The S-PLUS Fornax Project (S+FP): An extragalactic catalog covering sim 5 virial radii around NGC 1399 with galaxy properties
Pith reviewed 2026-05-16 02:12 UTC · model grok-4.3
The pith
A catalog of 119580 galaxies toward the Fornax cluster supplies multi-band photometry, photometric redshifts and derived physical properties out to five virial radii.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The central result is an extragalactic catalog of 119580 galaxies projected toward NGC 1399, the dominant galaxy of the Fornax cluster, extending roughly five virial radii in right ascension and three in declination. The catalog combines twelve-band S-PLUS optical photometry with GALEX ultraviolet, VHS-VISTA near-infrared and AllWISE mid-infrared measurements. Photometric redshifts are estimated at sigma_NMAD approximately 0.0219 with a practical lower limit of z approximately 0.03. Stellar masses, star formation rates and D4000 indices are obtained by machine-learning regression trained on SDSS spectroscopic matches, and the catalog completeness is reported as 72 percent from comparison to
What carries the argument
Supervised deep learning neural networks with dimensionality reduction for separating galaxies from stars and spurious objects, followed by machine-learning regression trained on SDSS data to estimate stellar masses, star formation rates and D4000 from the multi-band photometry.
If this is right
- The catalog enables density analyses and large-scale structure mapping across the Fornax region out to five virial radii.
- Stellar mass and star formation rate distributions can be used to test models of galaxy formation and evolution in a dense environment.
- Photometric redshifts above z approximately 0.03 provide target selection for future spectroscopic surveys.
- The multi-wavelength coverage supports studies of environmental effects on galaxy properties from the cluster core to the outskirts.
- Completeness of 72 percent allows statistical corrections for luminosity functions and clustering measurements.
Where Pith is reading between the lines
- Cross-correlation with X-ray or radio maps could identify previously undetected infalling groups or filaments feeding the cluster.
- The photo-z accuracy may permit stacking analyses of weak-lensing signals around the cluster if shape measurements are added from other surveys.
- Extending the same classification pipeline to neighboring clusters would test whether the reported completeness and redshift precision generalize beyond Fornax.
- Dwarf galaxy populations at the catalog's faint end could be isolated to study environmental quenching at low stellar masses.
Load-bearing premise
The deep learning classifiers and machine-learning matching to SDSS data separate galaxies and recover stellar masses, star formation rates and D4000 without significant systematic biases from training-set mismatches or photometric errors.
What would settle it
A large independent spectroscopic sample showing systematic offsets in stellar masses or star formation rates larger than the quoted uncertainties, or a completeness drop well below 72 percent when compared to deeper imaging, would falsify the reliability of the derived catalog properties.
read the original abstract
Observational extragalactic catalogs over wide sky areas are essential for uncovering the large-scale structure of the Universe. They allow, among others, cosmological studies and density analyses that impose strong constraints on models of galaxy formation and evolution. By taking advantage of the wide field images and the 12 optical bands of the Southern Photometric Local Universe Survey (S-PLUS), we aim at providing a catalog of galaxies located, in projection, towards the Fornax galaxy cluster, within $\sim$ 5 virial radii in right ascension (R.A.) and $\sim$ 3 virial radius in declination (Dec) around NGC,1399, the dominant galaxy of the cluster. Such a catalog will allow unprecedented large-scale structure studies in that sky region. Supervised deep learning algorithms have been developed, utilizing neural networks complemented with dimensionality reduction techniques, to classify and separate spurious objects, stars and galaxies in a photometric catalog previously built for the S-PLUS Fornax Project (S+FP). That catalog was built using a combination of SExtractor configurations optimized for galaxy detection and characterization. A catalog of 119,580 galaxies was obtained in the direction of the Fornax cluster containing photometric information in the 12 optical bands of S-PLUS complemented with GALEX (UV), VHS-VISTA (NIR) and AllWISE (MIR) data. We estimate photometric redshifts ({\sigma}_ NMAD $\sim$ 0.0219) with a lower limit of z_ lim $\sim$ 0.03. Stellar masses, star formation rates (SFRs) and D4000_N index estimates were obtained through a machine learning approach, by matching S-PLUS photometric data to SDSS spectroscopic data. The completeness of the catalog (72%) was calculated by comparing with mock catalogs ...
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript presents the S-PLUS Fornax Project catalog containing 119580 galaxies projected within ~5 virial radii of NGC 1399. The catalog combines 12-band S-PLUS optical photometry with GALEX UV, VHS-VISTA NIR and AllWISE MIR data. Supervised deep-learning classifiers separate galaxies from stars and spurious objects; photometric redshifts are reported with σ_NMAD ~0.0219 and z_lim ~0.03; stellar masses, SFRs and D4000 indices are derived via machine-learning matching to SDSS spectroscopic data. Completeness is stated as 72% from mock-catalog comparisons.
Significance. If the reported photo-z scatter and completeness hold under full validation, the catalog supplies a multi-wavelength resource for large-scale structure and galaxy-evolution studies across an extended Fornax volume. The quantitative metrics (σ_NMAD, 72% completeness) directly support its use for density analyses and cosmological constraints.
major comments (1)
- The central claim that the ML-derived stellar masses, SFRs and D4000 values are reliable rests on the supervised matching to SDSS data, yet the manuscript provides no cross-validation scatter, bias statistics or training-set mismatch tests for these quantities. This omission is load-bearing because any systematic offset would propagate directly into the catalog's scientific utility.
Simulated Author's Rebuttal
We thank the referee for their constructive review and positive assessment of the S+FP catalog. We address the single major comment below and will revise the manuscript to incorporate the requested validation details.
read point-by-point responses
-
Referee: The central claim that the ML-derived stellar masses, SFRs and D4000 values are reliable rests on the supervised matching to SDSS data, yet the manuscript provides no cross-validation scatter, bias statistics or training-set mismatch tests for these quantities. This omission is load-bearing because any systematic offset would propagate directly into the catalog's scientific utility.
Authors: We agree that the current manuscript does not provide sufficient quantitative validation for the machine-learning derived stellar masses, SFRs and D4000 values. While the text describes the supervised matching of S-PLUS photometry to SDSS spectroscopic targets, it omits explicit cross-validation metrics, bias statistics, and tests for training-set mismatches. In the revised version we will add a new subsection (or expand the existing methods section) that reports: (i) the training/validation split and cross-validation procedure, (ii) scatter and bias statistics (e.g., median absolute deviation, mean bias, and outlier fraction) for each derived quantity, and (iii) any checks for systematic offsets arising from differences in filter sets or depth between S-PLUS and SDSS. These additions will directly support the reliability claims and strengthen the catalog's scientific utility. revision: yes
Circularity Check
No significant circularity in catalog construction
full rationale
The paper constructs an observational catalog of 119580 galaxies from S-PLUS wide-field imaging by applying SExtractor for detection, supervised deep learning (neural networks with dimensionality reduction) for star/galaxy/spurious classification, and standard machine-learning matching to external SDSS spectroscopic data for stellar masses, SFRs and D4000. Photometric redshifts are estimated with reported σ_NMAD ≈ 0.0219 and completeness (72%) is validated against independent mock catalogs. None of these steps reduce a claimed prediction to its own fitted inputs by construction, invoke self-citation load-bearing uniqueness theorems, or smuggle ansatzes; all quantities derive from direct photometry plus external reference data, rendering the derivation chain self-contained.
Axiom & Free-Parameter Ledger
axioms (1)
- domain assumption Supervised learning trained on SDSS spectra generalizes reliably to S-PLUS photometry for stellar mass, SFR, and D4000 estimation.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.