ObsCore update discussion : adding Axes information in Obscore table

[François Bonnarel]

Dear all,
     my 2 cents below
Le 15/04/2015 19:50, Louys Mireille a écrit :
> Dear all ,
> Here is a summary of discussions on the Obscore data model update, 
> with several iterations
> with Doug Tody, François Bonnarel, Laurent Michel, Daniel Durand and 
> Pat Dowler.
> After the suggestion I made at the last interop in Banff,
> (see 
> http://wiki.ivoa.net/internal/IVOA/InterOpOct2014DM/obcore-ObsdatasetDM_compatibility.pdf 
> )
> we iterated with Doug Tody to enhance the axes' description of  the 
> data part exposed in a dataproduct distributed by Obscore.
>
> These add_ons are meant to enrich the query response of ObsTAP 
> serving  N-D datasets in order to expose the
> dimensions and the nature of axes involved in the data part.
>
> The way the data values are organized in files ( File format, single 
> or multiple extensions in FITS, etc, ..) is not taken into account here,
> as the goal is discovery  and selection of  candidate datasets from 
> the query response.
>
> These values are not meant to be exact but represent a very good 
> approximation of the dimension in each physical axis.
> So we are proposing the addition of
>
>   * s_dim1, s_dim2 = the coverage in sampling elements ( pixels) for
>     each spatial axis
>   * em_dim = the coverage in spectral elements along the energy axis
>   * t_dim = the coverage in the time axis, as number of time bins
>   * pol_dim = the coverage in the polarization axis, as number of
>     polarization states
>
> The nature of each axis is given by their prefix (s (spatial), em 
> (energy), t (time) and pol (polarization) s mentionned in the Obscore 
> previous version.
>
> These parameters are not describing the packaging (this is not 
> providing enough information for cutout yet.
> If the value of the <n>_dim is 1, it means that this is a degenerate 
> axis.
> One cannot have a <n>_dim =0 but NULL is allowed (interpreted as 1).
>
> Please note that the default polarization for any imaging data is I, 
> intensity.
>
> It is important to note that the <n>_dim values are the overall 
> characteristic of the dataset and NOT the value of its WCS axis
> (it might be true for simple cases, i.e. simple image but certainly 
> not for mosaic observation where the gap
>  and/or the presence/absence of a given chip is not specified) .
>
> Here are some examples of various datasets exposed with this strategy :
>
>   * MUSE data cube
>
>      s_dim1   = 300
>      s_dim2   = 300
>      em_dim   = 3463
>      pol_dim      = 1
>      pol_state = I
>      t_dim      = 1
>
>   * 2MASS: 2D image
>
>      s_dim1   = 300
>      s_dim2   = 300
>      em_dim   = 1
>      pol_dim  = 1
>      pol_state = I
>      t_dim = 1
>
>
>   * MEGACAM: mosaic image
>
>      s_dim1   = 20000
>      s_dim2   = 20000
>      em_dim   = 1
>      pol_dim  = 1
>      pol_state = I
>      t_dim = 1
>
>
>   * STIS spectroscopy (1D):
>
>      s_dim1   = 1
>      s_dim2   = 1
>      em_dim   = 1024
>      pol_dim  = 1
>      pol_state = I
>      t_dim = 1
>
>
>   * STIS spectroscopy (2D long slit):
>
>      s_dim1    = 1024
>      s_dim2    = 1
>      em_dim    = 1024
>      pol_dim   = 1
>      pol_state = I
>      t_dim = 1
>
>
>   * ALMA:
>
>      s_dim1    = 1000
>      s_dim2    = 1000
>      em_dim   = 3000
>      pol_dim  = 4
>      pol_state = I/U/V/Q
>      t_dim = 1
> The more detailed mapping of s, em , etc on NAXIS1, NAXIS2, etc. in 
> terms of WCS values could be handled in
> a future */getmetadata /*capability of SIAv2, either via ADQL / Obstap 
> or with a query by parameter, awaiting a
> more complete WCS solution for all astronomical observation.
> A discussion on WCS implementation for instance  at the IVOA interop 
> meeting in Sesto in June will help to go forward.
Yes , see my previous emails on these lists.
>
> There are suggestions to use an optional UCD tag to specify the 
> flavour of these axes , like */em_ucd/*, for instance , already in 
> ObsCoreDM
> to help to disentangle between frequency , wave and energy in the 
> query response.
>
Yes I agree
I think this is interesting and could be used in all axes to specify 
roughly the global nature of the mapping
( s_ucd = pos.gal to describe a dataset organized along galactic axes)
An additional optional tag for the projection could be added (eg s_proj)

The "use case" for this is very simple : to figure out what kind of 
datacube is to be retrieved !!
I think It is interesting  to know if the cube is basically equatorial, 
in tangential projection and the em axis is FREQ or that what is going 
to come is Galactic, AITOFF and the em axis is sampled as a wavelength.
 From this additional information the user can infer what the retrieved 
file will look like and prepare some usage scenarii. But we shoul let 
all this optionnal

Best regards
François


> A draft is currently updated along these lines.
>
> Your comments , and suggestions welcome .
> Mireille
>
> -- 
> Mireille Louys	, Maître de conférences
> Centre de Données ( CDS)		Icube&  Télécom Physique Strasbourg, Pôle API
> Observatoire de Strasbourg 		300, boulevard Sébastien Brant
> 11, Rue de l'Université			CS 10413
> 67000 Strasbourg 				F - 67412 ILLKIRCH Cedex
> http://astro.unistra.fr			http://www.telecom-physique.fr
> tel : 03 68 85 24 34

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ivoa.net/pipermail/dm/attachments/20150416/b6439834/attachment-0001.html>