[fitsbits] FIT Keyword data dictionaries

[Mark Calabretta]

On Thu 2000/11/02 13:56:14 CDT, William Pence wrote
in a message to: fitsbits at nrao.edu

>> The FITS standard states angles should be in degrees!
>
>Actually, the FITS Standard is more qualified in its recommendation on units
>(in section 5.3): "For angular measures given as floating point values and
>specified with reserved keywords, degrees are the recommended units."  This
>wording was carefully chosen to make it legal to specify angular quantities
>as strings in 'dd:mm:ss' or 'HH:MM:SS' format for non-reserved keywords.

I'd like to stress that sexagesimal notation is only to be used for
>>>non-reserved<<< keywords, certainly as far as FITS WCS is concerned.

Why?

The WCS documents state that degrees are required units for angles
associated with map projections though their actual representation as
a keyvalue is not specified.

I must admit that use of sexagesimal notation had not even occurred to me
as a possibility but considering that this is really a matter for the FITS
keyvalue parser you might argue that it shouldn't matter.

Our primary concern was that a decimal number specified for CRVALn (say) for
a map projection would be interpreted as degrees rather than radians:

   CTYPE2 = 'DEC--TAN'
   CRVAL2 =        -1.5160        / Degrees, not radians.

Use of sexagesimal notation would certainly eliminate confusion with
radians here:

   CTYPE2 = 'DEC--TAN'
   CRVAL2 = '-01:30:56.5'         / (Illegal usage) Doesn't even look like
                                  / radians now - good.

So far so good...

>In particular the 'RA' and 'DEC' keywords have often been used in the past
>with a string value such as
>
>   RA      = '12:30:56.5'
>   DEC     = '-63:56:45.67'

...however, this risks introducing confusion between DMS and HMS for an
angle measured as time - is that RA 12 hours or 12 degrees?  Degrees
are required in WCS as always:

   CTYPE1 = 'RA---PCO'          / We have a FITS WCS.
   CRVAL1 = '12:30:56.5'        / (Illegal usage) 12.5160 degrees!  Did you
                                / think it was HMS?

I don't think it would be reasonable to expect a software keyvalue parser to
distinguish between DMS and HMS in such a case, primarily because we allow
user-defined coordinate systems denoted by generic xLON,xLAT and yzLN,yzLT
identifiers.  The nature of the longitude measure (deg or time) for these
is not generally known in advance:

   CTYPE1 = 'ZJLN-TSC'          / We have a FITS WCS.
   CRVAL1 = '12:30:56.5'        / (Illegal usage) Don't know what system the
                                / "ZJ" refers to but do still know this is
                                / 12.5160 degrees.
   WCSNAME= 'ZJ coordinates'    / Clear as mud!

Therefore we insist on degrees without exception.

Use of sexagesimal notation for RA in degrees would inevitably cause
confusion for human readers of the header (but ok for software readers).
Considering that the main reason to use sexagesimal notation would be to
help human readers it would therefore be counterproductive.

Mark Calabretta
ATNF