RE: CUAHSI Observations Data Model

NOTE: The galeon mailing list is no longer active. The list archives are made available for historical reasons.

Probably the key difference between the OGC SWE and conventional earth
observations approaches is that SWE is based on the Observations and
Measurements (O&M) model, which ties observed values to a _feature_
rather than a location.

The reasoning behind this follows more or less directly from the General
Feature Model (GFM) as described in the OGC Abstract Model, and also in
ISO 19101 and ISO 19109.
In the GFM the "feature" is the central (meta-)concept, carrying all
characteristics of the domain of discourse.
In contrast, geometric objects (points, lines, polygons, etc) are
abstractions that may describe some aspects of a feature (e.g. location,
shape) but are *not* features in their own right.
ISO 19107 (the "Spatial Schema") makes it clear that a geometry may only
carry spatial of geometric properties.

As an example in plain language, a *point* does not have a
temperature/pressure/chemistry, etc, the *material at a point* does!

In the O&M model this all is explicit, through the so-called "feature of
interest" which is the feature instance whose properties are under
observation.
Note, however, that it is often important to distinguish between the
"proximate" feature of interest (pixel/scene, station, sounding, well,
transect etc) and the "ultimate" or "project" feature of interest
(ground cover, water body, aquifer, rock unit, atmosphere, etc).
The former usually embody the sampling strategy, while the latter embody
domain semantics.

In conventional earth-observation science, the existence of a sampling
feature is often elided: its position used as a proxy.
This is also followed through in the ISO "Coverages" model, which
assigns properties by location.
There is an implicit "feature" encompassing the entire coverage domain,
with the coverage range describing the distribution of the property
value within the bounds of this feature.
This implicit feature, if described at all, is in the coverage
"metadata" somewhere.
This is all fine in practice, and because O&M describes a *conceptual*
model there are many potentially conformant serializations or
implementations.
The OGC SWE suite of standards are relatively explicit about it.
But IMHO the key point is to recognise that there *is* a feature of
interest, not just points, lines and polygons.

Simon




  • 2007 messages navigation, sorted by:
    1. Thread
    2. Subject
    3. Author
    4. Date
    5. ↑ Table Of Contents
  • Search the galeon archives: