tehoretical SEDs in VO applications

Miguel Cerviño mcs at iaa.es
Thu May 4 05:42:34 PDT 2006


Dear all,

I just send this comments that may be discussed in the next meeting.
It is just a comparison about how different VO applications manage
a SED building following the SED data model requirements.
The applications compared are VOSpec, specview, topcat, and VOplot.
(Note that topcat and VOplot are general propose tools and VOSpec and
specview are specific for SEDs).

I had find that more or less the current SED data model is fine for  
theory,
and may be, model parameters should be specified in  <GROUP> ...</GROUP>
fields for characterization or provenance (it is an issue to be  
defined in
the theory group). Note that in the theoretical domain, the  
parameters that
define the model are equally relevant than the model results themselves.

However, for the case of applications, (visualization tools mainly) I  
had found some
inconsistence/problems with a VOTable designed in such a way, and  
also internal inconsistency
between different tools. I summarize them here:

a) problem with units nomenclature: There is no recommendation about  
the use of units in the VOTable
fields. In particular, VOSpec and specview use different ways to  
define the units for the flux
e.j. cm**2 vs. cm2 to define the $cm^2$ (in latex). Would it be  
possible to define any
recommendation about that? Otherwise the same VOTable would not be
used for both applications!

b) Metadata and saved files: As pointed before, metadata in  
theoretical models are fundamental to
understand and perform any posterior analysis or comparison with real  
data (or "pipeline" process
for datamining as example). Again, metadata are sometimes equal in  
importance than the simulation result. However:

   - the VOTable metadata are not visualized by VOSpec, sepcview and  
VOPlot
   - metadata otuside <GROUP> are visualized by Topcat, but not  
metadata in <GROUP>...</GROUP>
         (so may be the use of a <GROUP> for characterization of the  
model is not a good idea after all...)

About the VOTables that can be saved by VOSpec, specview and Topcat  
(VOPlot do not save VOTables)
it is found that metadata is systematically lost!! (it means, such  
type of applications are not completely suitable
to retrieve theoretical models thought the VO registry for their  
posterior use!). In particular:

   - specview: do not include niether the original VOTable  <GROUP>  
of <PARAM> but include their own
         <GROUP> and <PARAM>'s
  -  VOSpec  do not include the original metadata
  - and  VOTables saved by topcat include table metadata but
       all <GROUP> parameters are lost.


c) Columns to be plotted:

      c.1) specview needs 3 requirements to plot a VOTable:

        i.- the Table must have a name (only VOTables with <TABLE  
name="whatever"> looks to work
        ii.- utype in x and y coordinates in FIELD must be specified

       c.2) VOSpec needs two (non-standard) <PARAM...> lines to  
specify the columns and only
               two columns can be plotted (i.e. do not include for  
example a column with errors that would be
               useful for both observations and theoretical data.

   Finally, for theoretical SEDs, they are some times stored in a  
multicolumn format, I have not tested
with spectview, but it is not supported by VOSpec.

d) A final comment, it looks that only VOSpec is able to manage a  
Theoretical spectral acces protocol:
It means, perform the query to a server that gives back a table with  
the parameter space covered by the
models that the server provides, and perform a "second" query over a  
subsample of the parameter space covered.



I hope that this comments would help to improve the applications and  
for the theory group

best regards

	Miguel Cerviño




More information about the theory mailing list