Dear Bert In your example:
> varN:coordinates = "lat lon"; > varN:cell_methods = "lat: point lon: point"; if lat and lon are 1D coord vars, they aren't usually listed as coordinates (though it is not an error to do so). CF relies on the Unidata convention of associating 1D coords with data vars through the equality of dimension and coord var name. > If any one of the variables refers > to 'lat'/'lon' by means of a cell method other than 'point', e.g. > 'mean', then it must be specified what range/cell the cell_method > relates to i.e. 'bounds' should be required. That seems more like a > consistency *requirement* than merely a recommendation. Yes, that is the current recommendation. We could promote it to a requirement for the case where cell_methods is specified, I think, since both cell_methods and bounds are CF metadata. > If 'cell_methods' is not specified on one or more variables, then the > need for 'bounds' becomes more fuzzy. The discussion on the default > interpretation for 'cell_methods' in Section 7.3 suggests that one > might distinguish between extensive quantities like > 'cloud_area_fraction' and intensive quantities like > 'sea_surface_height'. The former might require it, whereas it would > be optional in the latter case. That's right. We decided to recommend that cell_methods always be included because it's hard to make the distinction intensive/extensive. > I would agree with a general recommendation to > include 'bounds' when 'cell_methods' is missing on one or more > variables. I think that's a good idea. I would support it. That goes beyond what we currently have, and would require at least a "defect" ticket, perhaps a substantive change. It would be a helpful change, to encourage data-writers to describe their data better by including cell_methods and bounds. Best wishes Jonathan _______________________________________________ CF-metadata mailing list [email protected] http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
