Classifying Objects by Medium-Band Observations
- a spectrophotometric 17-filter survey -

project: quality of multi-colour classification


Figure:  The redshift accuracy expected for galaxies and quasars in different simulated model surveys, each using 150 ksec WFI exposure time. The dashed line is a pure UBVRI survey, the solid black line is COMBO-17.


At present, the reliability of the classification and the quality of the estimated photometric redshifts is assessed using a variety of approaches. So far, there is available:

  • spectroscopy of ~20 stars, ~100 galaxies and ~10 quasars in CADIS
  • spectroscopy of ~100 X-ray sources in the CDFS
  • spectroscopy of 7 quasars from 2QZ in the S11 field
  • spectroscopy of ~50 cluster galaxies in Abell 901/2
  • spectroscopy of a few z~1 galaxies in the CDFS
  • and from Monte-Carlo simulations of the survey (see figure)
These sources of information have not yet produced a homogeneous and conclusive picture of the quality of the COMBO-17 classes and redshifts, and its dependence on magnitude, class, redshift and SED type. Spectroscopy of larger and more representative samples is needed to establish final quality maps. Two proposal for such spectroscopy with VIMOS in Aug/Sep 2003 are submitted as part of the ESO/GOODS effort and XMM-follow-up, and should provide several thousand galaxy spectra as well as several dozen QSO/Sy1 spectra to establish the desired quality maps.

The classification is known to work quite well on objects well matched by the template library, including stars of types A to M9, bright QSOs, and 90% of the galaxies. Exceptions to this rule are explicitely Sy-1 galaxies which at z<1.5 often show mixed spectra with contribution from both the stellar light of the host galaxy and the active nucleus. As the classification does not consider required "mixed templates" as of now, Sy-1-galaxies are not identified as such and receive completely wrong redshift measurements. Also, roughly 10% of the galaxies appear to have spectra not well matched by the Kinney et al. templates. These galaxies are even at the bright end found at redshifts wrong by more than +/-0.05. The fraction of problematic galaxies possibly increases in clusters which might point at E+A galaxies to be the problem (which can lead to a solution as well). Fixing these loose ends in a systematic manner might amount to new projects being introduced under the lead of different persons.

Contact person:  Christian Wolf, Oxford


Back to COMBO-17 science page


Last update Dec 5, 2002, CW