Annu. Rev. Astron. Astrophys. 1992. 30: 613-52 Copyright © 1992 by Annual Reviews. All rights reserved

3. MODELS

There are two distinct kinds of models that are needed to interpret results from faint-galaxy surveys. The first is a model for the statistical properties of galaxies - the relative numbers of galaxies with certain profile shapes, surface brightnesses, and spectral energy distributions - and how a population of galaxies with these properties appears at high redshift. Cosmological effects, effects of galaxy evolution, and clustering characteristics are included in the model.

The other kind of model pertains to the details of the detection process. As remarked earlier, this is an important consideration because at very low detection levels various kinds of systematic errors can predominate. The detection model takes into account the noise from the night sky background, the residual flat-field errors, the effect of seeing, and the operation of a specific algorithm for measuring image fluxes and sizes on such data. A direct way to address these issues is to make an explicit two-dimensional simulation of the sky (Chokshi and Wright 1988), including sources of noise and other image degradation (Yee 1991), examples of which are shown in Figure 8. The same image-reduction methods used on the real sky (examples also shown in Figure 8) may then be applied to the artificial sky to reveal and calibrate the sources of bias. Because of the specific nature of the detection model for each set of observations, we will not discuss the models in depth, but Figure 8 provides examples of the state-of-the-art in simulated images of the sky.

Figure 8. Examples of faint-galaxy fields and simulations, each 96 arcsec on a side. The upper left panel gives a simulated picture for high Galactic latitudes in the Gunn r band by Howard Yee (see Yee 1991). It assumes 15,000 sec exposure on a 4-m telescope with 15% overall efficiency, a sky brightness of µr = 21.4 mag arcsec-2, 0.4 arcsec pixels, and seeing FWHM = 1 arcsec. The upper-right panel labeled ``r'' is actually the KG3 filter of Hall and Mackay (1984) for their exposure of Selected Area 57 (also with 0.4 arcsec pixels) with an integration time of 4,200 sec on the Kitt Peak 4-m telescope. The lower-left panel is a simulation by Arati Chokshi in Cowie's K' band (see Chokshi and Wright 1988). The simulation assumes = 1, an exposure time of 105 sec with a 2.2-m telescope, an overall efficiency of 10%, µK = 14.0 mag arcsec-2, and 0.48 arcsec pixels. Seeing is roughly simulated with 2 pixels smoothing. The lower-right panel gives the comparison very deep K'-band exposure by Cowie et al. (1992) of SSA-22, comprising a total exposure of 1.4 x 105 sec.