Subject: LSST SWG meeting minutes, 1/9/03 in Seattle

From: strauss@astro.Princeton.EDU

Submitted: Tue, 18 Feb 2003 22:47:02 -0500 (EST)

Message number: 71 (previous: 70, next: 72 up: Index)

			LSST Science Working Group
			Meeting, Seattle Washington, January 9, 2003

  Minutes by Sidney Wolff and Michael Strauss
(Apologies these are so late!  Purely Michael's fault...)

Attending: 
Chuck Claver
Andy Connolly*
Kem Cook*
Daniel Eisenstein*
Peter Garnavich*
Ralph Gaume
Richard Green (by phone)
Al Harris*
Zeljko Ivezic
Steve Larson*
David Monet*
David Morrison*
Jeremy Mould
Knut Olsen
Connie Rockosi
Abi Saha (by phone)
Mike Shara*
Chris Smith
Michael Strauss*
Chris Stubbs*
Tony Tyson*
Wayne van Citters
Sidney Wolff
(* denotes member of the SWG)
(I may have missed a few people here; apologies if so). 

***********Novae and other variables
The first presentation was by Mike Shara, who had been charged with
describing the opportunities for studying variable stars, including
dwarf novae, RR Lyraes, W Ursae Majoris stars (contact binaries),
Miras, etc.  Mike stressed a unique opportunity for using novae as
standard candles and as proxy tracers of stellar populations outside
galaxies.  With the LSST it will be possible to study brighter novae
out to about a factor of 2 beyond the distance of the Virgo cluster.
Novae are exquisite standard candles; the relationship between decline
rate and Mv has a 1-sigma scatter of about 0.05 mag.  The magnitude at
peak depends on the mass of the white dwarf; more massive white dwarfs
are associated with bigger, faster bursts.

The decline lasts from days to months; no classical nova has erupted a
second time as far as observers are concerned.  The time scale between
bursts is definitely > 200 years and probably greater than thousands
of years.  There are 10-30 novae per year in a galaxy like the Milky
Way.  The physics of these objects is well-understood; and the
emphasis here would be using them as standard candles, to trace the
distribution of stars outside the luminous parts of galaxies.

The study of novae would benefit from observations with other
facilities but does not need them.  Since novae are blue and hot, the
preferred filter is B but V is also fine.  R would simply reduce the
limiting magnitude by about half a magnitude.  A color may be
necessary for classification, although that is only necessary at one
epoch.  Light curve coverage, perhaps every second night, is required.
For this problem it would be better to cover, say, 10 percent of the
sky nightly than 100 percent of the sky weekly.  Potentially, much of
the follow up could be provided on other telescopes by the cataclysmic
community after the novae are found by LSST (this brings up a broad
general question about the extent LSST should do its own follow-up; we
as a group have not yet come to a consensus on that).  It is, however,
necessary to obtain a measurement at maximum brightness.  The light
curves are independent of stellar population.  If we assume that 10
percent of stars lie outside galaxies and that there are 3000 large
galaxies out to to 2x the distance to Virgo (requiring that we go down
to about 25 mag), then LSST can expect to discover 10^4 novae per
year.
	
With this data set on novae it would be possible to study galactic
harassment, tidal interactions in groups, clusters, etc.  If the
reddening could be determined, then novae in galaxies could be used to
determine peculiar velocities of nearby galaxies, but the bright
background would present a challenge.

LSST could also observe all RR Lyraes from here to Andromeda. These
variables could be identified from their color and amplitude; enough
random observations (say, once every three nights) could be phased to
derive periods.  Macho has established feasibility of phasing.

Ivezic pointed out that SDSS has 700 square degrees of multi-epoch
data.  The meta-lesson from this is that if you work hard on
astrometric and photometric accuracy in your survey, everything else
falls into place, and, for example, looking for variables becomes
quite straightforward.  
  On short timescales, the variable object population (at least to
20th magnitude) is dominated by stars, mostly RR Lyraes in the halo.  If 
one waits longer than few months, variable point sources in the SDSS are 
dominated by quasars.    Cook pointed out that MACHO discovered many 
variable quasars, some with very short time scales.

*********Cadence of observations
Cadence will determine types of variables discovered; Ivezic argued
that it is possible to define a single cadence that is optimized for
the discovery of variables and that works for asteroids-i.e. one
cadence fits all.  While this cadence may not be optimum for any
specific problem, it is plenty good enough for most of the science
proposed to date for the LSST.

The specific proposal for a cadence for LSST made by Ivezic is one
pair of observations separated by perhaps 10 minutes and made in R
plus one other color.  Over that ten minutes, one observes of order 35
interlocking fields, giving measures of variability on a range of
timescales.  A few nights later in a given dark run, the pair is
repeated in R and a third color.  The argument is that colors are very
helpful in sorting out the properties of main belt asteroids, and this
cadence gives color information.  Moreover, this gives some color
information for variable objects of all sorts.  This type of spacing
allows the linking of observations of a moving object (Near-Earth
Asteroids (NEA), Main-Belt Asteroids (MBA), and Kuiper Belt Asteroids
(KBO)), without confusion.

  In particular, most of the moving objects one is likely to see are
main-belt asteroids.  "Connecting the dots" for the faster-moving
NEA's will be difficult with this background.  At the limiting
magnitude of LSST, the average separation of MBAs will be 2.3 arcmin;
since MBAs move 3-18 arcmin per day, their positions will be
scrambled.  Observations can be linked by obtaining two observations
in a single night to get the linear velocity.  To select the optimal
time between observations, consider KBOs.  With 50 mas astrometric
accuracy, one needs to wait 10 minutes as a lower limit to detect
motion.  Monet cautioned that 100 mas may be more likely with LSST
because of short exposures.  Ivezic then proposed that the scanning
pattern for LSST should be to go up and down perpendicular to the
ecliptic, covering 1/3 of the sky per night with 2-3 revisits per
month.

The scan pattern means that the overlap areas of the circular field of view 
are observed at 25 sec intervals; the repeated observations observe specific 
areas at a range of intervals from 24 seconds (assuming two 10-second 
exposures followed by 4 seconds of overhead, which is probably 
optimistic) to 22 minutes.

The details of the plan can be found at:
http://www.astro.Princeton.EDU/~ivezic/talks/AAS201lsst.ps

*****Requirements on filters, especially u band
Claver stressed the need to begin putting numbers on the LSST
requirements.  What slew times do we need?  How good does the image
quality have to be (PSF width, shape, anisotropy, uniformity, etc)?
How far into the blue do we need to go and are we willing to sacrifice
red performance to get it?  The optics of the telescope are strongly
affected by this decision.

The latter question resurrected the debate about the U-band.  U is
very useful for distinguishing white dwarfs, RR Lyraes, and low
redshift quasars.  On the other hand, LSST can use time samples to
find quasars; Sloan has shown that every quasar varies if you wait
long enough.  Similarly one can find white dwarfs from their proper
motions.  There appeared to be no argument for obtaining U as part of
the main survey.  It would be useful to measure U during the
multi-color survey of the static sky planned for the initial years of
operation of the LSST.  However, the primary emphasis remains on
observing at red wavelengths, and performance at long wavelengths
should not be compromised to optimize U performance.

  Stubbs returned to the issue of filters.  He argued that an
observation should be viewed as a multi-color measurement and that the
focal plane should be partitioned into two colors, so that
measurements are always made in two filters.  Abi Saha pointed out
that it will be necessary to trade multi-colors against sky coverage.
Kem Cook stressed the need to develop the arguments for obtaining
color curves; i.e., do we need light curves for variable objects as a
function of time?  Supernovae are one example where you do need full
light curves in several colors, so as to be able to solve
simultaneously for extinction, SN type, and K correction.   For some
science goals, repeat observations in one band most of the time, and
in multiple bands some smaller fraction of the time, may be
appropriate.  Note that Zeljko's scheme described above gives
time-variable information in R on timescales of 15 minutes, obtaining
a single color for objects on this timescale.  Every time this field
is redone, a different color is measured.  This is an interesting
compromise approach to this problem.  

  Jeremy Mould suggested that we needed to think in terms of data
products-brightness in a single filter as a function of time, color at
a single time, and color as a function of time and determine what the
science problems included in the design reference mission really
require.

  Al Harris suggested that it makes sense to use LSST for followup only
where the density of objects is high; this is true of faint asteroids,
where we can expect to see several asteroids at mag 24 per field of
view.  If objects are sparse, then follow up should be done with other
telescopes.

*********Tsunami's 
Dave Morrison said that there will probably be a workshop on the
tsunami hazard in ~6 weeks.  He passed out a written summary of where
things stand today.  Basically, there is good agreement about what
happens when an impact occurs in the deep ocean and how it propagates
in deep water, but not about what happens when the tsunami reaches a
coastline; how does the wave break, and how does it run up to the
shore?  There is some modest agreement that if you are more than 1 km
inland, you'll be OK, but of course, a lot of the world's population
lives within one km of the shore...

********KBO science
Gary Bernstein said that the role of the LSST in KBO science is not
discovering rare, bright KBOs, which are being found now, but rather
in obtaining enough statistics to use KBOs to construct a fossil
record of the outer solar system.  There are now ~800 KBO's known,
enough to *start* to see dynamical structure.  A reasonable goal is to
find 10^5 KBOs.  LSST gets much of the KBO science for free---those
that can be seen at 5 sigma in 20 sec exposure.  There are probably
20,000 such objects.

It will take a specialized cadence to find some of the rest: going to
26th will require coadding over 1 hour of exposure.  KBOs typically
move 1 arcsec per hour (4 arcsec at opposition); if observations are
made at quadrature, it would be possible to co-add and obtain the
equivalent of 3600 sec exposures.  With 3-4 such measurements, it is
possible to get orbits.  Such a campaign sustained for a year along
the ecliptic could yield a million objects, and would allow us to map
out the full ecliptic plane.  This would allow us to do things like
look for resonances (analogous to Trojans) due to Neptune.  Note that
a single observatory at a latitude of +/-30 degrees is not well suited
for observing the whole ecliptic. 

Once a KBO is found and its orbit is known, it is possible to use 20
sec exposures to recover them (if they are not *too* faint) learn
about them since their position will be known within a PSF.  Note that
detailed colors are not needed; the spectra are usually well-described
by power-laws, so one color is adequate.  Of course, one doesn't
necessarily want to observe these through an arbitrarily wide filter,
as the PSF of such an observation can be quite poor. No atmospheric
dispersion corrector will work over three degrees

Bottom line: KBO science doesn't drive filter choices or cadence but
long exposures and a full ecliptic survey would yield great benefits.

*********Astrometry
Dave Monet defined the key questions about astrometry with the LSST to
be: What is likely to be achieved with this telescope in terms of
astrometric accuracy?  And what is the first derivative of the
accuracy with respect to exposure time, image quality, etc?  LSST must serve as
its own precursor telescope since it reaches much fainter limits than
any existing or planned astrometric survey.  The LSST astrometric
catalog also needs to be self-improving.  A calibration plan is
needed.  In other astrometric news: Early experiments with OTCCDs are
encouraging in that positional measurements appear to be repeatable at
the expected levels.  The SDSS does astrometry to 40-50 mas accuracy.
It has been suggested that this is limited by coherent anomalous
refraction in the atmosphere, but Dave wants to check this; in
particular, it may be that most of the power is on large (>30')
scales, meaning that relative astrometry can be quite a bit more
accurate.  It will be important to characterize this as a function of
scale and exposure time, using SDSS and other wide-field data.  The
only need for absolute astrometry mentioned was asteroid work, where
the requirements are quite a bit looser.  

  On astrometric science: LSST, even with its high astrometric
accuracy, will not beat comparison with the POSS (i.e., with the much
longer time baseline), at least for objects brighter than about 20th
mag.  What LSST will do particularly well is parallax and astrometric
wobble studies; the various cadences suggested all do fine for that.
It will be possible to look for wiggles, i.e. of L and T dwarfs, to
detect low mass companions with periods maybe up to 20 years.

*********Photometric redshifts
Andy Connolly is working on photometric redshifts for weak lensing,
evaluating the performance of various combinations of filters; going
for 4-5 filters buys wavelength coverage and improves redshifts.  If
the data one gets is of too low S/N, the redshift errors not only
degrade, but become non-Gaussian. 

*********Putting it all together
Michael Strauss then wrote down some of the issues that need to be
quantified as part of the DRM:

*Sky coverage
*Filters
*Image quality
	Weak lensing:  
	All agree on short exposures to turn systematics into statistics
*Photometric accuracy
	Should detection of planetary transits be a requirement?
	(This requires good relative photometry).  Is the cadence
	suitable? Sloan controls systematics well and can beat down
	photometric errors by repeated measurements.  Can LSST do as
	well? 

*What is the typical interval of time between repeat observations of a
given area of sky? 
   Al Harris reported on an analysis of this question carried out by
Ted Bowell.  Ted was unfortunately not there to give us the details
(fogged out in Phoenix, while we enjoyed beautifully clear skies in
Seattle; go figure!); it differed quantitatively but not qualitatively
from Zeljko's analysis.  We hope to have further discussion of this at
the next meeting.  In particular, Zeljko argued for a single cadence
for all types of moving objects, while Ted was less convinced (in
absentia). 

As we saw from the variable star discussion, a cadence of 1-2 day
repeats is necessary for the variable star science.  But the full sky
is not needed.  Note that this is much too fast for SN, who would
prefer repeats on 3-8 days, over the full light curve. 

*For astrometric and photometric calibrations, one gains by maximum
overlap of fields.  We need to think about this in the context of the
"presurvey" as well, when our astrometric and photometric standards
will be laid down. 

  The meeting ended somewhat inconclusively on these discussions.
There is a general sense that we've had enough discussions on general
science goals and order-of-magnitude estimates of what we need, and it
is time to get specific.  We don't really have quantitative statements
of what deliverables and requirements we have for each of the science
goals, and we have thus far not really tackled the question about
whether we want to do one single survey in a single mode, or several
different surveys in different modes.  Michael posted (lsst-general
57) a strawman proposal of what the different science programs might
want, but did not try to stitch them all together. 

   Our next meeting will be a phone conference on Thursday, February
20 (2 PM Eastern Time), followed by a face-to-face meeting in Tucson
on March 18-19.  The push here will be to try to make progress on the
above; let's get some of this stuff written down, quantitative and
specific!  


LSST LSST LSST LSST LSST    Mailing List Server   LSST LSST LSST LSST LSST LSST
LSST
LSST  This is message 71 in the lsst-general archive, URL
LSST         http://www.astro.princeton.edu/~dss/LSST/lsst-general/msg.71.html
LSST         http://www.astro.princeton.edu/cgi-bin/LSSTmailinglists.pl/show_subscription?list=lsst-general
LSST  The index is at http://www.astro.princeton.edu/~dss/LSST/lsst-general/INDEX.html
LSST  To join/leave the list, send mail to lsst-request@astro.princeton.edu
LSST  To post a message, mail it to lsst-general@astro.princeton.edu
LSST
LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST LSST