Simple Spectral Access use cases

[Anita Richards]

Belated response - spectral model for interferometry data. This is an
attempt to highlight features consistent with Doug's comments below; it is
in no way a complete model for data cubes; for a more thorough
interferometry model see developments led by Peter Lamb in the radio-vo
and data model groups.

Doug Tody <dtody at nrao.edu> wrote:
>
> Radio data cubes are perhaps best dealt with as images with a spectral
> dimension, as they are currently represented.  It is clear though that
> a 1D spectrum could be extracted from a data cube (Ray was thinking of
> implementing a service to do this).  With future instruments (e.g. EVLA)
> with thousands of spectral channels this could be a powerful technique.
In our questionnair to interferometry data providers, most said that it
seemed much more promising to provide a VO interface to local tools to
extract such spectra, than to make VOs understand the many and volatile
ways of handling visibility data or even interferometry image cubes.
However, that is not in competition with Ray's suggestion for a generic
tool where cobes exist in suitable standard formats.

> Since such spectra can be generated on the fly most of the pecularities
> of radio data might be hidden.  The biggest problem might be how to
> characterize the observable.  Are there any other things to worry about
> that you can think of?
>
> That is, if a simple 1D spectral model includes
>
>     wavelength|frequency|energy
>     flux (observable - need to characterize, eg continuum subtracted etc.)
>     error
>
> what else is needed to describe cm or mm radio data?  Thanks.

Taking as read the information already in RSMv0.8 - other/expansion:
(much of this is in the STC models but maybe not all)

At some stage it is vital to know the velocity convention accurately - an
alarming number of astronomers don't bother to worry about this until they
do what I did and get a lovely continuum observation instead of a line.
+ Vlsr v. Vhelio, need to know date obs and what sort of Vlsr for full
  accuracy.  Nice ?Starlink? program called 'rv' supplies conversions in
  given celestial directions.
+ Vradio v. Vopt v. Vrel.
http://wiki.astrogrid.org/bin/view/Astrogrid/AstronomicalFrequency
is an attempt at a jargon buster with links to rigorous definitions

Interferometer data can have a position accuracy of mas but much older
maser data may not have been phase-referenced ie poor absolute position
accuracy, confounded by the fact that may sources have significant proper
motions so you can't just re-measure the position.

Masers can be very temporally variable and move about in the spectrum
(i.e. epoch is important; many published atlases give a list of
'peak' and 'full extent' but these may change dramatically so cannot
always be used as identifying characteristics.)

Aperture size, e.g. the resolution of a single dish or the beam size of an
interferometer for data in Jy/beam, or the region over which the flux
density was measured for information extracted from a datacube. Also
spatial filtering, see response to questionnaire below.

Polarisation - much maser data are highly polarised, typified by Zeeman
splitting (also other species, e.g. JCMT obs. of CO polarisation).  Hence
spectra/data cubes are often available in any or all of
circular correlations LL RR LR RL (or RR LL RL LR - a source of
confusion...) or the analogous
linear XX YY XY YX etc.
or the Stokes parameters  IQUV
It would be straightforward I think to translate Zeeman splitting LL-RR
into line of sight magnetic field but the user would have to convince
themselves the Zeeman pair was valid...

Single dish often published in Tant as flux units, conversion at least
approx. is possible but not always given in the same paper (cf Mag to
Jy...).

Alternatively maser data may be published as brightness temperatures -
Orion KL recently reached 10^17 K at 22 GHz for a water maser flare!

Multiple lines which may overlap e.g. OH 1665/1667 MHz; a much bigger
issue at mm freqs.  Hence there may not be a linear conversion from a
calibrated but unanalysed spectrum, to velocities - you need to assign the
lines first.

I hope the ISO people have described the use of templates, this becomes
more and more important as you get into crowded regions.  Ultimately you
want access to chemical and atomic databases.... I see there is a talk on
this at ADASS.

Even for radio, one useful thing might be to segregate typical OH/IR
twin-peaked profiles, from the narrower but less regular profiles typical
of star-forming regions (and some thinner-shelled Miras etc...)

Interpretation of HI absorption may require access to accompanying
continuum images to distinguish between absence of HI and mere absence of
background.

---------------------------
Simple Spectral Access Use Cases

Data Providers (e.g., data centers, archives)

    - Data provider name
MERLIN/VLBI National Facility, Jodrell Bank Observatory, University of
Manchester

    - Spectral data collections from this source
Spectra and SEDs in 1 2 or 3D at radio frequencies

    - Characteristics of data (number of spectra, size, irregularly
sampled, nonsampled (as in HE/UV) multiple flux arrays, variance arrays,
etc.)

MERLIN data are taken in discrete frequency bands, within which the
frequency channels are uniformly sampled.  At present the processed data
is all bandpass-corrected.  Access to the bandpass (sensitivity per
channel) is rarely needed apart from reprocessing visibility data or
planning observations but it could be supplied as look-up tables.

MERLIN has a bandwidth of 16 MHz max (until 2005 when e-MERLIN gives us 2
GHz) or less for narrow channels, a resolution of 0.5 kHz in 1.6 GHz (~100
m/s or 10^7 inverse fractional resolution) is possible but only over ~0.25
MHz - this gives a very narrow velocity coverage of tens or 100's of km/s
which may not cover all the emission from an object.

Also, some objects like nearby Mira CSEs and Galactic star-forming regions
may be poorly spatially sampled, both straightforwardly (i.e. partial
mosaics) and by the interferometric filtering of spatial frequencies.
Comparing spectra taken by e.g. VLA MERLIN VLBI gives an idea of the scale
of the emission regions but beware temporal variability.

    - Current data storage format
Processed visibility data FITS in AIPS (on line but not yet public
interface)
Images FITS on-line, published
High-res spectral raw data proprietary format on-line, will be converted
to AIPS FITS probably in medium term.

    - Is the data available online?

Present situation covering the above Q's:

MERLIN continuum and some HI data are stored on disc as processed,
calibrated visibilities.  The observatory log and FITS images are
published via CDS and are coming into the AstroGrid prototype. At present
a user would need to download images or request and download uv data in
order to extract spectra, SEDs and spectral index plots. In the near
future we hope to impliment an interface to AIPS to allow automatic
extraction of SEDs in 1 2 or 3 D. These would be returned as ascii tables
(VOTable should be OK) or FITS images as appropriate.

The data needed to make high (spectral) resolution datacubes are stored as
raw visibilities, and in the near future all we can do is tell people what
is available after a certain amount of off-line work.  However quite a lot
of these data have been analysed and published including lists of maser
components which are used to generate ID spectra at multiple spatial and
spectral locations.  One thing I would like to investigate is linking from
the observatory log to published data from journals.  This is best done
via Vizier in most cases, which means encouraging and expanding the degree
to which journals publish tables directly in Vizier or possibly provide
links.

We also host the HIJASS survey which, along with HIPASS, is the whole of
the sky in single dish mosaics in HI. It is possible to extract spectra
http://www.atnf.csiro.au/research/multibeam/release/
- David Barnes is the person to ask for details.

As part of my VO radio responsibilities I am investigating how to get a
number of other radio archives published, including single dish spectra
such as from the Puschino observatory.  The EVN VLBI online archive at
present is a list of observed positions and calibration information; a few
spectra are available but mostly for diagnostic purposes rather than
astrophysical analysis links to published data should be considered.

    - Other comments

Data Consumers (e.g., analysis packages, e.g., VO demonstrations that
might want to add the ability to fetch and display a spectrum)

    - Name of application, package, demo, etc.
Interface to AIPS/AIPS++ (managed by data providers!) to extract spectra
from interferometry and multi-beam-mosaiced data

    - Summarize capabilities of software
Specialised to handle this type of data (can handle other data with
limitiations); in particular, extracting images or direct info (incl.
spectra) from visibility data;
making spectral index maps;
measuring flux in apertures/per source intelligently to generate data for
SED or high-res spectra;
continuum subtraction from images, data cubes or (sometimes) uv data;
handles polarisation;
understands/converts velocity conventions, Jy/beam v. Jy/pixel etc,
precesses/corrects positions with required high accuracy
should propagate errors correctly.

I believe that primary beam corrections etc. can also be applied to
correct flux measurements.

    - Desired characteristics of input data
Calibrated visibility data or images

    - Desired input data format or formats (e.g., graphics, FITS, XML)
FITS at present. Also ascii/VOTable for access to extracted published
spectra.

    - Other comments

Will expand on interferometry jargon if required (tried to keep this
brief)

best wishes

Anita
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Dr. Anita M. S. Richards, AVO Astronomer
MERLIN/VLBI National Facility, University of Manchester,
Jodrell Bank Observatory, Macclesfield, Cheshire SK11 9DL, U.K.
tel +44 (0)1477 572683 (direct); 571321 (switchboard); 571618 (fax).