[CF-metadata] stations and trajectories

From: John Caron <caron>
Date: Fri, 24 Jun 2005 11:28:33 -0600

Jonathan Gregory wrote:

>Dear John
>
>
>
>>Its not clear to me how a program can
>>automatically see that this is a station dataset, and figure out what
>>the station dimension is, etc. What do you think?
>>
>>
>
>One answer is, does it matter? I mean, if you want to plot it, or do some
>computation with it, you will probably tell the program which variable you
>want. One way to tell, however, would be by the station dimension
>not having a coordinate variable, or having one with no units.
>If you wanted to be more positive about it, you could require the station
>coordinate var to have a particular standard_name, but for that we would need
>to decide whether we needed a convention for its values.
>
>Best wishes
>
>Jonathan
>
>
Hi Jonathan:

The goal of the "data type" layer of the Netcdf-Java library is to
automaticallly identify things like station datasets and to add various
convenience routines such as getting all the data for stations in a
geographic area, etc. So we are trying to build a common interface that
can work on a broad set of datasets from all over, including of course,
CF. This interface assumes that you can identify automatically what data
is associated with the stations, which is associated with the station
observations, etc. For this purpose Ive been leaning towards using
dimensions to group variables together along the lines outlined in the
document. Obviously there are any number of ways of doing that, but
there is some minimum information we need.

If theres anyone interested in giving input to this effort, please
contact me, we are about to start an email discussion group.

Thanks!
Received on Fri Jun 24 2005 - 11:28:33 BST