Hi John, Rich, all,
Thanks for this. Just thought of another question - is it possible to find out
what the min-max range of "natural" coordinates is, or would I have to perform
the transformation (theoretically for every timestep) and search the returned
3D array? I suppose that I could do the transformation for one timestep only
and assume the range won't be very different for other timesteps since the free
surface (or surface pressure) isn't likely to vary too much?
Rich - yes, I probably could use the IDV routines although that means another
dependency for the sake of one routine, plus I need to look at how to
instantiate the IDV objects correctly in all cases. Just requires a bit more
[mailto:netcdf-java-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John Caron
Sent: 04 May 2011 17:47
Subject: Re: [netcdf-java] Questions about handling dimensionless coordinates
On 5/4/2011 9:51 AM, Jon Blower wrote:
I'd like to add support for dimensionless vertical coordinates in ncWMS and I
have a few questions:
How common is it for the transformation between dimensionless coords and
natural coords to be time-dependent? Under what circumstances would this
occur? Does it happen when the coordinate depends on a pressure field that
varies with time?
yes, thats the common case that ive seen: surface_pressure(time) for
hybrid/sigma coordinates. Im not sure about the ocean vertical coordinates.
If the VerticalTransform exists, can I assume that getCoordinateArray() will
always return a valid array, i.e. that nj4 has automatically found all the
auxiliary fields necessary to create the transform (such as pressure)?
yes, it will fail fast (at construction) if theres a problem.
I can see how to get the units of the transformed vertical coordinate (i.e. the
"natural" coords, via VerticalTransform.getUnitString()) but how can I get the
"positive" direction? In other words, if the natural coordinates are in
metres, how do I know whether this is depth or height?
hmm, i never thought of that problem. im guessing a domain expert could look at
and tell us the right answer, which we could then embed into the code.
What would be the most "natural" interpolation method in the vertical when
transforming a gridded field from dimensionless coordinates into levels -
nearest-neighbour or linear? (Or something else?)
dunno - id guess linear or log, depending on the transform. perhaps the
domain expert could look at the above list and give some advice.
Dr Jon Blower
Technical Director, Reading e-Science Centre
Environmental Systems Science Centre
University of Reading, UK
Tel: +44 (0)118 378 5213
netcdf-java mailing list
For list information or to unsubscribe, visit: