On 7/21/2011 10:15 AM, Ken Knapp wrote:
CF-SAT Listers-
Based on the various ways that bands/channels can be differentiated, I
lean toward the first solution that Tom described (shown below).
On 7/20/2011 3:37 PM, Tom Whittaker wrote:
One simply used a name like
"band" solely as a dimension and defined a standard_name for this so
that code could recognize it as such. Several other variables were
then used to define wavelength, etc., again using standard_names to
alert applications as to what they were.
The problem with the other is that there so many ways to differentiate
channels. Rudi brought up polarization. Other issues include:
- You'd also need frequency as a std variable (since microwave
channels are generally titled that way).
- Channels with same frequency but variable widths [where the center
frequency is the same, but the range od sensitivity changes, like with
SSMIS (http://en.wikipedia.org/wiki/SSMIS) and AMSU which is even more
complex (http://en.wikipedia.org/wiki/Advanced_Microwave_Sounding_Unit)
-Similar frequency but with differing optical path cell pressure
(something sits in the optical path on the satellite ... a la SSU)
http://www2.ncdc.noaa.gov/docs/podug/html/c4/sec4-2.htm
So I would recommend not creating coordinate axis types because there
are so many ways to differentiate "bands".
Thoughts?
-Ken
If one adds a dimension, say, "band", to a variable, an application
might want two different things:
1) some "user readable" coordinates that allow a user to select the band
that they want. this comes up in GUIs like the IDV, and in web services
like WMS. We dont really care what the semantics are, as long as the
user understands what they are selecting.
2) more sophisticated apps might want to do various calculations, based
on understanding what the value of wave length, wave number,
polarization, etc.
For the first case, we need some (any) standard coordinate axis for the
band dimension. I think probably wavenumber and wavelength would
suffice, ie all data could use one or the other.
For the second we can use as many variables as are needed / wanted. One
can make them into auxilary coordinates simply by adding the variables'
names to the data "coordinates" attributes. This doesnt help generic
applications much unless those variables are given a standard name in
order to describe their meaning. So the cf-satellite group could come up
with a set of standard names for this purpose, eg "polarization",
"wavelength_width", etc.
So i would say the proposal could define new axis types "wavenumber" and
"wavelength", and require all conforming files to use one or the other
as a coordinate (regular or auxiliary) for the band dimension (the
actual dimension name can be anything). Define other variables as your
community needs, clarify the semantics using a standard_name, and
recommend that these are used as needed as auxiliary coordinates.
john
