Hi Ag,
Do the two vertical modes of measurement each have the same number of points on
the axes? Or if they do not, then would it be acceptable to utilize a single
vertical dimension of size equal to the larger of the two and to allow missing
value flags for the unused space on the shorter vertical axis?
If the answer to either of these is yes, then the problem reduces to the "sigma
coordinate problem". To encode sigma coordinate data in netCDF one creates an
additional *dependent* variable (not a standard netCDF coordinate variable)
whose vertical dimension is the length above and whose values are the
(time-varying) heights (or depths). I believe that you'll find this is
documented in the CF conventions -- together with the conventions that link the
measured variables to these time-varying (or space-varying) Z coordinates.
Hope this helps.
- steve
P.S. There is a specific function in the Ferret program called ZAXIS_REPLACE to
allow you to shift between sigma coordinates and fixed coordinates when working
with the data. I'm sure there are equivalent techniques in IDL and Matlab and I
presume GrADS, too.
===================================================
"Stephens, A (Ag)" wrote:
> Dear all,
>
> I have been trying to convert a colleague to using NetCDF and conforming to
> the Climate and Forecasts Metadata Convention. However, she has some sodar
> data that has an interesting quirk to it.
>
> The instrument sits on the ground and looks vertically up, measuring the
> profile of wind speed and direction. It therefore has two coordinate
> variables: time and altitude.
>
> There are two modes in which the instrument can measure which vary the
> altitude coordinates so during the course of a day the sodar may be run in
> each mode for different periods. This produces output with two altitude
> dimensions.
>
> The easiest option for NetCDF would be to output to a new file each time the
> mode changes. However, it is useful to be able to plot the entire day onto
> one graph. Does NetCDF have a suitable solution to this problem?
>
> My hunch would be to create two altitude dimensions, say 'altitude1' and
> 'altitude2' and two lots of wind parameters that vary with each altitude.
> Alternatively, you can define the altitudes for each time step throughout
> the measurement period (but this is inefficient).
>
> I would hope there was a more elegant solution which would allow plotting
> packages to view the entire dataset as continuous rather than separate
> variables defined on separate dimensions.
>
> Any ideas?
>
> Thanks,
>
> Ag
> -------------------------------------------------------------
> Ag Stephens Ph : +44 (0)1235 446220
> Data Scientist, Fax: +44 (0)1235 446314
> British Atmospheric Data Centre,
> Rutherford Appleton Laboratory,
> Chilton, Didcot, OX11 0QX, U.K.
> Email: A.Stephens@xxxxxxxx Web: http://badc.nerc.ac.uk
> -------------------------------------------------------------
--
Steve Hankin, NOAA/PMEL -- Steven.C.Hankin@xxxxxxxx
7600 Sand Point Way NE, Seattle, WA 98115-0070
ph. (206) 526-6080, FAX (206) 526-6744
