SN Survey Image Simulations

A Simulated Roman High-Latitude Time-Domain Imaging Survey

WFI Simulations of Supernovae in the Deep Field

SN image sim lightcurve


Wang et al. (2022) have created the first simulations of realistic images with the Nancy Grace Roman Space Telescope that include artificial Type Ia supernovae (SNe Ia) injected as point sources in the images. These authors have created an image time series over two years of a nominal High-Latitude Time-Domain Survey, containing about 1,050 SNe Ia, covering a 1 square degree subarea of a planned 5 square degree deep survey. These images are released here publicly, along with input catalogs of all injected sources. The goal of this project is to help inform decisions on survey strategies toward the use of Roman for a measurement of dark energy through SNe Ia as probes from redshifts z ~ 0.1 to 2.

Details on the Generation of the Simulated Images

Although at present there is not a single defined survey strategy for the SN core survey, Wang et al. (2022) relied on an initial study of multiple strategies presented by Hounsell et al. (2018) isolating a single, representative survey strategy from that study. These simulations follow the same architecture as used by Troxel et al. (2021) to simulate the High-Latitude Imaging Survey, and include multiple effect that can cause systematic bias in cosmological analysis, such as background light, Poisson noise, reciprocity failure, dark current, classical nonlinearity, and interpixel capacitance. In this simulation, SNe Ia are overlaid as point sources next to galaxy positions as a time series of images. Recent infrastructure for characterizing the relation between SN properties and host-galaxy properties, as part of the SNANA package (Kessler et al. 2009), were utilized.

Simulated images were generated in the WFI bands R062, Z087, Y106, J129, H158, F814, K213, and W149 for all 18 Sensor Chip Arrays (SCAs). The simple pointing strategy for these simulations does not include a dither strategy to deal with the gaps between chips. Furthermore, the presence of bright objects in the simulations will limit the field coverage, as 12 diffraction spikes due to the Roman secondary mirror support struts will be visible for most stars in the simulations.

To create the simulated images, the WFI focal plane was arranged at the RA, Dec location of the Akari Deep Field South (center, RA=71deg, Dec=-53deg) with a given position angle. Each SCA has a fixed relationship to the overall focal plane, allowing its footprint to be determined from the information about the center point. Each image consists of a single SCA with 4088 x 4088 pixels. The simulated images were rendered using the GalSim software package (Rowe et al. 2015) specifically, the galsim.roman module in GALSIM v2.3.2. Each image was processed through a series of modifications, simulating physical effects, such as dark current, detector nonlinearity, and read noise. The effects of persistence, however, were not included. Rather than include the high-resolution estimate of the Roman spider pattern in GALSIM, a binning of 8 pixels from the high-resolution pupil image was used to simulate the point-spread function (PSF). For brighter stars, a binning of 4 or 2 pixels was used to prevent artifacts.

Wang et al. (2022) chose to follow the Imaging:High-z strategy from Hounsell et al. (2018) with updates to the reference survey by Rose et al. (2021). Though this strategy has both a deep tier and medium/wide tier, these image simulations focus solely on the deep tier (owing to computational limitations). The deep tier consists of the YJHF bands and exposure times of 300, 300, 300, and 900 s, respectively. The number of SNe Ia predicted by this strategy is ~8,000 using cutoffs for generation at 0.05 < z < 3.0.

Please see Wang et al. (2022) for details on the insertion of galaxies, stars, and supernovae into the simulated images.

Description of the Simulation Products

The 3 main products of this simulation are images, galaxy input catalogs, and SN lightcurve catalogs. The images are created in FITS format and contain header keyword information to make them compatible with HST MultiDrizzle software. The headers were created using Galsim's World Coordinate System (WCS) code and have WCS information in them. For each pointing, 18 separate images are created, one for each WFI SCA.

While there are unchanging input catalogs for galaxy, star, and SN photometry, an additional "truth file" is created for each image containing every galaxy drawn on the image, containing each object's ID from the original input catalog, as well as their positions (both WCS and pixel). A similar truth file for stars and SNe was created. The magnitude drawn on each image for SNe is also included.

A Single Static Pointing

Release of the full simulation is pending.

In the meantime, the release of a single static pointing is available here. This subset serves as a useful introduction to the full simulation, in that SNe and other objects appear in the same positions for each detector for all of the epochs. Images are only in the Y filter with a 300 s exposure time. This subset contains ~50 SNe Ia over the course of 200 days at the nominal 5-day survey cadence.

Below are the files and their descriptions:

Images simulated at various roll angles

Release of the full simulation is pending.

In the meantime, a second dataset with multiple pointings is available here. This subset provides a more realistic set of challenges, since objects do not remain fixed between images. However, this simulation does not represent a realistic pointing strategy. Images are in the YJHF filters with 300, 300, 300, and 900 s exposure time, respectively. This subset contains ~100 SNe Ia over the course of 200 days at the nominal 5-day survey cadence. Images were simulated using GalSim and take into account various detector effects. Additionally, coadded mosaics are created using Astrodrizzle, although users may want to create your own coadds.

Below are the files and their descriptions:

  • galaxia_akari.fits.gz - Input catalog for stars. Contains the following information for all stars:
  • RA, Dec star position (in radians)
    YJHF magnitudes (null)

  • rotate_update_truth_gal.fits.gz - Input catalog for galaxies. Contains the following information for all galaxies:
  • gind Index number for the galaxy
    RA, Dec galaxy position (in radians)
    int_e1, int_e2 Intrinsic shape of the galaxy (using GalSim's Shear class, distortion definition)
    g1, g2 Shear applied to galaxy (using GalSim Shear, reduced shear definition)
    rot Angle of galaxy (in radians)
    size Half-light radius (in arcseconds)
    z Redshift
    logsfr, logssfr, logmass Estimated star formation rate, specific star formation rate, and mass
    bflux, dflux Percentage of flux in bulge (bflux), disk (dflux * (1 - bflux)), and knots ((1 - dflux) * (1 - bflux))
    pind Index to catalog from which photometric properties of galaxy are taken (id_sim in 3dhst_sim_input_cat_v1.5.fits)
    RZYJHFW magnitudes (null)
    id_3dhst Index to 3DHST catalog ("All Fields" version from
    COEFF_SPECBASIS* Coefficients for constructing SED using EAZY

  • 3dhst_sim_input_cat_v1.5.fits.gz - CANDELS catalog from which photometric information of galaxies was drawn. Contains the following information that is used for the simulations:
  • RZYJHFW magnitudes (null)
    id_3dhst Index to 3DHST catalog ("All Fields" version from
    logsfr, logssfr, logmass Estimated star formation rate, specific star formation rate, and mass (using 3dhst information)
    COEFF_SPECBASIS* Coefficients for constructing SED using EAZY

    Also contains the following information for the original galaxies, which is not used for the simulations because they are taken from elsewhere:

    RA, Dec galaxy position (in radians)
    a, b, kron_radius, fwhm major axis, minor axis, Kron radius, FWHM

  • WFIRST_AKARI_FIXED_HEAD.FITS - Input catalog for SNe. Contains the following information on all SNe (more details in the SNANA manual:
  • SNID SN index number
    RA, Dec SN position (in radians)
    NOBS Number of SN observations (datapoints in lightcurve)
    PTROBS_MIN, PTROBS_MAX Indices to first and last observation of SN in lightcurve file (1-indexed for whatever reason)
    REDSHIFT_* Various redshift versions of SN ( REDSHIFT_FINAL used for images)
    HOSTGAL_OBJID host galaxy ID ( gind in static_update_truth_gal.fits)
    PEAKMJD Date of the SN peak brightness
    SIM_PEAKMAG_* Peak magnitude in the various bands

  • WFIRST_AKARI_FIXED_PHOT.FITS.gz - Input lightcurves for SNe. Each line contains the following:
  • MJD Modified Julian Date
    BAND Filter
    FLUXCAL SN Flux after noise added by SNANA (not used for the images)
    SIM_MAGOBS SN magnitude before noise added by SNANA (not used for the images)

  • Truth files for images - Contains a limited set of truth information about objects drawn on each image along with reference to the input catalogs:
    • Files with no suffix describe galaxies
    • Files with "_sn" suffix describe SNe
    • Files with "_star" suffix describe stars

    rotate_Y_truth.tar.gz rotate_J_truth.tar.gz rotate_H_truth.tar.gz rotate_F_truth.tar.gz

  • paper_rotate.fits - Contains the RA and Dec (in degrees), angle (in degrees), date (in MJD), and filter for each pointing
  • rotate_*_images_*.tar.gz - The images themselves (download scripts linked in below):

    • Filter in name
    • Next number refers to line in paper_rotate.fits
    • Number between 1--18 in individual tarfile indicates SCA

  • rotate_*_coadds.tar.gz - Coadded images created using Astrodrizzle (tarfiles linked in below):

    • Name includes position of coadd (lower left corner)

  • rotate_coadd_truth.tar.gz - Truth files for the coadds. Contains the following:
    • rotate_update_coaddlist.fits.gz - Describes which images contribute to each coadd (1-indexed from paper_rotate.fits)
    • rotate_update_index_*_.fits.gz - Describes galaxies and stars in each coadd. Galaxies are labeled with gal_star = 0 and stars are labeled with gal_star = 1
    • rotate_update_full_index.fits.gz - Describes all galaxies and stars along with which coadd they are located in

    Note that the gzipped tarballs of the images are generally 1.3 GB in size. A curl script for downloading the image dataset is here. A wget script is here.

    Note that the gzipped tarballs of the coadds are generally 29 GB in size. A curl script for downloading the coadd dataset is here. A wget script is here.