Selection Effects and Robust Measures of Galaxy Evolution

A variety of subtle, and not-so-subtle selection effects influence the interpretation of galaxy counts, sizes and redshift distributions in the Hubble Deep Field. Comparison of the different HDF catalogs available in the literature and on the world-wide-web reveals generally good agreement, although the effects of different isophotal thresholds and different splitting algorithms are readily apparent. As the basic source detection and photometry algorithms are similar for the different catalogs, the selection effects are likely to affect them all. Through simulations, we explore the utility of image moments for inferring the true sizes of galaxies. The truncation of galaxy profiles at a fixed isophote has serious consequences, which limit constraints on the size distribution to galaxies with isophotal magnitudes I<27.5. Present-day L* spirals would be undetected in the HDF above redshifts z ~ 1.2, and present-day ellipticals would disappear at z ~ 1.8. The Lyman break provides a way to identify high-redshift galaxies at very faint magnitudes. However, galaxies that are at redshifts high enough to vanish from the HDF F300W or F450W filters also suffer severely from photometric biases. For example, at fixed total apparent magnitude and physical scale length, a galaxy at z=4 will have a mean surface brightness 1.2 mag fainter than a galaxy at z=2.75. This lower surface brightness will result in an apparent decrease in the number density of objects, and the inferred luminosity density, even for models where there is no intrinsic evolution. We illustrate the effects of these biases on the estimates of the number of Lyman ``dropouts'' in the HDF and on the luminosity density at z>2.