Euclid preparation

Abstract

The various Euclid imaging surveys will become a reference for studies of galaxy morphology by delivering imaging over an unprecedented area of 15 000 square degrees with high spatial resolution. In order to understand the capabilities of measuring morphologies from Euclid-detected galaxies and to help implement measurements in the pipeline of the Organisational Unit MER of the Euclid Science Ground Segment, we have conducted the Euclid Morphology Challenge, which we present in two papers. While the companion paper focusses on the analysis of photometry, this paper assesses the accuracy of the parametric galaxy morphology measurements in imaging predicted from within the Euclid Wide Survey. We evaluate the performance of five state-of-the-art surface-brightness-fitting codes, DeepLeGATo, Galapagos-2, Morfometryka, ProFitand SourceXtractor++ , on a sample of about 1.5 million simulated galaxies (350 000 above 5σ) resembling reduced observations with the Euclid VIS and NIR instruments. The simulations include analytic Sérsic profiles with one and two components, as well as more realistic galaxies generated with neural networks. We find that, despite some code-specific differences, all methods tend to achieve reliable structural measurements (< 10% scatter on ideal Sérsic simulations) down to an apparent magnitude of about I E = 23 in one component and I E = 21 in two components, which correspond to a signal-to-noise ratio of approximately 1 and 5, respectively. We also show that when tested on non-analytic profiles, the results are typically degraded by a factor of 3, driven by systematics. We conclude that the official Euclid Data Releases will deliver robust structural parameters for at least 400 million galaxies in the Euclid Wide Survey by the end of the mission. We find that a key factor for explaining the different behaviour of the codes at the faint end is the set of adopted priors for the various structural parameters.

Astronomia ex machina: a history, primer and outlook on neural networks in astronomy

“…As we move into the hidden layers, this scaling issue only gets worse. Also, since MLPs must take an unrolled image as an input, they disregard any spatial properties of their training images, and so either need a substantial amount of training data to classify or generate large images, 12 or an expert to extract descriptive features from the data in a preprocessing step. We can see this issue writ large in the previous section-most of the MLP applications described in §3 require an expert to extract features from the data for the network to then train on!…”

“…Fussell & Moews [229] achieved this with a stacked GAN architecture [231], and Holzschuh et al [230] use the related StyleGAN architecture [189] to the same end. Bretonnière et al [12] use a flow-based model 36 [233,234] to conditionally simulate galaxy observations. They found that their approach could produce more accurate simulations than the previous analytical approach, at the cost of inference time.…”

“…[11]) and entirely data-driven simulation (e.g. [12,13]). Most of the criticism of machine learning techniques, and deep learning 2 in particular, is levelled at the perceived ‘black box’ nature of the methodology.…”

Self-consistent Atmosphere Representation and Interaction in Photon Monte Carlo Simulations

“…A number of publications have demonstrated a diverse set of applications for an ab initio simulator in astronomy using PhoSim 1 . PhoSim has been used to (1) to test the design of Rubin/LSST (Xin et al 2015;Angeli et al 2016;Thomas et al 2016) and JWST , (2) to plan for future observations (e.g., Chang et al 2012Chang et al , 2013aChang et al , 2013bBard et al 2013Bard et al , 2016Mandelbaum et al 2014;Thomas & Kahn 2018;Sanchez et al 2020;Bretonniere et al 2023;Merlin et al 2023), (3) for advanced image processing algorithm development (e.g., Meyers & Burchat 2015;Li et al 2016;Carlsten et al 2018;Nie et al 2021aNie et al , 2021b, (4) to understand physical effects (Beamer et al 2015;Xin et al 2018;Walter 2015), and (5) for advanced AI/machine learning development by simulating training sets ). In addition to published studies, hundreds of users have used PhoSim for informal studies.…”

New insights and predictions on high-redshift galaxies from wide-area infrared surveys

“…In the case of Euclid data alone, 30 out of the 65 contaminants cannot be included in these diagrams because they do not have a well-constrained (I E − Y E ) colour. 4, this means that in the worst-case scenario, 30 additional contaminants could actually survive this selection, and so the contamination fraction would be 0.11 instead of 0.01. Therefore, the purity of the apparent z > 6 sample would still improve from applying colour selection criteria, although possibly not as drastically as presented in Table 2.4.…”