Dark Energy Survey Year 3 Results: Measuring the Survey Transfer Function with Balrog

Abstract

We describe an updated calibration and diagnostic framework, Balrog, used to directly sample the selection and photometric biases of Dark Energy Survey's (DES) Year 3 (Y3) dataset. We systematically inject onto the single-epoch images of a random 20% subset of the DES footprint an ensemble of nearly 30 million realistic galaxy models derived from DES Deep Field observations. These augmented images are analyzed in parallel with the original data to automatically inherit measurement systematics that are often too difficult to capture with traditional generative models. The resulting object catalog is a Monte Carlo sampling of the DES transfer function and is used as a powerful diagnostic and calibration tool for a variety of DES Y3 science, particularly for the calibration of the photometric redshifts of distant "source" galaxies and magnification biases of nearer "lens" galaxies. The recovered Balrog injections are shown to closely match the photometric property distributions of the Y3 GOLD catalog, particularly in color, and capture the number density fluctuations from observing conditions of the real data within 1% for a typical galaxy sample. We find that Y3 colors are extremely well calibrated, typically within ∼1-8 millimagnitudes, but for a small subset of objects we detect significant magnitude biases correlated with large overestimates of the injected object size due to proximity effects and blending. We discuss approaches to extend the current methodology to capture more aspects of the transfer function and reach full coverage of the survey footprint for future analyses.

Dark Energy Survey Year 3 results: Cosmology from cosmic shear and robustness to data calibration

“…III B. To estimate the statistical connection between the wide-field and deepfield photometry pðcjĉÞ, the fitted light profile of each Deep Field galaxy is drawn into real DES Y3 science images multiple times in random positions with BALROG [107]. Processing these renderings with the DES photometry and shape measurement pipeline delivers the mapping of galaxies with noisy wide-field riz and successful shape measurement to the deep ugrizJHK s color space.…”

“…The subset of these DES Deep Field galaxies with BALROG [107] wide-field realizations that pass the weaklensing selection and have external high-quality redshift information forms the redshift sample. It is constructed from both spectroscopic and multiband photometric redshifts as detailed in Sec.…”

“…(ii) Redshift calibration methodology is summarized in Myles, Alarcon et al [81]. This framework utilizes external training data from narrow-band photometric and spectroscopic sources and DES deep observations with overlapping near-infrared data, presented in Hartley, Choi et al [106], mapped to the wide field with improved image injection measurements (Everett et al [107]). The full scheme incorporates new, independent methods, a two-step reweighting with self-organizing maps (Buchs et al [108]), clustering redshifts (Gatti, Giannini et al [82]) derived from correlations with the DES foreground REDMAGIC lens samples, and small-scale shear ratios (Sánchez, Prat et al [83]), to accurately constrain the redshift distributions.…”

Dark Energy Survey Year 3 results: Marginalisation over redshift distribution uncertainties using ranking of discrete realisations

“…This method makes use of three sets of observations: the full DES-Y3 wide field sample, the DES-Y3 Deep Fields (Hartley, Choi et al 2020) sample, and compilation of spectroscopic redshift surveys. Galaxies from the wide sample are grouped into phenotypes using the Self-Organised Maps (SOM) method of dimensional reduction (see e. DES data and recovers their properties, see Everett et al 2020) is then used to quantify the probability of a given Deep Fields galaxy appearing to have a given phenotype when observed in the wide field. A second SOM dimensional reduction is then applied to the Deep Fields galaxy observations, with the spectroscopic sample used to characterise the true redshift distribution for each deep phenotype.…”

Galaxy Morphological Classification Catalogue of the Dark Energy Survey Year 3 data with Convolutional Neural Networks

“…In this study, we compare our CNN predictions with four different resources: (1) the Galaxy Zoo 1 (GZ1) catalogue using the galaxies that were not present in the training set (Section 2.3); (2) visual classifications carried out by TC, CC, and AAS 3 (Section 4.1); (3) VIPERS unsupervised spectral classification (Siudek et al 2018, Section 4.2), and (4) non-parametric methods using the structural measurements from Tarsitano et al (2018) (Section 4.3). In DES Y3 GOLD catalogue, a quantity with 'FRACDEV' (Everett et al 2020) The architecture starts from an input of dimension 50×50×3, and is followed by three convolutional layers with kernel sizes of 3, 3, 2 and channel sizes of 32, 64, 128, respectively plus a max-pooling layer after each. Two dense layers with 1,024 hidden units are following the third convolution layer.…”