Hi Jonathan, Tim, Doug, et al:
Jonathan Gregory wrote:
> Dear Tim
>
> So for the case of multiple analysis and forecast times, where all combinations
> exist, you have two multivalued time dimensions, one for forecast reference
> time (= base time, analysis time, initialisation time) and one for forecast
> period (lead time). The former has units of "UNITS since TIME" and the latter
> is a plain time-unit. Is that right?
thats what i understand also
>
> In http://mailman.cgd.ucar.edu/pipermail/cf-metadata/2006/001008.html we
> proposed instead to deal with this 2D case by introducing an index dimension
> and collapsing it into 1D, with the forecast reference time and the forecast
> period being 1D auxiliary coord vars. That obscures the 2D structure, which is
> a loss. The reason for proposing it was that it reduced the number of different
> structures that would have to be used. The 2D structure can only be used if
> all 2D combinations exist.
the index dimension method makes a compact file, but id say that its hard to figure out without documentation.
>
> In http://mailman.cgd.ucar.edu/pipermail/cf-metadata/2007/001470.html, the
> method that Doug uses, 2D is also collapsed into 1D, but instead of introducing
> a plain index dimension, the 1D coordinate is validity time (forecast reference
> time + forecast period). That is a better idea, I think, because it's a useful
> coordinate variable rather than just an index. Again forecast reference time
> and forecast period are 1D auxiliary coord vars.
this is good because its got only 1D coordinates, which are meaningful. the downside is that the coordinate values may not be unique, so its not really a coordinate variable. this can mess with some software that assumes one can uniquely name things with the coordinate values.
>
> I agree that the CF standard should have examples for each of the recommended
> structures. That would need a proposal on trac and maybe John was offering to
> make one.
yes, if theres general agreement on how to proceed.
>
> What do you think about using Doug's/Paco's collapse of 2D into 1D? To get all
> the 24-h forecasts, for instance, you'd have to search the 1D auxiliary coord
> var of forecast period and use it to select particular indices of the time
> dimension. That is a bit more awkward, but the structure has an advantage in
> generality.
looks like the main defect is that again, the coordinate values may not be unique.
im not that concerned with the search problem. one must read in the coordinate values and figure out what they correspond to.
-----------------------------
the main elements of any solution are the 3 types of time coordinates, with standard names:
forecast_reference_time
forecast_period
time
such that forecast_reference_time + forecast_period = time.
there are 3 main difficulties to answer:
1. in the general case, forecast_period may be 2D (not common though), and time is often 2D. is this allowable as auxiliary coordinates?
2. there may be missing cases, akin to ragged arrays. possible solutions are an "index dimension" (section 8.2), or to use missing values, or ?. (I notice that existing examples I have seen find a way to avoid this)
3. solutions that try to stick with only 1D variables will, in general, have repeated coordinate values, making them technically not coordinate variables. will we allow this?
>
> Best wishes
>
> Jonathan
> _______________________________________________
> CF-metadata mailing list
> CF-metadata at cgd.ucar.edu
> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
Received on Fri Jan 09 2009 - 18:21:48 GMT
