Re: TASS Database, plans and progress.

[aah@nofs.navy.mil]

Tom wrote:
>Yep.  I think this is a problem we will have to work out over time.  I would 
>figure that we start at the equator, and slowly work away from it creating
>our own standard star fields as we go.  Over time we have our own equivalent
>of Landolt standards.  If we are looking at 45 degrees, then there might be
>more error in going to a Landolt standard than in using something closer.  I
>think only time and a lot of data will tell.
  This is not advised.  The UBVRI system is currently defined by the set of
Landolt standards; you measure all stars with respect to those standards.
If you create your own set of standards, you are on a different system.  It
might be close to Landolt, but it might also be systematically off by several
hundredths, especially for odd-ball stars.  The preferred observing sequence
is to periodically go off and measure Landolt stars, making sure you get
red and blue standards and high/low airmass standards, in order to transform
all other star measures.  This means that photometric nights will require
a fair amount of non-survey frames, and is one reason why I suggested that
the Flagstaff mark IV site do quite a bit of such observations since we
(normally) get more photometric weather than other sites.  Use Flagstaff to
standardize the fields; use other sites to do the differential photometry
to find variables.

Chris wrote:
>I'll also guess that if we start with a 2.5M entry seeded catalog and start
>processing observations the catalog will grow without limit.
  I'd agree, if we don't also pare down the catalog periodically and remove
those objects with only 1/2/few observations.  I'll say it again:  keep the
multiply-observed objects on-line and do what science you can with that
dataset, and put the singly-observed objects in a different database for
those who want to search for moving objects or other single events.  You can
see how you will get swamped with the current setup.
>This mean [transform] error is computed once per what?  Once per site per
> night for each
>filter? or just once per site per nite?  I assume it is just to delta between
>the "true" magnitude and the observed, transformed magnitude averaged over some
>group of standards.
  Once per site per filter per photometric night.  For non-photometric nights,
you would have to use Tycho stars to set the zero points and the slope
coefficients from previous photometric nights to color-correct.  That will
give you reasonably close (but not exact) magnitudes for stars on poor
nights.  And then, don't use the non-photometric nights in the mean
magnitude determinations of tass_cat.  Therefore, I see two branches (plus
some comments) in the processing:
(1) photometric night.  Use all Landolt standards, determine color terms
    and zero points, calculate standard deviation of fit to Landolt standards.
    Create new transform entry.  Use this entry for calculating magnitudes
    for all stars observed that night.  Mark these observations as 'good'.
(2) non-photometric night.  Use Tycho stars to set zero points in each
    frame, and previously determined color terms to correct the magnitudes.
    Mark these observations as 'poor'.
(3) Any observation that is not detected in two or more different filters
    on the same night from the same site has to fall into the 'poor' category
    since you cannot color-correct.  You can set zero points from either
    Landolt or Tycho stars for these entries.
(4) What is the difference between a photometric night and a non-photometric
    night?  Usually it is the quality of the fit to the Landolt standards.
    If that fit is good to +-0.03mag or so, I'd call it photometric.  You
    can get picky, and say that the night was photometric for a few hours
    before/after some clouds came in, but I'd leave that determination for
    the experienced observer and just do a black/white division.
Then
(a) For tass_cat, use all observations for determining coordinate match to
    the seeds and for updating the coordinates.  Use only 'good' observations
    for updating the magnitudes.  Note that this means you have a number
    of 'nobs' fields that will be required: nobs(radec), nobs(V), nobs(R),
    nobs(I).  These are needed in order to determine the sigmas from the
    mean magnitude and sum-of-squares fields.  Note also that you should
    use the raw magnitude error + transformation error as a weight when
    adding a new observation into the means.  Logically complex but necessary
    to do it right.  If you look at my 'master' program, I actually did
    an iteration pass to further remove highly deviant points; Mike uses a
    median least squares solution which also works.
(b) Variable star searches can be even more complex.  The easiest is just
    to look for objects that have large sigmas in the tass_cat.  Harder
    searches try to use non-photometric nights by calculating differential
    magnitudes with respect to local secondary standards (like my difcal
    program), or else calculate the Welch-Stetson index.
(c) First things first: do a good job on the tass_cat mean catalog, compare
    with known stars/standards in the strip.  When you are sure that you
    are getting good results with respect to 'constant' stars, then branch
    out to the variable search and make sure you detect all of the known
    variable stars that you should, and can create light curves that match
    published data.  Finally, extend the search to locate new variables and
    study them.  You have to take logical steps.
Arne