7 wbs 1.7 simulation and survey planning -...

16
7 WBS 1.7 Simulation and Survey Planning 7 WBS 1.7 Simulation and Survey Planning ................................................................. 1 7.1 Overview of Simulation and Survey Planning .................................................... 2 7.2 WBS 1.7.1 Survey Strategy (Observing plans and observing simulations) ........ 2 7.3 WBS 1.7.2 Calibration Algorithms ..................................................................... 3 7.3.1 WBS 1.7.2.1 Photometric Calibrations ...................................................... 3 7.3.2 WBS 1.7.2.2 Camera Calibration/Response System ................................... 4 7.3.3 WBS 1.7.2.3 Point Spread Function Calibration Plan ............................... 9 7.3.4 WBS 1.7.2.4 Astrometric Calibration Plan................................................. 9 7.3.5 WBS 1.7.2.4 Photometric Redshift (Photo-z) Calibrations....................... 10 7.4 WBS 1.7.3 Survey Data Simulation................................................................... 10 7.4.1 WBS 1.7.3.1 DECam Instrument Model Specifications ........................... 10 7.4.2 WBS 1.7.3.2 Catalog Level Simulations .................................................. 11 7.4.3 WBS 1.7.3.3 Image Level Simulation Algorithm Implementation ........... 13 7.4.4 WBS 1.7.3.4 Image Level Simulation Production .................................... 14 7.4.5 WBS 1.7.3.5 Simulation Storage .............................................................. 14 7.5 WBS 1.7.4 Mock Data Reduction Challenge (“Truth tables”) ........................ 15 7.6 WBS 1.7.5 Survey Planning Milestones ........................................................... 15 1

Upload: others

Post on 06-Jun-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

7 WBS 1.7 Simulation and Survey Planning 7 WBS 1.7 Simulation and Survey Planning................................................................. 1

7.1 Overview of Simulation and Survey Planning .................................................... 2 7.2 WBS 1.7.1 Survey Strategy (Observing plans and observing simulations) ........ 2 7.3 WBS 1.7.2 Calibration Algorithms ..................................................................... 3

7.3.1 WBS 1.7.2.1 Photometric Calibrations...................................................... 3 7.3.2 WBS 1.7.2.2 Camera Calibration/Response System................................... 4 7.3.3 WBS 1.7.2.3 Point Spread Function Calibration Plan ............................... 9 7.3.4 WBS 1.7.2.4 Astrometric Calibration Plan................................................. 9 7.3.5 WBS 1.7.2.4 Photometric Redshift (Photo-z) Calibrations....................... 10

7.4 WBS 1.7.3 Survey Data Simulation................................................................... 10 7.4.1 WBS 1.7.3.1 DECam Instrument Model Specifications........................... 10 7.4.2 WBS 1.7.3.2 Catalog Level Simulations .................................................. 11 7.4.3 WBS 1.7.3.3 Image Level Simulation Algorithm Implementation ........... 13 7.4.4 WBS 1.7.3.4 Image Level Simulation Production.................................... 14 7.4.5 WBS 1.7.3.5 Simulation Storage .............................................................. 14

7.5 WBS 1.7.4 Mock Data Reduction Challenge (“Truth tables”)........................ 15 7.6 WBS 1.7.5 Survey Planning Milestones........................................................... 15

1

Page 2: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

7.1

7.2

Overview of Simulation and Survey Planning The DECam project will produce both simulated and real data for input into the Data Management system. The catalog and image simulations will help us develop our science analysis codes and data reduction pipelines prior to the start of survey observations. Moreover, the simulations will allow us to validate our analytical forecasts of dark energy parameter constraints, as well as to extend our analyses to comprehensively characterize and incorporate the various sources of statistical and systematic errors inherent in each of our 4 key project techniques. Yearly cycles of new catalog and image simulations are followed by science analysis and data reduction challenges carried out using the simulation outputs, and these are already in progress, in coordination with the Data Management project. The level of scale and sophistication involved in each round of simulations will improve, in order to meet the requirements set in conjunction with the science analysis goals of the key project science working groups, and with the data reduction/pipeline development testing goals of the data management project. In 2004 we produced at Fermilab the Level 0 Image Simulations (ImSim0), where we reformatted existing SDSS images to produce 220 GB of simulated DES images, in order to provide an early jump start for the data management effort to test processing pipelines. In 2005-6 we produced at Fermilab and the University of Chicago our Level 1 Catalog Simulations (CatSim1), consisting of catalogs of stars and galaxies covering 500 deg2 of sky, based on the parent Hubble Volume N-body simulation, and our Level 1 Image Simulations (ImSim1), consisting of 500 GB of simulated science images, generated using shapelets-based rendering of astronomical objects and incorporating models of DES observing and of DECam optics and detectors. These simulated data were used successfully to test the data management processing pipelines during the Data Challenge 1 effort. We have now produced Level 2 Simulations (CatSim2/ImSim2), where we have generated 3.2 TB of simulated science and calibration image data, covering 4 tilings in each of the griz filters for 500 deg2 of sky. These simulated data have been used to successfully feed the Data Challenge 2 effort. For our future simulation rounds (Levels 3-5, starting in 2007) we plan to produce 5000-deg2 full-DES catalogs, using higher-resolution parent cosmological simulations, with different cosmologies and input physics, produced by Barcelona and other DES collaboration members. For each simulation round we also plan to produce simulated images corresponding to ~1 year of DES observations, as well as improve our treatment of object rendering, instrument models, and systematic effects. These simulations will be used to test recovery of input cosmologies from both the catalog and image levels using the 4 DES key project techniques, as well to provide data to stress test the full data processing system.

WBS 1.7.1 Survey Strategy (Observing plans and observing simulations)

This task involves scientist effort to upgrade the DES observing simulations code to add new functionality, including simulations of supernova observations and options to

2

Page 3: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

explore more observing strategies, in particular to optimize for weak lensing. This task also involves writing the document that develops the DES observing plans and procedures for the supernova survey and for the near-IR Y-band imaging, as well as writing the document that specifies the final DES survey observing plans and procedures.

7.3

7.3.1

WBS 1.7.2 Calibration Algorithms The following tasks concern scientist effort on the various types of calibrations that DES needs to carry out, specifically calibrations of photometry, point spread function, astrometry, and photometric redshifts (photo-z’s).

WBS 1.7.2.1 Photometric Calibrations

This task involves scientist effort to derive an optimal plan for the photometric calibration of the DES survey imaging data. To accomplish this, we will carry out the following work:

• We will use catalog and image level simulations of the DES to test and refine the DES absolute and relative photometry algorithm and code. In particular, the DES will rely heavily on observations of massively overlapping “tiles” (each a single pointing of the DECam mosaic camera) and the use of matrix-based techniques to derive accurate relative photometry solutions for and to tie together all the DES pointings. Moreover, once such an accurate relative photometry map is derived for the DES, then an absolute calibration may be achieved by using standard stars to tie the DES observations to the AB magnitude system.

• We will compile and deliver the final SDSS Stripe 82 and Southern ugriz standard star catalogs that will greatly aid DES photometric calibrations, via the fact that the same scientists working on those catalogs in the SDSS are responsible for this item in the DES. In particular, the Stripe 82 data provides 1% calibrated standard stars over nearly 300 deg2 of sky area, while the Southern ugriz standards provide a network of standards spread out across the Southern Celestial Hemisphere.

• We will write the document that specifies the DES flatfielding procedure, given the expected data from the DES calibration and system response measurements. In particular, we expect to employ the so-called Stubbs tunable laser calibration system, which will provide detailed pixel-by-pixel system throughput vs. wavelength data that we will employ to flatfield DECam data. We will also work out a procedure to remove scattered light and pupil ghost reflections; likely we will rely on using star flat observations and pointing-by-pointing overlaps for this purpose.

• Finally, we will write the documents that specify the calibration plans and procedures to achieve the 2% requirement and 1% goal for relative and absolute photometric calibrations for DES.

3

Page 4: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

7.3.2 WBS 1.7.2.2 Camera Calibration/Response System The camera calibration/response system is composed of two subsystems: a standard dome flat subsystem, which will be used for standard processing of each night’s data and will be adequate for achieving the photometric calibration requirements for galaxy photometry and photo-z’s, and a tunable laser subsystem, which will necessary for achieving the photometric calibration requirements of the DES Supernova program. We discuss each in turn.

7.3.2.1 Dome Flat Subsystem The DECam will make use of a dome flat subsystem of a type that is standard at astronomical observatories. The purpose of the dome flat subsystem is to measure variations in the relative response of the DECam system (which includes the optical transmission of the filters and the optics, the reflectance of the Blanco primary, and the quantum efficiency of the CCDs) over the full DECam focal plane. In order to be effective, the dome flat subsystem must illuminate the entire DECam focal plane uniformly. In order to achieve this, the DECam dome flat subsystem will be composed of a Lambertian dome flat screen (at which the telescope will be pointed for calibrations) and a set of dome flat lamps. To provide uniform illumination of the flat field screen, three sets of dome flat lamps will be attached the top ring of the telescope support structure, each set separated by 120° from the other two sets. The Blanco currently employs such a dome flat subsystem (see Figure 7-1 and Figure 7-2). We plan to use the current Blanco flat field screen, but it is possible that we will use different dome flat lamps. For wide-band filters (like those to be employed by the DES), the effective wavelength of the filter depends on the spectral energy distribution of the object observed. Since system response depends on effective wavelength, we will want to use lamps that yield a similar effective wavelength as our objects of interest (primarily, elliptical galaxies at a typical redshift of z~0.5). To do this, either we can either use incandescent lamps with dome flat lamp filters or we can use LEDs1 to achieve the appropriate effective wavelength.

7.3.2.2 Tunable Laser Subsystem In order to derive a full system response function – the fraction of photons detected as a function of wavelength – for each pixel in the DECam focal plane, a more elaborate subsystem is required. This is the task of the tunable laser calibration subsystem. Here, as in the description of the dome flat subsystem, we take the full system to include the optical transmission of the filters and the optics, the reflectance of the Blanco primary, and the quantum efficiency of the CCDs. In essence, the tunable laser system permits one to monochromator the full optical path from the telescope’s entrance pupil to the CCD detectors over the whole focal plane in one fell swoop. We base our design on that described by Reference 2. The tunable laser subsystem will consist of the following components:

4

Page 5: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

• A laser system whose output light is capable of being tuned from 400nm to 1100nm ( see Figure 7-3).

• A Lambertian flat field screen (e.g., the current Blanco flat field screen; see Figure 7-2.

• Optical fiber that will carry the laser light up through the telescope structure to illuminate the flat field screen.

• A NIST-calibrated photo-diode, mounted the top ring of the telescope support structure, which will measure the full illuminated region of the flat field screen (and thus monitor the amount of light incident on the telescope pupil).

• Photo-current integration electronics and A/D converter for the for photo-diode Following the prescription of Stubbs et al. (2007), the data-taking sequence is:

1. Project light from laser onto flat field screen via the optical fibers. 2. Measure the flux reflected from screen, incident on telescope pupil, with the

NIST-calibrated photo-diode. 3. Compare the flux detected by instrument to the incident flux (as measured by

photo-diode). Ideally, each exposure would continue until the CCDs reach about one-third to one-half of full well. For Opotek laser used by Stubbs et al. (2007), this should take roughly 20 to 30 sec to reach 60,000 to 90,000 electrons.

4. Repeat at a succession of wavelengths to determine system response as a function of wavelength over full focal plane.

With these data, a wavelength-dependent response function can be constructed for each pixel. As an example, a succession of images from test data obtained by Stubbs for the Mosaic II camera is shown in Figure 7-4. Such data for the DECam will be important for the absolute color calibration of the DES photometry; for comparing the photometry of Type Ia supernovae at different redshifts, which is essential for building a well-calibrated SNIa Hubble Diagram; and for comparing DES photometry with photometry from other instruments.

5

Page 6: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

Dome Flat Screen

Figure 7-1 The Blanco Telescope with the current prime focus cage attached. The dome flat screen is indicated. During the process of obtaining dome flats, the telescope is pointed at the center of the dome flat screen, which is illuminated by dome flat flats attached to the perimeter of the top ring of the telescope support structure (see Fig. 6.18-2). (Credit: Tim Abbott, CTIO.)

6

Page 7: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

(Third set behind PF camera)

Dome Flat Lamps

Figure 7-2 The top ring of the Blanco’s support structure. Note the position of the three sets of dome flat lamps (the third set is obscured by the prime focus camera). (Credit: Tim Abbott, CTIO.)

7

Page 8: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

Figure 7-3 The Opotek tunable laser used in Chris Stubbs’ prototype tunable laser calibration system. (Credit: Chris Stubbs.)

8

Page 9: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

Figure 7-4 A succession of images from Mosaic II camera test data obtained by Stubbs using his tunable laser calibration. Note the fringing inherent in the Mosaic II response function. (Credit: Chris Stubbs)

7.3.3

7.3.4

WBS 1.7.2.3 Point Spread Function Calibration Plan This task involves scientist effort to write the document that specifies the DES point spread function (PSF) calibration plans and procedures. PSF calibrations will be important for accurate object photometry and in particular will be vital to achieve the accurate galaxy shape measurements needed to achieve the DES weak lensing key project goals. In particular this task will involve close collaboration with the DES weak lensing science working group, which will be responsible for providing the PSF and galaxy shear measurement modules for the DES Data Management pipelines.

WBS 1.7.2.4 Astrometric Calibration Plan This task involves scientist effort to write the document that specifies the DES astrometric calibration plans and procedures. Accurate astrometry is needed in order to register and coadd the imaging data from separate DECam pointings and hence will be important so that the DES data can achieve optimal photometric depth. Moreover, accurate astrometric calibration and correction for the geometric distortions due to the DECam optics and to the atmosphere will be essential for accurate weak lensing measurements.

9

Page 10: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

7.3.5

7.4

7.4.1

WBS 1.7.2.4 Photometric Redshift (Photo-z) Calibrations This task involves scientist effort to derive and optimal plan for photo-z calibrations for the DES, via the following work:

• We will evaluate the photo-z algorithms for DES using simple DES mock catalogs and real galaxy imaging and redshift survey data sets. Those real data sets will include in particular the galaxy data from the SDSS, CNOC2/RCS BCS, DEEP2, VVDS surveys, which spans from bright magnitudes down to the DES photometric limit i≈24. Moreover, we will derive the requirements on the sizes of the spectroscopic training sets needed for accurate DES photo-z calibrations.

• We plan to further test the DES photo-z algorithm using DES catalog level simulations (described below) which will improve the fidelity (compared to that of the mock catalogs above) of the simulated data to real galaxy properties. In particular, we will optimize the photo-z’s for different galaxy populations (field galaxies, clusters, etc.) and for different key project science (e.g, weak lensing).

• We will moreover optimize the DES photo-z algorithm using DES image level simulations (again described below) which will provide additional realism by accounting for the various atmospheric and DECam instrumental effects (detector noise, PSF, sky, etc.). Here we will explore the optimal magnitudes and colors that should be used for photo-z calculations, as well as any additional quantities (e.g., concentration index and morphology) that may be employed to improve photo-z accuracy.

• Finally, based on the above work, we will write the document that specifies the final DES photo-z calibration plans and procedures.

WBS 1.7.3 Survey Data Simulation

WBS 1.7.3.1 DECam Instrument Model Specifications This task involves scientist effort to specify the DECam optics, detector, and other instrumental model details to be used for the catalog and image simulations. The most relevant optics-related instrumental effects we are including in the simulations are point spread functions (PSFs), the pupil ghost, and geometric distortions. The DECam optics are modeled using a code developed by S. Kent. The optics-only PSFs in the DES griz filters have been computed by V. Scarpine using this code, with 1885 PSFs distributed over a grid of focal plane positions. Each PSF image covers 64x64 pixel2 (0.025”/pixel). In addition to the optical corrector contribution, there are additional instrumental PSF contributions, as itemized in the DES Science Requirements Document v6.5, Table 3.3. Currently we are modeling all of these other PSF contributions together simply as a Gaussian of 0.4” FWHM, which is convolved with the optics-only PSF. Note the largest of these additional contributions is CCD diffusion, with a 0.31” FHWM,

10

Page 11: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

compared to the optical-corrector contribution of 0.27” FHWM. The total instrumental PSFs in each filter are then represented by shapelet fits of order n=6 (28 coefficients). Future tasks will include explicitly modeling the PSF effects due to optical misalignments (e.g., focal plane tilts/offsets) and defocusing. To approximately model the pupil ghost present in the optical corrector design, we are currently using a Hubble-Reynolds law: I(R) = 0.06 I Rc2 / (R + Rc)2 , where R = radius, Rc = 0.6 deg, and I = sky background. Finally, to model the geometric distortions present in the optical design, we now use a purely radial r2-dependent scale change, resulting in a difference of 0.2% in scale and 0.4% in area between the center and the edge of the DECam focal plane, as indicated by S. Kent’s analysis of the corrector distortions. We plan to improve the fidelity of both our pupil ghost and geometric distortion models using more detailed optics calculations, and will implement the improvements in our Level 4 simulations. Our instrumental detector model includes the usual CCD effects of bias, read noise, gain, bad columns, flatfields, etc. We use a flux scale of 498, 494, 369, 380 electrons/sec in griz for a magnitude = 20 object in each filter. We currently use a bias value with mean of about 1000 electrons and set the dark current to zero. The read noise has a mean of 10 electrons, and the gain has a mean of 2 electrons/ADU; the read noise and gain values for each of the 2 amplifiers on each of the 62 DECam CCDs in the image simulations are specified in https://www.darkenergysurvey.org/the-project/simulations/imsim0/readnoiseMap. A mean of 20 bad columns per CCD is used, as specified in https://www.darkenergysurvey.org/the-project/simulations/imsim0/badColumnMap. The flatfield is currently the same for each CCD, and taken to be a replicated version of a LBNL CCD test flatfield image, available at https://www.darkenergysurvey.org/the-project/simulations/imsim0/desFlat.fit.gz, which includes the bright “glowing edges” present on the DECam CCDs. Bright star artifacts are also added, including saturated bleeding columns and a simple model (taken from the Terapix SkyMaker code) of the diffraction spikes and diffuse halos around bright stars. We also include cross talk between the 2 amplifiers of each CCD, with a cross talk coefficient of 0.001. Finally, we include cosmic rays, drawn randomly from a library of 1000 real cosmic rays taken from actual Fermilab DECam CCD test data, and added at the rate observed in the lab of 70 cosmic rays per 100 sec per 2kx4k CCD. For all these detector-related instrument model parameters, we plan to update them to reflect the actual distributions of real DECam CCDs as the data become available from the ongoing CCD testing effort at Fermilab.

7.4.2 WBS 1.7.3.2 Catalog Level Simulations This task involves scientist effort to assemble and collate the simulated galaxy and stellar object catalogs for the different DES simulation rounds. Our galaxy catalogs are currently based on dark matter particles from large parent cosmological N-body simulations. For our Level 1-3 simulations, we have used the publicly available Hubble Volume Simulation, with cosmological parameters Ω⊂ = 0.3,

11

Page 12: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

ΩΛ = 0.7, h=0.7, and σ8 = 0.9, as the parent N-body box. We extract a 500 deg2 sky area from the Hubble Volume, out to a maximum redshift z = 1.4, and use the ADDGALS method (R. Wechsler) to assign galaxies to the Hubble Volume dark matter particles. Each simulated galaxy is assigned the properties of a real SDSS galaxy, and the procedure reproduces the local luminosity-color-density correlations from the SDSS. “Primary” galaxies, based on Hubble Volume dark matter particles, are assigned down to 0.4 L*, resulting in ~15 million objects over the 500 deg2 catalog area. “Faint” or “background” galaxies, not based on dark matter particles, are then assigned in order to fill in from 0.4 L* down to an apparent magnitude limit R=25, resulting in ~40 million additional objects over the 500 deg2 area. The luminosity function model used is that of Blanton et al. for the SDSS, with simple passive luminosity evolution applied (M* brightens by 1.3 mag per unit redshift) irrespective of galaxy type. No redshift evolution beyond the local SDSS color-environment correlations is included. Starting with our Level 3 simulations, we will begin to use the so-called “Intermediate” N-body simulations (produced by the Barcelona DES group), which have a factor of 10 better mass resolution cf. the Hubble Volume. We will again assign galaxies using the ADDGALS method, but now extended to match the galaxy properties of fainter and higher-redshift galaxy surveys, e.g., DEEP2, in order to more realistically account for galaxy evolution with redshift. The simulated galaxies are assigned shape parameters using shapelet coefficients. The master shapelet coefficient distribution used has been derived from a set of SDSS main sample galaxies by E. Sheldon using B. Kelly’s shapelet fitting code. The simulated galaxies are assigned shapelet coefficients from this master catalog of about 36,000 SDSS galaxies. The shapelet decompositions are carried out to order n=15, corresponding to 136 shapelet coefficients. Visual inspection of the reconstructed shapelet fits showed a ~5% rate of bad fits/reconstructions, and these have been removed using cuts on total flux and bounds on negative flux for the shapelet reconstructions. Three classes of stars are included in the simulated stellar catalogs: bright USNO-B stars, faint simulated stars, and SDSS photometric standard stars. Real USNO-B stars are needed in order to test the astrometric module in the DES Data Management (DESDM) image processing pipeline. USNO-B stars with r < 20 are included. Simple linear conversions are used to transform from USNO-B photographic magnitudes to SDSS griz magnitudes; these conversions are only crude approximations as USNO-B photometry is inherently poor, but this is acceptable as the USNO-B stars are needed to calibrate astrometry, not photometry. At fainter magnitude r > 20 we use simulated stars, and we construct such a sample by using the Besancon stellar population synthesis model for our galaxy, specifically via the web form available at http://bison.obs-besancon.fr/model/. Finally, standard stars are needed to test the photometric calibration module in the DESDM pipeline. This sample has been derived from multi-epoch observations of the SDSS Southern Equatorial Stripe (Stripe 82) by D. Tucker, via selection of stars with 5 or more r-band observations and standard deviation < 0.05 mag per observation in r. A total of about 32,000 standard stars in 3 fields are used for the current simulations.

12

Page 13: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

7.4.3 WBS 1.7.3.3 Image Level Simulation Algorithm Implementation This task involves computing professional and scientist effort to update and improve the image and catalog simulation code, primarily to implement new simulation features. Our image simulations are generated using a Java-based image simulation package developed by C. Stoughton and N. Kuropatkin. The Java code is supplemented by a tcl/Fortran-based code written by H. Lin to prepare the observing and object information input to the Java package. The simulated images correspond to the 3-deg2 DECam “pointings,” which are observed over a set of simulated nights, each with different photometric zeropoints and extinction coefficients in the griz filters. The tcl/Fortran code produces files listing the observing information (pointing name and coordinates, date and time, airmass, filter, exposure time, seeing FWHM, sky brightness, photometric zeropoint, and observation type) for the simulated DECam pointings. The photometric zeropoints are drawn from a Gaussian with mean = 0 and σ = 0.1 magnitude, and the extinction coefficients are drawn from Gaussians with mean = 0.15, 0.1, 0.05, 0.025 for griz and σ = 0.03. In addition, even for nominally “photometric” nights, we also add small (σ ~ 0.01 mag) pointing-to-pointing zeropoint offsets, using real field-to-field offsets taken from photometric, repeat imaging data on SDSS Stripe 82. A subset of the nights, approximately 25%, are intended as “non-photometric,” and for these we draw zeropoint offsets from the SDSS-II Supernova Survey data, again on Stripe 82, but now taken under generally non-photometric, but still usable conditions. These zeropoint offsets are intended to serve as realistic tests for the photometric calibration module in the DESDM reduction pipeline. The seeing is drawn from the CTIO Mosaic-II seeing distribution from the Supermacho project, and a Moffat-profile atmospheric PSF is adopted. The sky brightness is also drawn from a Gaussian with mean = 742, 1850, 2630, 8970 electrons/pixel/100 sec in griz, and width 30% of the mean. The tcl/Fortran code then prepares the input object catalogs, in binary fits table format, for each simulated pointing. The observing information and the properties (RA/Dec, magnitudes, redshifts, shapelet coefficients) of the objects to be simulated are then read in by the Java simulation package. The simulated galaxies and stars are then rendered using shapelet coefficients, to order n=15 (136 coefficients) for galaxies and to order n=6 (28 coefficients) for stars. Shear due to gravitational lensing can be applied to the galaxies, using operations in shapelet space. The galaxies are finally convolved, again in shapelet space, with the stellar PSF, which itself is a convolution of a Moffat-profile (®=3.5) atmospheric PSF with the instrumental PSF (both shapelets of order n=6). Sky background, detector noise, and all the other instrumentral effects described under WBS 1.7.3.1 are then applied by the Java simulation code. We plan to make improvements and add new features to the image simulation and related code; some examples include: spatially and temporally correlated PSF patterns, due to atmospheric and/or instrumental effects; spatially and temporally correlated photometric zeropoint variations and color terms; and improved treatment of diffraction spikes and halos around bright stars.

13

Page 14: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

7.4.4

7.4.5

WBS 1.7.3.4 Image Level Simulation Production This task involves computing professional effort to oversee the production of the simulated images. Each simulated 3-deg2 DECam pointing consists of 62 CCD images, in multi-extension fits (MEF) files, each 1 GB in size. We primarily use computing nodes on FermiGrid and other Open Science Grid (OSG) sites for production of the simulated images. Compute time is 3-5 hours on typical grid nodes for the science images. In addition, we also produce related calibration images, specifically biases, flatfields, and standard star observations. To illustrate, for ImSim2 and for the images generated for DESDM Data Challenge 3 (DC3), we simulate 500 deg2, tiled 4 times in each of 4 filters over 10 observing nights. The data volume is broken down as follows:

• 2500 science images = 2.5 TB (500 deg2 x 4 filters x 4 tilings) • 240 standard star images = 240 GB (10 nights x 6 fields x 4 filters) • 500 calibration images = 500 GB (10 nights x [10 biases + 10 flats x 4 filters]) • Total data volume = 3200 images = 3.2 TB uncompressed

These data sets have been generated using a small amount of grid computing resources, averaging about 50 grid compute nodes over a 2-week production period. For each of the ImSim3 and subsequent image simulation rounds, we plan to cover 5000 deg2 x 2 tilings x 4 filters, or about 1 year’s worth of DES data. Together with an additional 20% in calibration images, the raw data volume will be about 20 TB per ImSim round, about a factor of 6 larger than those for ImSim2 and DC3. For these later simulations rounds, we plan to use additional grid resources, over 100 nodes simultaneously, as well as to extend the image production run over a longer 3 month period.

WBS 1.7.3.5 Simulation Storage This task accounts for the M&S costs of tapes, machines, and disks needed to store the catalog and image simulations, including both the raw data we produce, as well as the corresponding reduced data that will be produced by the DESDM data processing pipelines. Our tape storage needs are based on:

• 20 TB (5000 deg2 raw images, 2 tilings x 5 filters) • 40 TB (2x raw, expansion for reduced images) • 80 TB (2x reduced, intermediate images for coadd) • 20 TB (final coadded images) • 40 TB (suite of 4 full N-body simulations, 10 TB each) • 200 TB total

14

Page 15: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

The disk space needs are based on spinning 20 TB of raw images, plus room for testing and for catalog simulations and N-body light cone outputs. Typical annual M&S costs account for a 200 TB increment of tape ($70k) and for 30 TB of machines and disks (including replacements; $50k for 5 file server machines at 6 TB disk per machine) .

7.5

7.6

WBS 1.7.4 Mock Data Reduction Challenge (“Truth tables”) This task involves scientist effort in generating the “truth table” files, containing the true values of the positions, magnitudes, redshifts, shapes, and other properties for the objects in the DES image simulations. The truth tables are ascii files in comma separated value (csv) format (commas separate the data column values), to facilitate ingestion into the DESDM object database.

WBS 1.7.5 Survey Planning Milestones The level 2 survey planning milestones consist of the completion of each year’s level of catalog and image simulations; these milestones are spaced at approximately 6 month intervals. The completion of the final survey observing plan is an additional level 2 milestone. Further level 4 milestones indicate completion of test and validation catalogs and images, and of the completion of 50% of the catalog and image data for each simulation level.

15

Page 16: 7 WBS 1.7 Simulation and Survey Planning - Illinoisweb.hep.uiuc.edu/home/jjt/des_docs/tdr/simulationandplanning.pdf · DES observing and of DECam optics and detectors. These simulated

References 1 Marshall & DePoy 2005, astro-ph/0510233. 2 Stubbs & Tonry (2006, ApJ, 646, 1436) and Stubbs et al. (2007, ASP Conf. Ser., Vol. 364, p. 373).

16