Hawaii Two-0 Progress Report

Current status of Hawaii Two-0 data collection including NEP archival data (AKARI & HEREOS) and CDFS archival data. The percentages below reflect time awarded through the end of 2020 and account for the descoped plan removing Y-band.

Awarded time in S2020A (3 nights DEIMOS) will be executed in June+July (Keck). Note that the 3 awarded HSC nights  in S2020A were cancelled on March 24 when all MKO telescopes were shut down until the end of April due to the coronavirus pandemic. See the update for S21B (below) about HSC nights requested by IPMU in support of the H20 program.


NEP HSC Photometery


NEP Keck Spectroscopy





H20 Observing Runs


Subaru HSC — Lost 4 nights due to earthquakes

Subaru HSC — Lost 3 nights due to earthquakes
2 successful nights (Jan 1&2)

Subaru HSC — Lost 1.25 nights due to weather, 1.25 successful nights; May 30 & 31 (2nd half), July 3-5 (2nd half)

Subaru HSC — 2 successful nightsOctober 23 & 24

Keck II DEIMOS — Lost 0.75 nights due to weather, 1.25 successful nights, August 28 & 29

Subaru HSC — Update March 24: 3 nights cancelled due to the shutdown of all MKO telescopes until the end of April due to the coronavirus pandemic,  April 17, 18, 20, 21, 26, 28 (2nd half)

Keck II DEIMOS — awarded 3 nights, June 27 & 28, July 15

Subaru HSC — 2 successful nights with HSClost 1.5 night due to weather

Keck II DEIMOS — 2 successful nights1 night lost to weather 


CFHT MegaCam — awarded 20 hrs of queue time (PI Zalesky). First data collected in March 2021.

Subaru HSC — awarded 17.4 hrs of queue time (PI Zalesky)

Keck II DEIMOS — awarded 3 nights (Keck offered to optimize the time by giving us 4x 3/4 nights on July 7-10)



HSC Coverage Maps

EDFF HSC Coverage

The solid black line indicates the boundary of the EDFF H20 coverage. Total enclosed area is ~10 square degrees.

NEP HSC Coverage

The dashed white circle indicates the boundary of the NEP H20 coverage. Total enclosed area is ~10 square degrees.

Archival Data

Through diligent searching of the archives we already have NEP HSC data from the HEROES  and AKARI-NEP projects, and CDFS data from a collection of various small projects. These archival data make up for nearly the entirety of the HSC time lost in 2018.

All data are being re-reduced at the IfA (by A. Repp & C. McPartland) using the latest photometric/astrometric calibrations from Pan-STARRS and Gaia, along with the latest version (6.7) of the HSC data reduction pipeline.



Data Reduction Progress


  • The HSC Data Reduction Pipeline has been successfully installed and tested on the UH Cray supercomputer
  • Co-I Chambers will ensure the team has access to the latest Pan-STARRS data for astrometric and flux calibration.
  • Generation of HSC mosaic will be done by Co-Is Repp & McPartland
  • The IRAC image mosaic pipeline of Co-I Capak has recently been optimized for cluster/supercomputer environments. The runtime of the pipeline analysis is now only a few hours, rather than a few weeks with the original version.
  • IRAC photometry will be measured using IRClean (Co-I Capak lead) that was successfully applied in COSMOS and SPLASH Laigle et al. (2016).
Extraction of Data from the Archives
  • We have discovered a significant amount of HSC data in the Subaru archive covering the two H20 fields:
    • NEP: HEROES and AKARI-NEP programs (~50% of total data needed)
    • CDF-S: various small programs (~10% of total data needed)
  • With these data in hand, we have nearly made up for the HSC time lost in 2018. See Archival Data on the Progress Page for more details.
Preliminary Tests
Purchase of Data Storage/Reduction Server
  • We have recently purchased a dedicated server for H20 data storage and reduction. This will allow us to streamline the data reduction process by minimizing latency due to data transfer.
  • 77 TB hard drive, 96 CPU cores, 512 Gb RAM
  • Currently waiting on delivery of a second dedicated server with similar specifications

Summary of Progress by Semester

Semester 1: S18A

  • 4 nights of HSC time lost to EARTHQUAKE
  •  Kick-off meeting to decide observing strategy for Hawaii Two-0 project
  •  Consult with IfA transient experts on optimizing survey strategy for supernova detection
  •  Prepare Spitzer SLS data in order to construct full SEDs with HSC data
  •  Setup HSC pipeline on UH Cray
  •  Secure funding for two IfA grad students to work on H20 program
  •  Preliminary image reduction and catalog
  •  Prepare and execute first Subaru HSC observations in NEP
  • Preliminary NEP photo-z catalog including stellar masses
  • First selection of high-z candidates in NEP for follow-up spectroscopy

Five color (grizY) image of archival HSC data in the NEP

Semester 2: S18B

  • 3 nights of HSC time lost to EARTHQUAKE
  •  Present IfA talk on first semester progress and data availability
  •  Confirm depth requirements of the H20 Observations using COSMOS HSC data
  •  Identify z=4-7 dropouts in COSMOS HSC data to estimate H20 final outcomes
  •  Finalize Spitzer data reduction pipeline
  •  Prepare and execute first Subaru HSC observations in NEP
  •  Preliminary NEP photo-z catalog including stellar masses
  •  First selection of high-z candidates in NEP for follow-up spectroscopy
  •  Prepare and execute first Subaru HSC observations in CDFS
  •  Produce preliminary stacked images and photometric catalogs for CDFS
  •  Produce preliminary photo-z and stellar mass catalog for CDFS
  •  First selection of high-z candidates in CDFS for follow-up spectroscopy (see the Science Page)
  • Finalize first data release for NEP

Left: Color selection for r-band droupouts. Right: Distribution of r-band dropouts (red points) in the COSMOS field.

Semester 3: S19A

  • 4 nights of HSC time awarded, May 29&30 (2nd half), July 26—31 (1st half)
  •  First selection of high-z candidates in NEP for follow-up spectroscopy
  •  Preliminary NEP photo-z catalog including stellar masses
  •  Prepare updated images and catalogs for NEP & CDFS
  •  Preliminary measurements of Galaxy Two-Point Correlation Function
  •  Preliminary measurements of Galaxy-Dark Matter bias (see figure at right, and the Science Page)
  • Prepare and execute first Subaru HSC observations in NEP
  • Execute first round of CDFS spectroscopic follow-up
  • Finalize first data release for CDFS

Beck et al. “Hawaii Two-0: A preliminary analysis of high-redshift galaxy bias” (accepted by MNRAS, arXiv:1909.12854, ADS Link)

Semester 4: S19B

  •  Submitted IfA observing time proposal for 4 (+1) nights with Subaru HSC and 3 (+2) nights with Keck II DEIMOS
  • Awarded 2 nights with Subaru HSC (October 23 & 24) and 2 nights with Keck II DEIMOS (August 28 & 29)
  •  Purchased new data reduction/storage server
  •  Re-reduction of H20+archival NEP & CDFS HSC data. See coverage maps of NEP & CDFS
  •  Started spectroscopic follow-up. Preliminary reduction/analysis is promising with continuum detections in ~70% of sources and emission line detections in ~20% of sources. All sources with measured redshifts are consistent with expectations from the dropout selection technique (i.e. there are no low-redshift interlopers).
  •  Beck et al. “Hawaii Two-0: A preliminary analysis of high-redshift galaxy bias” (submitted to MNRAS October 2019, arXiv:1909.12854, ADS link)
  •  Data collection in the Spitzer SLS Fields was completed Summer 2019. Full mosaics have been constructed and vetted in the last month! See SLS data for more details. These data are available to anyone at the IfA by contacting the H20 team.
  • Following an assessment of progress through the first four semesters of our LCP, we have decided it would be prudent to descope the LCP by forgoing the HSC Y-band observations. This descope brings the total number of HSC nights down from 30N to 18N. Considering the 6 nights of successful data collection through the LCP, and an equivalent of 6 nights of archival data, we have a strong chance of reaching 100% completion of a four band (griz) imaging program for the H20 LCP.  The HSC griz imaging in conjunction with the SLS 3.4+4.5 micron imaging will be sufficient for identifying massive galaxies at z~3-6.5, and thus does not significantly degrade our original science plan. The Y-band data will eventually be obtained by Euclid, along with JH-band observations, that will allow us to identify massive galaxies out to redshifts z~14 (if they exist!).
  • Finalize first release of calibrated images from NEP & CDFS

Example of current HSC i-band data quality in the NEP.

Examples of Lyman alpha detections in dropout sources in the NEP. Left: HSC i-band cutout image with fixed physical scale of 70 kpc at the source redshift. Right: Zoom in of the Lyman alpha line in the DEIMOS spectra.

*** NEW Spitzer SLS Data ***

The Spizer SLS data collection was completed in Summer 2019, and full mosaics were constructed and vetted in October 2019. Below, we show the IRAC channel 1 & 2 data in the CDFS and NEP. For a closer inspection of the images, here is a zoomable image of the Spitzer ch1 data in the NEP.



Semester 5: S20A

  • Awarded 3 nights with Subaru HSC (2nd half of April 17, 18, 20, 21, 26, 28) and 3 nights with Keck II DEIMOS (June 27 & 28, July 15)
  • Continue spectroscopic follow-up
  •  Tractor/Farmer photometry meeting in January
  •  Updated HSC pipeline to version 7.9.1
  •  Generated bright star masks following Coupon et al. 2018
  •  Reduce available NEP HSC images
  •  Preliminary testing of forced photometry measurements using full depth Sptizer data from SLS
  • Complete Farmer forced photometry in NEP & CDFS
  • Expect to have a vetted catalog of photometry for the full depth region of NEP (~1 million sources) by May 2020
  • Catalog and image data for NEP & CDFS are available by contacting the H20 team

gri composite image showing a ~11×11 arcmin region of the NEP

NEP Depths

Preliminary photometry of the region of NEP where we already have depth shows that we are meeting or exceeding the survey requirements for depth.

An illustration of typical results using the Farmer

The depths probed by our Spitzer and HSC imaging lead to densely crowded fields. To obtain accurate photometry of each source and mitigate the issues presented by overcrowding and blending, we use The Tractor (Lang 2016 a,b). The Tractor offers precise photometry by modeling every source with physically motivated prescriptions rather than by assigning apertures, as done with historically popular tools like SExtractor (Bertin & Arnouts 1996). Resolved sources, typically galaxies, are modeled with exponential or De Vaucoulers profiles (de Vaucouleurs 1948), while point sources are modeled as point spread functions. In each case, models are optimized by fitting for the various structural parameters in a given band designated for modeling and forcing the models on to the other photometric bands while allowing the flux to vary. The best fit model is the one that minimizes the difference between the model flux and total measured flux summed over the area of the detected source, as depicted below. The top row shows the initial guess of the solution obtained during the modeling step, where structural parameters are free to vary, while the bottom row shows the final solution in which only the flux is optimized.

Sources detected nearby to one another in crowded regions must be modeled simultaneously to properly account for blending (e.g., other Fig). The models for these sources are optimized using a joint likelihood.

Preliminary Photo-z Tests

We have successfully completed preliminary testing of photometric redshifts using the software package EAZYpy

HSC Data Reduction Improvements

We have recently re-reduced all of the available HSC data in the NEP field using the newest version (v7.9.1) of the HSC pipeline. There are notable improvements in background subtraction and artifact rejection. The two images below show the same patch of sky reduced with v6.7 (left) and v7.9.1 (right) of the HSC pipeline. Here, we see the satellite trail in the old image is completely removed, and the scattered light around bright stars is signifcantly reduced.

Semester 6: S20B

  •  Reduce available CDFS HSC images
  •  Conduct spectroscopic follow-up of high-z dropout galaxies
    • See below for examples of Ly-alpha detections from DEIMOS spectroscopy
  • Prepare final release candidate for CDFS images and catalog
  •  ADAP proposal submitted for H20 cosmology analysis
  • H20 Image Viewer
Constraints on σ8 from recent cosmological surveys, and expectations for the proposed analysis.  The blue band illustrates the constraints on σ8 from the DES cosmic shear analysis (Troxel et al. 2018), with the redshift extent indicating the approximate redshift range probed by DES.  Note that if DES were to measure σ8 in redshift bins, the constraints would be less tight. The orange point with errorbars represents the value of σ8 inferred from the high-redshift primary CMB measurements from Planck.  The blue points with errorbars represent a forecast for the proposed analysis. For illustrative purposes, we assume a fixed value of Ωm.

Three color (giz) composite images from H20 HSC data of the Euclid Deep Feild Fornax (EDFF/CDFS). Visit https://h20.ifa.hawaii.edu/ to see more images.

Highlights from S20B DEIMOS Spectoscopy

Left: gri color image of the source from the H20 HSC data. Right: Zoom in on Lyman-α line detections in DEIMOS spectra of high-dropouts observed in S20B.


  • First CFHT data collection in March 2021
  • Purchased third data reduction server which will allow to build a fully operational compute cluster.

Institute for Astronomy

We are one of the largest university astronomy programs in the world.