On Behalf Of Sean Forde
Hi Ben,
I have CCed the GMLJP2 list on this to see if anyone there has anything to add.
I think I have a basic grasp of how GML can be used to specify the relationships among
multiple "images" in a JPEG 2000 file. I envision this in the case of
hyperspectral satellite imagery. But I have a couple questions regarding how this would
connect to GALEON.
First, how might GMLJPEG2K be used in terms of Web Coverage Services?
I see several possible uses for GMLJP2 in a WCS. First, GMLJP2 is a natural fit
as an output format for a WCS. Since it contains GML and multiple images, it
meets or exceeds the capabilities of the standard WCS output formats (NITF,
HDF, DTED, etc). Compression makes it small, GML makes it rich and
self-describing.
Agreed
A second use is as a source data format for a WCS. That is, a WCS can use a
GMLJP2 file as a data store from which it can serve coverages.
Agreed
On the back-end, GMLJP2 is simply a good format for the exchange of rich geographic data.
A WCS that runs off a database, for example, may "ingest" the GMLJP2 file into
its internal data structures, and reproduce GMLJP2 coverages in response to GetCoverage
requests. Internally it stores the dat a in some proprietary way, but externally it
accepts and produces GMLJP2 data.
Agreed
Would each image in the JPEG2K be a separate coverage?
Some WCSes may want to do things this way. I am not sure if it is possible to
present a GMLJP2 file, which is essentially a collection of coverages, as a
single coverageOffering in a WCS.
I think this is a WCS question.
Second, if you had the output of a weather forecast model instead of
hyperspectral satellite imagery, would that work in the GMLJPEG2K you have
worked out? In this case, you would have to deal with multiple parameters (
e.g., pressure, temperature, humidity, wind speed and direction, etc.), a third
spatial dimension, and a set of forecast times for each.
Multiple parameters is handled now. We have already created examples with
(pressure, temperature) vs spatial location. I expect that that the other cases
can be handled also, although they were not experimented with in the
trial. Let me see if we can construct some examples of each.
To the extent that this is may be expressed in GML, it may be expressed in a
GMLJP2 file. I would guess that there are several different ways to do this
within GMLJP2, but I haven't sat down and done it so I don't know for sure.
I guess there's actually a third question that is sort of a combination of the two issues. Namely, in the
case of forecast model output, one could specify an almost limitless set of possible "coverages."
For example one might ask for a coverage showing variations in pressure with longitude and height as the
"axes." Or another one that's commonly used to study phenomena like El Nino is a longitude by time
"coverage."
The limitless aspect is handled in GML by application schema. We need to look
at the case of non-spatial domain description (time is handled) since this had
a distinguished role in ISO 19123 on which GML is based.
These are some of the issues we face in the GALEON WCS experiment and I
wondered if your group has considered them and come up with any conclusions or
solutions.
Ben, we have not considered these use cases in great depth, and so cannot offer
any useful conclusions. I do think that this is fundamentally a GML question.
We need to answer the question of how to express these relationships in GML,
and then we can apply this to a GMLJP2 file.
Perhaps at the upcoming OGC meeting in Bonn we schedule some time to sit down
and work this out. I think we need domain experts from GALEON and some GML
experts, get them in a room, and see what materializes. Does this sound like a
good idea to you?
Yes that would be appropriate.
