We demonstrate how galaxy morphologies can be represented by weighted sums of ‘eigengalaxies’ and how eigengalaxies can be used in a probabilistic framework to enable principled and simplified approaches in a variety of applications. Eigengalaxies can be derived from a Principal Component Analysis (PCA) of sets of single- or multi-band images. They encode the image space equivalent of basis vectors that can be combined to describe the structural properties of large samples of galaxies in a massively reduced manner. As an illustration, we show how a sample of 10,243 galaxies in the Hubble Space Telescope CANDELS survey can be represented by just 12 eigengalaxies. We show in some detail how this image space may be derived and tested. We also describe a probabilistic extension to PCA (PPCA) which enables the eigengalaxy framework to assign probabilities to galaxies. We present four practical applications of the probabilistic eigengalaxy framework that are particularly relevant for the next generation of large imaging surveys: we (i) show how low likelihood galaxies make for natural candidates for outlier detection (ii) demonstrate how missing data can be predicted (iii) show how a similarity search can be performed on exemplars (iv) demonstrate how unsupervised clustering of objects can be implemented.

We introduce an empirical methodology to study how the spectral energy distribution (SED) and galaxy morphology constrain each other and implement this on ∼8000 galaxies from the HST CANDELS survey in the GOODS-South field. We show that the SED does constrain morphology and present a method which quantifies the strength of the link between these two quantities. Two galaxies with very similar SEDs are around 3 times more likely to also be morphologically similar, with SED constraining morphology most strongly for relatively massive red ellipticals. We apply our methodology to explore likely upper bounds on the efficacy of morphological selection using colour. We show that, under reasonable assumptions, colour selection is relatively ineffective at separating homogeneous morphologies. Even with the use of up to six colours for morphological selection, the average purity in the resultant morphological classes is only around 60 per cent. While the results can be improved by using the whole SED, the gains are not significant, with purity values remaining around 70 per cent or below.

No abstract