Avoiding (photo- z z ) Catastrophe
A. J. Battisti, E. da Cunha, S. Jin, J. A. Hodge · 2025
Read Avoiding (photo- z z ) Catastrophe by A. J. Battisti, E. da Cunha, S. Jin, J. A. Hodge (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-01251.txt.
Introduction
Battisti and co-authors warn in a 2025 paper about a sharp failure mode in estimating galaxy distances from images alone. Codes that fit a galaxy's colors can often pin down its redshift well. At times, though, two very different fits match the same colors, and the code picks the wrong one by a wide margin, a failure the field calls catastrophic.
The team points to one dusty galaxy that was first read as a far, early source but sits at a middle distance. The case is striking because rich data across many wavelengths still led the codes astray. They urge three fixes. First, fit the whole span of light from ultraviolet to radio. Second, let the dust model bend in more ways. Third, build in a floor for doubt that reflects the limits of the models and the data.
Catalog record
| CREATOR | A. J. Battisti, E. da Cunha, S. Jin, J. A. Hodge |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 7 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-01251
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Avoiding (photo- z z ) Catastrophe
A. J. Battisti, E. da Cunha, S. Jin, J. A. Hodge
Abstract
Spectral modeling codes that estimate photometric redshifts (photo- z z ) are a powerful and often reliable method for determining redshifts of galaxies. However, there are notable instances where degeneracies in spectral energy distribution (SED) colors lead to ‘catastrophic’ failures. We highlight the case of COSBO-7, a dusty, intermediate- z z galaxy that masqueraded as a high- z z source, because it demonstrates a unique scenario where photo- z z codes run into issues despite extensive multi-wavelength photometry. We advocate that photo- z z fitting should aim to: (1) use the entire available SED (UV–radio) whenever possible to help break color degeneracies, (2) allow flexible dust attenuation prescriptions, both in terms of the attenuation curve slope and a varying 2175Å absorption feature, and (3) implement uncertainty floors to account for limitations in spectral models and also on the photometry itself.
I
Introduction
Galaxies at high-
z
are often identified as candidates through photometric redshifts (photo-
z
or
z_{\mathrm{phot}}
) derived from their spectral energy distribution
(
M. Salvato et al. 2019
, SED; )
1
1
1
We ignore machine-learning
z_{\mathrm{phot}}
methods.
. However, photo-
z
codes can have instances of ‘catastrophic’ failure, often defined as
|z_{\mathrm{phot}}-z_{\mathrm{spec}}|/(1+z_{\mathrm{spec}})>0.15
(
H. Hildebrandt et al. 2012
)
, with low-
z
, dusty galaxies masquerading as high-
z
candidates being a common case
(
J. A. Zavala et al. 2023
;
S. Jin et al. 2024
, e.g.,)
. Given the investment of telescope time used for redshift confirmation, it is important to thoroughly test photo-
z
solutions before pursuing those avenues, even in instances where high quality JWST data are available.
II
The Case of COSBO-7
COSBO-7 is a strongly-lensed galaxy in COSMOS/PRIMER
(
N. Scoville et al. 2007
;
J. S. Dunlop et al. 2021
)
, first presented in the COSBO survey
(
F. Bertoldi et al. 2007
)
with a IR+radio-based photo-
z
of
z_{\mathrm{phot}}=3.1^{+0.5}_{-0.4}
. Later,
C. Ling et al. 2024
found that photo-
z
based on 10 bands of JWST/NIRCam+MIRI across numerous codes favored
z_{\mathrm{phot}}\gtrsim 7
and also presented an ALMA Band 3 line detection that could be CO(7–6) at
z=7.455
. Unexpectedly, the low-
z
solution was realised to be correct after detections of CO(7–6) and [CI](2–1) at
z=2.625
from ALMA Band 6 (instead of [CII];
S. Jin et al. 2024
), with the Band 3 line being CO(2–1).
We adopt photometry of COSBO-7 from
C. Ling et al. 2024
and
S. Jin et al. 2024
.
We use the
MAGPHYS+photo
-
z
code
(
A. J. Battisti et al. 2019
)
, and we refer readers to the aforementioned paper for details. Briefly,
MAGPHYS+photo
-
z
is a modified version of
MAGPHYS
(
E. da Cunha et al. 2008
;
E. da Cunha et al. 2015
)
that combines physically-motivated models for stellar and dust emission to fit ultraviolet-to-radio photometry and determine galaxy
z_{\mathrm{phot}}
likelihoods and physical properties simultaneously.
MAGPHYS+photo
-
z
assumes a
z_{\mathrm{phot}}
prior that peaks at
z\sim 2
and decreases at both lower and higher redshift (similar to cosmic SFR density,
P. Madau & M. Dickinson 2014
). It assumes a 10% uncertainty floor and does not perform any ‘tuning’ to optimize residuals/zeropoints.
Figure 1:
Upper panel:
MAGPHYS+photo
-
z
best-fit solutions for three filtersets: F090W–F1800W (green), MIPS24–VLA1.4GHz (orange), and all filters (black).
Lower panel:
Likelihood distributions for
z_{\mathrm{phot}}
(median and 1
\sigma
listed) and a few galaxy properties (some uncorrected for lensing magnification,
\mu
).
E_{b}^{\prime}
describes the 2175Å feature strength
(
A. J. Battisti et al. 2019
)
.
For JWST-only filters (green), there is a degeneracy where the F090W upper-limit can be interpreted as either a Lyman break or a moderate 2175Å dust absorption feature, resulting in two
z_{\mathrm{phot}}
peaks (first panel, second row). For filters longer than JWST (orange), the
z_{\mathrm{phot}}
solution is very broad but single-peaked at low-
z
. Combining all filters (black) provides the tightest constraint on
z_{\mathrm{phot}}
.
Figure
1
presents the
MAGPHYS+photo
-
z
results for COSBO-7 assuming three filtersets: (1) near-IR–mid-IR, (2) mid-IR–radio, and (3) near-IR–radio (i.e., all filters).
Case (1) is the same filterset used in
C. Ling et al. 2024
and case (2) is roughly comparable to
F. Bertoldi et al. 2007
. We find similar results as previous studies, with Case (1) favoring a high-
z
solution, although a secondary low-
z
solution is evident. Case (2) favors a broad single-peaked, low-
z
solution. Case (3) favors a narrow single-peaked, low-
z
solution. We note our Case (3)
z_{\mathrm{phot}}
disagrees with
z_{\mathrm{spec}}
by 3
\sigma
, but that it is not a catastrophic failure (
|z_{\mathrm{phot}}-z_{\mathrm{spec}}|/(1+z_{\mathrm{spec}})=0.06
).
III
Discussion
We outline three aspects that can help reduce instances of catastrophic failures:
First, it is beneficial to use photometry spanning as wide of a wavelength range as possible to break color degeneracies (upper limits also useful). A well-known degeneracy occurs between the Lyman break and Balmer break features, but a similar issue can arise due to the 2175Å absorption feature (e.g., Fig 1). Sharp SED transitions can also arise due to emission lines
(
J. A. Zavala et al. 2023
, e.g.,)
, which are difficult to self-consistently model with photometry, and wider coverage can mitigate their impact.
Second, adopting flexible dust attenuation prescriptions that include a varying 2175Å feature can provide more reliable
z_{\mathrm{phot}}
posteriors than a fixed attenuation curve. The average shape of dust attenuation curves likely evolve with redshift
(
A. J. Battisti et al. 2022
, e.g.,)
and there is evidence that a moderate 2175Å feature is typical in SFGs up to
z\sim 3
(
A. J. Battisti et al. 2020
, e.g.,)
, and has also been seen up to
z\sim 7
(
J. Witstok et al. 2023
, e.g.,)
.
Third, adopting uncertainty floors is important for numerous reasons: (1) formal uncertainties are often underestimated, (2) zeropoint variations between instruments are common, (3) template models used in SED-fitting codes are not perfect representations of reality. The last point is often forgotten and uncertainty floors help account for both limitations in spectral models and the photometry itself.
Finally, we note COSBO-7 shows a sharp SED transition between F1800W and MIPS24, presumably due to PAH emission. Including MIRI filters into our Case (2) fit gives a tighter constraint of
z_{\mathrm{phot}}=2.93^{+0.09}_{-0.20}
from
MAGPHYS+photo
-
z
(not shown), highlighting that the templates describing PAH features impact the results.
For this reason, using flexible IR templates
(
B. T. Draine & A. Li 2007
, e.g.,)
may improve
z_{\mathrm{phot}}
accuracy and reduce potential bias.
The JWST data presented in this article were obtained from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute. The specific observations analyzed can be accessed via
https://doi.org/10.17909/qb2s-1t93 (catalog doi: 10.17909/qb2s-1t93)
.