Photometric Redshift Estimation with Galaxy Morphology using Self-Organizing Maps

TL;DR

This paper demonstrates that combining multi-band photometric and morphological data with Self-Organizing Maps improves photometric redshift estimation accuracy for galaxies, using a sample from the CANDELS survey.

Contribution

It introduces a novel approach using Self-Organizing Maps to incorporate galaxy morphology into photometric redshift estimation, enhancing accuracy over traditional methods.

Findings

Photometric and morphological data yield comparable redshift estimates to SED modeling.

Self-Organizing Maps effectively map multi-dimensional galaxy observations.

Morphological data improves the robustness of redshift predictions.

Abstract

We use multi-band optical and near-infrared photometric observations of galaxies in the Cosmic Assembly Near-Infrared Deep Extragalactic Legacy Survey (CANDELS) to predict photometric redshifts using artificial neural networks. The multi-band observations span over 0.39 microns to 8.0 microns for a sample of about 1000 galaxies in the GOODS-S field for which robust size measurements are available from Hubble Space Telescope Wide Field Camera 3 observations. We use Self Organizing Maps (SOMs) to map the multi dimensional photometric and galaxy size observations while taking advantage of existing spectroscopic redshifts at 0 < z < 2 for independent training and testing sets. We show that use of photometric and morphological data led to redshift estimates comparable to redshift measurements from SED modeling and from self-organizing maps without morphological measurements.