Full calibration of the tomographic redshift distribution from the HSC PDR3 Shape Catalog with DESI

arXiv:2511.18133 · astro-ph.CO · Submitted 2025-11-22 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Astrophysics Radio. Generated commentary on the latest astrophysics papers.

Vera: I'm Vera, and with me are Jocelyn and Subrahmanyan, guest researcher.

Jocelyn: Today's paper: "Full calibration of the tomographic redshift distribution from the HSC PDR3 Shape Catalog with DESI".

Vera: This paper details the full calibration of the four tomographic redshift bins within the Hyper Suprime-Cam (HSC) weak lensing catalog using Data Release 1 and 2 from the Dark Energy…

Jocelyn: First, who's behind it and why it matters.

Title: Vera: So, continuing our discussion on "Full calibration of the tomographic redshift distribution from the HSC PDR3 Shape Catalog with DESI," I was looking at the initial descriptions of what they were measuring—specifically regarding those bins and source densities. It seems like they're focused on getting a solid average raw angular density for specific populations.

Jocelyn: Right, and that mention of achieving an average raw angular density for HSC sources of about zero point one arcmin−two in the bin really grounds the paper in concrete observable numbers. It tells us exactly what scale they're aiming for in their measurements.

Subrahmanyan: Because galaxy clustering measurements are fundamentally dependent on source density; if you can't reliably estimate or sigma eight due to poor sampling, all the subsequent physics is just guesswork.

Vera: It’s fascinating how they describe down-sampling that fifth tomographic bin by twenty-five percent—it sounds like a necessary step to hit that target density while keeping the measurement statistically sound enough for comparison.

Jocelyn: And it seems like they are using DESI DR2 QSOs alone for these specific measurements, which adds another layer of systematic consideration we need to keep track of when interpreting the results.

Subrahmanyan: That reliance on a single source type, even one as powerful as QSOs, dictates the potential biases we must account for when extrapolating those findings to general galaxy populations.

Vera: It makes me think about how sensitive these measurements are; they’re trying to calibrate something so fundamental—the redshift distribution—using multiple distinct tracers within the same survey volume.

Jocelyn: Precisely, Vera; it suggests that the consistency across different tracers, or even within different bins for the same tracer, is what will ultimately validate their "full calibration" claim.

Subrahmanyan: We're looking at a highly interconnected system here: galaxy clustering depends on redshift bins, which depend on source density measurements, all feeding into cosmology.

Summary: Vera: Moving onto the summary section of "Full calibration of the tomographic redshift distribution from the HSC PDR3 Shape Catalog with DESI," what really jumps out to me is how they are presenting these cross-correlation functions in Figure ten comparing different spectroscopic bins against those final tomographic bins.

Jocelyn: And seeing that figure must be telling us a lot about the reliability of the QSO tracers across various scales, right? Are we seeing clean signals everywhere?

Subrahmanyan: The ability to cross-correlate multiple tracers like this is key because it lets them build up a statistical picture that’s far more constrained than any single tracer could provide alone.

Vera: It seems they’re showing how the QSO data can act as a critical bridge, linking the known spectroscopic measurements with the photometric bins that are inherently fuzzier in terms of redshift certainty.

Jocelyn: And it really highlights that even though they are measuring something like a zero point one arcmin−two population, they are also displaying measurements at *both* scale cuts, which is good for showing robustness across observational parameters.

Subrahmanyan: The comparison between those two scale cuts, as noted in the text, suggests that while the QSO power might be similar at these redshifts—which is somewhat unexpected given typical expectations—it’s a powerful confirmation of their methodology.

Vera: What I found particularly interesting was the discussion about how the QSO population's sparsity limits signal at small scales, especially below the halo virial radius, yet they still report a significant detection for "Bin five."

Jocelyn: That persistence in detecting 'Bin five' even with those limitations—that’s what proves the sensitivity of their calibration technique, Vera. It’s not just theory; it’s measurable on the sky.

Subrahmanyan: It speaks to the fact that even when statistics are tight, if the underlying physical signal is strong enough, these methods can extract meaningful cosmological constraints.

Improvements: Vera: Okay, shifting focus to potential improvements suggested by "Full calibration of the tomographic redshift distribution from the HSC PDR3 Shape Catalog with DESI," I noticed they mention assuming the photometric sample galaxy bias reported in section three point three remains valid at high redshifts. That assumption feels big.

Jocelyn: It does sound like a major simplifying assumption, Jocelyn; if that bias model breaks down at z about two then the whole calibration gets shaky, doesn't it?

Subrahmanyan: Absolutely; the evolution of galaxy bias with redshift is one of the most active and difficult areas in modern cosmology because it connects galaxy properties to underlying dark matter halos.

Vera: But they also point out that for these high redshifts, the statistical power of QSOs is broadly similar for both scale cuts, which mitigates some worry about needing a perfect bias model across every single measurement point.

Jocelyn: And they use the example of the QSOs revealing a photometric redshift scatter towards higher redshifts in this bin— z about two point zero - two point three. That's a concrete observational result showing how the data behaves!

Subrahmanyan: This specific redshift scatter is valuable because it allows them to constrain not just the mean redshift, but also the intrinsic width of the photometric distribution, which feeds back into better tomographic bin definitions.

Vera: They mention that while QSO clustering calibration is novel, recent DES Y6 works use eBOSS QSOs for comparable sensitivity. That’s a really reassuring benchmark for the community reading this paper.

Jocelyn: Knowing that other major groups are finding comparable results using related tracers gives confidence to the scientific community that this approach has real footing, even if it's pushing boundaries.

Subrahmanyan: It suggests a maturing field where these cross-calibration techniques are becoming standard tools for extracting cosmological parameters from multi-messenger or multi-survey data sets.

Conclusion: Vera: Wow, we’ve covered a lot of ground discussing "Full calibration of the tomographic redshift distribution from the HSC PDR3 Shape Catalog with DESI." If I were to summarize the core impact, it's that these techniques are really maturing us into a new era of redshift mapping.

Jocelyn: I agree, Vera; what sticks with me is how robustly they demonstrated QSO sensitivity for high-redshift calibration, especially when dealing with those outlier populations in photometric redshift catalogs.

Subrahmanyan: From a big picture view

J. Choppin de Janvry, S. Gontcho, A Gontcho, U. Seljak, A. Baleato Lizancos, E. Chaussidon, W. d’Assignies, J. DeRose, S. Heydenreich, E. Paillas

Lawrence Berkeley National Laboratory, Cyclotron Road, Berkeley, California 94720, United States · University of California, Berkeley (Department of Physics) · University of Virginia (Department of Astronomy) · University of California (Sproul Hall) · Institute de Física d’Altes Energies (IFAE), The Barcelona Institute of Science and Technology · Brookhaven National Laboratory, Physics Department · University of California, Santa Cruz (Department of Astronomy and Astrophysics) · Universidad Diego Portales (Instituto de Estudios Astrofísicos) · University of Arizona (Steward Observatory) · University of Pittsburgh (Department of Physics & Astronomy and Pittsburgh Particle Physics, Astrophysics, and Cosmology Center - PITT PACC) · Boston University (Department of Physics) · Università degli Studi di Milano (Dipartimento di Fisica Aldo Pontremoli) · INAF-Osservatorio Astronomico di Brera · University College London (Department of Physics & Astronomy) · Institut d’Estudis Espacials de Catalunya (IEEC) · Institute of Space Sciences, ICE-CSIC · Universidad Nacional Autónoma de México (Instituto de Física) · Universidad de los Andes (Departamento/Observatorio Astronómico) · University of Portsmouth (Institute of Cosmology and Gravitation) · Fermi National Accelerator Laboratory · NSF NOIRLab · Southern Methodist University (Department of Physics) · University of California, Irvine (Department of Physics and Astronomy) · Sorbonne Université, CNRS/IN2P3, Laboratoire de Physique Nucléaire et de Hautes Energies (LPNHE) · Universitat Autònoma de Barcelona (Departament de Física) · Institució Catalana de Recerca i Estudis Avançats · Perimeter Institute for Theoretical Physics · University of Waterloo (Waterloo Centre for Astrophysics) · University of Texas at Dallas (Department of Physics) · Southwest Research Institute (NOIRLab)

astro-ph.CO

Submitted: 2025-11-22

Updated: 2026-08-25

Comments: 42 pages, 11 figures, 5 tables

Journal ref: JCAP05(2026)073

DOI: 10.1088/1475-7516/2026/05/073

Code: https://github.com/cosmodesi/pycorr

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 92/100

The gist: This paper details the full calibration of the four tomographic redshift bins within the Hyper Suprime-Cam (HSC) weak lensing catalog using Data Release 1 and 2 from the Dark Energy Spectroscopic

Key concepts

Tomographic Redshift Bins
These are specific groupings used to divide the redshift distribution of sources into distinct bins. The paper calibrates these bins using measurements from the HSC PDR3 Shape Catalog and DESI data to create a solid average raw angular density for each bin.
Source Density
This refers to the number of sources per unit area. Achieving an average raw angular density of about zero point one arcmin-two in a bin is a concrete observable number used by the researchers to ground their measurements and ensure statistical soundness for subsequent galaxy clustering physics.
Cross-correlation Functions
These are measurements comparing different spectroscopic bins against the final tomographic bins. Cross-correlating multiple tracers allows researchers to build a statistically constrained picture that is more reliable than using any single tracer alone.
Galaxy Bias Evolution
This is the study of how the bias of galaxies changes with redshift. The hosts discuss assumptions about this model, noting that its validity at high redshifts is a major point of concern in cosmological studies.

Terminology

Summary

This paper details the full calibration of the four tomographic redshift bins within the Hyper Suprime-Cam (HSC) weak lensing catalog using Data Release 1 and 2 from the Dark Energy Spectroscopic Instrument (DESI). Accurate characterization of the normalized sample redshift distributions projected along the line of sight n(z) is critical because differences in tomographic calibration can lead to significant cosmological parameter shifts in weak lensing analyses.

Methodology and Data Sources

The study employs the clustering redshifts technique, which leverages angular cross-correlations between spectroscopic samples and the photometric galaxies in each tomographic bin. Unlike previous HSC calibrations, this work includes z > 1.2 redshift sources such as emission line galaxies (ELG) and quasars (QSO) sources, allowing for a complete calibration of all the redshift bins. The researchers utilize several DESI target classes to perform these measurements:

  • Bright Galaxy Survey (BGS)

  • Luminous Red Galaxies (LRG)

  • Emission Line Galaxies (ELG), including LOP and VLO samples

  • Quasi-Stellar Objects (QSO)

Modeling and Systematic Mitigation

To improve accuracy, the authors investigate common possible systematics that can affect the clustering redshifts calibration, specifically focusing on galaxy bias and magnification effects. The researchers propose corrections to reduce these impacts by relaxing the assumption of linear bias to instead assume no redshift evolution of the cross-correlation coefficient, which allows them to leverage smaller clustering scales. They also model magnification effects (g times mu and mu times g) caused by matter distribution along the line of sight, using a magnification matrix to solve for corrected measurements.

Furthermore, the redshift distributions are modeled using B-splines to regularize and smooth out measurements that may statistically fluctuate. The study also addresses galaxy bias evolution by measuring the auto-correlation of the photometric sample and applying corrections to account for how clustering changes across redshift.

Results and Comparative Analysis

The calibration results show that the shifts in redshift are considerably smaller than the shifts obtained in the HSC Year 3 cosmic shear analyses. The analysis utilized two different scale cuts—0.3–3 h-1 Mpc and 1–5 h-1 Mpc—to verify that fiber collisions did not introduce bias, finding that the consistency between these cuts supports their findings. The specific results for the tomographic bins are:

  • Bin 1: Shows a small shift towards low redshifts.

  • Bin 2: Demonstrates good agreement with the photometric calibration.

  • Bins 3 and 4: Exhibit a shift towards higher redshifts relative to previous work.

For the two high-redshift bins, the shifts compared to Rau+2022 are z 3 = -0.039+0.020-0.021 and z 4 = - 0. 48 (with asymmetric errors). These results offer competitive measurements exclusively from clustering redshifts when compared to Cosmic Shear Ratio (CSR) and Galaxy Clustering-Weak Lensing (GC-WL) analyses.

High-Redshift Probing with Quasars

The study highlights the utility of the QSO sample for probing high-redshift source distributions beyond z about 1.6. Despite lower number densities, quasars provide significant potential for high redshift calibration and can assist in breaking degeneracies in redshift populations. This capability is essential for identifying outlier populations confused by the chosen photometric redshift algorithm, such as the potentially degenerate high-redshift population identified in the first two tomographic bins.

Improvements for AI systems

1. Dirichlet-Prior B-Spline Density Estimators for Latent Space Modeling

  • What the improved AI system can do: In probabilistic machine learning tasks (e.g., Bayesian Neural Networks or Variational Autoencoders), this system can model complex, multi-modal probability density functions (PDFs) in latent space with guaranteed non-negativity and smoothness. By utilizing B-spline parametrization constrained by a Dirichlet prior, the system eliminates the negative density artifacts common in standard kernel density estimation and provides superior uncertainty quantification in low-signal/high-noise regimes by regularizing the distribution through optimized spline knots.

2. Cross-Correlation Latent Alignment (CCLA) for Heterogeneous Sensor Fusion

  • What the improved AI system can do: This system enables high-precision property estimation (such as depth, material composition, or chemical concentration) in low-fidelity, high-density sensors (e.g., standard RGB camera streams or low-resolution satellite imagery) by leveraging sparse, high-fidelity anchor data (e.g., LiDAR or hyperspectral scans). Instead of relying on direct regression—which is prone to spectral/resolution gaps—the system learns the angular cross-correlation between feature densities in both domains, allowing it to reconstruct the true underlying distribution of properties even when the primary sensor has significant information deficits.

3. Systematic Selection and Magnification Matrix (SSMM) Layers

  • What the improved AI system can do: In computer vision pipelines for autonomous robotics or remote sensing, this layer corrects for observation-induced density biases—systematic shifts in object/feature counts caused by sensor saturation, lens distortion, or atmospheric interference. By integrating a magnification matrix into the feature extraction process, the AI can solve an inverse problem to decouple the true physical density of objects from the systematic selection effects of the sensing hardware, preventing biased counting and spatial distribution errors in high-stakes environments.

Abstract

The calibration of tomographic redshift distributions is essential for cosmological analysis of weak lensing data. In this work, we calibrate all four tomographic bins of the Hyper Suprime Camera (HSC) weak lensing catalog with the Dark Energy Spectroscopic Instrument (DESI) Data Release 1 and 2 using the clustering redshifts technique. We include z > 1.2 redshift sources such as emission line galaxies (ELG) and quasars (QSO) sources in our calibration, which were not available in the previous HSC calibration (Rau et al. 2022), allowing a complete calibration of all the redshift bins. We find the first tomographic bin exhibits a small shift towards low redshifts. The second bin is in good agreement with the photometric calibration, while third and fourth bin exhibit a shift towards higher redshifts. However, these shifts are considerably smaller than the shifts obtained in the HSC Year 3 cosmic shear analyses. We evaluate the impact of galaxy bias and magnification effects from all the samples on the measurements, finding them to be small, and we propose corrections to reduce them further. Specifically, we relax the assumption of linear bias and only assume no redshift evolution of the cross-correlation coefficient, allowing us to leverage smaller clustering scales. We model the redshift distributions with splines and compare our results to previous analyses as well as to other parameterizations found in literature. For the two high-redshift tomographic bins, we find the shifts to higher redshifts with respect to the measurements performed in Rau+2022 to be Δz 3=-0.043+0.022-0.020 and Δz 4=-0.050+0.012-0.012.

Sources

Related papers