For a rampling discussion on C&R and the level-1 specifications, see Coma Reliability and Completeness .
Reliability
The Herc extended source database consists of 700 objects (including all repeats!), most of which are real galaxies (a lot of them!), some are false detections (e.g., stars, double stars, meteor streaks) and the remainder are unknown sources (generally too faint to classify). Each source has been visually inspected and classified as such. The results are given below.
The reliability is measured as follows:
Reliab = Ngalaxies / (Nfalse + Ngalaxies)
There are far more galaxies than stars or artifacts, and the galaxies also are much more numerous than the "unknown" (faint) sources. The reliability reflects this, >95% for J brighter than 14, H brighter than 14, and K brighter than 13.5 or so. The reliability can probably be increased another 2% or 3% by increasing some of the key scoring thresholds, for example the "wsh" (see Star - Galaxy Discrimination) without adversely sacrificing completeness (see below). In addition, another 1 to 2% may be gained from elimination of artifacts (meteor streaks and features associated with very bright stars). Still, the 99% level-1 specification will still be a challenge to meet in this field as things currently stand.
The normalized galaxy counts (number per sq. degree) is given below.
Completeness
The detection repeatibility can provide a good estimate of the completeness since our sample consists of galaxies with a variety of shapes, morphology and size. Since there are five scans, we expect at most five detections (although it is possible for more detections, since there is also in-scan overlap between coadds, but this area is small). A small fraction of the total area is lost near the edges of the field since the scans have a slight cross-scan position difference (a few arc seconds -- the largest departure is about 25 arcsec) -- this affect has been accounted for.
