Dear John > As the radiation wavelength and incidence angle cannot vary over the image > to obtain reliable measurements, the radiation_wavelength and > nominal_incidence_angle should be specified as attributes.
That would of course be possible in netCDF, but it is not consistent with the approach of CF. CF uses attributes to describe the structure of the data storage, and to point to variables which contain geophysical metadata (especially coordinates of various kinds), but CF attributes do not contain geophysical information (see Appendix A). CF always uses coordinates for geophysical information, even if single-valued. There are various reasons for that which maybe have already come up in this discussion, including: * It keeps the convention simple by limiting the number of attributes that are defined by CF. Names of geophysical quantities (like radiation_wavelength and nominal_incidence_angle) are entries in a lexicon (the standard name table) rather than being defined by the convention itself. * Geophysical quantities (including radiation_wavelength and nominal_incidence_angle) have units. An attribute can't have an attribute, and it is inflexible if the units have to be defined, rather than allowing any physically equivalent unit for a quantity. (We have this inflexibility in the grid_mapping attributes of Appendix F, unfortunately.) * In most cases, there may arise a need for them to be multivalued. In some cases it is obvious. We could use an attribute for the height above the ground, for example, if it is was single-valued, but it's obvious that we also need multivalued coordinate variables for this, and it's simpler to use the same machinery in both cases. I'm sure you're right that in your application you don't need these quantities to be multivalued, but why rule it out? Other related applications might want them to be multivalued. Best wishes Jonathan _______________________________________________ CF-metadata mailing list [email protected] http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
