(Note: I apologize if any of the following reflects my misunderstanding something about the CF standard -- please feel free to educate me.)
The other problem being that a measurement like sea water temperature can take place in a LOT of different environments (deep, shallow, pole, equator, or next to an active vent or in a lab). To me this suggests that associating "likely" (sensible) values with the CF name is indeed most perilous, unless the CF name explicitly limits the number of environments.
I have always thought there were 4 logically distinct limits:
- the acceptable maximum/minimum associated with a particular measurement method (e.g, a particular instrument or model); numbers outside these bounds are instrumentation/calculation/program errors
- the max/min associated with a particular deployment/location; numbers outside these bounds deserve inspection
- the max/min measured in a particular file
- the max/min for plot limits and color scales.
I've always wished all 4 were available in all data files.
As noted, the last of these is particularly important to the people doing the plotting, but it not knowable to most of the people generating the netCDF file (except in a closed environment, with a small group having end to end control). I would not recommend trying to use it to drive a general-purpose plotting program (or better, use it if present, but be prepared to fall back to the other three).
As just one example, I loved setting flag 'plot limits from -1 to 9, even though the flag range was 0 and 1; this put the flag at the very bottom of the plot. Not something everyone would want to do.
John
At 5:06 PM +0000 12/21/06, Bryan Lawrence wrote:
>Steve is so right!
>
>The ozone hole was ignored for years because someone decided to omit
>non-sensible values: the problem is that the folk who choose "sensible"
>don't choose the colour scale at the viz end ... so you can't rely on
>undershoots and overshoots being obvious to the end-user at first
>parse... here be dragons!
>
>A safe alternative which might save time in visualising large datasets
>would be for the *actual* minimum and maximum to be optionally stored as
>attributes somehow ... this would be very useful for data mining as well
>as visualisation ...
>
>Bryan
>
>On Thu, 2006-12-21 at 09:39 -0700, Steve Emmerson wrote:
>> Be careful. I've seen lots of sea surface temperatures of 38 celsius
>> and higher in my work as a research associate at the University of Miami.
>>
>> Jon Blower wrote:
>> > Dear all,
>> >
>> > As some of you know, we at Reading have developed an online,
>> > interactive visualization tool for CF-compliant NetCDF data
>> > (http://lovejoy.nerc-essc.ac.uk:8080/ncWMS/godiva2.html). In order to
>> > create images of the data that are stored in the NetCDF files, the
>> > tool needs to know a sensible range of possible data values. For
>> > example, for sea temperature a sensible range is -5 to 35 C.
>> >
>> > Many data files specify the valid_min and valid_max attributes: these
>> > define the range of all possible values - outside this range, data are
>> > interpreted as missing values and are thus "hard" boundaries. This
>> > range is often set very wide to ensure that no true data values are
>> > interpreted as missing.
>> >
>> > I think it would be useful if there were a similar pair of attributes
>> > (say "suggested_min" and "suggested_max") that are interpreted by
>> > visualization tools as clues to generating a sensible colour scale,
>> > but values outside this range are not interpreted as missing.
>> >
>> > Do others think this is sensible, and is there anything similar existing
>> > in CF?
>> >
>> > Thanks,
>> > Jon
>> >
>> _______________________________________________
>> CF-metadata mailing list
>> CF-metadata at cgd.ucar.edu
>> http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
>_______________________________________________
>CF-metadata mailing list
>CF-metadata at cgd.ucar.edu
>http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
--
----------
John Graybeal <mailto:graybeal at mbari.org> -- 831-775-1956
Monterey Bay Aquarium Research Institute
Marine Metadata Initiative: http://marinemetadata.org || Shore Side Data System: http://www.mbari.org/ssds
Received on Thu Dec 21 2006 - 17:08:32 GMT
