On Wed, 2005-06-08 at 17:22 -0700, John Caron wrote:
> thanks, i hadnt seen that section before in this context.
>
> i see that you have a string valued coordinate variable. Is the
> intention that be a unique identifier ?
>
> In the case im trying to deal with, we have a unique id, a description
> and optionally a wmo id associated with each station. In general, one
> could have arbitrary information you want to associate with each station.
>
I don't think that netCDF is a particularily good format for storing
time series. netCDF is designed for gridded data. I think the problem
with time series is that there is not necessarily a shared time
coordinate. Mostly there will be time-value pairs. So for each time
series you would need an index and an id and you would end up with a
data set that has many dimensions (unlimited if the time series are
allowed to grow).
At one stage I needed to come up with a way of storing lots of time
series of varying length. In the end I used a data base (sqlite). I am
still not too happy with this solution. One side effect of using a data
base is querying the data set using SQL is rather nice. I suspect HDF5
might help or netCDF4 once it is released.
Cheers
magi
--
Magnus Hagdorn
School of GeoSciences
The University of Edinburgh
Grant Institute
West Mains Road
Edinburgh EH9 3JW
Scotland
PHONE: (+44) 131 650 8523
FAX: (+44) 131 668 3184
email: Magnus.Hagdorn at glg.ed.ac.uk
web: http://www.geos.ed.ac.uk/homes/mhagdorn/
Received on Thu Jun 09 2005 - 01:05:54 BST
