081504 / degrees per pixel, increasing northward 

(etc)

These keywords would need to be repeated for all the other columns that 
contain images.

However, in their particular case, *all* the images contained in the 
table have exactly the same WCS keyword values, so it is redundant to 
repeat the same keywords for every column.  So, they would prefer to 
just list all the WCS keywords once, using the standard keywords that 
are used for simple FITS images:

WCSAXES   =                    2 / number of World Coordinate System axes
CRPIX1  =             86.88896 / reference pixel
CRPIX2  =            261.93919 / reference pixel
CRVAL1  =    284.6423916705557 / right ascension at reference pixel
CRVAL2  =    39.92821811003438 / declination (degrees) at reference 
pixel
CTYPE1  = 'RA---TAN'           / the coordinate type for the first axis 

CTYPE2  = 'DEC--TAN'           / the coordinate type for the second axis 

CDELT1  = -0.00110560439808150 / degrees per pixel, increasing eastward 

CDELT2  = 0.001105604398081504 / degrees per pixel, increasing northward 

(etc)

We were pleasantly surprised to find that Mark Calabretta's wcslib 
library already supports this extension to the WCS usage.  The standard 
image keywords can be used in binary tables to provide the default WCS 
values for all vector columns in the table.  These WCS values may be 
overridden in any particular column by providing the column-specific 
form of the WCS keywords for that column.  This only works for images in 
vector columns; it does not apply to 'pixel list' images in binary 
tables because there is no well-defined association between coordinate 
axis number and column number.

So, the question is, should we encourage projects to use this extension 
to the WCS usage for vector images in binary tables?  Are there any good 
reasons not to use this?  Do other WCS libraries besides wcslib support 
this?

There is one other complexity in their case:  accurate astrometry is 
essential, so they need to define the distortion corrections that should 
be applied to the nominal WCS keywords.  For this purpose, they are 
considering using the SIP convention (see 
http://fits.gsfc.nasa.gov/registry/sip.html) which has keywords like:

AP_0_1
BP_1_2
etc.

Alternate forms of these SIP keywords have not been defined for use with 
images stored as vector cells in binary tables, so they would either 
need to define a new set of keywords for this purpose (which would be 
difficult given the 8-character limit on keyword names), or just use the 
existing keywords that were designed for simple FITS images.

Bill Pence
-- 
____________________________________________________________________
Dr. William Pence                       William.Pence at nasa.gov
NASA/GSFC Code 662       HEASARC        +1-301-286-4599 (voice)
Greenbelt MD 20771                      +1-301-286-1684 (fax)