2004ApJ...600L.167M

We use extensive multiwavelength photometric data from the Great Observatories Origins Deep Survey to estimate photometric redshifts for a sample of 434 galaxies with spectroscopic redshifts in the Chandra Deep Field-South. Using the Bayesian method, which incorporates redshift/magnitude priors, we estimate photometric redshifts for galaxies in the range 18<R_AB<25.5, giving an rms scatter σΔz/1+z_spec≤0.11. The outlier fraction is less than 10%, with the outlier-clipped rms being 0.047. We examine the accuracy of photometric redshifts for several special subclasses of objects. The results for extremely red objects are more accurate than those for the sample as a whole, with σ=0.051 and very few outliers (3%). Photometric redshifts for active galaxies, identified from their X-ray emission, have a dispersion of σ=0.104, with 10% outlier fraction, similar to that for normal galaxies. Employing a redshift/magnitude prior in this process seems to be crucial in improving the agreement between photometric and spectroscopic redshifts.