Subject: Cadence strawman
From: Daniel Eisenstein
Submitted: Wed, 19 Feb 2003 16:38:42 -0700
Message number: 73
(previous: 72,
next: 74
up: Index)
Hi all,
At our Seattle meeting, we had a fair bit of dicussion about the
cadence of observations. I wanted follow-up on that; if we can
make a strawman plan here, we can think more concretely about a DRM.
I wrote the following as a first-person strawman, knowing that a) a lot
(most? all?) of the contents came from others and b) probably some things
are wrong. I'm simply trying to boil down the arguments I heard to try
to make the discussion more concrete.
One thing I took away from Zeljko's talk (which is posted as lsst-general/62)
is that we can separate the cadence discussion into two parts: what one does
inside of a few hours, and what one does outside of a few hours (i.e.
inter-day).
Taking the short timescale question, Zeljko argued for sub-minute
scale exposures with some form of CR split, with a return to the
field ~15 minute later. I recall some question as to the timescale
for that return, with an option of a few minutes. In any case, I
will call the complete sub-hour set of exposures a "visit".
A longer gap offers the ability to detect the motion of KBOs in a
single visit, albeit with only limited proper motion precision.
There was a claim that detecting KBOs in a single visit was not
necessary, because one would be able to recover them by playing
connect the dots in the database of one-time detections. This
didn't sound impossible to me, and yet I *would* design for proper
motion detection in a single visit if there were no clear penalty.
It would be much cleaner and computationally far easier.
My impression was that the argument against longer gaps is that one
makes it harder to link up faster moving asteroids. However, it
was not clear to me that this took into account the possibility
of finding fast movers in the CR splits. If one does a CR-split
on both halves of the visit, then with 4 exposures (at least;
PANSTARRS++ would have more) one could detect the sub-minute
motion in both images and correlate it even over multiple arcminutes.
It therefore seems to me that if we believe we can reliably solve
the match-up problem up to 100", then 15 minutes ought to be safe.
15 minutes is about 100 times longer than 10 seconds, so if we
can detect motions between 1" and 100", we will be sensitive to
proper motions over a range of 10^4. Indeed, those numbers sound
conservative to me, so I ask if we could do 30 minutes.
It seems to me that one wants to set the gap time scale as long
as one can get away with, because it improves the predicted position
in future nights, thereby increasing the length of the window for returning
to hammer down orbits.
So, a question to the pundits and/or simulators: does the above scheme
work? If so, is there an argument for going to ~30 minutes?
Is there any argument against a long enough gap to detect KBOs?
The above was probably crafted for a single-aperture design, but it
seems to me that it could apply equally well to PANSTARRS. Obviously
PANSTARRS *could* do all of its CR-splits strictly simultaneously,
but there's no reason it couldn't drive two of the telescopes 30 seconds
behind the other two.
Let me say that other aspects of Zeljko's scheme also sounded very good
to me. The ability to do the second half of a visit in a different filter,
and therefore to get a single color for objects that vary on >hour scales
is *very* attractive. Yes, we would have to pick both filters to be
"asteroid-friendly" but that's a separate discussion.
Also, the ability to get cadences of intermediate time-scales (few minutes)
in the overlap regions sounded useful. If one felt that some aspect of
asteroid recovery or variability had a rare pathology with the "standard"
cadence, one would be able to test this over a non-negligible (~10%)
fraction of the sky.
If a visit consists of 2 pointings, each about 25 seconds (with overheads
and a CR-split), with a unique field of 6 square degrees, then one covers
432 unique square degrees per hour, or ~4000 square degrees in a 9 hour
night.
--------------------------------------------------------------------------
Once we've settled on the routine for a single visit, then we have the
building block out of which to build the day-to-day cadence. I will
proceed assuming that the visit has a gap of at least 15 minutes.
Some major questions are:
1) How much time does one have to return to a particular field before
the asteroids have move too much to match up?
2) How many times per lunation do NEOs need to be imaged?
3) What must be the typical time-scale for basic SNe light curves?
4) How many months must SNe be followed for?
5) How do we arrange the filter cadence?
First, let me propose to remove one aspect from the discussion. Mike Shara
made an excellent case for the utility of novae, and my understanding
was that imaging on consecutive nights was strongly preferred.
Similarly, orphan afterglows strongly prefer consecutive nights,
and really would prefer multiple consecutive nights. I expect that
doing these applications right will essentially require hitting the
same fields every night. This is not possible to do over the entire
sky with a single-aperture design (or PANSTARRS in a "locked" mode).
Hence, I will suggest that we plan to reserve (e.g.) 2 hours per night to
hit some fraction (e.g. 800 square degrees) of the sky every night,
rotating filters as possible. The "focus" area would change from season
to season (say, replacing about 1/3 monthly), so that the whole visible
sky would be covered over the course of about a decade. With this, let's
remove "day-scale" transients (including fine study of SNe) from the
asteroid/SN cadence question.
If that leaves 8 hours for the "standard" program, then that means
3500 square degrees per night. That means it takes about a week
(plus weather) to cover the visible sky. Apparently, we could
get to all fields about 3-3.5 times a month (with no implication here
of a uniform cadence).
For question 1, since (100"/1") times 15 minutes is a day, I can only
presume that returning next day is sufficient to match up all asteroids
(phew). Once we have a second visit, I assume that the third visit can
be scheduled more comfortably.
The question is whether we can get away with longer. My understanding is
that supernovae would prefer a fairly regular cadence of 4-6 days.
I don't know if the asteroids can wait that long! Similarly, I don't
know if 3 visits per month is sufficient for basic supernovae light curves.
There seems to be a tension here that needs to be resolved!
If we could get ~6 visits a month, with at least one on 1-2 day return,
then everyone would be happy. This appears not to be feasible with
single-aperture, 7 degree FOV, with 50 second visits. Either one
changes one of those assumptions, e.g. integrating shorter or
splitting the aperture/FOV, or one decides to cover less of the sky
with the 6/month cadence. For example, if asteroids could deal with
a guarantee of 2 complete coverages per month with exposures on near-nights,
then one could cover 25-40% of the visible sky with an additional 3-4 visits
(with the weather risk going here). This yields a wedding-cake style of
survey: everything twice a month, a third ~weekly, and 4% daily.
Regarding filters, I think it would be a huge mistake not to do the
non-moving variable sky in at least 3 filters. Clearly, the supernovae
need it, but novae in galaxies will want colors for reddening, and any
serendipitous discovery will want it. Indeed, I would rather have 4 filters.
Zeljko's scheme of having one pointing in each visit be of a consistent
color and then varying the second filter seems sound to me (AB one night,
AC the next, and possibly adjusting the second filter with bright/dark time).
For example, in visiting the sky twice a month, one would get three
filters, with the "preferred" one twice.
However, it is not clear to me that these filters need to be the standard
R=4 set. I think we should consider whether a set of wider, possibly
overlapping filters, could work. For example, if one measures (gr) and (ri)
and (r) then one measures g-i more accurately than if one measured g, r,
and i separately (but one measures g-r and r-i worse). Obviously, we
are limited by the desire for image quality, but we should do more
experimentation, especially in the red. We may be able to get a wide
bandpass and a widely separated color.
Comments very encouraged.
Daniel
LSST LSST LSST LSST LSST Mailing List Server LSST LSST LSST LSST LSST LSST
LSST
LSST This is message 73 in the lsst-general archive, URL
LSST http://www.astro.princeton.edu/~dss/LSST/lsst-general/msg.73.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