On the "when" and "where"

[Rob Seaman]

Arnold Rots writes:

> may I respectfully suggest to consider a proper STC object to specify 
> the "when" and "where" of an event?  Since we do have a standard way 
> of expressing this kind of information in the VO context, it would 
> seem advantageous to use it.  For one thing, it will promote 
> interoperability with other parts of the VO.  For another, the 
> information specified in version 0.3 is not really adequate and a 
> proper STC object (to be precise, ObsDataLocation) would take care of 
> that.

Let me expand on Arnold's cogent suggestion.  Here is the v0.3 sketch 
of what a VOEvent will provide for spatial and temporal coordinates:

Coord1 (double, angle in degrees)
Coord2 (double, angle in degrees)
  Frame (string, default J2000)
   Time (double, modified Julian Day)
TimeUncertainty (double, fraction of a Julian Day)

As one would expect from the early version of the standard, this is a 
rather slim cartoon of the real needs for appropriately characterizing 
a given event.  For instance, solar system events are mentioned in the 
draft.  One imagines that we wouldn't restrict events to the two 
dimensional surface of the Earth centered celestial sphere, but rather 
also support a third dimension when appropriate - or, say, Jovian 
instead of Terrestrial centered coordinates.  My perception is that STC 
would provide this support in a straightforward fashion.

Personally, I'm more concerned about capturing appropriate temporal 
coordinates.  A scalar time (whether MJD or not) plus an uncertainty is 
insufficient.  This also currently appears to not be fleshed out that 
much more in STC:

     <TimeFrame>
       <Name>Time</Name>
       <TimeScale>TT</TimeScale>
       <TOPOCENTER/>
     </TimeFrame>

There still seems to be a focus on scalar time, but the addition of a 
timescale to correspond to the frame of the spatial coordinates is 
absolutely critical.  For instance, it is quite likely that there will 
be a major upheaval in the UTC standard in the next few years.  We must 
retain the flexibility of using timescales other than UTC.  (Note that 
this applies equally to derived representations such as MJD.)

One is confident that an appropriate representation can be found for 
capturing the celestial coordinates (plus errors and frame) along with 
a scalar timestamp (plus error and timescale) for a typical point 
source event.  We must also appropriately characterize less point like 
events - less point in time as well as space.  Some events will be 
observed to start, and then to end.  Examples from the draft include 
solar flares and volcanic activity on IO, but could also include any 
number of extra-solar phenomena stretching from stellar to 
extragalactic astronomy.

What about periodic events?  LSST should be a wonderful machine for 
discovering variable stars, for example.  One would imagine that the 
appropriate way to report this would be via a period and zero phase 
epoch.

I won't belabor my point.  I doubt that Arnold and I are the only folks 
to recognize that the general case of an astronomical event requires 
quite a rich semantic model.

Rob Seaman
NOAO
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/enriched
Size: 3180 bytes
Desc: not available
URL: <http://www.ivoa.net/pipermail/voevent/attachments/20050322/c717948b/attachment-0001.bin>