NOTE: The cf-satellite mailing list is no longer active. The list archives are made available for historical reasons.
Jonathan: With the growing interest in the CF conventions around the world by the satellite CF data producer and user communities coupled with the ubiquitous nature of wavelength-based satellite CF data sets, does it make sense to add a paragraph to Section 4 Coordinate Types to discuss Spectral Coordinates ? very respectfully, randy ++++ Dear Aleksandar > I know this will likely end up as a trac ticket but would like first > to gauge the community's opinion about defining a new coordinate type. > Satellite data originates as measurements at a number of intervals of > the electromagnetic spectrum. These intervals are commonly referred to > as bands or channels. Deciding on how to store the band information is > one of the key issues toward a standardized representation for > satellite data. > > The convention seems to allow storing of band information either as a > numerical coordinate variable or as a string auxiliary coordinate > variable. Yes, the CF standard could handle both of those, without any modification. A trac ticket may not be needed. I certainly think there is no problem at all for a numerical coordinate of band wavelength. You need only to propose a new standard name for it, if one is needed. There is already a generic standard name of radiation_wavelength, included for use as a coord variable just as in your first example. If you need something more specific, I would suggest sensor_radiation_wavelength. The coord values for this would be the central wavelengths, and you could also supply bounds specifying the range of wavelengths covered by the sensor. Although string-valued auxiliary coordinate variables are possible already, as used in your second example, I would argue they are less useful as metadata than numerical ranges. That is because the main use of CF is to support intercomparison of datasets, which is better-defined if numbers are used. If there are widely used definitions of named wavelength bands, required for intercomparison of many datasets, a standard_name could be defined with a number of permitted string values. I think this extension could probably be seen as a new standard_name, not requiring a change to the conventions, although it could be explicitly mentioned in section 6 like Roy is proposing for biological taxa. Best wishes Jonathan
cf-satellite archives: