End to End science with VO from major facilities

1
End to End science with VO(from major facilities)

• Gerry Gilmore

2
my assumptions

• The international virtual observatory is an
  outstanding idea it can make science more
  efficient and more effective, save money by
  removing duplication, and allow higher-quality
  calibrations to deliver better data. It needs,
  and deserves, long-term support.
• It could also fail horribly, by allowing bad
  science. What are the risks and preventions?

3
end to end

• Data must be understood its origins, limits,
  strengths, errors, systematics, etc
• Every experiment must be repeatable
• A reader/referee must know what happened well
  enough to repeat it (in principle)
• The data provider must be able to be credited,
  and to be blamed

4
end to end

• The process is in principle simple
• DATA ? vo stuff ? user ? understanding
• reliable science requires either that an
  astronomer understands the data
  acquisition/calibration/reduction processes
• or, more typically
• cites and uses the results from a well-checked
  public description of that process (eg, HST
  pipeline calibrations)

5
An effect of VO?

• No small research group will have the expertise
  to really understand the limitations of all the
  many datasets VO makes available
• And many simple queries will try to take data
  outside their range of calibration
• ? even major facilities can be mis-used

6
The multi-wavelength challenge
7
The multi-wavelength challenge

• Surface brightness maps vs source lists
• Upper limits, etc

8
Matching multi-wavelength data sets is possible
only for a very expert large team until VObut
how reliably can it be done?
9
Multi-wavelength matching can be very
interesting And is easy in simple
cases Example here The Orphan stream in the
Segue Field of Streams Belokurov etal
2006 Fellhauer etal 2006 Zucker etal 2006
10
Multi-wavelength matching can be very
interesting And is easy in simple cases Example
here The Orphan stream in the Segue Field of
Streams Belokurov etal 2006 Fellhauer etal
2006 Zucker etal 2006
11
(No Transcript)
12
Reference frame GAIA- is opticalradio plan now?
Limits can be intrinsic or extrinsic different
astrophysics, different methods?
13
Thermal IR and Sub-mm surveys new tools for huge
complex data sets and maps are already essential
14
The effect of VO

• VO will certainly democratise astronomy
• Powerful tools can dominate ? powerpoint
• VO needs to empower, not limit
• VO will break the multi-wavelength access
  barrier, and allow more complexity
• There are sure to be serious errors from this!
• Poor or inappropriately calibrated data can be
  used unknowingly in a complex system.
• This will be a huge challenge for referees.

15
Archive or Albatross?

• How much extant data is worth saving?
• How do we decide what to accept?
• Must balance resources old v new

16
Some early lessons

• Unique science exists in the archives
• Retrieving it is hard work (now)
• Lots of defective data are in the archives
• Accessing bad data will destroy AVO
• We need data standards as well as
  interoperability standards

17
An effect of VO?

• No small research group will have the expertise
  to really understand the limitations of all the
  many datasets VO makes available
• Will large expert data centres (CDS, IPAC,)
  become even more necessary how are these to be
  funded, if their role is international helpline
  support?
• Need they exist as entities? Linux model? Virtual
  institutes?

18
IS VO a free lunch for most?

• If so, it will fail.
• VO must retain active participation by most
  potential user communities if it is to be used
  and developed. How?
• Does this create a monster, and inhibit future
  individual creativity?
• Most great ideas, as for VO, come from a few
  exceptional people (PIs) too rigid a structure
  prevents this in future
• BUT software maintenance is expensive, and
  requires a different structure than does
  development

19
The 3 Cs credit, careers, citations

• Careers depend on credit, citations, name
  recognition
• What credit is available to reward the
  provider/calibrator/corrector of high-quality
  data used through VO? Citation!
• Citing what? calibration description
• And similarly, de-merits, or health warnings for
  poor data

20
the VO tomorrow

• Generic justification for public funding
• 1) VO is essential to allow effective public
  access to processed data ? longevity of research
  use
• 2) it also removes the need for duplication
• Stability implies significant continuing support
  and development ? VO career paths, VO management
  structures
• And decision making challenges there is no
  single PI Institution/group who decides?

21
end to end

• All this implies that only high quality, well
  calibrated, data sets are appropriate for an
  ordinary non-specialist, user
• if a dBase doesnt have a published description,
  and known quality, it should not be available to
  the unwary.
• And it must never be utilised unknowingly
• A typical Google search generates rubbish

22
end to end science from major facilities does VO
have a role?

• yes!
• What VO needs to deliver science of the quality
  appropriate to major facilities is the standard
  of quality-control which the major facilities
  implement themselves
• A default elite data set, with optional add-ons
  for the expert. NOT vice-versa.

23
end to end

• ? Extreme deduction 1
• private publication of data is a high-risk
  potential disaster such data should never be
  accessed by the VO by default
• ? Extreme deduction 2
• The default request to VO should access ONLY data
  from the major facilities and observatories
• (have a directory of approved datasets, in the
  same way that papers are refereed? I trust IVOA
  to do this reliably.)
• Corollary 1 experts can switch in other info
  only by making extra efforts, passing barriers
• Corollary 2 the SQL (or whatever) query should
  automatically be delivered to the scientist in a
  format for inclusion in the paper, along with
  relevant references to all accessed data sets

24
end to end science from major facilities does VO
have a role?

• yes!
• and, in fact, arguably ONLY from major facilities
  by default
• and someone needs to house the career VO
  infrastructure system