Dark Energy Survey Year 1 Results: Photometric Data Set for Cosmology
A. Drlica-Wagner, I. Sevilla-Noarbe, E. S. Rykoff, R. A. Gruendl, B. Yanny, D. L. Tucker, B. Hoyle, A. Carnero Rosell, G. M. Bernstein, K. Bechtol, M. R. Becker, A. Benoit-Levy, E. Bertin, M. Carrasco Kind, C. Davis, J. de Vicente, H. T. Diehl, D. Gruen, W. G. Hartley, B. Leistedt, T. S. Li, J. L. Marshall, E. Neilsen, M. M. Rau, E. Sheldon, J. Smith, M. A. Troxel, S. Wyatt, Y. Zhang, T. M. C. Abbott, F. B. Abdalla, S. Allam, M. Banerji, D. Brooks, E. Buckley-Geer, D. L. Burke, D. Capozzi, J. Carretero, C. E. Cunha, C. B. D'Andrea, L. N. da Costa, D. L. DePoy, S. Desai, J. P. Dietrich, P. Doel, A. E. Evrard, A. Fausti Neto, B. Flaugher, P. Fosalba, J. Frieman, J. Garcia-Bellido, D. W. Gerdes, T. Giannantonio, J. Gschwend, G. Gutierrez, K. Honscheid, D. J. James, T. Jeltema, K. Kuehn, S. Kuhlmann, N. Kuropatkin, O. Lahav, M. Lima, H. Lin, M. A. G. Maia, P. Martini, R. G. McMahon, P. Melchior, F. Menanteau, R. Miquel, R. C. Nichol, R. L. C. Ogando, A. A. Plazas, A. K. Romer, A. Roodman, E. Sanchez, V. Scarpine, R. Schindler, M. Schubnell, M. Smith, R. C. Smith, M. Soares-Santos, F. Sobreira, E. Suchyta, G. Tarle, V. Vikram, A. R. Walker, R. H. Wechsler, J. Zuntz
DDES-2015-0118FERMILAB-PUB-17-180-AE
Draft version May 4, 2018
Typeset using L A TEX twocolumn style in AASTeX61
DARK ENERGY SURVEY YEAR 1 RESULTS: PHOTOMETRIC DATA SET FOR COSMOLOGY
A. Drlica-Wagner, I. Sevilla-Noarbe, E. S. Rykoff,
3, 4
R. A. Gruendl,
5, 6
B. Yanny, D. L. Tucker, B. Hoyle, A. Carnero Rosell,
8, 9
G. M. Bernstein, K. Bechtol, M. R. Becker,
12, 3
A. Benoit-L´evy,
13, 14, 15
E. Bertin,
13, 15
M. Carrasco Kind,
5, 6
C. Davis, J. de Vicente, H. T. Diehl, D. Gruen,
3, 4
W. G. Hartley,
14, 16
B. Leistedt, T. S. Li, J. L. Marshall, E. Neilsen, M. M. Rau,
19, 7
E. Sheldon, J. A. Smith, M. A. Troxel,
22, 23
S. Wyatt,
21, 1
Y. Zhang, T. M. C. Abbott, F. B. Abdalla,
14, 25
S. Allam, M. Banerji,
26, 27
D. Brooks, E. Buckley-Geer, D. L. Burke,
3, 4
D. Capozzi, J. Carretero, C. E. Cunha, C. B. D’Andrea, L. N. da Costa,
8, 9
D. L. DePoy, S. Desai, J. P. Dietrich,
31, 19
P. Doel, A. E. Evrard,
32, 33
A. Fausti Neto, B. Flaugher, P. Fosalba, J. Frieman,
1, 35
J. Garc´ıa-Bellido, D. W. Gerdes,
32, 33
T. Giannantonio,
26, 27, 7
J. Gschwend,
8, 9
G. Gutierrez, K. Honscheid,
22, 23
D. J. James,
37, 24
T. Jeltema, K. Kuehn, S. Kuhlmann, N. Kuropatkin, O. Lahav, M. Lima,
41, 8
H. Lin, M. A. G. Maia,
8, 9
P. Martini,
22, 42
R. G. McMahon,
26, 27
P. Melchior, F. Menanteau,
5, 6
R. Miquel,
44, 29
R. C. Nichol, R. L. C. Ogando,
8, 9
A. A. Plazas, A. K. Romer, A. Roodman,
3, 4
E. Sanchez, V. Scarpine, R. Schindler, M. Schubnell, M. Smith, R. C. Smith, M. Soares-Santos, F. Sobreira,
48, 8
E. Suchyta, G. Tarle, V. Vikram, A. R. Walker, R. H. Wechsler,
12, 3, 4 and J. Zuntz (DES Collaboration) Fermi National Accelerator Laboratory, P. O. Box 500, Batavia, IL 60510, USA Centro de Investigaciones Energ´eticas, Medioambientales y Tecnol´ogicas (CIEMAT), Madrid, Spain Kavli Institute for Particle Astrophysics & Cosmology, P. O. Box 2450, Stanford University, Stanford, CA 94305, USA SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA Department of Astronomy, University of Illinois, 1002 W. Green Street, Urbana, IL 61801, USA National Center for Supercomputing Applications, 1205 West Clark St., Urbana, IL 61801, USA Universit¨ats-Sternwarte, Fakult¨at f¨ur Physik, Ludwig-Maximilians Universit¨at M¨unchen, Scheinerstr. 1, 81679 M¨unchen, Germany Laborat´orio Interinstitucional de e-Astronomia - LIneA, Rua Gal. Jos´e Cristino 77, Rio de Janeiro, RJ - 20921-400, Brazil Observat´orio Nacional, Rua Gal. Jos´e Cristino 77, Rio de Janeiro, RJ - 20921-400, Brazil Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA 19104, USA LSST, 933 North Cherry Avenue, Tucson, AZ 85721, USA Department of Physics, Stanford University, 382 Via Pueblo Mall, Stanford, CA 94305, USA CNRS, UMR 7095, Institut d’Astrophysique de Paris, F-75014, Paris, France Department of Physics & Astronomy, University College London, Gower Street, London, WC1E 6BT, UK Sorbonne Universit´es, UPMC Univ Paris 06, UMR 7095, Institut d’Astrophysique de Paris, F-75014, Paris, France Department of Physics, ETH Zurich, Wolfgang-Pauli-Strasse 16, CH-8093 Zurich, Switzerland New York University, CCPP, New York, NY 10003, USA George P. and Cynthia Woods Mitchell Institute for Fundamental Physics and Astronomy, andDepartment of Physics and Astronomy, Texas A&M University, College Station, TX 77843, USA Faculty of Physics, Ludwig-Maximilians-Universit¨at, Scheinerstr. 1, 81679 Munich, Germany Brookhaven National Laboratory, Bldg 510, Upton, NY 11973, USA Austin Peay State University, Dept. Physics-Astronomy, P.O. Box 4608 Clarksville, TN 37044, USA Center for Cosmology and Astro-Particle Physics, The Ohio State University, Columbus, OH 43210, USA Department of Physics, The Ohio State University, Columbus, OH 43210, USA Cerro Tololo Inter-American Observatory, National Optical Astronomy Observatory, Casilla 603, La Serena, Chile Department of Physics and Electronics, Rhodes University, PO Box 94, Grahamstown, 6140, South Africa Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK Kavli Institute for Cosmology, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK Institute of Cosmology & Gravitation, University of Portsmouth, Portsmouth, PO1 3FX, UK Institut de F´ısica d’Altes Energies (IFAE), The Barcelona Institute ofScience and Technology, Campus UAB, 08193 Bellaterra (Barcelona) Spain Department of Physics, IIT Hyderabad, Kandi, Telangana 502285, India Excellence Cluster Universe, Boltzmannstr. 2, 85748 Garching, Germany Department of Astronomy, University of Michigan, Ann Arbor, MI 48109, USA Department of Physics, University of Michigan, Ann Arbor, MI 48109, USA Institut de Ci`encies de l’Espai, IEEC-CSIC, Campus UAB, Carrer de Can Magrans, s/n, 08193 Bellaterra, Barcelona, Spain Kavli Institute for Cosmological Physics, University of Chicago, Chicago, IL 60637, USA Instituto de Fisica Teorica UAM/CSIC, Universidad Autonoma de Madrid, 28049 Madrid, Spain Astronomy Department, University of Washington, Box 351580, Seattle, WA 98195, USA Santa Cruz Institute for Particle Physics, Santa Cruz, CA 95064, USA
Corresponding author: Alex [email protected] a r X i v : . [ a s t r o - ph . C O ] M a y Australian Astronomical Observatory, North Ryde, NSW 2113, Australia Argonne National Laboratory, 9700 South Cass Avenue, Lemont, IL 60439, USA Departamento de F´ısica Matem´atica, Instituto de F´ısica, Universidade de S˜ao Paulo, CP 66318, S˜ao Paulo, SP, 05314-970, Brazil Department of Astronomy, The Ohio State University, Columbus, OH 43210, USA Department of Astrophysical Sciences, Princeton University, Peyton Hall, Princeton, NJ 08544, USA Instituci´o Catalana de Recerca i Estudis Avan¸cats, E-08010 Barcelona, Spain Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr., Pasadena, CA 91109, USA Department of Physics and Astronomy, Pevensey Building, University of Sussex, Brighton, BN1 9QH, UK School of Physics and Astronomy, University of Southampton, Southampton, SO17 1BJ, UK Instituto de F´ısica Gleb Wataghin, Universidade Estadual de Campinas, 13083-859, Campinas, SP, Brazil Computer Science and Mathematics Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831 Institute for Astronomy, University of Edinburgh, Edinburgh EH9 3HJ, UK
ABSTRACTWe describe the creation, content, and validation of the Dark Energy Survey (DES) internal year-one cosmologydata set, Y1A1 GOLD, in support of upcoming cosmological analyses. The Y1A1 GOLD data set is assembled frommultiple epochs of DES imaging and consists of calibrated photometric zeropoints, object catalogs, and ancillary dataproducts—e.g., maps of survey depth and observing conditions, star-galaxy classification, and photometric redshiftestimates—that are necessary for accurate cosmological analyses. The Y1A1 GOLD wide-area object catalog consistsof ∼
137 million objects detected in coadded images covering ∼ in the DES grizY filters. The 10 σ limitingmagnitude for galaxies is g = 23 . r = 23 . i = 22 . z = 21 .
8, and Y = 20 .
1. Photometric calibration ofY1A1 GOLD was performed by combining nightly zeropoint solutions with stellar locus regression, and the absolutecalibration accuracy is better than 2% over the survey area. DES Y1A1 GOLD is the largest photometric data set atthe achieved depth to date, enabling precise measurements of cosmic acceleration at z (cid:46) Keywords: surveys, catalogs, techniques: image processing, techniques: photometric, cosmology: ob-servations INTRODUCTIONThe Dark Energy Survey (DES; DES Collaboration2005, 2016) is a photometric survey utilizing the DarkEnergy Camera (DECam; Flaugher et al. 2015) on theBlanco 4m telescope at Cerro Tololo Inter-American Ob-servatory (CTIO) in Chile to observe ∼ of thesouthern sky in five broadband filters, g, r, i, z, Y , rang-ing from ∼
400 nm to ∼ The primary goal of DES is to studythe origin of cosmic acceleration and the nature of darkenergy through four key probes: weak lensing, large-scale structure, galaxy clusters, and Type Ia supernovae.More generally, DES provides a rich scientific data setand has already had a significant impact beyond cos-mology (e.g., DES Collaboration 2016).Precision measurements of dark energy with DES relyon an unprecedented survey data set and a comprehen-sive understanding of the survey performance. It is nec-essary to identify, characterize, and mitigate the influ-ences of variable observing conditions, data processingartifacts, photometric calibration nonuniformity, and as-trophysical foregrounds. For example, photometric cali-bration must be accurate and uniform to avoid introduc-ing noise and bias into photometric redshift estimates.Studies of galaxy clustering depend on a detailed knowl-edge of survey coverage, galaxy detection efficiency, and The DECam filter throughput is publicly available at . the accuracy of recovered galaxy properties. Further-more, detailed modeling of the point-spread function(PSF) and instrument response is required to performgalaxy shape measurements on objects that are fainterthan the the detection limit of a single DES image. Thescale and complexity of assembling, characterizing, andvalidating the DES data motivate a collaborative effortthat draws upon and enables a wide range of scientificanalyses.Here we describe the creation, composition, and val-idation of the DES first-year (Y1) data set in supportof cosmological analyses (shown schematically in Fig-ure 1). While this data set is currently proprietary to theDES Collaboration, this document is intended to serveas a reference for these data products when they becomepublicly available. Observing for DES Y1 spanned from2013 August to 2014 February and covered ∼
40% of theDES footprint, averaging three to four visits per band.The resulting images were processed through the DESdata management (DESDM) system (Ngeow et al. 2006;Mohr et al. 2008; Sevilla et al. 2011; Mohr et al. 2012;Desai et al. 2012; Morganson et al. 2018) an assembledinto the DES year-one annual data set (Y1A1). Y1A1consists of reduced single-epoch images and object cata-logs (known as “Y1A1 FINALCUT”), along with multi-epoch coadded images and associated multi-band cata- Note that the DES Y1 cosmology data set described here isdistinct from the forthcoming DES public data release, which willinclude data from the first 3 years of DES. logs (known as “Y1A1 COADD”). Photometric calibra-tion of Y1A1 was performed globally on a CCD-to-CCDbasis, and maps of the survey coverage and depth wereassembled with the mangle software suite (Hamilton &Tegmark 2004; Swanson et al. 2008). The Y1A1 dataset covers ∼ in any single filter with inhomo-geneous coverage and depth. In total, ∼ ofthe Y1A1 footprint has simultaneous coverage in all fiveDES filters.The desired precision of DES cosmological analysesmotivates further refinement of Y1A1. The resultingdata set, referred to as Y1A1 GOLD, is accompaniedby extensive validation and ancillary data products tofacilitate cosmological analyses. The primary compo-nents of Y1A1 GOLD are (Figure 1): (1) a multi-bandphotometric object catalog subselected from the Y1A1COADD object catalog; (2) an adjusted photometriccalibration to improve uniformity over the survey foot-print; (3) shape and photometry information from a si-multaneous multi-epoch, multi-object fit; (4) a set of an-cillary maps quantifying survey characteristics using theHEALPix rasterization scheme (G´orski et al. 2005); and(5) several value-added quantities for high-level analy-ses (i.e., a star-galaxy classifier and photo- z estimates).When creating the Y1A1 GOLD object catalog, sev-eral classes of non-physical, spurious, or otherwise prob-lematic objects were identified and either flagged or re-moved. The calibrated magnitudes of objects were alsocorrected for interstellar extinction using a stellar locusregression (SLR) technique. The ancillary data prod-ucts associated with Y1A1 GOLD accurately quantifythe characteristics of the survey, further mitigating theimpact of systematic uncertainties. A high-level sum-mary of the performance of Y1A1 GOLD is tabulatedin Table 1.Our purpose here is to document the production andperformance of the Y1A1 GOLD data set in supportof upcoming DES cosmology analyses. We start bydescribing the DES Y1 observations in Section 2 andbriefly reviewing the image reduction pipeline applied toproduce the Y1A1 data set in Section 3. In Section 4 wedescribe the photometric calibration of the Y1A1 data,and in Section 5 we describe the image coaddition pro-cess. In Section 6 we discuss the creation of uniqueobject catalogs, while in Sections 7 and 8 we describethe ancillary maps and value-added quantities producedto complement the Y1A1 GOLD catalog. We brieflyconclude in Section 9. DATA COLLECTIONDES has been allocated 105 nights per year on theBlanco telescope starting in 2013. The first year of DESobserving spanned from 2013 August 31 to 2014 Febru- ary 9 and consisted of both full and half nights. Detailson DES operation and data collection are provided byDiehl et al. (2014); here we briefly summarize some ofthe key details relevant to the creation of Y1A1 GOLD.DES consists of two observing programs: a shallowerwide-area survey and a deeper time-domain (“super-nova” or “SN”) survey (Figure 2). The DES wide-areasurvey footprint covers ∼ with 90s exposuresin griz and 45s exposures in Y . A single imaging passover this footprint, called a “tiling”, collects science dataover ∼
83% of the survey footprint owing to inefficien-cies in the pointing layout and camera footprint (e.g.,area not covered owing to gaps between CCDs, non-functioning CCDs, and problematic area near the edgesof the CCDs). The DECam pointings for each tilingare shifted relative to each other by a large fraction ofthe camera field of view in a dither pattern designedto maximize uniformity and distribute repeated detec-tions of a given object over the focal plane. During Y1,DES observed ∼ of the wide-area survey foot-print with three to four dithered tilings per filter. TheY1 footprint consisted of two areas: one near the ce-lestial equator including Stripe 82 (S82; Annis, Jamesand Soares-Santos, M. and Strauss, M. A. and others2014), and a much larger area that was also observed bythe South Pole Telescope (SPT; Carlstrom et al. 2011).During Y1, DES collected 17,671 wide-area survey ex-posures in a variety of observing conditions (Diehl et al.2014).The SN survey observes 10 fields in four filters ( griz )on a regular cadence to detect and characterize super-nova through difference imaging (Kessler et al. 2015).Longer exposure times ( ≥
150 s) and frequent repeatedvisits result in a significantly deeper survey in the SNfields. All 10 SN fields reside within the DES wide-area footprint, but only two were covered by wide-areaimaging in Y1 (Figure 2). Over the course of Y1, DEScollected a total of 2699 time-domain survey exposures.In addition to the wide-area and SN survey fields, twoauxiliary fields outside the DES footprint were observedto aid in the training of photometric redshifts and star-galaxy classification. Fields overlapping with COSMOS(Scoville et al. 2007) and VVDS-14h (Le F`evre et al.2005) were observed during the DES Science Verification(SV) period. These observations are deeper than mostof the Y1 wide-area survey.During DES operation, sets of biases and flat-field cal-ibration exposures were taken in each filter before eachnight of observing. Standard-star fields were observedat three different airmasses at the beginning and end ofeach night unless conditions were obviously nonphoto- Several exposures taken during engineering time earlier in2013 August were also included in the the Y1A1 data set. Data from the DECam Science Verification period is availableat: https://des.ncsa.illinois.edu/releases/sva1 . DECamObservationsImage ReductionPhotometricCalibrationImage CoadditionCoadd SourceExtraction Catalog ArtifactRemovalCalibrationAdjustmentAncillary MapCreation Star/GalaxySeparationPhoto- z EstimationMulti-EpochFittingY1A1 Processing Y1A1 GOLD
Figure 1.
Schematic of the constituents of the Y1A1 processing (left) and the additional Y1A1 GOLD products (right).
Table 1.
Y1A1 GOLD Data Quality Summary
Parameter Band Reference g r i z Y
Median PSF FWHM 1.25 (cid:48)(cid:48) (cid:48)(cid:48) (cid:48)(cid:48) (cid:48)(cid:48) (cid:48)(cid:48)
Section 7.2Sky Coverage (in all bands) 1786 deg Section 7.3Astrometric Accuracy 25 mas (relative); <
300 mas (external) Section 5.1Absolute Photometric Uncertainty (mmag) 14 4 2 15 32 Section 4.4Relative Photometric Uniformity (mmag) 19 22 20 20 18 Section 4.4Completeness Limit (95%) 23 . . . . σ ) a . +0 . − . . +0 . − . . +0 . − . . +0 . − . . +0 . − . Section 7.1Multi-Epoch Galaxy Magnitude Limit (10 σ ) a . +0 . − . . +0 . − . . +0 . − . . +0 . − . . . . Section 7.1Galaxy Selection ( i ≤
22) Efficiency > <
3% Section 8.1Stellar Selection ( i ≤
22) Efficiency > <
6% Section 8.1 a The quoted values correspond to the mode, 16 th percentile, and 84 th percentiles of the magnitude limit distribution. Usingthe median instead of the mode reduces the magnitude limit by ∼ .
05 mag. metric. Cloud cover was monitored continuously dur-ing observing by the RASICAM all-sky infrared camera(Lewis et al. 2010; Reil et al. 2014).DES images the sky whenever weather allows theBlanco dome to be open, resulting in some exposuresbeing taken in very poor conditions. Thus, data qual-ity monitoring is essential to select exposures that meetthe scientific requirements of the survey. The qualityof exposures is evaluated based on the PSF, sky bright-ness, and sky transparency. For each exposure, we define On DES half nights only two standard-star fields were ob-served at the midpoint of the night. t eff to be the ratio between the actual exposure timeand the exposure time necessary to achieve the samesignal-to-noise ratio (S/N) for point sources observed innominal conditions (Neilsen et al. 2015). To pass pre-liminary data quality cuts, wide-area survey exposuresmust have t eff > . r , i , and z band and t eff > . g and Y band. The median measured t eff for Y1 was t eff = 0 .
75 in the r , i and z band and t eff = 0 .
49 inthe g and Y band. In contrast, the preliminary dataquality cuts for SN exposures require FWHM < (cid:48)(cid:48) and The effective exposure time is defined as T eff = t eff T exp , where T exp is the shutter-open exposure time. -120°-60°0°+60°+120° -120°-60°0°+60°+120°-60°-30°0°+30° +60° -60° -30° 0°+30°+60°SN-ESN-XSN-SSN-C VVDS-14COSMOS S82SPT
Figure 2.
DES Y1A1 GOLD sky coverage in celestial coordinates (red) plotted in McBryde-Thomas flat polar quartic projection.Specific regions of Y1A1 GOLD footprint, including the SN and auxiliary fields, are explicitly labeled (see Section 2). Thenominal DES five-year footprint is outlined in black. that a 20th-magnitude simulated source have signal-to-noise ratio >
20 ( >
80) for the shallow (deep) exposures(Kessler et al. 2015). Of the exposures taken during Y1,82% of the wide-area exposures and 95% of the SN expo-sures passed their respective data quality criteria (i.e.,did not require re-observation). A number of additionalexposures were removed from the Y1 data set due to in-strumental artifacts (scattered light and internal reflec-tions from bright stars, contaminating light from air-planes, poor telescope tracking, shutter malfunctions,dome occultations, etc.). In total, the Y1A1 FINAL-CUT processing consists of 16,857 DECam exposures,including wide field, SN, auxiliary fields, and standardstars. IMAGE PROCESSINGThe DESDM system is responsible for reducing, cat-aloging, and distributing DES data. Earlier iterationsof the DESDM image processing pipeline are outlinedin Sevilla et al. (2011) and Mohr et al. (2012), anda more detailed summary with updates for the forth-coming DES three-year processing is available in Bern-stein et al. (2017a) and Morganson et al. (2018). Herewe briefly summarize the single-epoch image process-ing steps applied during the DES Y1A1 FINALCUTcampaign. The Y1A1 FINALCUT campaign resultedin ∼
20 TB of processed images and a catalog of ∼ Overscan and Crosstalk:
Each DECam CCD hastwo amplifiers for converting photo-carrier countsto analog-to-digital units (ADU). For each ampli-fier, the average in the overscan region was cal-culated and subtracted on a row-by-row basis.Crosstalk is manifested as low-level leakage of elec-tronic signals between different readout amplifiers and is observed at the level of ∼ − for pairs ofamplifiers on the same CCD and ∼ − − − for pairs of amplifiers on different CCDs on thesame electronic back plane. Crosstalk was cor-rected by applying a matrix operation to the si-multaneous readout of 140 amplifiers (includingthe amplifiers for the eight focus and alignmentCCDs). The elements of the crosstalk correctionmatrix were derived from the median amplifieroutput for each “victim” channel as a function ofthe “source” amplifier signal for a large number ofsky images. Crosstalk between the DECam CCDsis found to be nonlinear when the signal on thesource amplifier exceeds its saturation level – i.e.,the level at which the amplifier response becomesnonlinear (Figure 2 of Bernstein et al. 2017a).This crosstalk nonlinearity was incorporated intocrosstalk correction. There is no evidence for tem-poral variation in the crosstalk between CCDs onyear timescales, and a single crosstalk matrix wasused for the Y1A1 processing.2. Bias Correction:
A master bias frame was con-structed from the average of ∼
100 zero-secondexposures taken during the pre-night calibrationsequences over the course of the Y1 observing sea-son. This master bias was subtracted from eachCCD to remove any residual fixed pattern noisenot incorporated by the overscan correction.3.
Bad-Pixel Masking:
Bad pixel masks were createdfor each CCD by identifying outliers in sets of bi-ases and g -band flat-field calibration exposures.These bad pixels were masked and interpolatedbased on values in adjacent columns. The Y1A1processing campaign used a single static bad pixel Figure 3.
Processed DECam image from Y1A1 (top) andCCD layout (bottom). The three empty slots in the DECamimage correspond to CCD2, CCD31, and CCD61. CCD61failed during SV, while CCD2 failed part way through Y1.One amplifier of CCD31 has time-variable low-light-levelnonlinearity and this CCD was not processed for Y1A1. mask. Two CCDs have failed since commissioningand were removed from Y1 processing (Figure 3;Diehl et al. 2014). CCD61 failed on 2012 Novem-ber 7 and data from this CCD were not used inY1. CCD2 failed on 2013 November 30 and datafrom this CCD were only included for the earlymonths of Y1. Nonlinearity Correction:
Several ( ∼
10) CCD am-plifiers have a nonlinear response at low light levels(generally below 300 ADU/pixel). For DES, thisaffects the sky level in short ( ∼
15 s) standard-star CCD2 subsequently recovered on 2016 December 29. observations and wide-survey dark-sky g -band ob-servations (90 s). For most other filters/exposuretimes, the night sky alone is enough to give a suffi-cient number of counts per pixel to make the non-linearity correction negligible. The nonlinearity ef-fect can be several percent at very low light levels.At very high light levels ( > × ADU/pixel),there is also a small nonlinear behavior ( (cid:46) The Y1A1 dataprocessing did not correct for charge-induced pixelshifts (i.e., the “brighter-fatter” effect; Antiloguset al. 2014; Gruen et al. 2015), although correc-tions have been incorporated into more recent re-ductions of the DES data (Bernstein et al. 2017a).5.
Pupil Correction:
An additive correction was ap-plied for pupil ghosting in each exposure. As partof this process, “star flats” were created in eachfilter and CCD by taking multiple dithered expo-sures of a dense stellar field and fitting a cubicpolynomial to variations in the observed bright-nesses of stars. The pupil ghost correction wasconstructed on a CCD by CCD basis for each ex-posure from the star flat and the level of sky back-ground (including scattered light and the night-sky pupil image). The pupil correction was scaledand subtracted from each CCD individually. Thistechnique can leave gradients of several percent inthe sky background level (worst in z and Y band),which propagate into the reduced science imagesand are corrected during photometric calibration(Appendix A). Flat Fielding:
The response of DECam to thenight sky is more stable than nightly variationsin the illumination of the flat-field screen takenduring pre-night calibrations. Therefore, in Y1A1we created a single average flat-field frame foreach filter from ∼
100 individual flat-field expo-sures. The science exposures were divided by theaverage flat-field frames normalized on a CCD byCCD basis. The pupil and flat-field correctionsused for Y1A1 processing remove small-scale fluc-tuations in the background due to pixel-size vari-ations, i.e., tree rings, edge brightening, and tapebumps (Plazas et al. 2014). However, this cor- The other amplifier on CCD31 is stable and has been recoveredin more recent processing. More recent implementations of the data processing pipelinefit the additive correction over the full focal plane rather thanCCD by CCD (Morganson et al. 2018). rection is approximate and results in photomet-ric measurement residuals at the level of ∼ . Weight Plane Creation:
A weight plane image wascreated containing the inverse variance of the flat-fielded image value in each pixel. The varianceestimate summed the expected Poisson noise andread noise. Saturated pixels were flagged and setto zero in the weight plane. The weight plane isused to assign relative weights to images duringthe coaddition process.8.
Fringe Frame Correction:
Fringing is visible in z -and Y -band exposures. The fringing pattern isnearly identical in these bands but has a largeramplitude in the Y band. A set of templates wasconstructed from a stack of ∼ z - and Y -bandexposures from DES SV. These template imageswere median filtered and averaged on a pixel-by-pixel level to construct a fringe frame. In the re-duction pipeline, each CCD of the z - and Y -bandexposures had its median sky level measured, andthis sky level was used to scale the fringe frame,which was then subtracted on a CCD-by-CCD ba-sis. The scaling method was identical to that usedto scale and subtract the pupil pattern. The vastmajority of exposures have a fringe residual thatis < .
1% of the sky background level. Exposurestaken under the brightest conditions accepted for Y -band observing can have a fringe residual thatis ∼ .
4% of the sky background level.9.
Illumination Correction:
Light reflected from theflat-field screen fills the telescope pupil differentlythan the focused light of distant stars. To accountfor pixel-level differences in the throughput of theflat-field images, we applied a multiplicative cor-rection to the DECam response based on the starflats. After dividing by the star flats the residualdifference in response between CCDs is typically <
2% peak to peak (Appendix A.1).10.
Preliminary Astrometric Solution:
A world coor-dinate system (WCS) was installed in the imageheader at the time of observation using a fixeddistortion map derived from the star flats and anoptical axis read from the telescope encoders. Thepointing of each image was updated matching thecenters of bright stars measured with
SExtractor (Bertin & Arnouts 1996; Bertin et al. 2002) to theUCAC-4 catalog using
SCAMP (Bertin 2006). ThisWCS was replaced by a superior one during thecoaddition step (Section 5), and the astrometricaccuracy of the Y1A1 GOLD catalog is describedin Section 5.1. 11.
Artifact Removal:
Bright stars ( (cid:46)
16 mag) satu-rate the 90 s DES exposures in griz . Saturatedpixels are set to zero in the image weight mapplane. Brighter stars can produce charge overflowinto pixels in the CCD readout direction. Theseoverflow pixels are flagged in the mask plane, ze-roed in the weight plane, and interpolated in theimage plane. In addition, corresponding pixels onthe victim amplifier of the CCD are masked ow-ing to large nonlinear crosstalk. Extremely brightoversaturated stars can leave a secondary chargeoverflow in the readout register of the amplifier,conventionally called “edge bleeds” (see Fig. 5 inBernstein et al. 2017a). Edge bleeds can be lo-cated some distance from the bright star and arestrongest in the rows near the readout register.These rows are identified and masked.Energy deposited from cosmic-ray interactionswith the CCDs were detected on single images us-ing the findCosmicRays algorithm adopted fromthe LSST software stack. The cosmic-ray pix-els were flagged in the mask plane, zeroed in theweight plane, and interpolated in the image plane.Long streaks produced by rapidly moving objects(i.e., meteors and Earth-orbiting satellites) weredetected using a Hough transform algorithm andwere also masked (Melchior et al. 2016).12.
Single-Epoch Catalog Creation:
Object cata-logs were produced for each CCD using the
AstrOmatic package (Bertin 2006). Photometricfluxes were derived using
PSFEx and
SExtractor for fixed apertures, the PSF model, and a galaxymodel. The local sky background on each CCDwas estimated by
SExtractor . The single-epochY1A1 FINALCUT catalogs served as an input intothe photometric calibration described in Section 4. PHOTOMETRIC CALIBRATIONThe photometric calibration of Y1A1 was a multistepprocess largely following the procedure of Tucker et al.(2007). Photometric calibration was performed on thesingle-epoch Y1A1 FINALCUT images first on a nightlyand then on a global basis. An additional calibration ad-justment was derived from the stellar locus and appliedat catalog level. Below we briefly describe the steps inthe photometric calibration of Y1A1; a more detaileddiscussion of the Y1A1 photometric calibration can befound in Appendix A.4.1.
Nightly Photometric Calibration
A preliminary photometric calibration of the Y1A1data was performed on a nightly basis. Standard-star https://lsst-web.ncsa.illinois.edu/doxygen/x_masterDoxyDoc/namespacelsst_1_1meas_1_1algorithms.html fields were observed at various airmasses at the begin-ning and end of each night. These images were re-duced and the centroids of stars were matched to a setof primary standard stars from Sload Digital Sky Sur-vey (SDSS) DR9 (Smith et al. 2002). The DES sec-ondary standards were then transformed to an initialDES AB photometric system via a set of transformationequations derived from SDSS DR9 and supplementedby UKIDSS DR6 (Appendix A.4). This tied the DESflux calibration of the secondary standards to SDSS andto the AB magnitude system (i.e., Padmanabhan et al.2008).The transformed nightly standards were used to fit aset of nightly photometric equations to model the spa-tial and temporal dependence of the DECam instrumentthroughput (Equations (A1)–(A5)). These equationstrack the accumulation of dust on the Blanco primarymirror, the relative throughput of the atmosphere atCTIO, and variations in the throughput and shape ofthe filter response at the location of each CCD. Thenightly photometric equations produce an initial photo-metric calibration for all exposures taken on photometricnights. The relative calibration scatter for the nightlysolution on a typical photometric night is ∼ .
02 magrms. This nightly photometric calibration was used toanchor the relative global calibration of non-photometricexposures described in the next section. A more detaileddescription can be found in Appendix A.1.4.2.
Global Calibration
We implemented a global calibration module (GCM)to derive calibrated zeropoints for all exposures, includ-ing those taken under non-photometric conditions, andto improve on the relative calibration accuracy achievedby the nightly photometric solution. The GCM pro-cedure follows that of Glazebrook et al. (1994) and isdescribed in more detail in Appendix A.2. Briefly, theY1A1 data were split into regions of contiguous, over-lapping images where at least one image had been pre-viously calibrated. The calibrated images served as areference against which other images in the groupingwere calibrated. To be calibrated by the GCM, an im-age needed to either overlap a calibrated image or havean unbroken path of overlapping images to a calibratedimage.Following the prescription of Glazebrook et al. (1994),we estimate the rms magnitude residual for each CCDimage from overlap with other CCD images. The rmsdistribution over all CCD images is a measure of theinternal reproducibility uncertainty on small scales (thescales of overlapping CCD images) and is a measure ofthe precision of the overall GCM solution. We find therms to be ∼ tertiary standard stars observed under pho-tometric conditions and calibrated by the nightly pho-tometric equations. The tertiary standards were chosensuch that they would anchor the global solution on scales > Photometric Calibration Adjustment
The global calibration is found to be uniform at the ∼
2% level in each band over the majority of the Y1A1survey footprint (discussed in Section 4.4). However,non-uniformity in the colors of objects can severely im-pact DES science by introducing a spatial dependenceon object selection and photo- z estimation. The SLRtechnique uses the distinct shape of the stellar locus incolor-color space to provide a relative calibration of ex-posures in different bands (e.g., Ivezi´c et al. 2004; Mac-Donald et al. 2004; High et al. 2009; Gilbank et al. 2011;Desai et al. 2012; Coupon et al. 2012; Kelly et al. 2014).To correct for residual spatial non-uniformity in the cal-ibration and account for Galactic reddening (includinguncertainties in the amplitude of reddening and possi-ble variations in the effective Milky Way dust law), wehave applied a secondary adjustment to the calibrationof the coadd object catalogs derived from the stellar lo-cus. Gradients in stellar population are subdominant toother calibration uncertainties in Y1A1 given the DESfilter bandpasses and high Galactic latitude of the survey(e.g., High et al. 2009; Kelly et al. 2014). We followedthe procedure of Drlica-Wagner et al. (2015) and applieda modified version of the BigMACS
SLR code (Kelly et al.2014) coupled with an empirical stellar locus to derivezeropoint adjustments to improve the color uniformityof stars across the Y1A1 footprint. The SLR adjust-ment was tied to the i -band magnitude derived from theGCM, dereddened using the Schlegel et al. (SFD; 1998)dust map with a reddening law from O’Donnell (1994).The SLR zeropoint adjustments were interpolated to thepositions of each object in the catalog and were applieddirectly to the magnitudes of objects derived from thecoadded images. In this way, the calibrated magnitudesof the Y1A1 GOLD catalog are already corrected forinterstellar extinction . After the SLR adjustment, thecolor of stars was found to be uniform at the ∼
1% levelacross the footprint, which was verified using the red se- https://code.google.com/p/big-macs-calibrate/ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) r-band2.8 mmag Figure 4.
Internal reproducibility uncertainty for the Y1A1 r -band photometric zeropoints calculated by comparing the rmscalibrated magnitudes of stars in overlapping CCDs. The mode of the rms internal calibration uncertainty is 2 . quence of galaxies. More detail on the SLR calibrationadjustment can be found in Appendix A.3.4.4. Photometric Calibration Accuracy
To quantify the accuracy of photometric calibration,we would like to characterize the statistical distributionof ∆ m = m meas − m true , where m meas and m true arethe measured and true magnitude of catalog objects, re-spectively. The characterization of the ∆ m distributioncan be split into two components: (1) an “absolute”calibration accuracy that represents a linear shift of the∆ m distribution (e.g., the mean of the distribution), and(2) a “relative” calibration accuracy that represents thespread of the ∆ m distribution (e.g., standard deviationof the distribution). In reality, values of m true are notavailable, and we must make use of the calibrated mag-nitudes from other surveys or synthetic models, whichhave their own associated uncertainties. We describeseveral calibration validation studies below and summa-rize the results in Table 2.The absolute calibration of the Y1A1 GOLD is tiedto SDSS through the DES secondary standard stars. Asan independent cross-check on the absolute photometriccalibration, we examined the CALSPEC standard star,C26202 (Bohlin et al. 2014). We calculated syntheticmagnitudes for C26206 by convolving Hubble Space Tele-scope (HST) spectra ( stisnic 006 ) with the focal-plane-averaged DECam filter throughput including at-mospheric attenuation at an airmass of 1.3 (Berk et al.1999). The predicted magnitude of C26202 in each of theDES grizY bands is g = 16 . r = 16 . i = 16 . z = 16 . Y = 16 . uncertainty. We derive an absolute offset (in mag) of δg = 0 . δr = 0 . δi = 0 . δz = 0 . δY = 0 . σ ( g ) = 0 . σ ( r ) = 0 . σ ( i ) = 0 . σ ( z ) = 0 . σ ( Y ) = 0 . σ ( g ) = 0 . σ ( r ) = 0 . σ ( i ) = 0 . σ ( z ) = 0 . σ ( Y ) = 0 . − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band -80 0 80 ∆ r (mmag) N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ r (mmag) 0.0040.0080.0120.016 N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ r (mmag) r-band22 mmag Figure 5.
Comparison of stellar magnitudes from the DES Y1A1 GCM and those estimated from APASS/2MASS transformedinto the DES filter system (Appendix A.4). The sky plot (left) shows the median magnitude offset for stars binned into ∼ . HEALPix pixels. The GCM calibrated magnitudes are consistent with the transformed values from APASS/2MASSwith a half-width of σ = 22 mmag (calculated between the 16th and 84th percentiles). Similar figures for other bands areshown in Appendix A. other CALSPEC standards (LDS749B and WD0308-565), the absolute calibration of Y3 is believed to beaccurate at the ∼
1% level. The relative calibration ofY3 was performed over the contiguous Y3A1 footprintusing an independent forward global calibration method(FGCM) and is found to be uniform at the 0 .
7% level(Burke et al. 2018). We checked the absolute calibra-tion of Y1A1 GOLD by matching stars against theirY3 counterparts over the Y1A1 GOLD footprint. Wefound that the absolute offset between Y1A1 GCM andY3A1 FGCM was δg = 0 . δr < . δi = 0 . δz = 0 . δY = 0 .
05, while the relative cal-ibration spread was σ ( g ) = 0 . σ ( r ) = 0 . σ ( i ) = 0 . σ ( z ) = 0 . σ ( Y ) = 0 . IMAGE COADDITIONImage coaddition allows DES to detect fainter objectsand mitigates the impact of residual transient imagingartifacts (e.g., unmasked cosmic rays, satellite streaks,etc.). Combining multiple dithered exposures also posi-tions objects at different points on the focal plane, mit-igating systematics associated with the non-uniform re-sponse of the instrument.DESDM produced image coadds from the weightedaverage of overlapping single-epoch images. The pix-els of the input images were remapped onto a uniformpixel grid using
SWarp with the
LANCZOS3 kernel (Bertinet al. 2002; Bertin 2010). The remapped pixel grid wasdefined on coadd tiles spanning 0 .
73 deg × .
73 deg andcomprising 10 × remapped pixels (a pixel scale of0 . (cid:48)(cid:48) / pix, comparable to the physical pixel scale of DE- Cam). For each tile, one coadded image was producedfor each photometric band.Before performing image coaddition, several imagequality checks were run to identify and blacklist CCDimages with severe imaging artifacts. CCD images af-fected by strong scattered light artifacts were identifiedby a ray tracing algorithm using the Yale bright starcatalog (Hoffleit & Jaschek 1991), the telescope point-ing, and a detailed model of the DECam optics, fil-ter changer, and shutter assemblies. Several exposureshave excess noise in one or more of the DECam CCDbackplanes. These CCD images were identified throughvisual inspection and through the detection of a largenumber of spurious catalog objects. In addition, CCDimages that were affected by bright meteor trails andairplanes were identified through visual inspection. Lessthan 1% of CCD images were blacklisted and removedfrom the coadd process.When DESDM created coadded images, the PSFs ofthe individual input images were not homogenized. Thisdecision was motivated by studies of SV data wherePSF homogenization was found to produce correlatedsky noise, which made it difficult to properly estimatethe photometric uncertainties of galaxies. While non-homogenized PSF coaddition yields better-behaved pho-tometric uncertainties, it can introduce sharp PSF dis-continuities on the ∼ . ◦ PSFEx ; Bertin 2006). Some of these issues can beaddressed by using quantities measured in the Y1A1 FI-NALCUT catalog (Section 6); however, studies that de-pend sensitively on morphological characterization (i.e.,1
Table 2.
Photometric Calibration Validation
Technique Band g r i z Y (mmag) (mmag) (mmag) (mmag) (mmag)
Absolute Photometric Offset
GCM vs. C26202 14 4 2 15 32GCM vs. Y3 FGCM 23 < Relative Photometric Uniformity
GCM vs. APASS/2MASS 19 22 20 20 18GCM+SLR vs. APASS/2MASS+SFD 25 24 20 18 15GCM vs. Y3 FGCM 14 7 8 13 15
Note —Summary of photometric calibration performance for the Y1A1 GOLD data set. SeeSection 4.4 for more details. weak lensing analyses) perform their own simultaneousfit of the individual single-epoch images (Section 6.3). In addition to the main survey, there are several re-gions where the DES Y1 imaging is considerably deeperthan the nominal three to four tilings. Coadds havebeen created in these regions using different numbersof input images to achieve different photometric depths.The Y1A1 GOLD coadd catalog thus contains four dif-ferent samples:1. WIDE: The WIDE coadd data sample is built fromexposures in the S82 and SPT regions of the Y1wide-area survey footprint and has a depth of threeto four tilings. One of the SN fields, SN-E, resideswithin the SPT region; however, to maintain uni-formity the WIDE data set only includes imagesthat were taken as part of the DES wide-area sur-vey (the SN-E exposures are included in the otherdata sets that follow).2. D04: The D04 sample is constructed by coaddingimages in the SN, COSMOS, and VVDS-14hfields with the goal of reaching an effectivedepth roughly comparable to the WIDE sam-ple. Quantitatively, exposures were selected togive (cid:80) exp j t eff ,j T exp ,j (cid:39) T wide , where t eff ,j is theeffective exposure time scale factor for exposure j (Section 2), T exp ,j is the shutter-open time forexposure j , and T wide is the wide-area exposuretime in Y1 (90s in griz and 45s in Y ). Whenselecting exposures for the D04 and D10 samples,we attempted to apply data quality selectionsbased on FWHM and t eff . For the D04 sample,exposures in the grizY bands were generally re-quired to pass the wide-area survey data qualityrequirements (Section 2) and have FWHM < . (cid:48)(cid:48) Studies with PSF homogenization are ongoing, and PSF-homogenized coadds have been used for several DES science anal-yses using SV data (Hennig et al. 2017; Klein et al. 2017).
However, in several cases these requirements wererelaxed to better approximate the desired depth.While the D04 sample was designed to mimic thedepth of the WIDE survey, the longer exposuretimes for the auxiliary and SN fields result in adata set that is on average ∼ . MAG AUTO σ limiting magni-tude for galaxies (Section 7.1) in the D04 sampleis g = 23 . r = 23 . i = 22 . z = 22 . Y = 20 . z algorithms and object classification (e.g.,Hoyle et al. 2017).3. D10: The D10 sample is constructed in the SN,COSMOS, and VVDS-14h fields by coadding im-ages to an effective depth of 10 exposures. The10-exposure depth is intended to mimic the ex-pected main survey depth at the end of DES. Sim-ilar to D04, general criteria requiring survey qual-ity, FWHM < . (cid:48)(cid:48) riz and FWHM < . (cid:48)(cid:48) gY were applied. The median MAG AUTO σ limitingmagnitude for galaxies (Section 7.1) in the D10sample is g = 24 . r = 24 . i = 23 . z = 22 . Y = 20 . < . (cid:48)(cid:48) riz -band and FWHM < . (cid:48)(cid:48) g -band (no FWHM requirement is placed on Y -band). Exposures are still required to pass thesurvey quality cuts, but no restriction is placed onthe number of exposures that go into the coadd.The median MAG AUTO σ limiting magnitude forgalaxies (Section 7.1) in the DFULL sample is g = 24 . r = 23 . i = 23 . z = 23 . Y = 21 . ∼
10% of the area having a limiting magni-tude greater than 25 in griz . Astrometric Accuracy
Astrometric calibration places the DES exposuresonto a consistent reference frame with each other andwith external catalogs. We used
SCAMP (Bertin 2006)to find an astrometric solution including corrections foroptical distortion towards the edges of the focal plane.During Y1A1 FINALCUT processing, initial astromet-ric calibration was performed on individual exposures.Starting with an approximate initial solution providedby the telescope control system, the
SExtractor win-dowed image coordinates of bright stars in the DES ex-posures were extracted and matched against the UCAC-4 stellar catalog (Zacharias et al. 2013).When building coadd tiles, an additional astrometricrefinement process was performed to remap the DES in-put images against each other and against the 2MASScatalog (Skrutskie et al. 2006). The single-epoch cata-logs from all exposures overlapping a tile were input to
SCAMP , and a simultaneous best fit was obtained treatingexposures from each filter as separate instruments. Thisbest-fit astrometric solution was used when combiningimages. After astrometric refinement, the median inter-nal astrometric precision of the Y1A1 wide-area coaddimages is ∼
25 mas (3 σ -clipped rms dispersion aroundthe mean for stars with S/N > This has been confirmed bycomparisons between Y3 DES data and
Gaia
DR1 (GaiaCollaboration 2016) where the median astrometric un-certainty is found to be ∼
150 mas (DES Collaboration2018). OBJECT CATALOGS6.1.
Coadd Catalog Creation
Catalogs of unique astrophysical sources were assem-bled from the coadded images. The goal of the DESDMcatalog production was to assemble the most inclusivecatalog of sources while maintaining a low contamina-tion fraction. The production of catalog subsamples The median depth of the DFULL sample in g - and r -band iscomparable to that of D10 owing to the fact that few additionalexposures passed the survey quality and FWHM requirements out-side of the deep SN fields. The r -band depth is 0 .
05 mag shallowerin DFULL owing to a slightly larger area with more varied dataquality. Bernstein et al. (2017b) show that using
Gaia
DR1 the astro-metric solution for a single DECam exposure can be made accurateto within 3 − that are complete to a given threshold is left to subse-quent science analyses. Source detection, morphologicalcharacterization, and multi-band photometric flux mea-surements were performed using SExtractor (Bertin &Arnouts 1996; Bertin et al. 2002). Source detectionused a
CHI-MEAN combination of the coadded imagesin r + i + z (Szalay et al. 1999; Bertin 2010). The CHI-MEAN detection image was designed to minimize dis-continuities between regions with different numbers ofexposures (see Appendix B). In contrast, flux and shapemeasurements were performed on each band individu-ally using
SExtractor in dual mode (i.e., analyzing theimage for an individual band simultaneously with thedetection image). The local background was estimatedvia 16 ×
16 pixel boxes with 3 σ clipping of bright pixelsand median filtering of the boxes. The image was con-volved with a 3 × . σ per pixelwas applied over the convolved image to detect objects.Source localization was derived from the barycenter ofthe object in the i, z, Y, r, g single-band coadd images(in order). Coadd object positions in world coordinates(J2000 epoch) were computed using the astrometric so-lution found during image coaddition (Section 5.1).The depth and PSF of the DES imaging result inoverlapping isophotes for objects in crowded regions,e.g., galaxy clusters, star clusters, and dense stel-lar regions around the LMC. Incomplete deblendingof overlapping objects affects the measured shapesand photometric properties of cluster galaxies, whichimpacts weak lensing and cluster cosmology science. SExtractor attempts to deblend each detected objectinto sub-components using a multi-thresholding algo-rithm (Bertin & Arnouts 1996). An object is sepa-rated into two (or more) new objects if the intensityof the new object is greater than a fraction of the to-tal intensity set by the
DEBLEND MINCONT parameter,while the number of deblending thresholds is set bythe
DEBLEND NTHRESH parameter. The Y1A1 process-ing campaign adopts 0.001 and 32, respectively, for thetwo parameters. These values were optimized based onSV data to balance completeness and purity for clustergalaxies. More aggressive deblending techniques for theDES data have been explored in Zhang et al. (2015).
SExtractor was used to measure object photometryvia several methods (see Sevilla et al. 2011).1. Fixed aperture fluxes (
FLUX APER ) were measuredfor 12 circular apertures with different radii from0 . (cid:48)(cid:48)
25 to 9 (cid:48)(cid:48) .2. Elliptical aperture fluxes (
FLUX AUTO ) were calcu-lated using the second-order moments of each ob-ject to derive the elongation and orientation of thebest-fit ellipse (Kron 1980). The ellipse scaling fac-tor was derived from the first-order moment of theradial distribution.3 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right AscensionInternal
17 25 33Astrometric Uncertainty(mas) N o r m a li z e d A r e a ( a . u . )
15 20 25 30 35Astrometric Uncertainty (mas)0.040.080.120.160.20 N o r m a li z e d A r e a ( a . u . )
15 20 25 30 35Astrometric Uncertainty (mas)
Internal25 mas − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right AscensionExternal
175 275 375Astrometric Uncertainty(mas) N o r m a li z e d A r e a ( a . u . )
150 200 250 300 350 400Astrometric Uncertainty (mas)0.0040.0080.0120.016 N o r m a li z e d A r e a ( a . u . )
150 200 250 300 350 400Astrometric Uncertainty (mas)
External330 mas
Figure 6. (Top): Relative internal astrometric error in milliarcseconds derived by comparing the positions of stars in theindividual DES exposures that go into the Y1A1 coadds. (Bottom): Relative external astrometric error derived by comparingthe position of stars in DES and 2MASS (without correcting for proper motion). The color scales represent the astrometricuncertainty in milliarcseconds, while the legends of the right panels report the modes of the distributions. The SN exposuretimes are significantly longer than the wide-area survey exposure times leading to a fainter saturation threshold. This reducesthe number of non-saturated bright stars and increases the astrometric uncertainty estimated by this technique (a more accurateestimate of the astrometry in the SN fields can be found in Kessler et al. 2015).
3. PSF model fluxes (
FLUX PSF ) suitable for point-like sources were fit to the measured PSF shape.As mentioned in Section 5, PSF discontinuities inthe Y1A1 coadd images can degrade the quality ofthe PSF model fluxes.4. Exponential model fluxes (
FLUX MODEL ) suitablefor galaxies were fit by convolving a one compo-nent exponential model with a local model of thePSF. These fluxes were fit both individually ineach band and by fixing the model shape basedon the detection image (
FLUX DETMODEL ).Among the morphological measurements performedby
SExtractor , two are designed to separate point-likeobjects (i.e., stars) from spatially extended sources (i.e.,galaxies). The first is the
CLASS STAR variable whichuses a neural network to assess the “stellarity” of anobject (Bertin & Arnouts 1996). The second variable,
SPREAD MODEL , is derived from the Fisher’s linear dis-criminant between a model of the PSF and an extended source model convolved with the PSF (Desai et al. 2012;Bouy et al. 2013; Soumagnac et al. 2015). The applica-tion of these variables to star-galaxy separation is de-tailed in Section 8.1.As stated previously, catalog quantities were also de-rived for individual single-epoch exposures that com-pose the coadded images. Objects detected on the in-dividual exposures were associated with sources in thecoadd catalog using a 1 (cid:48)(cid:48) matching radius. While shal-lower, the single-epoch catalogs are important for prob-ing the temporal domain. Additionally, the photome-try of the single-epoch catalogs is not subject to thePSF discontinuities present in the coadds. For this rea-son, we calculated a number of photometric and mor-phological quantities from the average of single-epochmeasurements weighted by their associated statisticaluncertainties (the names of these quantities are prefixedby “
WAVG ”). In particular, the weighted-average spread-model quantity (
WAVG SPREAD MODEL ) has been shown to4
Table 3.
Y1A1 GOLD Catalog Selection
Selection Description
NITER MODEL { GRIZ } > g, r, i, z -bands. SPREADERR MODEL { GRIZ } > SPREADERR MODEL = 0 indicates afailure in the photometric fit. yield better star-galaxy separation (Drlica-Wagner et al.2015) for stellar objects, and the weighted-average PSFmagnitudes (
WAVGCALIB MAG PSF ) have been found toyield more precise stellar photometry than the corre-sponding coadd quantities. In addition, uncertaintiesfor the
WAVG quantities are calculated directly from thevariance in the measurements from individual exposuresand thus avoid any systematics introduced in the coad-dition process.6.2.
Y1A1 GOLD Catalog Selection
We assembled the Y1A1 GOLD object catalog as ahigh-quality subselection of the objects extracted fromthe Y1A1 coadd images. When selecting the Y1A1GOLD catalog, we sought to remove spurious, non-physical objects while minimally decreasing the sta-tistical power of any scientific investigation (Table 3).Specifically, we required that objects be observed, butnot necessarily detected, at least once in each of the g , r , i , and z bands. We also required that all objects have SPREADERR MODEL > g , r , i , and z bands toeliminate objects with unphysical SPREADERR MODEL val-ues indicative of a failure in the
SExtractor photomet-ric fit. We also identify several classes of objects thatare extremely unusual and flag them for exclusion frommost cosmological analyses (Table 4). In addition to ob-jects flagged by
SExtractor , we specifically identify (1)objects with extremely blue ( { g − r, r − i, i − z } < −
1) orextremely red ( { g − r, r − i, i − z } >
4) colors, (2) brightstars that saturate some of the single-epoch inputs tothe coadd image, (3) objects that have a large ( > (cid:48)(cid:48) )offset in the windowed centroid derived from the g and i bands. Finally, we require that objects reside withinthe Y1A1 GOLD footprint (Section 7.3) and flag anyobjects that reside in poor-quality or potentially prob-lematic (“bad”) regions (Section 7.4).6.3. Multi-Epoch, Multi-Object Fitting
The Y1A1 coadded images provide deeper and moresensitive object detection than individual single-epochimages. However, the coaddition process averages across Objects that are not detected in a specific band have a sen-tinel value of
SPREADERR MODEL = 1. multiple images, resulting in a discontinuous PSF andcorrelated noise properties. Precision measurementsthat rely on an accurate PSF determination, such asgalaxy shape measurements for cosmic shear, require ajoint fit of pixel-level data from multiple single-epochimages.We used the ngmix code (Sheldon 2014; Sheldon &Huff 2017; Jarvis et al. 2016) to reanalyze pixel-leveldata from multi-epoch postage stamps of each object inthe Y1A1 GOLD coadd catalog. We used PSFEx (Bertin2011) to model and interpolate the PSF at the locationof each object, and then we generated an image of thePSF using the python package, psfex . We then usedthe ngmix code to fit this reconstructed PSF image to aset of three free, independent Gaussians.We used ngmix in “multi-epoch” mode to simultane-ously fit a model to all available epochs and bands. Inthis mode, a model is convolved by the local PSF in eachsingle-epoch image, and a χ sum is calculated over allpixels in a postage stamp. This is repeated for eachepoch and band, and a total χ sum is calculated. Wethen find the parameters of the model that maximizethe likelihood.We took this procedure one step further, performingsimultaneous multi-epoch, multi-band, and multi-objectfit, which we call “MOF”. We first identified groups ofobjects using a friends-of-friends algorithm (e.g., Huchra& Geller 1982; Berlind et al. 2006). We then fit themembers of the group using the following procedure:1. Perform an initial model fit to each object, mask-ing the light from neighbors using the ¨uberseg al-gorithm (Jarvis et al. 2016).2. Fit the model to each object again, this time sub-tracting the light from neighbors using the modelsfrom the previous fit.3. Repeat the previous step until all fits converge, ora maximum of 15 iterations was reached.This fit was performed simultaneously in the g, r, i, z bands using all available imaging epochs and assumingthe same spatial model for all bands and epochs. Anexample of this procedure is shown in Figure 7.We found that fitting a galaxy model with fully freebulge and disk components was highly unstable, so weadopted the following approach, inspired by the “com-posite” model used in the SDSS. We first fit the diskand bulge models separately, represented by an exponen-tial and De Vaucouleurs’ profile (de Vaucouleurs 1948),respectively. We then determined the linear combina- https://github.com/esheldon/ngmix https://github.com/esheldon/psfex Table 4.
Y1A1 GOLD Catalog Flags
Flag Bit Selection Description1
FLAGS { GRIZ } > SExtractor { g − r, r − i, i − z } < − { g − r, r − i, i − z } > NEPOCHS G = 0) AND (
MAGERR AUTO G < . MAG MODEL I − MAG AUTO I ) < − . | α J ,g − α J ,i | > (cid:48)(cid:48) OR | δ J ,g − δ J ,i | > (cid:48)(cid:48) )AND ( MAGERR AUTO G < .
05) Objects with large astrometricoffsets between bands tion of these models that best fit the data, M tot = f dev M dev + (1 − f dev ) M exp (1)where M dev is the bulge model, M exp is the disk model,and f dev represents the fraction of light in the bulgecomponent. This total model is unlikely to be a goodfit of the data, and we only use it as a starting point fora more refined model. We formed a new model that hasthe best f dev determined as above, as well as the sameratio of scale lengths for the bulge and disk components.This new model has free parameters for the center, ellip-ticity, overall scale, and fluxes. A common center, scale,and ellipticity were used for all bands, but the flux foreach band was left free.For computational efficiency, each component of thismodel was approximated by a sum of Gaussians (Hogg& Lang 2013). This choice made convolution with thetriple-Gaussian PSF model very fast. A fast approxima-tion for the exponential function was also used to speedup computations (Sheldon 2014).We imposed uninformative priors on all parametersexcept for the ellipticity and the fraction of light presentin the bulge, f dev . For both of these parameters, weapplied priors based on fits to deep COSMOS imag-ing data, provided as postage stamps with the GalSim project . We defined convergence to be when the fluxfrom subsequent fits to objects did not change morethan one part in a thousand, and structural parame-ters such as scale and ellipticity did not change by morethan a part in a million. For incorporation into theY1A1 GOLD catalog, we converted MOF fluxes to mag-nitudes and applied the SLR adjustment discussed inSection 4.3. 6.4. Catalog Completeness
We assessed the completeness and purity of the Y1A1GOLD catalog by comparing it against data from the https://github.com/GalSim-developers/GalSim Figure 7.
A group of objects fit using the MOF algo-rithm. In the top row we show: (left) the sky-subtracted im-age, (center) the models for neighboring sources, and (right)the SExtractor segmentation map. In the bottom row weshow: (left) the sky-subtracted image after also subtractingthe light from neighbors, (center) the fraction of light as-signed to the central object (100% in white and 0% in black),and (right) the weight map. Note that a bad column and aflagged object are identified in the weight map. The maskedobject was not fit, and thus its light was not subtracted.
Canada-France-Hawaii Lensing Survey (CFHTLenS)W4 field (Erben et al. 2013; Hildebrandt et al. 2012),which overlap the S82 region of Y1A1 GOLD. The DESdata in this overlap region has a typical 10 σ limitingmagnitude of g ∼ . r ∼ . i ∼ . z ∼ . i and z bands and ∼ . g and r bands (Table 1). Inthis region, CFHTLenS is (cid:38)
18 19 20 21 22 23 24 25
Magnitude in DES System F r a c t i o n o f O b j e c t s g-bandr-bandi-bandz-band Figure 8.
Completeness (solid circles) and contamination(dashed triangles) of the Y1A1 GOLD coadd object cata-log determined by comparison to the CFHTLenS W4 field.Object matching was performed within a 1” radius andCFHTLenS magnitudes were transformed to the DES sys-tem using the equations in Appendix A.4. Statistics werecalculated for the subset of objects that were unmasked inboth surveys and have been truncated at the 5 σ limitingmagnitude of CFHTLenS (Erben et al. 2013). tude of CFHTLenS objects into the DES system (Ap-pendix A.4) and removed objects residing in maskedregions of either survey. We associated objects betweenthe two catalogs based on a spatial coincidence of 1 (cid:48)(cid:48) and required a matching magnitude within 2 mag. Wethen calculated the detection completeness as the frac-tion of CFHTLenS objects that are matched to Y1A1GOLD objects as a function of the CFHTLenS magni-tude transformed into the DES system. The contam-ination of the Y1A1 GOLD catalog is assessed as thefraction of Y1A1 GOLD objects that are unmatched toCFHTLenS objects as a function of magnitude. We findthat the 95% completeness limit of the Y1A1 GOLDcatalog is g = 23 . r = 23 . i = 22 .
9, and z = 22 . (cid:46) > ∼
1% of objects at DES depth.We also note that Y1A1 object detection was performedon a combined r + i + z detection image and no S/Nthreshold was applied to the measurements in individualbands when calculating completeness. ANCILLARY MAPSSeveral ancillary maps were produced to character-ize the coverage, sensitivity, observing conditions, andpotentially problematic regions of Y1A1 GOLD as afunction of sky position. Generating ancillary mapsfor Y1A1 GOLD was a multi-step process: we createda vectorized representation of the survey coverage and limiting magnitude using mangle (Hamilton & Tegmark2004; Swanson et al. 2008), we rasterized the mangle maps with
HEALPix for ease of use, we estimated observ-ing conditions over the survey footprint, and we subs-elected a nominal high-quality footprint. Finally, weflagged sky regions where the true survey performancedeviates from that estimated by the ancillary data prod-ucts (i.e., the regions around bright stars, astrometricfailures, etc.). Each of these steps is described in moredetail below.7.1.
Maps of Survey Coverage and Depth
Quantifying survey coverage and limiting magnitudeas a function of sky position is essential for statisti-cally rigorous cosmological analyses. To accurately trackcharacteristics of the DES survey at the sub-CCD level,DESDM produces mangle masks (Hamilton & Tegmark2004; Swanson et al. 2008) as part of the Y1A1 COADDpipeline. These masks are an accurate representationof the coverage, sensitivity, and overlap of DECam ex-posures including dead CCDs, gaps between CCDs,masked regions around bright stars, and bright streaksfrom Earth-orbiting satellites.During coadd production, mangle masks were createdat the level of coadd tiles (Figure 9). The steps are thefollowing:1. Polygons were created using the four corners ofeach input CCD image and assigned a weight equalto the median value of pixels in the CCD weightplane.2. Satellite streaks were represented by polygons, andthe area of these polygons was removed from thesingle-epoch CCD polygon.3. Polygons were trimmed to fit the tile boundaries.4. Polygons were subdivided into disjoint regionswith the balkanize command. Following theweighted-average scheme chosen for image coad-dition, the total weight of a balkanized polygon isthe sum of the weights of the individual polygons.5. Regions around bright stars and bleed trails areremoved from the mangle mask. While the preciselocation of these artifacts is image dependent, it iscomputationally simpler to mask the stacked mapwith the largest shape covering a bright star orbleed trail rather than removing these regions fromeach single-epoch polygon.6. The mangle coadd weight map was converted intoa 10 σ limiting magnitude map for a 2 (cid:48)(cid:48) diameteraperture: m lim = m ZP − . (cid:32) (cid:115) π ( D/ ω w tot (cid:33) , (2)7 D E C ( d e g ) σ ) 0.2 0.4 0.6 0.8 1.0Coverage Fraction Figure 9.
Coverage and depth maps for a single Y1A1 coadd tile. (Left) Vectorized mangle weight map for an r -band coaddtile. Satellite trails, star masks, and chip gaps are stored at full resolution. (Center) Pixelized 10 σ limiting magnitude mapfor galaxies using HEALPix at nside = 4096. (Right) Pixelized map of the coverage fraction at HEALPix nside = 4096. Thistile is located on the border of the Y1A1 footprint and has been chosen for illustrative purposes due to its variable depth andincomplete coverage. where m ZP = 30, is the tile zero-point, D = 2 (cid:48)(cid:48) , ω pix = 0 . (cid:48)(cid:48)
263 is the pixel size, and w tot is the totalweight of the polygon. This definition of the mag-nitude limit corresponds to the MAG APER 4 quan-tity measured by
SExtractor .While the vectorized mangle masks are a very accu-rate representation of the DES survey coverage, theyare computationally unwieldy for many scientific anal-yses. To increase the speed and ease with which sur-vey coverage and limiting magnitude can be accessed,we generate anti-aliased
HEALPix maps of these quanti-ties (Figure 9). Pixelized maps of the survey coveragefraction were created at a resolution of nside = 4096(area = 0 .
73 arcmin ) by calculating the fraction ofhigher-resolution subpixels ( nside = 32768, area =0 .
01 arcmin ) that were contained within the mangle mask. Similarly, maps of the survey limiting magni-tude were generated at nside = 4096 by calculating themean limiting magnitude for subpixels ( nside = 32768).When calculating the limiting magnitude, subpixels thatwere not covered by the survey were excluded from thecalculation, while subpixels that have been masked (i.e.,bright stars, bleed trails, etc.) were assumed to havethe limiting magnitude of their parent polygon. The HEALPix resolution of nside = 4096 was chosen as acompromise between computational accuracy and easeof use. This resolution was found to have a negligibleeffect on the correlation function of simulated galaxieson scales larger than 0 . (cid:48) mangle coverage and depth maps into 10 σ limiting magnitude maps for galaxy photometry. We se-lected galaxies using the MODEST CLASS star-galaxy clas-sifier (Section 8.1) and trained a random forest model topredict the 10 σ limiting magnitude as a function of ob-serving conditions. The input vector for the random for- est included the PSF FWHM, sky brightness, airmass,and exposure time for each band being fit (Section 7.2).The training was performed on coarse HEALPix pixels( nside = 1024) that contained more than 100 galaxies.Once trained, the model was applied to the pixels atthe full mask resolution of nside = 4096. We derivedmagnitude limits for both coadd
AUTO magnitudes andthe multi-epoch composite model magnitudes derived bythe MOF (Section 6.3). We applied the SLR calibrationadjustment (Section 4.3) to the resulting depth mapsto correct for interstellar extinction and zeropoint non-uniformity. The median 10 σ limiting magnitudes for MAG AUTO are g = 23 . +0 . − . , r = 23 . +0 . − . , i = 22 . +0 . − . , z = 21 . +0 . − . , Y = 20 . +0 . − . , where the uncertaintiesrepresent the 16 th and 84 th percentiles of the distribu-tion. In comparison, the median 10 σ limiting magni-tudes for the MOF CM MAG magnitudes are g = 23 . +0 . − . , r = 23 . +0 . − . , i = 22 . +0 . − . , and z = 22 . +0 . − . . We findthat the depth estimates are accurate at the level of 6%-7%, but that 3%-4% of this measured uncertainty is dueto “pixelization noise” resulting from averaging over arange of depths when fitting the model on coarse pixels.An example of the resulting depth maps for r band canbe found in Figure 10, and figures for the other bandscan be found in Appendix C.7.2. Maps of Survey Characteristics
Variations in observing conditions can be a significantsource of systematic uncertainty in cosmological analy-ses. In a wide-area optical survey such as DES, variableobserving conditions can imprint spurious spatial cor-relations, noise, and depth fluctuations on the objectcatalogs that are used for galaxy clustering and cosmicshear analyses. By identifying and characterizing thesesystematic effects, it becomes possible to quantify andminimize their impact on scientific results. We followedthe procedure developed by Leistedt et al. (2016) to con-8 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) r-band23.2 mag Figure 10.
Sky map and normalized histogram for the r -band 10 σ limiting magnitude ( MAG AUTO ) derived in HEALPix pixelsover the Y1A1 GOLD footprint. The mode of the limiting magnitude distribution is shown inset in the right panel. Thederivation of the limiting magnitude is described in Section 7.1. Similar figures for other bands are shown in Appendix C. struct survey characteristic and coverage fraction mapsfor the Y1A1 GOLD data set using
QuickSip . Sincethe nonlinear transfer function between the stack of im-ages at any position on the sky and the final galaxycatalog is largely unknown, we created maps of manydifferent survey observables. For each band, we createdmaps of both weighted- and unweighted-average quan-tities of each image. The main quantities expected tobe used for null tests in cosmological analyses with theY1A1 GOLD catalog are the total exposure time, themean PSF FWHM, the mean airmass, and the sky back-ground. The inverse variance weighted averages of thesequantities are shown in Figure 11. Further modeling ofthe survey transfer function is important for DES cos-mology analyses, and several approaches have alreadybeen developed (e.g. Chang et al. 2015; Suchyta et al.2016). 7.3.
Footprint Map
The nominal footprint for the Y1A1 GOLD catalogis defined using an nside = 4096
HEALPix map. Fora pixel to be included in the Y1A1 GOLD footprint, itmust meet the following criteria simultaneously in the g, r, i, z bands:1. A mangle coverage fraction ≥ . mangle (Section 7.1).2. A coverage fraction of ≥ . ≥
90 s (Sec-tion 7.2).4. A valid solution from the SLR calibration adjust-ment (Section 4.3). https://github.com/ixkael/QuickSip These selection criteria reduce the total coadded areaof Y1A1 covered in any band, 1927 deg , to a nomi-nal WIDE+D04 Y1A1 GOLD footprint in g, r, i, z of1786 deg . Simultaneously applying the same criteria tothe Y band (with a minimum exposure time of 45 s) re-sults in a g, r, i, z, Y footprint of 1773 deg . These num-bers were calculated by summing the coverage fractionof pixels in the footprint.7.4. Bad Region Mask
Masks were developed to remove regions where sur-vey artifacts make it difficult to control systematic un-certainties when doing cosmological analyses. Since notall science topics require the same masks (e.g., stud-ies of galaxy evolution may not want to mask nearbygalaxies), the various masks are collected into a bitmapdefined in Table 5. Removing area associated with anyof these masks results in a WIDE+D04 footprint areaof 1506 deg in g, r, i, z and 1496 deg in g, r, i, z, Y .7.4.1. Catalog Artifacts Unphysical colors (bit=64):
This mask is designedto remove imaging artifacts that were not maskedbefore creating coadds. In particular, this mask re-moves regions where there are significant reflectedlight artifacts (both specular and diffuse) frombright stars, un-masked orbital satellite trails, andcoadd saturation artifacts. This mask is pixelizedat nside = 2048 and pixels with ≥ . .2. Astrometric discrepancies (bit=1):
We flag regionsthat have a high concentration of galaxies withlarge astrometric offsets between filters. We selectgalaxies with i <
MAGERR AUTO G < .
2, and9 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Airmass − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ − ◦ ◦ ◦ ◦ ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ − ◦ ◦ ◦ ◦ ◦ − ◦ FWHM (pix) − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension
114 210 306
Exposure Time (s) − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension
230 305 381
SkyBrightness (ADU)
Figure 11.
Survey characteristics of the Y1A1 GOLD data set estimated from the inverse variance weighted stack of single-epoch images in r -band at each position on the sky. Panels correspond to mean airmass (top left), PSF FWHM in pixels (topright), exposure time in seconds (bottom left), and sky brightness in ADU (bottom right). Table 5.
Y1A1 GOLD Bad Region Mask
Flag Bit Area Description( deg )1 30.1 High density of astrometric discrepancies2 119.5 2MASS moderate star regions (8 < J < < B < < J < − < V < . Note — Masked regions for the Y1A1 GOLD WIDE+D04 footprint.The masked area is calculated using the coverage fraction of thepixels that are removed from the footprint by each mask. Thecriteria defining each mask can be found in Section 7.4. windowed positions in g and i band differ by morethan 1 (cid:48)(cid:48) . This criterion has been found to selectobjects in regions of strongly variable background(e.g., the wings of bright stars, regions of poorsky subtraction, regions with scattered light, etc.).The resulting masked area in this case is 30 . .3. PSF model failures (bit=512):
There are severalregions where PSF modeling failed owing to vary-ing depth and a discontinuous PSF. Coadd tilespossessing poor PSF models are identified as hav-ing a large number of stars where the coadd PSFmagnitude differs from the weighted-average sin-gle epoch PSF magnitude by more than 0.2 mag.We flag
HEALPix pixels ( nside = 512) possessing >
20 stars with large discrepancies in PSF mag-nitudes. The total region masked in this way is7 . . 7.4.2. Bright Stars
Regions around saturated stars were masked at thepixel level as part of the image processing pipeline de-scribed in Section 3. However, catalog-level investiga-0tion revealed a residual increase in the number densityof objects surrounding the brightest stars. To avoid con-tamination from spurious objects in the halos of brightstars, we designed radial masks based on the brightnessof the contaminating stars and the number density ofsurrounding objects. These masks were developed fortwo bright star catalogs as described below.1.
Yale bright star regions (bit=32):
Masked regionswere determined from the positions and magni-tudes of stars in the Yale Bright Star Catalog (Hof-fleit & Jaschek 1991). The masking radius was de-termined from the V -band magnitude of each star,following the equation: R = 0 . − (0 . × V. (3)Minimum and maximum masking radii were im-posed at 0.1 deg and 0.4 deg, respectively. Theresulting masked area is 18 . .2. We mask regionsaround bright stars from the 2MASS catalog(Skrutskie et al. 2006) within a radius of R = 0 .
09 deg − (0 . × J, (4)assuming a minimum and maximum masking ra-dius of 0.01 deg and 0.05 deg, respectively. Manyof the bright stars in 2MASS overlap with thefaintest stars in the Yale Bright Star Catalog,and we find a comparable masking radius (al-beit derived using different bands). Because thefainter 2MASS stars may not be problematic forall science applications, we split the 2MASS starmask into stars with 5 < J < < J <
12. The masked areas are 38 . and119 . , respectively.7.4.3. Large Foreground Objects The Large Magellanic Cloud (bit=16):
The cen-ter of the Large Magellanic Cloud (LMC) is lo-cated ∼ ◦ < α < ◦ and − ◦ < δ < − ◦ . The LMC mask removed95 . .2. Bright galaxies (bit=4):
The Third Reference Cat-alog of Bright Galaxies (RC3; Corwin et al. 1994) contains galaxies subtending (cid:38) (cid:48) . Since galaxysize is highly correlated with magnitude, we con-tinue to use a magnitude-dependent masking for-mulation similar to that applied to bright stars.We masked a circular region around RC3 galaxieswith 10 < B <
16 with a magnitude-dependentselection: R = 0 .
269 deg − (0 . × B. (5)We imposed minimum and maximum maskingradii such that 0 .
03 deg < R < . . .3. Globular clusters (bit=256):
The high stellar den-sity of Milky Way globular clusters makes themdifficult regions for cosmology analyses. We identi-fied three globular clusters, NGC 1261, NGC 1851,and NGC 7089, and masked circular regions withradius 1.5 times the angular size reported by Sin-nott (1988). This resulted in a total masked areaof 0 . . VALUE-ADDED QUANTITIESThe astrometric, photometric, and morphological pa-rameters derived for each object are supplemented withadditional information important for astrophysical andcosmological analyses. These “value-added quantities”are built from the calibrated coadd object catalog andprovide additional information on an object-by-objectbasis. The two primary value-added quantities providedwith Y1A1 GOLD are: (1) a simple star-galaxy classi-fier, and (2) a set of photo- z estimates.8.1. Star-Galaxy Separation
As part of the Y1A1 GOLD catalog, we pro-duced a “
MODEST CLASS ” object classification with theprimary goal of selecting high-quality galaxy sam-ples.
MODEST CLASS is based on the i -band coaddquantity SPREAD MODEL I and its associated error,
SPREADERR MODEL I . SPREAD MODEL is a morphologi-cal variable defined as a normalized linear discriminantbetween the best-fit local PSF model and a slightly moreextended model composed of a circular exponential diskconvolved with the PSF (Desai et al. 2012; Soumagnacet al. 2015). The i band was chosen as the referenceband for object classification owing to its depth andsuperior PSF. Image-level simulations of the DES datasupport the conclusion that i band yields the best over-all performance for object classification, and this resultwas verified using deep HST imaging on the COSMOSfield.We used space-based imaging of COSMOS (Leau-thaud et al. 2007) and GOODS-S (Giavalisco et al. 2004)along with spectroscopic observations from VVDS (LeF`evre et al. 2005) that overlapped the Y1A1 GOLD foot-print as a truth sample for developing MODEST CLASS .1 Table 6.
Y1A1 GOLD
MODEST CLASS
Star-Galaxy Classification
Class Selection Description0
SPREAD MODEL I + (5 / × SPREADERR MODEL I < − .
002 Unphysical PSF fit (likely stars)1
SPREAD MODEL I + (5 / × SPREADERR MODEL I > .
005 ANDNOT ( | WAVG SPREAD MODEL I | < .
002 AND
MAG AUTO I < .
5) High-confidence galaxies2 | SPREAD MODEL I + (5 / × SPREADERR MODEL I | < .
002 High-confidence stars3 0 . < SPREAD MODEL I + (5 / × SPREADERR MODEL I < .
005 Ambiguous classification
Note — The high-purity and high-completeness galaxy samples are defined as
MODEST CLASS = 1 and
MODEST CLASS ∈ { , } , respectively.Similarly, the high-purity and high-completeness stellar samples are defined as MODEST CLASS = 2 and
MODEST CLASS ∈ { , , } , respectively. SPREAD_MODEL_I
SPREADERR_MODEL_I
GalaxiesStars
CLASS_STAR_I
SPREAD_MODEL_I + ( / ) SPREADERR_MODEL_I
Galaxies Stars
16 18 20 22 24 i (mag) SPREAD_MODEL_I + ( / ) SPREADERR_MODEL_I
GalaxiesStars l og ( C o un t s ) Figure 12.
MODEST CLASS star-galaxy selection for objects in a ∼
13 deg region centered on α , δ = (51 ◦ , − ◦ ). The leftpanel shows the distribution of SPREAD MODEL I and its error. The middle panel compares the distribution of the
SExtractor neural-network classifier,
CLASS STAR , to the
MODEST CLASS selection criteria. The right panel shows a tight stellar locus inthe i -band magnitude compared against the MODEST CLASS criteria. In all panels the black (red) lines correspond to the pure(complete) galaxy selection threshold applied on
MODEST CLASS . We defined star and galaxy samples optimized for“high completeness” and “high purity” by applyingthresholds on the combination of
SPREAD MODEL I and
SPREADERR MODEL I . The object classification schemeis defined in Table 6 and shown graphically in Figure 12.Following Drlica-Wagner et al. (2015), we validatedthe performance of the
MODEST CLASS star-galaxy clas-sifier on data from CFHTLenS (Erben et al. 2013;Hildebrandt et al. 2012). We matched CFHTLenScatalog objects to the Y1A1 GOLD data (Section 6)and selected high-quality samples of stars and galax-ies using the
CLASS STAR and
FITCLASS measurementsby CFHTLenS (Heymans et al. 2012). Specifically,our CFHTLenS stellar selection was (
FITCLASS =1) OR (
CLASS STAR > .
98) and our galaxy selectionwas (
FITCLASS = 0) OR (
CLASS STAR < . ∼
7% of matched CFHTLenS objects are unclassified The high-completeness and high-purity samples differ in theclassification assigned to ambiguous objects. according to this prescription, and these objects are notused for assessing the performance of
MODEST CLASS .We define the “efficiency” of a galaxy sample as thenumber of true galaxies that are also classified as galax-ies divided by the total number of true galaxies inthe sample (i.e., the true positive rate). Conversely,the “contamination” of a galaxy sample is defined asthe number of galaxies that are misclassified dividedby the total number of objects classified as galaxies(i.e., the false discovery rate). Similar definitions ap-ply to the stellar selections, and the performance of the
MODEST CLASS galaxy and star selections are shown inFigure 13. We find that a high-purity galaxy selectionhas an efficiency (cid:38)
98% and a contamination rate (cid:46) i <
22. In contrast, the high-completeness stellarselection has an efficiency of (cid:38)
86% with a contamina-tion of (cid:46)
6% for i <
22. We estimate similar perfor-mance for
MODEST CLASS through a comparison againstthe DEEP2-3 field in the first public data release of Hy-per Suprime Camera (Aihara et al. 2018).2
17 18 19 20 21 22 23 24 25
Magnitude ( i -band) F r a c t i o n o f O b j e c t s Galaxy SelectionEfficiency (pure)Contamination (pure)Efficiency (complete)Contamination (complete)
17 18 19 20 21 22 23 24 25
Magnitude ( i -band) F r a c t i o n o f O b j e c t s Stellar SelectionEfficiency (pure)Contamination (pure)Efficiency (complete)Contamination (complete)
Figure 13.
Performance of the
MODEST CLASS star-galaxyclassifier based on a comparison to deeper imaging fromCFHTLenS. (Top): The measured efficiency and contamina-tion for high-purity (solid) and high-completeness (dashed)galaxy samples. (Bottom): The efficiency and contamina-tion for high-purity (solid) and high-completeness (dashed)stellar samples (note that
MODEST CLASS is not optimized forstellar selection).
The
MODEST CLASS selection provides an initial base-line for object classification and is found to be sufficientfor characterizing the distributions of stars and galax-ies in Y1A1 GOLD (Figures 14 and 15). Multi-variatemachine-learning techniques and template-fitting algo-rithms have the potential to provide much better objectclassification (e.g. Fadely et al. 2012; Soumagnac et al.2015, etc.). Several advanced object classification tech-niques are currently being explored within DES and willbe detailed in future publications (Sevilla-Noarbe et al.2018). We emphasize that
MODEST CLASS has been opti-mized for galaxy selection. Several alternative selectionshave been suggested for more complete samples of stars(e.g. Bechtol et al. 2015; Drlica-Wagner et al. 2015).8.2.
Photometric Redshift Estimation
In this section we briefly summarize the approach tophoto- z estimation and validation for DES Y1 scienceanalyses. While photo- z estimates were provided as partof the initial Y1A1 GOLD data set, it was realized that
17 18 19 20 21 22 23 24 25
MAG_AUTO O b j e c t s p e r b i n Stars g-bandr-bandi-bandz-band
17 18 19 20 21 22 23 24 25
MAG_AUTO O b j e c t s p e r b i n Galaxies g-bandr-bandi-bandz-band
Figure 14.
Number counts of objects passing the
MODEST CLASS pure star selection (top) and complete galaxyselection (bottom) as a function of
MAG AUTO magnitude inthe g , r , i , and z bands. The impact of galaxy contamina-tion can be seen in the stellar number counts at magnitudesfainter than i (cid:38) individual cosmology analyses benefit from photo- z es-timation and validation customized to their distinct sci-ence samples. Therefore, we present a general overviewof the photo- z estimation and validation procedures, andwe refer the reader to upcoming publications dedicatedto photo- z estimation for distinct DES analyses (e.g.,Hoyle et al. 2017; Gatti et al. 2018; Davis et al. 2017;Cawthon et al. 2017).Photo- z estimates were generated with two distinctalgorithms: the machine-learning code DNF (De Vicenteet al. 2016), and a modified version of the template code
BPZ (Ben´ıtez 2000; Hoyle et al. 2017). These two codesare representative of common machine learning and tem-plate fitting photo- z estimation techniques. Both algo-rithms utilized spectroscopic data for training, and adetailed discussion of the spectroscopic sample can befound in Gschwend et al. (2017).For many cosmological analyses, we are interested inaccurately characterizing the statistical distribution ofgalaxies in tomographic bins of redshift and less inter-ested in predicting the redshift of any individual galaxy.Thus, we applied two independent techniques targeted3 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension × × Stellar Density(deg − ) − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension × × Galaxy Density(deg − ) Figure 15.
Density of objects with i <
22 passing the high-completeness star (left) and high-purity galaxy (right)
MODEST CLASS selections. The linear color scales represent the density of catalog objects and are the same for both panels. The density ofobjects has been corrected for the coverage fraction of each pixel (Section 7.1). at validating the statistical properties of our predictedphoto- z distributions (Hoyle et al. 2017; Davis et al.2017).1. We performed a direct validation of the color-redshift relationship by matching galaxies fromDES science samples to galaxies with multi-bandphotometry obtained within the COSMOS field(Laigle et al. 2016). This choice of validation datamitigated the impact of redshift or galaxy-type de-pendent selection biases, which can affect spectro-scopic surveys (e.g., Bonnett et al. 2016; Hartleyet al. 2018). However, the 30-band photo- z esti-mates from COSMOS have a larger intrinsic uncer-tainty than spectroscopically determined redshifts.In addition, validating performance on a ∼ field leads to large uncertainty due to cosmic vari-ance, which was estimated using the Buzzard suiteof ΛCDM simulations (S´anchez et al. 2017; Wech-sler et al. 2017; DeRose et al. 2017).2. A second, independent indirect validation tech-nique relies on the clustering-redshift technique(Newman 2008; M´enard et al. 2013; Schmidt et al.2013). We selected a luminous red galaxy sam-ple (redMaGiC; Rozo et al. 2016), which has well-determined photo- z estimates, as a reference anddivided this sample into redshift bins of width∆ z = 0 .
02. We then divided the full sample ofDES objects into tomographic redshift bins basedon predicted photo- z and cross correlated the datain each tomographic bin with each of the morefinely binned redMaGiC reference samples. Wemeasured the excess angular cross-correlation sig-nal, which is proportional to the redshift distribu-tion. We calibrated a constant redshift offset ineach tomographic bin between the photo- z predic-tions and the clustering signal. We estimated theerrors arising from the evolution of galaxy-darkmatter halo bias and discrepancies in the shape of the clustering reshift distribution by repeatingthe same analysis using the Buzzard simulations(Gatti et al. 2018; Cawthon et al. 2017).Both validation techniques possess associated uncer-tainties. The direct validation technique has compa-rable uncertainties from sample variance (since COS-MOS covers a ∼ region of the sky) and systematicuncertainty in matching the morphological and color-magnitude-error distribution of the galaxy sample. Incontrast, we find that the dominant systematic uncer-tainties for the indirect validation technique come fromthe clustering bias evolution of the binned source galaxysamples and incorrectness in the shape of the photo- z distribution. In addition, we are unable to performindirect clustering validation for tomographic bins with z (cid:38) z performance metric forcosmic shear analyses is the bias of the estimated meanof a redshift distribution in a tomographic bin with re-spect to the unknown true mean redshift in that bin(Bonnett et al. 2016). We characterized the photo- z accuracy from the photo- z bias distribution, defined asthe difference between the average measured photomet-ric redshift and the average true redshift distribution,∆ z = (cid:104) z true (cid:105) − (cid:104) z phot (cid:105) . Since the true redshift distri-bution is unknown, we employed the direct and indirectvalidation techniques described above to estimate (cid:104) z true (cid:105) and ∆ z in four tomographic bins with 0 . < z < .
3. Wefind that both techniques yield | ∆ z | (cid:46) .
02 with an un-certainty of comparable magnitude when applied to theBPZ estimates for the primary subsample of the Y1A1GOLD catalog used for cosmic shear analyses (Hoyleet al. 2017; Zuntz et al. 2017).We present several other results from the validationof the BPZ template code optimized over the redshiftrange 0 . < z < . n ( z ) distribution for the weak-lensing shear cata-4log derived from Y1A1 GOLD. The n ( z ) distributionis found to be in good agreement with the n ( z ) pre-dicted from COSMOS when cosmic variance and otherassociated systematic uncertainties are accounted for(Hoyle et al. 2017). We also show a comparison be-tween the redshift estimate from a random samplingof the 30-band COSMOS P ( z ) (Laigle et al. 2016) andthe median photo- z derived from DES Y1 using BPZ.Structure along the line of sight is visible in the higher-resolution COSMOS redshifts but is not resolved byDES. For the full weak-lensing subsample, the normal-ized median absolute deviation (NMAD) of the quantity( z DES − z COSMOS ) / (1+ z COSMOS ) is 0 . − .
09, depend-ing on the point estimate used to determine the DESBPZ redshift. When restricted to i ≤
22, the photo- z NMAD decreases to 0 . − .
07. Due to the strictselection requirements of the weak-lensing subsample,the NMAD for the full Y1A1 GOLD galaxy sample isslightly larger ( ∼ . CONCLUSIONDuring its first year, DES imaged ∼ of thesouthern sky in each of the g , r , i , z and Y photo-metric filters. These data have been processed, cali-brated, coadded, cataloged, and characterized to formthe DES Y1A1 GOLD cosmology data sample, whichcovers ∼ with a depth of three to four tilingsper band and a photometric calibration accuracy of (cid:46) ∼ ) anddepth (five to six tilings per band). Improvements tothe Blanco telescope infrastructure, data processing al-gorithms, and photometric calibration are expected toyield higher-quality data. In addition, many of the im-provements developed for Y1A1 GOLD have been inte-grated into the core DESDM processing pipeline (e.g.,Morganson et al. 2018) and into automated tools for science catalog creation (e.g., Fausti Neto et al. 2017).However, we anticipate that future data sets will still re-quire the construction and validation of a high-qualitydata sample to serve as the basis for cosmological analy-ses. On the longer term, we expect that a similar proce-dure for the assembling and validating cosmology datasamples will be necessary for future surveys, such asLSST. We hope that the production of Y1A1 GOLDwill help serve as a road map for assembling cosmology-ready data samples for future large photometric surveys.ACKNOWLEDGMENTSFunding for the DES Projects has been provided bythe U.S. Department of Energy, the U.S. National Sci-ence Foundation, the Ministry of Science and Educationof Spain, the Science and Technology Facilities Councilof the United Kingdom, the Higher Education Fund-ing Council for England, the National Center for Su-percomputing Applications at the University of Illinoisat Urbana-Champaign, the Kavli Institute of Cosmo-logical Physics at the University of Chicago, the Cen-ter for Cosmology and Astro-Particle Physics at theOhio State University, the Mitchell Institute for Fun-damental Physics and Astronomy at Texas A&M Uni-versity, Financiadora de Estudos e Projetos, Funda¸c˜aoCarlos Chagas Filho de Amparo `a Pesquisa do Estado doRio de Janeiro, Conselho Nacional de DesenvolvimentoCient´ıfico e Tecnol´ogico and the Minist´erio da Ciˆencia,Tecnologia e Inova¸c˜ao, the Deutsche Forschungsgemein-schaft and the Collaborating Institutions in the DarkEnergy Survey.The Collaborating Institutions are Argonne NationalLaboratory, the University of California at Santa Cruz,the University of Cambridge, Centro de InvestigacionesEnerg´eticas, Medioambientales y Tecnol´ogicas-Madrid,the University of Chicago, University College Lon-don, the DES-Brazil Consortium, the University ofEdinburgh, the Eidgen¨ossische Technische Hochschule(ETH) Z¨urich, Fermi National Accelerator Laboratory,the University of Illinois at Urbana-Champaign, theInstitut de Ci`encies de l’Espai (IEEC/CSIC), the In-stitut de F´ısica d’Altes Energies, Lawrence BerkeleyNational Laboratory, the Ludwig-Maximilians Univer-sit¨at M¨unchen and the associated Excellence ClusterUniverse, the University of Michigan, the National Op-tical Astronomy Observatory, the University of Not-tingham, The Ohio State University, the University ofPennsylvania, the University of Portsmouth, SLAC Na-tional Accelerator Laboratory, Stanford University, theUniversity of Sussex, Texas A&M University, and theOzDES Membership Consortium.Based in part on observations at Cerro Tololo Inter-American Observatory, National Optical AstronomyObservatory, which is operated by the Association ofUniversities for Research in Astronomy (AURA) un-der a cooperative agreement with the National ScienceFoundation.5 Redshift n ( z ) COSMOS z mc D E S z m e d i a n G a l a x i e s p e r R e d s h i f t B i n Figure 16.
Performance of the BPZ photo- z estimates for the weak-lensing subsample of Y1A1 GOLD. (Left): The n ( z )distribution generated from a random Monte Carlo sampling (i.e., Wittman 2009) of the BPZ P ( z ) distribution for each galaxy.(Right): Comparison between a random sampling of the 30-band COSMOS P ( z ) (Laigle et al. 2016) and the median of theDES BPZ P ( z ) for each galaxy in the overlapping sample. Points are colored by the local density of comparison galaxies in0 . × .
04 redshift bins.
Facility:
Blanco (DECam)
Software:
SExtractor (Bertin & Arnouts 1996),
PSFEx (Bertin 2011),
SCAMP (Bertin 2006),
SWarp (Bertin et al. 2002; Bertin 2010), mangle (Hamilton &Tegmark 2004; Swanson et al. 2008),
HEALPix (G´orskiet al. 2005), astropy (Astropy Collaboration 2013), matplotlib (Hunter 2007), numpy (Van Der Walt et al.2011), scipy (Jones et al. 2001), healpy , fitsio , ngmix (Sheldon 2014). APPENDIX A. PHOTOMETRIC CALIBRATIONIn this Appendix, we provide more details on the photometric calibration of Y1A1, including nightly calibration(Appendix A.1), global calibration (Appendix A.2), and a SLR adjustment (Appendix A.3). We note that the nightlyand global calibration steps followed on the procedure of Tucker et al. (2007) and were performed on the single-epochcatalog data before coaddition. In contrast, the SLR adjustment was performed on the weighted-average magnitudesof multiple single-epoch catalogs and is applied directly to the coadded object catalogs. A collection of transformationequations between DES and several other surveys is provided in Appendix A.4.A.1.
Nightly Photometric Calibration
The first step in DES Y1 photometric calibration used observations of standard-star fields to derive a set of calibrationcoefficients for each photometric night. A subset of the standard-star fields listed in Table A.1 were observed at differentairmasses at the beginning and end of each DES night or half night (Section 2). The DES nightly standard-star fieldsare predominantly located in the equatorial fields of SDSS Data Release 9 (DR9; Ahn et al. 2012), with the additionof several fields from the Southern u (cid:48) g (cid:48) r (cid:48) i (cid:48) z (cid:48) Standard Network (Smith et al. 2017). Devoted observations of thesestandard-star fields were supplemented by DES survey observations that overlapped the standard-star fields. For Y band, we used stars from the equatorial fields of the UKIRT Infrared Deep Sky Survey Data Release 6 (UKIDSS DR6;Lawrence et al. 2007) matched against SDSS stars. All nightly standard stars were transformed to an initial DES ABphotometric system via matching to objects in SDSS and UKIDSS (Appendix A.4). The spatial distribution of theDES standard-star fields is shown in Figure A.1.Due to its provenance, primarily from SDSS DR9, we refer to the set of DES nightly standards as secondary standards. The fundamental standard for SDSS was the F subdwarf star BD+17 ◦ primary standards (Smith et al. 2002). These primary standards were in turn used (indirectly) inthe ubercalibration of SDSS (Padmanabhan et al. 2008). Thus, the DES secondary standards tie the absolute fluxcalibration of DES to the SDSS primary standards, to BD+17 ◦ g = − . ( F g ) − a g − b g × (( g − r ) − ( g − r ) fid ) − k g × X (A1) r = − . ( F r ) − a r − b r × (( g − r ) − ( g − r ) fid ) − k r × X (A2) i = − . ( F i ) − a i − b i × (( i − z ) − ( i − z ) fid ) − k i × X (A3) z = − . ( F z ) − a z − b z × (( i − z ) − ( i − z ) fid ) − k z × X (A4) Y = − . ( F Y ) − a Y − b Y × (( z − Y ) − ( z − Y ) fid ) − k Y × X (A5)where λ ( λ = g, r, i, z, Y ) is the calibrated standard-star magnitude in the DES system, F λ is the observed PSF flux(counts/sec), a is the photometric zeropoint for the night, b is the instrumental color term coefficient, ( g − r ) , ( i − z ) ,( z − Y ) are the calibrated standard-star colors, ( g − r ) fid , ( i − z ) fid , and ( z − Y ) fid are fiducial reference colors (chosen http://healpix.sourceforge.net https://github.com/healpy/healpy https://github.com/esheldon/fitsio https://github.com/esheldon/ngmix Table A.1.
DES Nightly Standard-star Fields
Field Name RA Dec Exposure Time (sec)J2000 J2000 g r i z Y
Preferred Fields a SDSS J2140-0000 21:40:00 +00:00:00 15 15 15 15 20SDSS J2300-0000 23:00:00 +00:00:00 15 15 15 15 20SDSS J0000-0000 00:00:00 +00:00:00 15 15 15 15 20SDSS J0100-0000 01:00:00 +00:00:00 15 15 15 15 20SDSS J0200-0000 02:00:00 +00:00:00 15 15 15 15 20SDSS J0320-0000 03:20:00 +00:00:00 15 15 15 15 20SDSS J0843-0000 08:43:00 +00:00:00 15 15 15 15 20SDSS J0933-0005 09:33:00 − − − − − Supplemental Fields b SA E1-A 01:24:50 − − − − − − − − a The preferred fields (with the exception of C26202/HST and MaxVis)have photometric standard stars covering the entire DECam focalplane. This permits photometric zeropoints to be determined for everyCCD using a single exposure. b The supplemental fields have photometric standard stars covering a10 (cid:48) × (cid:48) region and are typically used for expanding the range ofairmasses when no suitable primary field is observable. These par-ticular supplemental fields come from the Southern ugriz StandardStars project (Smith et al. 2017). so that the effects of b are relatively small for a star of typical color within the DES footprint), k is the first-orderextinction coefficient, and X is the airmass of the observation. The values for the fiducial colors are ( g − r ) fid = 0 . i − z ) fid = 0 .
09, and ( z − Y ) fid = 0 . a and b coefficients were determined for each functioning science CCD, while a single value of k was assumedfor the full focal plane. The nightly values of a track the overall throughput of the DECam instrument at the locationof each CCD, and variations in a mostly track the gradual accumulation of dust on the Blanco primary mirror. Thenightly values of b track variations in the shape of the total filter response curve at the location of each CCD (includingatmospheric transmission). Under photometric conditions, the value of k should not vary across the focal plane, anda single value of k was fit for the full focal plane. Variations in the nightly values of k track the relative throughputof the atmosphere at CTIO. The median values of a and b are shown for each science CCD in Figure A.2, and thenightly variations of a , b , and k are shown in Figure A.3. The site-average values for the a , b , and k coefficients aretabulated in Table A.2.We note that, despite the use of star flats and pupil corrections, there are still minor variations in the zeropointsacross the focal plane in Figure A.2. These variations can be attributed to two main sources: (1) the DES starflatprocedure is subject to small flat/planar gradients across the focal plane, with the understanding that the photometriccalibration procedure will remove such gradients; and (2) CCD-to-CCD variations in quantum efficiency have not been8 -120°-60°0°+60°+120° -120°-60°0°+60°+120°-60°-30°0°+30° +60° -60° -30° 0°+30°+60° Figure A.1.
Standard stars used for the photometric calibration of DES Y1. Nightly standard stars fields (Table A.1) aremarked with green circles while other secondary standards from SDSS and Smith et al. (2017) are shown in pink. The grid workof tertiary standards is shown in blue. aa g -25.39 r -25.48 i -25.38 z -25.06 Y -23.98 bb g -0.004 r i z Y Figure A.2.
Median fit values from DES Y1 for the a and b coefficients as a function of position on the DECam focal plane.The color scale represents the offset in the median fit value for each CCD with respect to the median for the focal plane as listedin the bottom left of each panel. The spatial variations in photometric zeropoints across the DECam focal plane are typicallyless than 0.02–0.03 mag for Y1A1. fully accounted for in the Y1A1 image processing, and this is also reflected in the smaller scale between-CCD variationsin the zeropoints. The DES Y3 processing has largely corrected for variations in quantum efficiency while Burke et al.(2018) show that gradients in the star flats can be successfully removed by the photometric calibration.The code used to perform the nightly fits is called the Photometric Standards Module (PSM). We note that PSMnot only fits Equations (A1)–(A5), but also performs an automated initial culling of non-photometric data using theoutputs of RASICAM (Lewis et al. 2010; Reil et al. 2014). PSM also culls dome-occulted exposures (identified bya strong gradient in a across the focal plane) and performs iterative sigma-clipping to achieve a good solution for anight. For the Y1 data set, we also culled nights with rms fit residuals > .
025 mag or an atypical ( ∼ σ outlier) fitvalue for the first-order extinction. The typical relative calibration scatter from the PSM solution is ∼ .
02 mag rms.This scatter includes the contribution from stellar shot noise in the standard-star observations. https://github.com/DarkEnergySurvey/PSM a a g -band r -band i -band z -band Y -band b b k MJD k MJD
Figure A.3.
Nightly fit values for the a coefficient of CCD35 (top), the b coefficient of CCD35 (middle), and the k coefficientof the DECam focal plane (bottom) for nights in DES Y1. For the a and b coefficients, the trends in CCD35 are found to berepresentative of the trends in the other CCDs. A.2.
Global Calibration
In addition to deriving the nightly calibration coefficients for each photometric night in DES Y1, we would liketo calibrate exposures taken under cloudy conditions and exposures where the nightly solution failed (e.g., due tocontrails). We would also like to improve on the ∼
2% rms relative calibration uncertainty achieved by the PSMsolution. To achieve both of these goals, we applied a global calibration to simultaneously calibrate all overlappingCCD images in the Y1 data set. In addition to calibrating images that lacked a PSM solution, the global calibrationcan achieve a relative calibration between overlapping images at the level of 0 .
003 mag (0 . The GCM generalizes theprocedure of Glazebrook et al. (1994) by replacing overlapping image “frames” with arbitrarily shaped overlappingcatalog data sets (these are still conventionally referred to as “images”). The procedure is summarized briefly asfollows.1. For each filter, consider n data sets for which (1 , . . . , m ) are uncalibrated and ( m + 1 , . . . , n ) are calibrated. Inmost cases these data sets represent object catalogs from individual exposures or CCD images. However, thecalibrated data set consists of standard stars spanning the entire Y1A1 footprint (Figure A.1).2. Compile a list of all unique pairs of observations of a common star on two data sets.3. For a given pair of images, i, j , let ∆ ij = median pairs ( m i − m j ) , (A6) https://github.com/DarkEnergySurvey/GCM Table A.2.
DES Y1 Average PSM Fit Values
Coeff. Band Median Mean σ Mean error Number a g − − r − − i − − z − − Y − − b g − − < r < i z Y k g r i z Y Note —Statistics were calculated for nights in Y1 with a good PSM fit.The a and b values were calculated individually for each CCD, whilethe k values were calculated for the full focal plane. where m i is the magnitude of a star in image i , m j is the magnitude of the same star in image j , and the medianis calculated over matched pairs of stars. Note that ∆ ij = − ∆ ji .4. Let ZP i be a floating zero-point that can be applied to the data set from image i to produce calibrated magnitudes.For images that are already calibrated ( i > m ), we fix ZP i = 0.5. Let θ ij define a function that selects overlapping image pairs. We define θ ij = 1 if i = j or if i and j overlap;otherwise θ ij = 0.6. To find calibrated zeropoints for each image, we minimize the sum of squares, S = (cid:88) i,j θ ij (∆ ij + ZP i − ZP j ) . (A7)7. We derive a calibrated magnitude for each object detected on image i (where i < m ) by adding ZP i to the rawinstrumental magnitudes.In Figure A.4, we show a simple example of the GCM algorithm on two disconnected groups of three overlappingdata sets (i.e., images). In each group, one of the overlapping images has been previously calibrated and serves as thereference against which the other images in its grouping are calibrated. To be calibrated, an uncalibrated image needseither to overlap a calibrated image (e.g., the left group in Figure A.4) or to have an unbroken path of overlappingimages to a calibrated image (e.g., image 3 in the right group of Figure A.4). In the right panel of Figure A.4 we showthe matrix equation that minimizes Equation (A7) for this particular set of images (Glazebrook et al. 1994). Notethat, via this matrix equation, the zeropoints for the two calibrated images (images 5 and 6) have been fixed to a valueof zero (1 × ZP = 0 and 1 × ZP = 0), since no offset is applied to these previously calibrated images.Following the prescription of Glazebrook et al. (1994), we estimate the rms magnitude residual for each CCD image, i , from overlap with other CCD images, j , asrms i = (cid:115) (cid:80) j θ ij (∆ ij + ZP i − ZP j ) (cid:80) j θ ij . (A8)The rms distribution over all CCD images is a measure of the internal (reproducibility) errors on small scales (thescales of overlapping CCD images) and is a measure of the precision of the overall GCM solution.1In principle, the GCM method is very precise, but carries the caveat that any small systematic gradients in theflat fielding of individual images can cause low-amplitude gradients over large scales. We used the set of secondarystandard stars and a sparse gridwork of tertiary standard stars (Figure A.1) as an “anchor” to keep the GCM fit fromdrifting due to any small systematic gradients in the Y1A1 FINALCUT exposures. We note that the sparse gridworkof tertiary standards in the SPT region was extracted from stars in photometric exposures, calibrated by the nightlyPSM results. From the full set of PSM-calibrated exposures in the SPT region, we selected a sparse gridwork of thin(1-degree-wide) “struts” of constant right ascension and constant declination. This morphology was chosen so that thetertiaries would anchor the GCM solution on large scales ( > a coefficients for each CCD (Figure A.2). In this manner, each exposure was temporarily flat-fielded acrossthe focal plane to reduce exposure-scale photometric gradients. For the first pass, only exposures that were classifiedas having been observed under photometric conditions – as determined by RASICAM – were allowed in the GCM fit.The first pass yielded a set of zeropoint offsets – one per exposure – for all the (apparently) photometric exposuresin the SPT region. The second run of GCM was essentially identical to the first, but it removed outlier exposures– ones with particularly “noisy” or discrepant zeropoints. For both the first and second runs, the sparse gridworkof tertiaries and the handful of individual calibrated SV fields (Figure A.1) were used as the calibrated data set forthe Glazebrook et al. (1994) algorithm. Again, this yielded a set of zeropoint offsets – one per exposure – for all thephotometric exposures in the SPT region. These individual CCD zeropoint offsets were applied to all the CCD imagesin the set of photometric exposures included in the second-pass run of GCM, creating a set of “quaternary” standardstars covering nearly all of the SPT region. In the third and final run of the GCM for the SPT region, the catalogfrom each individual CCD image was treated – as in the case of GCM runs for S82, COSMOS, VVDS-14h, and theSN fields – as the unit to be calibrated. Furthermore, all CCD images from the SPT region – those from photometricexposures and those from non-photometric exposures – were included in the GCM fit. For this third pass of the GCM,the newly created quaternary standard stars were used as the calibrated data set. This third pass of the GCM forthe SPT region yielded a set of zeropoint offsets for each CCD image, which was used to calibrate the Y1A1 GOLDsingle-epoch CCD images in advance of the image coaddition process.A.3.
Photometric Calibration Adjustment
To correct for residual color non-uniformity in the photometric calibration and to account for Galactic reddening(i.e., Figure A.5), the GCM calibration was adjusted at the catalog level using SLR (Section 4.3). A reference stellarlocus was empirically derived from the globally calibrated DES Y1A1 stellar objects in the region of the Y1A1 footprintwith the smallest E ( B − V ) value from Schlegel et al. (1998). Corrections were computed for the WAVGCALIB MAG PSF magnitudes described in Section 6. Our stellar selection was based on the weighted average of the
SPREAD MODEL quantity for the matched objects ( | WAVG SPREAD MODEL R | < . S/N >
10 in i band and S/N > grzY ). We segmented the sky into equal-area pixels using the
HEALPix scheme (G´orski et al. 2005), starting with a relatively fine grid, nside = 512 ( ∼ .
01 deg ). If there werefewer than 200 stars in a pixel, then we appended neighboring pixels using the get all neighbors function from healpy , enlarging the pixel chunks until they contained at least 200 stars. To reduce computation time in high-densityregions near the LMC, when there were more than 2000 stars per pixel we randomly down-sampled. Approximately97% of the wide-area survey footprint was fit in chunks of 9 pixels containing a median of ∼
400 stars and yielding2 (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) − − − − (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) × (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) ZP ZP ZP ZP ZP ZP (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) = (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) ∆ + ∆ ∆ + ∆ ∆ ∆ + ∆ (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) Figure A.4.
A schematic of the GCM algorithm based on Fig. 1 of Glazebrook et al. (1994).
Left:
The stars in images 5 and6 have been previously calibrated while the stars in the other images are uncalibrated. The algorithm minimizes the zeropointoffsets from all the overlapping images. Images that have a connected path via overlapping images to a reference image can becalibrated to that reference image.
Right:
The corresponding matrix equation for this set of images. − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) Figure A.5.
Interstellar extinction, E ( B − V ), over the Y1A1 GOLD footprint taken from Schlegel et al. (1998). The DESfootprint was explicitly chosen to occupy a low-extinction region at high Galactic latitude. an effective resolution of ∼ . . We applied a modified version of the BigMACS
SLR code (Kelly et al. 2014) tocalibrate each star from the reference exposure with respect to the empirical stellar locus. The absolute calibrationwas set against the i -band magnitude derived from the GCM solution, which was dereddened using the SFD mapwith a reddening law of A I = 1 . × E ( B − V ) SFD . This extinction correction was derived following the prescriptionof Cardelli et al. (1989) with R V = 3 .
1, but updated for the DES i -band throughput using optical-NIR coefficientsfrom O’Donnell (1994) assuming a source spectrum that is constant in spectral flux density per unit wavelength, f λ ( erg cm − s − ˚ A − ). The flat SED was chosen to represent the wide range of stellar SEDs from the Pickles ATLAS(Pickles 1998) and SEDs of galaxies over the range of redshifts probed by DES (Arnouts & Ilbert 2011).Variations in the average metallicity of the stellar populations used for the SLR will introduce systematic shifts thatare not due to photometric variation or Galactic reddening (e.g., High et al. 2009). For DES Y1A1, these shifts arelargest for the g band, where they can have a 1-2% effect on the calibration. A larger effect can be found in the vicinityof the LMC, which we avoid for extragalactic science. The effect of metallicity variations can be much worse at lowerGalactic latitudes and in bluer filters (i.e., u band).The final product was an SLR correction map at a resolution of nside = 512 that we implemented with a bi-linearinterpolation to obtain magnitude and flux corrections for the full Y1A1 GOLD catalog. The resulting SLR-adjustedmagnitudes used in the Y1A1 GOLD catalog are thus already corrected for Galactic reddening. https://code.google.com/p/big-macs-calibrate/ Photometric Transformation Equations
We have derived transformation equations between various surveys and the DES system. We document thesetransformation equations here for reference.We define a transformation from SDSS/UKIDSS to the DES system to place the nightly standard star exposures onan initial DES AB photometric system (Section 4.1): g des = g sdss − . × ( g − r ) sdss + 0 .
01 (A9) r des = r sdss − . × ( g − r ) sdss + 0 .
02 (A10) i des = i sdss − . × ( i − z ) sdss + 0 .
02 (A11) z des = z sdss − . × ( i − z ) sdss + 0 .
01 (A12) Y des = Y ukidss + 0 . × ( z sdss − Y ukidss ) + 0 . . (A13)These transformation equations were derived in a hybrid manner: the color coefficients were determined by matchingdata from the DES SV data set with data from SDSS DR9 (or, in the case of the Y band, with a combination ofUKIDSS DR6 Y band and SDSS DR9) and fitting the result. The zeropoint for each relation was determined fromsynthetic AB photometry. We applied the DES, SDSS, and UKIDSS filter curves to the Pickles (1998) stellar libraryand measured the offset between the two synthetic magnitudes at zero color for each filter band. We note that the largezeropoint offset for the Y -band transformation is due to the fact that the UKIDSS data are in the Vega magnitudesystem, while the Y des is set to the AB magnitude system. These transformation equations are valid for stars with( g − r ) sdss < .
2. For an individual object, the transformation from SDSS/UKIDSS to DES will depend on interstellarextinction. The DES footprint occupies a region of low extinction, and we estimate that the median correction due toreddening in the g band is 0 . g -band correction of < (cid:38) g -band.We validate the relative calibration accuracy of Y1A1 GOLD by comparing the calibrated magnitudes of stars inthe Y1A1 GOLD catalog against those derived from a combination of APASS (Henden & Munari 2014) and 2MASS(Skrutskie et al. 2006). We selected stellar objects from the Y1A1 GOLD catalog using MODEST CLASS (Section 8.1)and perform a 2 (cid:48)(cid:48) match to the APASS and 2MASS catalogs. We then fit a set of transformation equations to mapfrom g apass , r apass , and J to a predicted magnitude in each of the DES filters: g des = g apass − . × ( g − r ) apass − . r des = r apass − . × ( r − i ) apass − . i des = r apass − . × ( r apass − J − . − .
391 (A16) z des = ( J + 0 .
81) + 0 . × ( r apass − J − . − . Y des = ( J + 0 .
81) + 0 . × ( r apass − J − . − . . (A18)Equations A16-A18 are derived from a global fit of the Y1A1 GOLD data set and are valid for stars for which r apass − J < .
81. We find a cleaner and tighter relation using the hybrid APASS/2MASS ( r apass − J ) colorrather than a purely APASS ( r apass − i apass ) color for these transformation equations. These transformation equationsexplicitly remove any absolute calibration offset between the two data sets and can be used to test for spatial non-uniformity between the GCM calibration and these external catalogs (Figure A.7). We note that the residual structureseen in the S82 region of Figure A.7 does not appear in comparisons with SDSS DR10 or DES Y3, suggesting thatthis structure is a feature introduced by APASS.To validate the completeness and contamination of the Y1A1 GOLD catalog, we perform a comparison with theCFHTLenS data in the W4 field. In this case, we are interested in the transformed magnitude of all objects, so weperform no stellar selection. We use matched objects to derive a set of transformation equations from the CFHTLenS g (cid:48) , r (cid:48) , i (cid:48) , z (cid:48) filters to the DES g, r, i, z system: g des = g CFHT + 0 . g CFHT − r CFHT ) + 0 .
058 (A19) r des = r CFHT − . g CFHT − r CFHT ) + 0 .
021 (A20) i des = i CFHT − . i CFHT − z CFHT ) + 0 .
062 (A21) z des = z CFHT − . i CFHT − z CFHT ) + 0 . . (A22)We find that these equations should be valid for objects with g − r < . i − z < . Calibration Validation
In this section we show ancillary plots of the performance and validation of the Y1A1 photometric calibration(Figures A.6 – A.9).4 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension g-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) g-band3.1 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) r-band2.8 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension i-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) i-band3.1 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension z-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) z-band3.1 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension
Y-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) Y-band3.6 mmag
Figure A.6.
Internal rms errors in the photometric zeropoint reproducibility per CCD for DES Y1A1. The zeropoint rms iscalculated by comparing the calibrated magnitudes of stars in overlapping CCDs. Note that these data include observationstaken in both clear and cloudy conditions. Typical internal reproducibility errors are ∼ ∼ . − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension g-band -80 0 80 ∆ g (mmag) N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ g (mmag) 0.0050.0100.0150.0200.025 N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ g (mmag) g-band19 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band -80 0 80 ∆ r (mmag) N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ r (mmag) 0.0040.0080.0120.016 N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ r (mmag) r-band22 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension i-band -80 0 80 ∆ i (mmag) N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ i (mmag) 0.0050.0100.0150.0200.025 N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ i (mmag) i-band20 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension z-band -80 0 80 ∆ z (mmag) N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ z (mmag) 0.0050.0100.0150.0200.025 N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ z (mmag) z-band20 mmag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension
Y-band -80 0 80 ∆ Y (mmag) N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ Y (mmag) 0.0050.0100.0150.0200.025 N o r m a li z e d A r e a ( a . u . )
50 0 50 ∆ Y (mmag)
Y-band18 mmag
Figure A.7.
Comparison of stellar magnitudes from the DES Y1A1 GCM and those estimated from APASS/2MASS trans-formed into the DES filter system (Equations A14-A18). The sky plots (left) show the median magnitude offset for stars binnedinto ∼ . HEALPix pixels. The GCM calibrated magnitudes are consistent with the transformed values from APASS/2MASSwith σ ∼
20 mmag (calculated between the 16th and 84th percentiles). Note that the GCM g -band calibration disagrees withAPASS/2MASS by ∼
4% in the eastern portion of the SPT region (RA < − − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension g-band -35 25 85SLR Shift (mmag) N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag)0.0040.0080.0120.016 N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag) g-band σ = 30 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band -35 25 85SLR Shift (mmag) N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag)0.0080.0160.0240.0320.040 N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag) r-band σ = 12 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension z-band -35 25 85SLR Shift (mmag) N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag)0.0080.0160.0240.0320.040 N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag) z-band σ = 11 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension
Y-band -35 25 85SLR Shift (mmag) N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag)0.0060.0120.0180.0240.030 N o r m a li z e d A r e a ( a . u . )
40 0 40 80SLR Zeropoint Shift (mmag)
Y-band σ = 13 Figure A.8.
Adjustment to the GCM photometric zeropoints derived from the SLR fit, after removing the contribution frominterstellar extinction using the SFD maps and reddening from O’Donnell (1994). The width of these distributions representsadjustments to the calibration uniformity and differences between the interstellar extinction derived from the stellar locus andinterstellar dust maps. The SLR adjustment is generally ∼
10 mmag (rms) over most of the area. A larger adjustment is madein the g band, which reflects the larger impact of reddening in the blue filters and a region of non-uniformity in the west of thefootprint. There is no adjustment to the GCM i band because the SLR fit is tied to the dereddened magnitudes of stars in thatband. B. CO-ADD SOURCE DETECTIONThe Y1A1 COADD source detection was performed on a normalized “detection image” formed from a nonlinearcombination of the r , i , and z coadded images. The original SWarp combination formula for computing the value of a7 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension (g-r) -20 20 60SLR Shift (mmag) N o r m a li z e d A r e a ( a . u . )
25 0 25 50SLR Color Shift (mmag) 0.0050.0100.0150.0200.025 N o r m a li z e d A r e a ( a . u . )
25 0 25 50SLR Color Shift (mmag) (g-r) σ = 21 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension (r-i) -20 20 60SLR Shift (mmag) N o r m a li z e d A r e a ( a . u . )
25 0 25 50SLR Color Shift (mmag) 0.0080.0160.0240.0320.040 N o r m a li z e d A r e a ( a . u . )
25 0 25 50SLR Color Shift (mmag) (r-i) σ = 12 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension (i-z) -20 20 60SLR Shift (mmag) N o r m a li z e d A r e a ( a . u . )
25 0 25 50SLR Color Shift (mmag) 0.0080.0160.0240.0320.040 N o r m a li z e d A r e a ( a . u . )
25 0 25 50SLR Color Shift (mmag) (i-z) σ = 11 Figure A.9.
Color uniformity of the SLR adjustment applied to the GCM zeropoints. The adjustment was largest for the( g − r ) color in the eastern portion of the SPT region. The color non-uniformity in this region was one of the specific motivationsfor the SLR calibration adjustment. After the SLR adjustment was applied, the color of stars was found to be uniform at the1–2% level across the footprint. pixel of the detection image is (Bertin 2010): χ = (cid:115) (cid:80) c ≤ n w c f c n , (B23)where f c is the background-subtracted pixel value, w c is the weight of the pixel in channel c , and n is the number ofvalid inputs. Compared to the standard χ combination proposed by Szalay et al. (1999), χ leads to a less skewed noisedistribution (if one assumes that input noise follows a Gaussian distribution), while maintaining identical detectioncapabilities. However, both estimators have a bias that depends on n , which leads to visible seams between regions witha different number of input images. This motivated the implementation of two new normalized image combinationschemes in SWarp , with a variable offset applied to the original (still assuming that the inputs are normally andindependently distributed).
CHI-MEAN is recentered on the mean (e.g., Evans et al. 2000):
CHI-MEAN = (cid:113)(cid:80) c ≤ n w c f c − µ (cid:112) n − µ , (B24)with µ = √ n + 1) / n/ , (B25)8while CHI-MODE is recentered on the mode of the distribution:
CHI-MODE = (cid:113)(cid:80) c ≤ n w c f c − √ n − (cid:112) n − µ . (B26)The left panel of Figure B.1 shows a comparison of the distributions obtained from the original χ , CHI-MODE and
CHI-MEAN estimators for Gaussian input noise. The right panel of Figure B.1 shows that the
CHI-MEAN estimatorgenerates the most seamless stacking results, and it was used to produce the Y1A1 COADD detection images.
Figure B.1. (Left) Normalized distribution of the value x of a detection image pixel for the original χ ( OLD CHI , top),
CHI-MODE (middle) and
CHI-MEAN (bottom) estimators when the inputs to the co-add are normally and independently distributed. n isthe number of input images. The means of the distributions are shown as vertical lines. (Right) Gamma-corrected close-up ofa χ ( OLD CHI , top),
CHI-MODE (middle) and
CHI-MEAN (bottom) detection image computed from a set of 8 input images withzero-mean Gaussian white noise. The number of inputs decreases by steps of 64 pixels from left (eight inputs) to right (oneinput).
CHI-MEAN detection images are virtually seamless, even at the transition between one and two input images.C.
CATALOG DEPTH MAPSIn this appendix we collect a set of figures documenting the 10 σ limiting magnitude of the Y1A1 GOLD catalogas described in Section 7.1. We include depth maps both for the MAG AUTO values derived from the coadded images(Figure C.1) and for the
CM MAG values derived from multi-epoch, multi-object fitting (Figure C.2).9 − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension g-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) g-band23.4 mag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) r-band23.2 mag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension i-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) i-band22.5 mag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension z-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) z-band21.8 mag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension
Y-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) Y-band20.1 mag
Figure C.1.
Sky maps and normalized histograms of the 10 σ limiting magnitude for galaxies fit with MAG AUTO . The modeof the limiting magnitude distribution is shown in the right panel of each row. The derivation of the limiting magnitude isdescribed in more detail in Section 7.1. − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension g-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) g-band23.7 mag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension r-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) r-band23.5 mag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension i-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) i-band22.9 mag − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ − ◦ ◦ ◦ D e c li n a t i o n − ◦ ◦ ◦ ◦ ◦ − ◦ Right Ascension z-band N o r m a li z e d A r e a ( a . u . ) N o r m a li z e d A r e a ( a . u . ) z-band22.2 mag Figure C.2.
Sky maps and normalized histograms of the 10 σ limiting magnitude for galaxies fit with the MOF CM MAG . Themagnitude range for these figures is the same as the
MAG AUTO magnitude limits shown in Figure C.1. The mode of the limitingmagnitude distribution is shown in the right panel of each row. Note that these magnitude limits include the deeper D10 coaddsof the SN fields. The derivation of the limiting magnitude is described in more detail in Section 7.1.
Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2012,ApJS, 203, 21Aihara, H., Armstrong, R., Bickerton, S., et al. 2018, PASJ,70, S8Annis, James and Soares-Santos, M. and Strauss, M. A.and others. 2014, ApJ, 794, 120Antilogus, P., Astier, P., Doherty, P., Guyonnet, A., &Regnault, N. 2014, Journal of Instrumentation, 9, C03048Arnouts, S., & Ilbert, O. 2011, LePHARE: PhotometricAnalysis for Redshift Estimate, Astrophysics SourceCode Library, ascl:1108.009Astropy Collaboration. 2013, A&A, 558, A33Bechtol, K., Drlica-Wagner, A., Balbinot, E., et al. 2015,ApJ, 807, 50Ben´ıtez, N. 2000, ApJ, 536, 571Berk, A., Anderson, G. P., Bernstein, L. S., et al. 1999, inProc. SPIE, Vol. 3756, Optical Spectroscopic Techniquesand Instrumentation for Atmospheric and SpaceResearch III, ed. A. M. Larar, 348–353Berlind, A. A., Frieman, J., Weinberg, D. H., et al. 2006,ApJS, 167, 1Bernstein, G., Abbott, T., Desai, S., et al. 2017a, PASP,129, 114502Bernstein, G. M., Armstrong, R., Plazas, A. A., et al.2017b, PASP, 129, 074503Bertin, E. 2006, in Astronomical Society of the PacificConference Series, Vol. 351, Astronomical Data AnalysisSoftware and Systems XV, ed. C. Gabriel, C. Arviset,D. Ponz, & S. Enrique, 112Bertin, E. 2010, SWarp: Resampling and Co-adding FITSImages Together, Astrophysics Source Code Library,ascl:1010.068Bertin, E. 2011, in Astronomical Society of the PacificConference Series, Vol. 442, Astronomical Data AnalysisSoftware and Systems XX, ed. I. N. Evans,A. Accomazzi, D. J. Mink, & A. H. Rots, 435Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393Bertin, E., Mellier, Y., Radovich, M., et al. 2002, inAstronomical Society of the Pacific Conference Series,Vol. 281, Astronomical Data Analysis Software andSystems XI, ed. D. A. Bohlender, D. Durand, & T. H.Handley, 228Bohlin, R. C., Gordon, K. D., & Tremblay, P.-E. 2014,PASP, 126, 711Bonnett, C., Troxel, M. A., Hartley, W., et al. 2016,PhRvD, 94, 042005Bouy, H., Bertin, E., Moraux, E., et al. 2013, A&A, 554,A101 Burke, D., Rykoff, E. S., Allam, S., Annis, J., et al. 2018,AJ, 155, 41Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ,345, 245Carlstrom, J. E., Ade, P. A. R., Aird, K. A., et al. 2011,PASP, 123, 568Casali, M., Adamson, A., Alves de Oliveira, C., et al. 2007,A&A, 467, 777Cawthon, R., Davis, C., Gatti, M., et al. 2017, submitted toMNRAS, arXiv:1712.07298Chang, C., Busha, M. T., Wechsler, R. H., et al. 2015, ApJ,801, 73Corwin, Jr., H. G., Buta, R. J., & de Vaucouleurs, G. 1994,AJ, 108, 2128Coupon, J., Kilbinger, M., McCracken, H. J., et al. 2012,A&A, 542, A5Davis, C., Gatti, M., Vielzeuf, P., Cawthon, R., et al. 2017,submitted to MNRAS, arXiv:1710.02517de Vaucouleurs, G. 1948, Annales d’Astrophysique, 11, 247De Vicente, J., S´anchez, E., & Sevilla-Noarbe, I. 2016,MNRAS, 459, 3078DeRose, J., et al. 2017, in prep.DES Collaboration. 2005, arXiv:astro-ph/0510346—. 2016, MNRAS, 460, 1270—. 2018, arXiv:1801.03181Desai, S., Armstrong, R., Mohr, J. J., et al. 2012, ApJ, 757,83Diehl, H. T., et al. 2014, Proc. SPIE, 9149, 91490VDrlica-Wagner, A., Bechtol, K., Rykoff, E. S., et al. 2015,ApJ, 813, 109Erben, T., Hildebrandt, H., Miller, L., et al. 2013, MNRAS,433, 2545Evans, M., Hastings, N., & Peacock, B. 2000, StatisticalDistributions, Wiley Series in Probability and Statistics(Wiley)Fadely, R., Hogg, D. W., & Willman, B. 2012, ApJ, 760, 15Fausti Neto, A., da Costa, L., Carnero Rosell, A., et al.2017, submitted to Astronomy and Computing,arXiv:1708.05642Flaugher, B., Diehl, H. T., Honscheid, K., et al. 2015, AJ,150, 150Gaia Collaboration. 2016, A&A, 595, A2Gatti, M., Vielzeuf, P., Davis, C., et al. 2018, MNRAS,arXiv:1709.00992Giavalisco, M., Ferguson, H. C., Koekemoer, A. M., et al.2004, ApJL, 600, L93Gilbank, D. G., Gladders, M. D., Yee, H. K. C., & Hsieh,B. C. 2011, AJ, 141, 94 Glazebrook, K., Peacock, J. A., Collins, C. A., & Miller, L.1994, MNRAS, 266, 65G´orski, K. M., Hivon, E., Banday, A. J., et al. 2005, ApJ,622, 759Gruen, D., Bernstein, G. M., Jarvis, M., et al. 2015,Journal of Instrumentation, 10, C05032Gschwend, J., Rosell, A. C., Ogando, R., et al. 2017,submitted to Astron. Comput.Hambly, N. C., Collins, R. S., Cross, N. J. G., et al. 2008,MNRAS, 384, 637Hamilton, A. J. S., & Tegmark, M. 2004, MNRAS, 349, 115Hartley, W., et al. 2018, in prep.Henden, A., & Munari, U. 2014, Contributions of theAstronomical Observatory Skalnate Pleso, 43, 518Hennig, C., Mohr, J. J., Zenteno, A., Desai, S., et al. 2017,MNRAS, 467, 4015Hewett, P. C., Warren, S. J., Leggett, S. K., et al. 2006,MNRAS, 367, 454Heymans, C., Van Waerbeke, L., Miller, L., et al. 2012,MNRAS, 427, 146High, F. W., Stubbs, C. W., Rest, A., Stalder, B., &Challis, P. 2009, AJ, 138, 110Hildebrandt, H., Erben, T., Kuijken, K., et al. 2012,MNRAS, 421, 2355Hodgkin, S. T., Irwin, M. J., Hewett, P. C., et al. 2009,MNRAS, 394, 675Hoffleit, D., & Jaschek, C. . 1991, The Bright star catalogueHogg, D. W., & Lang, D. 2013, PASP, 125, 719Hoyle, B., Gruen, D., Bernstein, G. M., et al. 2017,submitted to MNRAS, arXiv:1708.01532Huchra, J. P., & Geller, M. J. 1982, ApJ, 257, 423Hunter, J. D. 2007, Computing In Science & Engineering,9, 90Irwin, M. J., et al. 2009, in prep.Ivezi´c, ˇZ., Lupton, R. H., Schlegel, D., et al. 2004,Astronomische Nachrichten, 325, 583Jarvis, M., Sheldon, E., Zuntz, J., et al. 2016, MNRAS,460, 2245Jones, E., Oliphant, T., Peterson, P., et al. 2001, SciPy:Open source scientific tools for PythonKelly, P. L., von der Linden, A., Applegate, D. E., et al.2014, MNRAS, 439, 28Kessler, R., Marriner, J., Childress, M., et al. 2015, AJ,150, 172Klein, M., Mohr, J. J., Desai, S., Israel, H., et al. 2017,ArXiv e-prints, arXiv:1706.06577Kron, R. G. 1980, ApJS, 43, 305Laigle, C., McCracken, H. J., Ilbert, O., et al. 2016, ApJS,224, 24 Lawrence, A., Warren, S. J., Almaini, O., et al. 2007,MNRAS, 379, 1599Le F`evre, O., Vettolani, G., Garilli, B., et al. 2005, A&A,439, 845Leauthaud, A., Massey, R., Kneib, J.-P., et al. 2007, ApJS,172, 219Leistedt, B., Peiris, H. V., Elsner, F., et al. 2016, ApJS,226, 24Lewis, P. M., Rogers, H., & Schindler, R. H. 2010, inProc. SPIE, Vol. 7735, Ground-based and AirborneInstrumentation for Astronomy III, 77353CLi, T. S., DePoy, D. L., Marshall, J. L., et al. 2016, AJ,151, 157MacDonald, E. C., Allen, P., Dalton, G., et al. 2004,MNRAS, 352, 1255Melchior, P., Sheldon, E., Drlica-Wagner, A., et al. 2016,Astronomy and Computing, 16, 99M´enard, B., Scranton, R., Schmidt, S., et al. 2013, ArXive-prints, arXiv:1303.4722Mohr, J. J., Adams, D., Barkhouse, W., et al. 2008, inSociety of Photo-Optical Instrumentation Engineers(SPIE) Conference Series, Vol. 7016, Society ofPhoto-Optical Instrumentation Engineers (SPIE)Conference Series, 0Mohr, J. J., Armstrong, R., Bertin, E., et al. 2012, Proc.SPIE, 8451, 84510DMorganson, E., Gruendl, R. A., Menanteau, F., et al. 2018,accepted to PASP, arXiv:1801.03177Neilsen, E., Bernstein, G., Gruendl, R., & Kent, S. 2015,“Limiting magnitude, τ , T eff , and image quality in DESYear 1”, Tech. Rep. FERMILAB-TM-2610-AE-CDNewman, J. A. 2008, ApJ, 684, 88Ngeow, C., Mohr, J. J., Alam, T., Barkhouse, W. A., et al.2006, in Proc. SPIE, Vol. 6270, Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference Series,627023O’Donnell, J. E. 1994, ApJ, 422, 158Padmanabhan, N., Schlegel, D. J., Finkbeiner, D. P., et al.2008, ApJ, 674, 1217Pickles, A. J. 1998, PASP, 110, 863Plazas, A. A., Bernstein, G. M., & Sheldon, E. S. 2014,PASP, 126, 750Reil, K., Lewis, P., Schindler, R., & Zhang, Z. 2014, inProc. SPIE, Vol. 9149, Observatory Operations:Strategies, Processes, and Systems V, 91490URozo, E., Rykoff, E. S., Abate, A., et al. 2016, MNRAS,461, 1431Rykoff, E. S., Rozo, E., & Keisler, R. 2015, submitted toAJ, arXiv:1509.008703