Units within Data Central

Units and their representation is one of the more complex parts of the metadata Data Central tracks, due to the existence of differing, contradictory standards for representing units, and a large legacy of existing data which does not match a particular standard chosen. Data Central uses standard tools (such as astropy) to read, convert, write and generally handle units in a VO-compliant way as possible, and aim to fix all unit issues before any public data release. In general, Data Central will not modify existing data releases to fix issues with units.

In terms of actually representing units, we suggest reading VOUnits, the IVOA Recommendation on units within the Virtual Observatory, as it provides useful information about how existing unit standards (or conventions), both within astronomy and outside, specify how to encode units. The VOUnits document has nothing particularly unusual, though it does note some corner cases worth taking note of, such as treating the string 'Pa' as Pascals rather than petayears (which would happen if naively if 'P' was always treated as representing Peta).

In general, there is nothing particularly strange you need to do to produce data with VO-compliant units, but there are two things you should probably check: the capitalisation and pluralisation of the units; and encoding dimensionless or unitless quantities.

Capitalisation and Pluralisation of Units

Astropy by default reads only the singular form of a unit (e.g. “second” over “seconds” or “Angstrom” over “Angstroms”), as generally this is the preferred form in most unit standards. Data Central handles plural forms for existing data releases, but for maximum compatibility, the singular form should be used for new data (we are unaware of any software which rejects the singular form).

Similarly, lower case (except for units named after a person) is to be used, with capitalisation of the first letter if the unit is named after a person (e.g. “Angstrom” over “angstrom”). All caps (every letter being upper case) will throw errors in astropy, and therefore should be avoided in new data.

Encoding Dimensionless or Unitless Quantities

Neither the FITS standard nor the IAU Style Guide discuss the representation of dimensionless/unitless quantities, so there is no specific representation in standard FITS. There are some differing conventions mentioned in the VOUnits document, with the VOUnits document recommending using an empty string.

For normalised quantities, such as normalised intensity, we suggest using the empty string for the unit, and describe the quantity in the comment (e.g. have the FITS comment for the field be “Normalised Intensity” for something stored as normalised intensity).