Combining Rubin and Roman photometry and its implications for photometric redshift estimation
Abstract
We perform a joint Roman--Rubin photometry analysis using the LSST Science Pipelines (LSP) on simulated images, by using the Roman F158 band as the reference band and applying forced photometry on the other bands. We find that the LSP CModel photometry for Roman bands is on average unbiased, but with a bright-end systematic uncertainty floor of 7-8%. The pgauss photometry as part of the Metadetection weak-lensing shape measurement pipeline also exhibits a similar floor despite a substantially different flux-bias behavior. However, this systematic uncertainty floor becomes negligible for the color residuals and both algorithms recover unbiased colors for Roman as well as Rubin. We also show that not accounting for correlated pixel noise in Roman IMCOM coadded images results in underestimation of flux errors by a factor of ~4. Based on the measured photometry, we build a machine-learning algorithm that maps the true galaxy properties to empirical flux errors, providing a convenient method to reproduce realistic photometry errors without having to rerun expensive image simulations and photometry pipelines. We apply this empirical error model to a new mock galaxy catalog with improved realism to investigate the impact of different survey depths and configurations on photometric redshift (photo-z) estimation performance. We find that 1) Rubin survey depth significantly impacts the photo-z precision, 2) additional systematic uncertainties on photometry could significantly degrade the photo-z performance especially towards high redshifts, therefore eroding the Roman--Rubin synergy, and 3) Roman photometry becomes important above z~1.5 while adding Rubin photometry to that of Roman improves photo-z performance across all redshifts. Our results provide a foundation for understanding how realistic uncertainties in galaxy photometry may propagate into the Roman--Rubin synergy in photo-z analyses.
Related
- Photometric Redshift Estimation with Hybrid Quantum Neural Network
- Optimizing Photometric Redshift Training Sets. I. Efficient Compression of the Galaxy Color–Redshift Relation with UMAP
- Tabular Foundation Models Dramatically Outperform Bespoke Photometric Redshift Estimators
- Learning JWST. I. A Foundation Model for New Population Discoveries and Morphology-Aware Photometric Redshift Measurements in the JADES Survey
- Anisotropic redshift distributions in photometric galaxy clustering and their cosmological impact
BibTeX
@misc{shin2026combining,
title = {{Combining Rubin and Roman photometry and its implications for photometric redshift estimation}},
author = {Tae-hyeon Shin and Rachel Mandelbaum and Chun-Hao To and Axel Guinot and Andrew Hearin and Arun Kannawadi and Chihway Chang and Federico Berlfein and Kaili Cao and Nihar Dalal and Michael Gabe and Christopher Hirata and Katherine Laliotis and Sidney Mau and Michael A. Troxel and Masaya Yamamoto and Roman HLIS Cosmology Project Infrastructure Team},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.22547},
eprint = {2610.22547},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.22547},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Combining Rubin and Roman photometry and its implications for photometric redshift estimation
AU - Tae-hyeon Shin
AU - Rachel Mandelbaum
AU - Chun-Hao To
AU - Axel Guinot
AU - Andrew Hearin
AU - Arun Kannawadi
AU - Chihway Chang
AU - Federico Berlfein
AU - Kaili Cao
AU - Nihar Dalal
AU - Michael Gabe
AU - Christopher Hirata
AU - Katherine Laliotis
AU - Sidney Mau
AU - Michael A. Troxel
AU - Masaya Yamamoto
AU - Roman HLIS Cosmology Project Infrastructure Team
PY - 2026
DA - 2026/10/15
PB - arXiv
DO - 10.48550/arxiv.2610.22547
UR - https://arxiv.org/abs/2610.22547
AB - We perform a joint Roman--Rubin photometry analysis using the LSST Science Pipelines (LSP) on simulated images, by using the Roman F158 band as the reference band and applying forced photometry on the other bands. We find that the LSP CModel photometry for Roman bands is on average unbiased, but with a bright-end systematic uncertainty floor of 7-8%. The pgauss photometry as part of the Metadetection weak-lensing shape measurement pipeline also exhibits a similar floor despite a substantially different flux-bias behavior. However, this systematic uncertainty floor becomes negligible for the color residuals and both algorithms recover unbiased colors for Roman as well as Rubin. We also show that not accounting for correlated pixel noise in Roman IMCOM coadded images results in underestimation of flux errors by a factor of ~4. Based on the measured photometry, we build a machine-learning algorithm that maps the true galaxy properties to empirical flux errors, providing a convenient method to reproduce realistic photometry errors without having to rerun expensive image simulations and photometry pipelines. We apply this empirical error model to a new mock galaxy catalog with improved realism to investigate the impact of different survey depths and configurations on photometric redshift (photo-z) estimation performance. We find that 1) Rubin survey depth significantly impacts the photo-z precision, 2) additional systematic uncertainties on photometry could significantly degrade the photo-z performance especially towards high redshifts, therefore eroding the Roman--Rubin synergy, and 3) Roman photometry becomes important above z~1.5 while adding Rubin photometry to that of Roman improves photo-z performance across all redshifts. Our results provide a foundation for understanding how realistic uncertainties in galaxy photometry may propagate into the Roman--Rubin synergy in photo-z analyses.
ER -