4.3. The importance of multiple large HST fields
Large scale structure is an important and frequently neglected source of systematic uncertainty ([Somerville et al. (2004)]) 6. Comparison of the left-hand and right-hand panels of Fig. 7 illustrates this point powerfully; the broad features of the evolution of early-type galaxy stellar mass density are discernable using the 30' × 30' Chandra Deep Field South alone, but correction of this result for cosmic variance using the other two COMBO-17 fields yields a significantly more convincing picture. Yet, while this kind of correction for cosmic variance may work when there is significant overlap between populations of interest (although it is debatable how far one should push such an idea), it is not a priori clear that rarer and/or optically-obscured phases of galaxy evolution such as AGN or IR-luminous mergers will be well modelled with such techniques. Furthermore, short-timescale astronomical phenomena, such as AGN or galaxy mergers, have lower number density and are potentially very strongly clustered leading to large uncertainties from number statistics and cosmic variance. Yet, these phases of galaxy evolution, where galaxies undergo important and potentially permanent transformations in the cosmic blink of an eye, require HST-resolution data in order to explore their physical drivers.
When HST could be viewed as an essentially endless resource, a piecemeal approach was perfectly optimal: more and/or larger HST fields could be justified on a case-by-case basis, depending on the science goals of interest. Yet, faced with an unclear future for HST, it is not obvious that this approach is optimal. HST perhaps should be thought of as a fixed lifetime experiment, where the primary goal could become the creation of an archival dataset which will support 10-20 years of top-class research. In the creation of such an archival dataset, important questions will need to be addressed: availability of resources will need to be balanced against number statistics and cosmic variance, arguably at least 2 HST passbands will be required to allow some attempt at morphological k-correction, and deep multi-wavelength data will be required for study of black hole accretion and obscured star formation, naturally driving the fields into one of a small number of low HI and cirrus holes (see Papovich, these proceedings).
I wish to warmly thank the GEMS and COMBO-17 collaborations - Marco Barden, Steven Beckwith, Andrea Borch, John Caldwell, Simon Dye, Boris Häußler, Catherine Heymans, Knud Jahnke, Martina Kleinheinrich, Shardha Jogee, Daniel McIntosh, Klaus Meisenheimer, Chien Peng, Hans-Walter Rix, Sebastian Sanchez, Rachel Somerville, Lutz Wisotzki, and last but by no means least Christian Wolf - for their permission to present some GEMS and COMBO-17 results before their publication, for useful discussions, and for their friendship and collaboration. It is a joy to be part of these teams. I wish also to thank Eelco van Kampen and his collaborators for their efforts to construct mock COMBO-17 catalogs, for their permission to share results from these catalogs in this article, and for useful comments. Chris Conselice, Emmanuele Daddi, Sadegh Kochfar, and Casey Papovich are thanked for useful and thought-provoking discussions on some of the topics discussed in this review. This work is supported by the European Community's Human Potential Program under contract HPRN-CT-2002-00316, SISCO.
6 It is interesting to note that a × 10 increase in area in a single contiguous field gives only a × 2 reduction in cosmic variance, because the various parts of the single field are correlated with each other. Back.