Re: spectral data

From: Anita Richards <amsr-at-jb.man.ac.uk>
Date: Fri, 2 May 2003 11:49:45 +0100 (BST)

In response to Frank Valdes' comprehensive and very readable posting,
> Incorporating spectra in the Next Phase of the VO ..
I think there are two more things which flow out of this: - errors
- conversions and conventions

These are covered in e.g. Greisen et al 2002 Representations of spectral coordinates in FITS - is this published yet?

and we need to think about two sorts of conversion 1) Accuracy needed for answering queries - here we need to know the bandpass shape etc. etc. if the user wants high accuracy. 2) Accuracy for Registry Resource metadata and crude conversions - we can probably ignor filter/instrumetn-specific and other details. In other cases also, e.g. line ratios from a single dataset, useful results can be obtained without all the information one would have in an ideal world.

I take issue with one comment:

> However, the 4D proposal is to not make the extension too general, and
> hence, complex. Instead, simply add two parameters to cover the vast
> majority of observational parameter space of interest to astronomers.

Unfortunately - or fortunately - that is a very dangerous thing to say. Polarization is one example of extra axes in image data, let alone other sorts of data like time-series along moving coordinates or visibility data.

Just one example in the image domain: Monitioring of SiO maser emission. Experiments currently published produce: Multiple epochs of RA-Dec-Frequency datacubes (4 axes)

in Stokes I Q U V			       (another 4 axes?)
in 2 or more lines of SiO	               (another >2 axes?)

I am not suggesting we start by figuring out how to represent this but we need to be able to easily add dimensions, and I think we need to tackle polarization after we have a working consensus on spectral coordinates.

In addition, some data use an extra dimension or dimensions for uncertainties, masks etc.

Regarding Ivo Busko's document on Generic Spectral Data Structures, I guess this is just for handling certain types of data? as it does not seem to cover the 3 or 4 D cases, e.g. datacubes, described by Frank Valdes. This is not a criticism, as I am sure Specview is very good in its domain, but it will not meet all the near future needs of the VO.

thanks
a

Received on 2003-05-02Z10:49:51