I have been following the "conventions" discussion on this mail
and would like to add my few cents worth re variable names etc.
I fall somewhere between Tim Holt and Harry Edmon on names and
attributes. I think we need to look at what the purpose of the
'name' is. In other words, I think we are trying to find one
solution to three problems....
1) If we want to be able to identify a variable as ocean_temperature
or ocean_salinity to an application such as a plot program
then a simple data type or name is sufficient as long as it uniquely
identifies the variable. Whether it was measured by CTD or a
current meter or whatever is irrelevant.
2) Temperature is not only Temperature. What are the units and
which defining temperature scale (IPTS-1948, IPTS-1968 or IPTS-1990)
For salinity also ... PPT or IPS-1978? Most users will only want to
know they have temperatures or salinities of the same type. A robust
application sould be able to make the appropriate conversions required
if provided with these pieces of information.
1) The unique identifier datatype (ie. ocean_temperature),
2) The units (such as degrees C)
3) The defining scale (ie. is it on IPTS-68 or IPTS-90)
4) other required attributes...
In physical oceanography the 3d piece of information is critical!
The scale which defines temperature and salinity MUST be known
if one is to use the data to compute any other parameters such as
density.
Item 4 can be critical depending on the variable. For example,
conductivity measurements (required to compute salinity) are usually
reported as Siemens/meter (or mS/cm). It is impossible to compute
salinity properly without knowing the value of conductivity at
a known point we call C(35,15,0) which was used when the instrument
was calibrated. This absolute value is not agreed upon internationally
so salinity is defined in terms of conductivity ratio which side-steps
that issue but requires us to drag whatever value we pick around with us.
3) If we want to be able to use netCDF files to archive our datasets
in a self-describing way, we need a great deal of additional meta-data
to describe them. This is where I think Tim Holt is headed with
his XMIDAS format and is a great example of the potential of netCDF
to solve a nasty problem. However I think its fair to say there will
always be variable attributes which are application specific which
most of us will have little use for. Maybe what is needed is an
attribute which is something like 'contributors_datatype' which
can handle things like Tim's "SBE_Sea_Surface_Temperature1" which
still has data_type_name of 'Temperature', units 'Degrees C' and
defining_scale of 'IPTS-90'
4) I also am curious as to what is meant by Sigma Coordinate Systems..
Rich Schramm.
(Sorry - no fancy logo)
