[Esip-preserve] A Note on Using a Classification Hierarchy to Allow Users to Select the Level of Detail They're Interested In

[Bruce Barkstrom]

In fiddling around with Protege, it occurred to me that we could use
the class/subclass/subsubclass ... hierarchy to allow users to select
how much detail they wanted to see on a particular collection of
documentation.

For example, in describing what I mentioned as a "sampling strategy",
there's a level of detail in which each datum has a categorization
into the source of the measurement in terms of several possible
descriptions:
   - a point in space (longitude, latitude, altitude) and an instant in time
(t),
      e.g. an instantaneous measurement of the solar irradiance or an in
situ
      measurement of temperature at a surface weather station
   - a point in space and an interval in time, e.g. a daily average
measurement
      of the solar irradiance at 1 Astronomical Unit
   - a region in space and an instant in time, e.g. the measurement of
      surface properties within the Point Spread Function of a single pixel
      of a remotely sensed image or the interpretation of a single IR
wavelength's
      radiance as being due to the vertical profile of the weighting
function
   ...
   - the concentration of O3 within a stratospheric volume

This level should probably also include what would be known as the "convex
hull" - the mathematically equivalent structure that one would obtain by
wrapping
all of the data values in a file within a Saran-Wrap covering.  This might
need to
include a four-dimensional convex hull if we need to include time as well as

space.

The space may also need to be expanded to include directions.  For example,
a photo's sampling of space (at an "instant" of time) would need to
incorporate
the description of the directions in which the light that impinged on the
lens were going.  If this is the case, then the potential description may
expand to at least six dimensions (long, lat, time, view zenith, view
azimuth).

Likewise, the sampling description may need to include frequency or
wavelength.  An IR instrument doing temperature sounding will have a
markedly different spatial sampling sensitivity at 20 microns than a channel
sampling at 10 microns.

However, not all users will want or need this level of description.  At the
next level up is probably a summary of the sampling envelope and of
the general pattern one might expect, as in two examples
   - global gridded monthly data (coverage of the whole Earth with
      a time sampling of 1 month) - probably with a picture or notation
      as to the nature of the projection being used in the grid - both
      horizontally and vertically
   - radiances in a raster image that includes both clear and cloudy
      scenes (where the description might include the width of the swath
      in the cross-track direction and the length of the swath in the
along-track
      direction) - noting that a discussion of "resolution" needs care as I
      mentioned yesterday

I strongly suspect that this categorization needs work - and should probably
have its presentation experimentally tested on a group of users made from
both the Earth science researcher community and from some other communities,
such as resource managers and students at various educational levels.

Finally, this approach gets us away from having to make a binary choice
between ignoring the finer levels of detail or not.  I'll grant that if you
haven't
had much exposure to the agonies science teams and instrument developers
have in getting the fine level of detail properly defined, this may seem
like too
much detail.  On the other hand, it you've been exposed to it, you'll be
exceedingly
unhappy about not including it somewhere in the metadata and documentation,
and perhaps even unhappy enough to convene NRC panels to look at it.

Bruce B.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.esipfed.org/pipermail/esip-preserve/attachments/20101212/f1bed1b5/attachment.html>