Dear Jonathan, and all, ----- Original Message ----- > It has > previously been suggested that different components should be in the > same > data variable, with a dimension for component, but I prefer your > solution, > because the components might have different units (e.g. in the case of > modulus > and direction), and they might be on different horizontal grids (e.g. > for > an Arakawa C-grid), which makes it undesirable or impossible to store > them in > a single data variable.
Of course, I agree with the restriction on units. I am however in doubt if it makes any sense to define vector objects from components that do not share the spatial, height, or time dimensions. A (2d) vector can be instantiated from (minimum) 2 components, if given at a given space/time location. As of today, CF allows Arakawa C-grid constructs where components do not share the same grid. But if I was to reunite these (independent) components into a vector object, I would define 2 vector objects. One for the x vectors (having value 0 for all the y components) and one for the y vectors (having value 0 for all the x components). The two vector fields would not share the same grid. But at least the two components of each vector would share the same locations. Thus, although I am unsure we should impose it from the convention, e.g. by allowing to define some attributes from the "umbrella" level (such as :coordinates), I see little use for defining vector objects that do not share all their dimensions. All the best, Tho,qs (on a French keyboqrd!) > Furthermore, your solution is more flexible since it > allows you to have several different decompositions in the same file, > as you > say. It is also backwards-compatible; the variables containing > components are > described exactly as before, and we have the extra umbrella variable > showing > some associations between them, which should not cause a problem to > any > existing software. > > We would need more standard names, as you say, and these could be > proposed as > required. The new standard names would be permitted only with the > umbrella > variables, or with ancillaries of the umbrella variables. > > Best wishes > > Jonathan -- ========================================== Thomas Lavergne Norwegian Meteorological Institute P.O.BOX 43, Blindern, NO-0313 OSLO, Norway Phone: (+47) 22963364 Fax: (+47) 22963380 Email: t.laver...@met.no OSISAF HL Portal: http://osisaf.met.no ========================================== _______________________________________________ CF-metadata mailing list CF-metadata@cgd.ucar.edu http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata