If things go insanely well in a run, we may discover more supernovae than we can follow properly. If this is the case, we need to have some kind of criteria for choosing which supernovae we will continue to study and which to dump. This file lists these criteria in no particular order and makes no claims to be complete. Feel free to add stuff yourself!
Some other considerations are the redshifts we would like to get, i.e., taking only faint ones in an attempt to get high-z for a melange of magnitudes for a good distribution. Also an attempt should be made to select targets that are visible homogeneously throughout the night.
We would like to observe supernovae far from the core of their host galaxies. By doing so, we minimize the amount of host galaxy that contaminate our photometry and spectroscopy. Keep in mind that the angular extent of a given distance decreases as you go to higher redshifts (at least the redshifts that we care about!) We may also be biasing ourselves towards following progenitor systems in galactic halos or in spiral arms.
This information is listed as one of the scores in tiles. Sometimes the score is miscalculated by the computer but can be determined by hand, most easily with the slice plots.
We would like to observe supernovae that show the largest percentage increase from the reference. (Of course assuming that the reference signal isn't just noise.) This selects candidates that are most likely to be real transient events and reduces the galactic background in our photometry and spectroscopy. Statically, we expect that most of our supernovae will come from bright galaxies so any of these guys may be strange. Could be super-luminous, gravitationally lensed, or just have a low surface-brightness host.
This information is listed as one of the scores in tiles.
We would like to observe supernovae with no apparent host galaxy. An intergalactic supernova would be quite an exciting find, and not all that implausible. And galactic background will not be a problem. Targeting such an object may be risky because it does not match our canonical supernova profile, and may just turn out to be crap. There may be just a low surface-brightness host.
If the scanner is doing his/her job, the identification as a spontaneous creation is part of the 'keep' process in 'tiles'. Their score will be visible with cansynop. Otherwise, I would check that the "distance from host" score is a large number and I would look at the tiles output to look for a host by eye.
The supernova color can be used to estimate its redshift/extinction/stretch. I have a file called ~akim/filtdat/system/sncols.dat which has the predicted color at different redshifts calculated from supernova spectra at a range of epochs.
This information is not readily available. Conversion of counts to magnitudes can be made with the transformation coefficients in $DEEPHOME/calibration/calibresult where the apertures are for 4(?) full-widths. A guess of the supernova epoch is also necessary.
We would like to select candidates that are not close to big honking objects whose light or other detrimental effects effect the region of the supernova.
The effects of a bad neighbor should be apparent on tiles in the form of a bad column or row or a wash of background light. Finding charts or a full image of the field can be looked at.
We should try to choose supernovae that are on fields that have already been well studied. We can get free references, redshifts, and even HST images of the supernova!
We should check if a known galaxy cluster or other object is on that field.
The BTC chip is warped and is not flat with respect to the sky. This could potentially effect our determination of RA and DEC for our candidates. This is particularly important for spectroscopy where accurate pointing is a must.
Use the chip number and coordinates to determine the candidate's distance from the optical center.
High galactic latitudes lower the number of stars in the field and reduces the galactic extinction. However, we may need to have some candidates at low latitude in order to have a objects distributed through the night.
Calculate the galactic latitude from the RA and DEC.
I don't see any strong correlation between color and stretch. However there is a slight correlation between color and redshift. The big problem is how much we trust the calibration for the night. I suspect that there are real correlations can be made using host color, but I doubt that the quality of data necessary for such assesments will not be available at the run. Particularly because the conditions at one of the nights of reference taking was not photometric.