HI,
I agree with Simon's description - but it would not be difficult to
generalize the current coverage concept to allow the domain to be other than
a spatial-temporal region.
Ron
Sent: May 12, 2005 6:53 PM
Yes.
Perhaps the most important differences between the GML/ISO concept of
"coverage" and netCDF and allies are:
1. netCDF etc support rather generalised mappings between arrays.
Thus the "domain" of the data (the independent variable, if you like) is not
really contrained.
In contrast, a geospatial coverage is a specialisation in which the domain
must be spatio-temporal, i.e. between 1 and 4 dimensions.
As Ron points out, JP2K is even more restricted in this sense - it only
supports 2D domains.
2. There is explicit support for registering the domain of a geospatial
coverage to real-world locations - the most well-known is where the domain
is a rectified grid, but other geometries may be used which are tied to a
spatial reference system.
3. As Ron points out, an "Observation" is the act of collecting some values
associated with a target.
The "result" of an observation is essentially parallel to the "range" in a
coverage.
The "target" of an observation is parallel to the "domain".
The Observation viewpoint focusses on the observation event, and is
primarily a means to access metadata associated with property values - how
they were obtained, when, by whone, etc.
This is of interest when trying to assess data reliability, and for data
insertion.
Futher down the chain, Observation/result values will often be collated into
"coverages" in preparation for some analysis, anomaly or feature-detection,
etc.
But it is very much the case that the structure of a "coverage" when
expressed as a map of the range over the domain is the same as the structure
of a an observation which maps a result to a target.
See https://www.seegrid.csiro.au/twiki/bin/view/Xmml/ObservationsAndMeasurements
for some detail.
Simon
