Kettleborough, Jamie writes:
> Hello Steve, Balaji,
>
> I think there are two problems here...
>
> Problem 1) Representation of geo-referenced probability distribution
> functions, e.g. the probability distribution of bath tub vortex strength
> as a function of latitude. (Not necessarily about models, not
> necessarily about forecasts). In particular the representation based on
> a set of sample points, where not all sample points have the same
> 'weight'.
>
> In CF context this could be time sampled, BUT we also have to take into
> account the fact that there are other ways of sampling - which include
> running lots of models.
Hi Jamie, I think we can sample across a _dataset_ quite adequately to
generate the kinds of PDF fields you'd like to see... and this dataset
will be created by aggregation servers of one kind or another that sample
metadata across a set of related netCDF files. The 'relatedness' is
described by some netCDF attributes that indicate that some datasets
belong to an 'ensemble' that may be sampled for generating PDFs. This is
not a novel procedure... this is how we all generate long timeseries now
from a set of files that each contain a time window.
> Problem 2) model ensemble aggregations - which can be a special case of
> the first problem - but only after the models have been re-gridded in
> some way onto the same grid (space and time).
At present, yes. There is every hope that aggregation servers in the near
future will be able to work across diverse grids, and at least perform
regridding on the fly. As I said, we already treat time this way.
> You think that we can not deal with Problem 2) in a general sense under
> NetCDF-3 and CF-1. Do you think we can still make progress with Problem
> 1? (I think/hope we can - I know I still have some use cases to write
> out.) This case can be accommodated under the NetCDF-3 data model - it
> does, however, introduce that 5th dimension.
I hope I've understood your PDF example above. If I haven't, then yes,
please do write out that use case!
Thanks,
--
V. Balaji Office: +1-609-452-6516
Head, Modeling Systems Group, GFDL Home: +1-212-253-6662
Princeton University Email: v.balaji at noaa.gov
Received on Wed Feb 28 2007 - 15:10:55 GMT
