Another thought on this, guys:
What is the merit of setting upper bounds on non-detections for transients that later appear and vanish, and storing those in DB...? For ESSENCE we do forced-centroid photometry at locations of interest, and store the resulting fluxes and uncertainties. chris. Quoting Tim Axelrod <[EMAIL PROTECTED]>: > Hi Kem, > > Thanks, that is very helpful to have. If I understand what you've > done, the rough averaging of the stellar density gives about 80000 / sq > deg, vs the 10000 I assumed. This underscores the point that the > sizing of the time-dependent database will depend *critically* on > whether we have a galactic plane exclusion zone, and if so, how big it is! > > So I agree with nearly everything, but do have one significant > disagreement as well. First, I am in complete agreement that we want > lightcurves for all stellar objects, not just those above some threshold > in a difference image. Where I disagree (I think) is about extending > this to objects that are detectable only in image stacks. This is > obviously what's needed for the static properties of galaxies (and stars > as well), but seems pretty useless for variability studies. I'm > inclined to say that unless an object has a useful SNR in a single > image, it will not have a lightcurve, but only static properties. > Obviously this is one extreme strategy, and yours is the other. In the > middle one can do photometry of sequences of stacks, each of which is > made by N individual exposures, where N = 4 or 10 or whatever. I > suppose there might be some interesting science there, but I do have my > doubts. > > So, I think we have a factor of two or so uncertainty based on how we > answer that question, and probably at least another factor of two > arising from how far into the plane we choose to go. > > Cheers, > Tim > > Kem Cook wrote: > > > Keywords: DataAccWG > > > >Hi All, > > > >I am responsible for the large number of stars which Jacek has in the > >current sizing estimate. > > > >I have had a look at various stellar density studies. A recent > >theoretical model of the Milky Way is given in Robin etal > >(astroph/0401052) which also presents a number of recent observational > >results for comparison (and more-or-less all of the observational results > >have higher stellar densities). I also think Hall etal (ApJS 140:185, > >1996) is relevant. > > > >I am assuming a depth of a single exposure is about 24.5 in r. I have > >estimated the galactic latitude and longitude coverage of the WL survey > >and other surveys (NEA, KBO, SN) from my experience with the LSST > >operations simualtor. I estimate about 6,000 square degrees in the > >galactic latitude range of +/-10-20, about 6,000 degrees in the range > >-20-40, and about 10,000 degrees in the range -40-90. This is > >conservative since I have no fields within 10 degrees of the galactic > >plane. From Robin, I estimate about 300,000 stars per square degree at > >10-20 degrees, 5,000 stars per square degree for 20-40 and 3,000 per > >square degree for 40-90. These are some sort of geometric mean for the > >rapidly declining density as we move away from the plane. For a sanity > >check, Hall etal get densities of 25,000 per square degree averaged over a > >range of latitudes from 35 to 63--clearly higher than my Robin etal model > >estimates. This sums to 2,000,000,000 stars in the database for single > >exposure depths. > > > >I conservatively assume that with 10 years of co-additions, we will see > >about twice as many stars. I also assume that the mean properties of all > >detected stars for a data releaese are part of the deep database. > > > >Now, there is a choice which needs to be made. We can release light > >curves for all detected stars, or we can only release light curves for > >stars which are detected in the difference images (ie variable stars). If > >we only release variable objects, this will decrease the number by about a > >factor of 20 (5% variability). This would be the cheap way to do things, > >but might impose some serious loading on the archive site by at least two > >sets of users: low amplitude pulsation studies and planetary transit > >studies. We will have to set some limit (3 sigma? 5 sigma?) for > >detections in the difference images which will be greater than a > >detectable signal in the folded light curve of a periodic variable. So, > >either we release light curves of all detected objects and let people > >analyze them as they will, or we face the prospect of having people > >generate light curves of all detectable objects in order to do this > >analysis. > > > >This sizing has not explicitly accounted for AGN and QSO variabliity, but > >this is well included in the 5% variability assumption for the 'stellar' > >objects. > > > >I think we may actually detect more stars in single exposures than I have > >estimated based upon the fact that we will likely have nearby, resolved > >galaxies in some of our fields and we will be observing the ecliptic, > >probably as it passes through the galactic plane. I also think that a > >factor of two is conservative going from single exposures to DR20 depths. > > > >cheers, > > > >Kem > > > >_______________________________________________ > >LSST-data mailing list > >[email protected] > >http://www.lsstmail.org/mailman/listinfo/lsst-data > > > > > ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program. _______________________________________________ LSST-data mailing list [email protected] http://www.lsstmail.org/mailman/listinfo/lsst-data
