Pith. sign in

REVIEW 3 major objections 5 minor 70 references

Estimating Fe and Mg Abundances in the Milky Way Dwarf Galaxies Using Subaru/HSC and DEIMOS

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Broad- and narrow-band photometry, calibrated by 610 spectra, yields reliable iron and magnesium abundances for 6,713 giants in four Milky Way dwarf galaxies and exposes radial abundance gradients that spectroscopy alone could not reach.

desk verdict Practical photometric abundance maps for four dSphs, but the headline outer-region gradients rest on extrapolated ML predictions and need radius-stratified validation. read the letter →

arxiv 2506.15952 v1 pith:UTBJRRZ4 submitted 2025-06-19 astro-ph.GA

Jihye Hong , Evan N. Kirby , Tiffany M. Tang , Masashi Chiba , Yutaka Komiyama , Lauren E. Henderson , Itsuki Ogami , Timothy C. Beers This is my paper · ORCID
classification astro-ph.GA
keywords dwarfspheroidalgalaxiesphotometricmetallicityestimationnarrow-bandphotometryRandomForestregressorstellarabundancesradialabundancegradientsstarformationhistory
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that photometry alone, when calibrated by a modest set of spectra, can map the iron and magnesium content of thousands of individual giant stars in four Milky Way dwarf galaxies. Using Subaru/HSC images in the broad g and i bands plus a narrow band centered on the magnesium triplet, the authors train a Random Forest on 610 stars with Keck/DEIMOS spectra and apply it to 6,713 photometrically selected giants. The resulting abundance maps reach three to nine half-light radii, far beyond the spectroscopic footprint, and reveal radial metallicity gradients that earlier studies could not detect. These gradients, together with the shapes of the abundance distributions, are interpreted as evidence for different star formation histories: extended enrichment in Fornax, a shorter enrichment episode in Sculptor, and brief intense bursts in Ursa Minor and Draco. If the calibration holds, the method turns wide-field photometry into a chemical survey tool for the dwarf galaxy population.

What carries the argument

The central object is a Random Forest regressor, an ensemble of decision trees whose predictions are averaged across trees, trained separately for [Fe/H] and [Mg/H]. Its input features are the absolute magnitudes $M_{g,0}$, $M_{i,0}$, $M_{NB515,0}$ and the colors $(g-i)_0$, $(g-NB515)_0$, and $(NB515-i)_0$, where the narrow $NB515$ band covers the Mg b triplet and is sensitive to surface gravity, allowing giant members to be separated from foreground dwarf stars and carrying much of the magnesium information. The model's per-star uncertainty comes from the tree-to-tree spread, and per-star prediction errors come from out-of-bag leaf-node root-mean-square errors; the most important features are $(NB515-i)_0$ for [Fe/H] and $(g-NB515)_0$ for [Mg/H].

What would settle it

Obtain medium-resolution spectra of a few dozen outer-region giant candidates in Fornax or Sculptor at predicted [Fe/H] near -2 and compare the measured abundances with the photometric predictions; a systematic offset comparable to the +0.74 dex overestimate already seen for the most metal-poor training stars would indicate that the outer gradients are calibration artifacts rather than real abundance patterns.

Watch

Extended reading notes

Core claim

The central claim is that a Random Forest regressor trained on three HSC magnitudes and three color indices can predict spectroscopic-quality [Fe/H] and [Mg/H] for giants in Fornax, Sculptor, Ursa Minor, and Draco. Tested by five-fold cross-validation on 610 DEIMOS stars, the model achieves root-mean-square errors of 0.265 dex for [Fe/H] and 0.263 dex for [Mg/H], with mean out-of-bag errors of 0.289 and 0.306 dex. Applied to 6,713 HSC giants after rejecting twenty percent of candidates with large tree-to-tree model uncertainty, the predictions reproduce the central spectroscopic metallicity distribution functions and the [Mg/Fe] versus [Fe/H] patterns. The enlarged sample reveals negative radial abundance gradients in [Fe/H] and [Mg/H] extending out to roughly seven, nine, three, and four half-light radii for Fornax, Sculptor, Ursa Minor, and Draco respectively, with metal-poor and magnesium-rich stars increasingly common in the outskirts.

Load-bearing premise

The load-bearing premise is that a machine-learning model trained on 610 mostly central stars with measured spectra can be trusted to estimate iron and magnesium for all selected giants out to several half-light radii, including outer stars whose colors lie beyond the training distribution and whose ages may differ from the training stars.

Editorial extensions

If this is right

  • Radial abundance gradients in these dwarf galaxies can now be measured to several half-light radii, revealing that metal-poor, magnesium-rich stars increasingly dominate the outskirts.
  • The Fornax abundance distribution, with its extended metal-rich tail and a [Mg/Fe] bump near [Fe/H] = -1, supports a long, centrally concentrated star formation history with a gradual rise in Type Ia supernova enrichment.
  • Sculptor's bimodal metallicity distribution and its central metal-rich concentration support a short, intense star formation episode in the periphery followed by more prolonged central enrichment.
  • The shallow gradients of Ursa Minor and Draco are consistent with brief, intense star formation episodes that nearly extinguished star formation early, with Draco's slightly steeper gradient indicating a somewhat longer decline.
  • The metal-poor end of the photometric distributions is systematically overestimated by about 0.74 dex for [Fe/H] below -2.5, so the most metal-poor photometric tail must be interpreted with caution.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the calibration transfers, the same pipeline could be applied to other dwarf galaxies already imaged by HSC, producing homogeneous abundance maps across the entire Milky Way satellite system without new spectroscopy.
  • An untested but natural extension would be to break the age-metallicity degeneracy with isochrone-based stellar ages, which could separate first-generation populations from later enrichment episodes and directly test the bursty star formation interpretation.
  • A direct observational test would be to take medium-resolution spectra of outer-region giants in Fornax or Sculptor, where the predicted [Fe/H] values fall near -2, and compare the measured abundances with the photometric predictions; a systematic offset similar to the metal-poor training bias would indicate the outer gradients are calibration artifacts.
  • Adding a metallicity-sensitive filter such as NB395, which covers the Ca II H and K lines, would likely reduce the metal-poor overestimation and sharpen the inferred low-metallicity tails, a testable upgrade to the current method.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper trains a Random Forest regressor on 610 DEIMOS spectroscopic measurements of [Fe/H] and [Mg/H] matched to Subaru/HSC photometry (g, i, NB515) for four Milky Way dwarf spheroidal galaxies (Fornax, Sculptor, Ursa Minor, Draco), and applies it to 6,713 HSC-selected giant candidates. It reports radial gradients in [Fe/H], [Mg/H], and [Mg/Fe] extending to roughly 7 half-light radii in Fornax, and uses these gradients plus abundance distribution shapes to infer differing star formation histories: extended formation in Fornax, shorter enrichment in Sculptor, and brief intense bursts in Ursa Minor and Draco. The paper also compares the photometric abundance distributions and [Mg/Fe] trends with prior spectroscopic results for the central regions.

Significance. If the photometric abundance estimates are reliable, the paper would deliver a substantial enlargement of the sample of chemically characterized stars in classical dSphs, reaching radii previously inaccessible to spectroscopy. The use of the NB515 narrow-band filter to constrain Mg and surface gravity is well motivated, and the authors are honest about known biases, including metal-poor overestimation and regression toward the mean. Machine-checked aspects are limited, but the paper does provide quantitative test metrics, out-of-bag error estimates, and a model-uncertainty exclusion criterion. However, the headline scientific claim—the outer-region radial gradients—rests on applying a model trained on predominantly central spectroscopic stars to stars far beyond the training footprint, and this is not yet validated by radius-stratified tests. The central-region agreement with earlier spectroscopy is a genuine strength, but it does not by itself establish the extrapolated gradients.

major comments (3)
  1. [§2.2, §4, Figs. 10–12] The central claim of new radial gradients beyond the spectroscopic footprint is not adequately validated. The Random Forest is trained on 610 DEIMOS stars that are strongly concentrated in the central regions (Fig. 10), and the paper itself concedes in §4 that the outermost [Fe/H] values near −2 'lie beyond the coverage of the DEIMOS sample and may therefore result from extrapolation.' Since §3.1 reports a mean residual of +0.74 dex for the 15 OOB stars with [Fe/H] ≤ −2.5, a radial trend in this bias could produce the apparent gradients even if the underlying metallicity distribution were flat. The authors should add a radius-stratified validation: for example, train on central stars only and test on the relatively outer DEIMOS stars, or bin the OOB/test residuals by galactocentric radius and show that the bias does not grow outward. They should also show how the metal-poor bias, if applied to the outskirts, would change the slopes in Figs. 11–12.
  2. [Table 1, §3.1, Fig. 8] The predictive power for [Mg/H] is weak: the test-set R² is 0.497 and the RMSE is 0.263 dex. Since [Mg/Fe] is computed by subtracting two predicted quantities, its accuracy is not independently validated, and §3.1 documents a −0.315 dex bias for stars with [Mg/H] > −0.8. Given that parts of the discussion (e.g., the α-knee and SFH interpretations in §4) rely on [Mg/Fe] trends, the authors should either quantify how the measured biases propagate into the claimed [Mg/Fe] gradients or soften the conclusions until a validation of [Mg/Fe] gradient recovery is provided.
  3. [Figs. 11 and 12] The radial gradient slopes are presented without uncertainties. Statements such as 'the slope ... does not differ significantly from the slope measured from the center outward' (Fornax) and comparisons between galaxies (e.g., 'Dra ... falls between those of UMi and Scl') are not supported without error bars on the slopes. The authors should report slope uncertainties, ideally from bootstrap resampling that accounts for the finite size and correlated nature of the photometric samples, and use these to test whether the inner and outer slopes are actually consistent.
minor comments (5)
  1. [§2.2] The description of the model uncertainty cut ('samples with model uncertainty greater than the maximum model uncertainty observed in the training sample were excluded') means that the threshold is defined by the training set; this should be stated more explicitly because it affects the comparison of training and test estimates.
  2. [§5] The text 'more ten times larger than our spectroscopic dataset' should read 'more than ten times larger'; the typo appears in the summary.
  3. [§4, Fig. 9] The statement that predictions 'beyond the dashed lines ... continue to follow similar trends' is made without a quantitative comparison. A simple metric, such as median [Mg/Fe] or [Fe/H] inside versus outside the DEIMOS footprint in matched bins, would strengthen the claim.
  4. [§3.4, Table 2] Because the model compresses the abundance range, the mean and standard deviation of the HSC predictions in Table 2 cannot be interpreted as unbiased estimates of the underlying stellar population; the authors should explicitly caution that the quoted widths are affected by regression toward the mean, particularly for [Mg/H].
  5. [§1] The filter name is typeset variously as 'N B515', 'NB515', and 'N B515'; using a consistent notation (e.g., NB515) throughout would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the photometric abundance estimates and radial gradients are supervised outputs of an RF trained on DEIMOS abundances, but the gradients are not fitted targets and the paper's acknowledged outer-region extrapolation is a validity concern, not a definitional circularity.

full rationale

The derivation chain is an empirical calibration, not a self-referential loop. The RF regressors map HSC photometry (three magnitudes and three colors) to [Fe/H] and [Mg/H] using 610 DEIMOS spectroscopic labels; the 6713 HSC predictions are evaluated on photometry that was not used to set the abundance targets, and the radial gradients (Figs. 11-12) are least-squares slopes computed from those predicted abundances rather than quantities fitted to produce the gradients. No equation in the paper defines the predicted abundance in terms of the gradient, and no fitted parameter is renamed as a prediction. The paper explicitly flags the main threat in Section 4: photometric [Fe/H] near -2 in the outer regions 'lie beyond the coverage of the DEIMOS sample and may therefore result from extrapolation,' and Section 3.1 documents a +0.74 dex mean residual for the 15 out-of-bag stars with [Fe/H] <= -2.5. These are external-validity and extrapolation risks for the novel gradient claim, not circularity: the model output is not equivalent to its input by construction. Self-citations (Kirby et al. 2010, 2011a,b; Henderson et al. 2025; Komiyama et al. 2018b; Ogami et al. 2025) supply training spectra, selection methods, and comparison data, but none is invoked as a uniqueness theorem or as the sole justification for a conclusion; the central-region agreement is checked against independent spectroscopic studies. Accordingly, no circular step meets the evidentiary bar of exhibiting a specific reduction of a result to its inputs.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. Its central claim rests on a chain of domain assumptions: photometric filters map to abundances, spectroscopic labels are unbiased, the trained relation extrapolates to large radii, and age effects are negligible. The Random Forest itself is a flexible nonparametric fit with hyperparameters chosen by cross-validation, and sample membership depends on hand-chosen cuts. These are not free physical constants, but they do shape the result.

free parameters (4)
  • Random Forest hyperparameters (n_estimators, min_samples_leaf) = 500, 1
    Section 2.2; chosen by 5-fold cross-validation to minimize RMSE; they control smoothness of the abundance maps but are not physically motivated.
  • Model uncertainty exclusion threshold = maximum in-sample OOB model uncertainty, retaining 80% of HSC sample
    Section 2.2; 20% of the HSC sample is removed based on this threshold, changing sample composition and thus gradient estimates.
  • Proper motion sigma-clipping widths = sigma = 3 (Fnx), 2 (Scl), 1.5 (UMi, Dra)
    Section 2.1, Equation 1; these ellipse sizes determine member selection and differ per galaxy, affecting which stars enter the sample.
  • Color-color polygon boundaries = dashed polygons in Figure 1, different for Fornax
    Section 2.1; the polygon explicitly excludes a foreground feature for Fornax and is a hand-drawn selection boundary that affects the sample and possibly the measured gradients.
assumptions (5)
  • domain assumption The NB515 narrow-band filter measures the Mg b triplet and is sensitive to surface gravity, so color-color cuts separate red giant members from foreground dwarfs.
    Section 2.1; this is the basis for defining the giant candidate polygon and for interpreting NB515 as a magnesium abundance indicator.
  • domain assumption The spectroscopic DEIMOS abundances used as training labels are accurate and unbiased across the full metallicity range.
    Section 2.1; the Random Forest inherits any systematic errors in [Fe/H] and [Mg/H] from Kirby et al. (2010) and Henderson et al. (2025).
  • ad hoc to paper The photometric abundance relation learned by the Random Forest in the central, spectroscopically sampled regions holds in the outer regions beyond the DEIMOS footprint.
    Section 2.2 and Section 4, Figures 11-12; this is required for the central claim of extended radial gradients, and the paper only partially acknowledges the extrapolation risk.
  • domain assumption The age-metallicity degeneracy does not seriously bias photometric abundance predictions; combining all four galaxies without age information performed as well as age-dependent alternatives.
    Section 2.2; the authors tested Fornax-only and age-metallicity relation models but proceeded with the combined model.
  • domain assumption Stars in each galaxy share the same distance modulus from McConnachie (2012), and Gaia proper motion filtering removes foreground contamination.
    Section 2.1; absolute magnitudes and membership depend on these assumed distances and proper-motion cuts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Estimating Fe and Mg Abundances in the Milky Way Dwarf Galaxies Using Subaru/HSC and DEIMOS." pith.science (2026). https://pith.science/paper/UTBJRRZ4

@misc{pith2026250615952,
  author       = {Pith},
  title        = {Pith review of: Estimating Fe and Mg Abundances in the Milky Way Dwarf Galaxies Using Subaru/HSC and DEIMOS},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UTBJRRZ4}},
  note         = {Machine review of arXiv:2506.15952}
}
read the original abstract

We investigate the chemical abundance distributions of the Fornax, Sculptor, Ursa Minor, and Draco dwarf galaxies using Subaru/HSC photometric data. The HSC dataset, which includes broadband g and i filters and the narrowband NB515 filter, offers sensitivity to iron and magnesium abundances as well as surface gravity, enabling the identification of giant stars and foreground dwarfs. For analysis, we selected a total of 6713 giant candidates using a Random Forest regressor trained on medium-resolution (R ~ 6000) Keck/DEIMOS spectroscopic data. Our analysis reveals the extent of radial metallicity gradients in the galaxies. Such trends, not detectable in earlier studies, are now captured owing to the substantially enlarged sample size and areal coverage provided by the HSC data. These results are also consistent with chemical abundance patterns previously observed in the central regions through spectroscopic studies. Furthermore, we infer that Fornax underwent extended star formation, whereas Sculptor formed both metal-poor and metal-rich stars over a shorter time. Ursa Minor and Draco appear to have experienced brief, intense star formation episodes leading to nearly extinguished star formation. This study underscores the critical role of the expanded HSC dataset in revealing chemical gradients that were previously inaccessible. Future work incorporating additional spectra of metal-poor stars and age-sensitive isochrone modeling will enable more accurate maps of chemical abundance distributions.

Figures

Figures reproduced from arXiv: 2506.15952 by the authors.

Figure 1
Figure 1. Left: The equatorial coordinates in R.A. and Decl of the dwarf galaxies from the HSC catalog are represented by black points, corresponding to Fnx, Scl, UMi, and Dra from top to bottom, respectively. Middle: Proper motions in in µαcosδ and µδ for stars with Gaia DR3 astrometry. The green dot and the ellipse represent the mean proper motion and three times the standard deviation calculated via sigma clipping. Stars m… view at source ↗
Figure 2
Figure 2. Color-magnitude diagrams for Fnx, Scl, UMi, and Dra, shown from left to right. The gray dots represent the HSC photometry data, while the color-coded dots based on [Fe/H] represent the DEIMOS-HSC sample. The black solid lines indicate the faintest absolute magnitudes that define the final prediction and training samples. some samples, referred to as the OOB sample, from the train￾ing of that tree. Because these OOB … view at source ↗
Figure 3
Figure 3. Comparison of measured and predicted abundances for the OOB and test samples. Top: The upper panel shows the [Fe/H] predictions for the OOB sample, color-coded by g0. Each symbol corresponds to a galaxy: circles for Fnx, triangles for Scl, squares for UMi, and stars for Dra. Test sample results are shown in black. The number of samples and RMSE values from 5-fold CV are given at the top. The black dashed diagonal li… view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: The feature importance used in the abundance prediction for HSC sample. The unfilled bars represent [Fe/H] predictions, while the filled bars represent [Mg/H] [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Color-magnitude diagrams of the HSC data with Fnx, Scl, UMi, and Dra shown from left to right, color-coded by [Fe/H] predictions [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Top: Color-color diagrams of the DEIMOS data with Fnx, Scl, UMi, and Dra shown from left to right, color-coded by [Mg/H]. Bottom: Same as the top panel, but for [Mg/H] predictions of HSC data. et al. (2011) based on photometric data from Coleman et al. (2005) and Walke…
Figure 7
Figure 7. Figure 7: Top: The distribution of [Fe/H], [Mg/H], and [Mg/Fe] for the DEIMOS sample, shown from left to right. The combined distribution of all DEIMOS samples is shown as a filled gray histogram. The filled blue histogram represents the abundance prediction results of the test …
Figure 8
Figure 8. Figure 8: The [Fe/H] vs. [Mg/Fe] distributions for the DEIMOS and HSC samples. From left to right, the panels correspond to Fnx, Scl, UMi, and Dra. Black points represent the DEIMOS measurements, and black error bars indicate their representative mean errors derived from statist…
Figure 9
Figure 9. Figure 9: Observed trends of [Mg/Fe] vs. [Fe/H] for spectroscopy (DEIMOS, thin solid line) and photometry (HSC, thick bars). The trend of the predictions for DEIMOS-HSC stars in the DEIMOS spatial footprint is indicated by the dashed line. The samples are divided into 0.1 dex bi…
Figure 10
Figure 10. Figure 10: Top row: The spatial distributions of stars in Fnx, Scl, UMi, and Dra are shown from left to right. In each panel, four color-coded radial regions extend from the galaxy center to the outskirts. Each region corresponds to 0.5, 1.0, and 1.5 times the half-light radius …
Figure 11
Figure 11. Figure 11: Left: Radial distributions of the [Fe/H] predictions for the HSC sample. The black dots represent the DEIMOS sample. From top to bottom, the panels correspond to Fnx, Scl, UMi, and Dra, with the number of stars indicated in the top-right corner of each panel. The vert…
Figure 12
Figure 12. Figure 12: Same as [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 17 canonical work pages

  1. [1]

    O., Chiti, A., Limberg, G., et al

    Barbosa, F. O., Chiti, A., Limberg, G., et al. 2025, arXiv e-prints, arXiv:2504.03593, doi: 10.48550/arXiv.2504.03593

    arXiv doi:10.48550/arxiv.2504.03593 2025

  2. [2]

    2006, A&A, 459, 423, doi: 10.1051/0004-6361:20065720

    Battaglia, G., Tolstoy, E., Helmi, A., et al. 2006, A&A, 459, 423, doi: 10.1051/0004-6361:20065720

    doi:10.1051/0004-6361:20065720 2006

  3. [3]

    2014, arXiv e-prints, arXiv:1403.5237, doi: 10.48550/arXiv.1403.5237

    Benitez, N., Dupke, R., Moles, M., et al. 2014, arXiv e-prints, arXiv:1403.5237, doi: 10.48550/arXiv.1403.5237

    arXiv doi:10.48550/arxiv.1403.5237 2014

  4. [4]

    L., Cassisi, S., et al

    Bettinelli, M., Hidalgo, S. L., Cassisi, S., et al. 2019, MNRAS, 487, 5862, doi: 10.1093/mnras/stz1679

    work page doi:10.1093/mnras/stz1679 2019

  5. [5]

    2015, arXiv e-prints, arXiv:1511.05741, doi: 10.48550/arXiv.1511.05741

    Biau, G., & Scornet, E. 2015, arXiv e-prints, arXiv:1511.05741, doi: 10.48550/arXiv.1511.05741

    arXiv doi:10.48550/arxiv.1511.05741 2015

  6. [6]

    2001, Machine Learning, 45, 5, doi: 10.1023/A:1010933404324

    Breiman, L. 2001, Machine Learning, 45, 5, doi: 10.1023/A:1010933404324

    doi:10.1023/a:1010933404324 2001

  7. [7]

    Breiman, L., Friedman, J., Olshen, R., & Stone, C. J. 1984, Chapman and Hall/CRC

    work page 1984

  8. [8]

    J., Moles, M., Crist´obal-Hornillos, D., et al

    Cenarro, A. J., Moles, M., Crist´obal-Hornillos, D., et al. 2019, A&A, 622, A176, doi: 10.1051/0004-6361/201833036

    doi:10.1051/0004-6361/201833036 2019

Show all 70 references
  1. [9]

    G., & Huang, W

    Cohen, J. G., & Huang, W. 2009, ApJ, 701, 1053, doi: 10.1088/0004-637X/701/2/1053 —. 2010, ApJ, 719, 931, doi: 10.1088/0004-637X/719/1/931

    2009 doi

  2. [10]

    G., Da Costa, G

    Coleman, M. G., Da Costa, G. S., & Bland-Hawthorn, J. 2005, AJ, 130, 1065, doi: 10.1086/432662 de Boer, T. J. L., Tolstoy, E., Hill, V ., et al. 2012a, A&A, 539, A103, doi: 10.1051/0004-6361/201118378 —. 2012b, A&A, 544, A73, doi: 10.1051/0004-6361/201219547 de los Reyes, M. A...

    2005 doi

  3. [11]

    2013, The Messenger, 154, 32

    Edge, A., Sutherland, W., Kuijken, K., et al. 2013, The Messenger, 154, 32

    2013

  4. [12]

    P., & Sanders, J

    Fallows, C. P., & Sanders, J. L. 2022, MNRAS, 516, 5521, doi: 10.1093/mnras/stac2550 Ferreira Lopes, C. E., Guti´errez-Soto, L. A., Ferreira Alberice, V . S., et al. 2025, A&A, 693, A306, doi: 10.1051/0004-6361/202451491

    2022 doi

  5. [13]

    2018, PASJ, 70, S3, doi: 10.1093/pasj/psx079

    Furusawa, H., Koike, M., Takata, T., et al. 2018, PASJ, 70, S3, doi: 10.1093/pasj/psx079

    2018 doi

  6. [14]

    2020, ApJS, 247, 7, doi: 10.3847/1538-4365/ab6441

    Han, S.-I., Kim, H.-S., Yoon, S.-J., et al. 2020, ApJS, 247, 7, doi: 10.3847/1538-4365/ab6441

    2020 doi

  7. [15]

    R., Lian, J., et al

    Hasselquist, S., Hayes, C. R., Lian, J., et al. 2021, ApJ, 923, 172, doi: 10.3847/1538-4357/ac25f9

    2021 doi

  8. [16]

    J., Tibshirani, R

    Hastie, T. J., Tibshirani, R. J., & Friedman, J. H. 2009, Springer 16 HONG ET AL

    2009

  9. [17]

    E., Kirby, E

    Henderson, L. E., Kirby, E. N., de los Reyes, M. A. C., Gerasimov, R., & Manwadkar, V . 2025, ApJ, 983, 117, doi: 10.3847/1538-4357/adbe7d

    2025 doi

  10. [18]

    2014, A&A, 572, A82, doi: 10.1051/0004-6361/201424645

    Hendricks, B., Koch, A., Walker, M., et al. 2014, A&A, 572, A82, doi: 10.1051/0004-6361/201424645

    2014 doi

  11. [19]

    2019, A&A, 626, A15, doi: 10.1051/0004-6361/201833950

    Hill, V ., Sk´ulad´ottir, ´A., Tolstoy, E., et al. 2019, A&A, 626, A15, doi: 10.1051/0004-6361/201833950

    2019 doi

  12. [20]

    N., Chiba, M., et al

    Hirai, Y ., Kirby, E. N., Chiba, M., et al. 2024, ApJ, 970, 105, doi: 10.3847/1538-4357/ad500c

    2024 doi

  13. [21]

    C., Xiao, K., et al

    Huang, Y ., Beers, T. C., Xiao, K., et al. 2024, ApJ, 974, 192, doi: 10.3847/1538-4357/ad6b94

    2024 doi

  14. [22]

    2016, in Astronomical Society of the Pacific Conference Series, V ol

    Jin, S., Irwin, M., Tolstoy, E., Lewis, J., & Hartke, J. 2016, in Astronomical Society of the Pacific Conference Series, V ol. 507, Multi-Object Spectroscopy in the Next Decade: Big Questions, Large Surveys, and Wide Fields, ed. I. Skillen, M. Balcells, & S. Trager, 241

    2016

  15. [23]

    2018, PASJ, 70, 66, doi: 10.1093/pasj/psy056

    Kawanomoto, S., Uraguchi, F., Komiyama, Y ., et al. 2018, PASJ, 70, 66, doi: 10.1093/pasj/psy056

    2018 doi

  16. [24]

    N., Cohen, J

    Kirby, E. N., Cohen, J. G., Guhathakurta, P., et al. 2013, ApJ, 779, 102, doi: 10.1088/0004-637X/779/2/102

    2013 doi

  17. [25]

    N., Cohen, J

    Kirby, E. N., Cohen, J. G., Smith, G. H., et al. 2011a, ApJ, 727, 79, doi: 10.1088/0004-637X/727/2/79

    doi

  18. [26]

    2011b, ApJ, 727, 78, doi: 10.1088/0004-637X/727/2/78

    Guhathakurta, P. 2011b, ApJ, 727, 78, doi: 10.1088/0004-637X/727/2/78

    doi

  19. [27]

    N., Guhathakurta, P., Simon, J

    Kirby, E. N., Guhathakurta, P., Simon, J. D., et al. 2010, ApJS, 191, 352, doi: 10.1088/0067-0049/191/2/352

    2010 doi

  20. [28]

    2018a, PASJ, 70, S2, doi: 10.1093/pasj/psx069

    Komiyama, Y ., Obuchi, Y ., Nakaya, H., et al. 2018a, PASJ, 70, S2, doi: 10.1093/pasj/psx069

    doi

  21. [29]

    2018b, ApJ, 853, 29, doi: 10.3847/1538-4357/aaa129

    Komiyama, Y ., Chiba, M., Tanaka, M., et al. 2018b, ApJ, 853, 29, doi: 10.3847/1538-4357/aaa129

    doi

  22. [30]

    2019, A&A, 625, A2, doi: 10.1051/0004-6361/201834918

    Kuijken, K., Heymans, C., Dvornik, A., et al. 2019, A&A, 625, A2, doi: 10.1051/0004-6361/201834918

    2019 doi

  23. [31]

    Lemasle, B., de Boer, T. J. L., Hill, V ., et al. 2014, A&A, 572, A88, doi: 10.1051/0004-6361/201423919

    2014 doi

  24. [32]

    2010, A&A, 523, A17, doi: 10.1051/0004-6361/200913413

    Letarte, B., Hill, V ., Tolstoy, E., et al. 2010, A&A, 523, A17, doi: 10.1051/0004-6361/200913413

    2010 doi

  25. [33]

    K., & Koch, A

    Lianou, S., Grebel, E. K., & Koch, A. 2011, A&A, 531, A152, doi: 10.1051/0004-6361/201116998

    2011 doi

  26. [34]

    2021, Journal of Machine Learning Research, 22, 1

    Lu, B., & Hardin, J. 2021, Journal of Machine Learning Research, 22, 1

    2021

  27. [35]

    2024, A&A, 686, A266, doi: 10.1051/0004-6361/202348093

    Lucchesi, R., Jablonka, P., Sk´ulad´ottir, ´A., et al. 2024, A&A, 686, A266, doi: 10.1051/0004-6361/202348093

    2024 doi

  28. [36]

    R., Ostheimer, J

    Majewski, S. R., Ostheimer, J. C., Kunkel, W. E., & Patterson, R. J. 2000, AJ, 120, 2550, doi: 10.1086/316836

    2000 doi

  29. [37]

    2010, ApJ, 722, 1879, doi: 10.1088/0004-637X/722/2/1879

    Maoz, D., Sharon, K., & Gal-Yam, A. 2010, ApJ, 722, 1879, doi: 10.1088/0004-637X/722/2/1879

    2010 doi

  30. [38]

    2006, MNRAS, 371, 643, doi: 10.1111/j.1365-2966.2006.10671.x Mart´ınez-Delgado, D., Alonso-Garc´ıa, J., Aparicio, A., & G´omez-Flechoso, M

    Marcolini, A., D’Ercole, A., Brighenti, F., & Recchi, S. 2006, MNRAS, 371, 643, doi: 10.1111/j.1365-2966.2006.10671.x Mart´ınez-Delgado, D., Alonso-Garc´ıa, J., Aparicio, A., & G´omez-Flechoso, M. A. 2001, ApJL, 549, L63, doi: 10.1086/319150

    2006

  31. [39]

    C., Crain, R

    Mason, A. C., Crain, R. A., Schiavon, R. P., et al. 2024, MNRAS, 533, 184, doi: 10.1093/mnras/stae1743

    2024 doi

  32. [40]

    McConnachie, A. W. 2012, AJ, 144, 4, doi: 10.1088/0004-6256/144/1/4

    2012 doi

  33. [41]

    W., & Venn, K

    McConnachie, A. W., & Venn, K. A. 2020, AJ, 160, 124, doi: 10.3847/1538-3881/aba4ab

    2020 doi

  34. [42]

    1997, ARA&A, 35, 503, doi: 10.1146/annurev.astro.35.1.503 Mendes de Oliveira, C., Ribeiro, T., Schoenell, W., et al

    McWilliam, A. 1997, ARA&A, 35, 503, doi: 10.1146/annurev.astro.35.1.503 Mendes de Oliveira, C., Ribeiro, T., Schoenell, W., et al. 2019, MNRAS, 489, 241, doi: 10.1093/mnras/stz1985

    1997 doi

  35. [43]

    2012, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference

    Miyazaki, S., Komiyama, Y ., Nakaya, H., et al. 2012, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference

    2012

  36. [44]

    8446, Ground-based and Airborne Instrumentation for Astronomy IV , ed

    Series, V ol. 8446, Ground-based and Airborne Instrumentation for Astronomy IV , ed. I. S. McLean, S. K. Ramsay, & H. Takami, 84460Z, doi: 10.1117/12.926844

    doi

  37. [45]

    2018, PASJ, 70, S1, doi: 10.1093/pasj/psx063

    Miyazaki, S., Komiyama, Y ., Kawanomoto, S., et al. 2018, PASJ, 70, S1, doi: 10.1093/pasj/psx063

    2018 doi

  38. [46]

    2020, ApJ, 905, 109, doi: 10.3847/1538-4357/abc486

    Miyoshi, T., & Chiba, M. 2020, ApJ, 905, 109, doi: 10.3847/1538-4357/abc486

    2020 doi

  39. [47]

    2024, ApJ, 971, 107, doi: 10.3847/1538-4357/ad5445

    Ogami, I., Komiyama, Y ., Chiba, M., et al. 2024, ApJ, 971, 107, doi: 10.3847/1538-4357/ad5445

    2024 doi

  40. [48]

    2025, MNRAS, 536, 530, doi: 10.1093/mnras/stae2527 ¨Ohman, Y

    Ogami, I., Tanaka, M., Komiyama, Y ., et al. 2025, MNRAS, 536, 530, doi: 10.1093/mnras/stae2527 ¨Ohman, Y . 1934, ApJ, 80, 171, doi: 10.1086/143595

    2025 doi

  41. [49]

    Okamoto, S., Ferguson, A. M. N., Arimoto, N., et al. 2024, ApJL, 967, L24, doi: 10.3847/2041-8213/ad4358 OpenAI,. 2024, ChatGPT-4o (May 13 version) [Large Language Model], https://chat.openai.com/chat

    2024 doi

  42. [50]

    B., Erkal, D., & Li, T

    Pace, A. B., Erkal, D., & Li, T. S. 2022, ApJ, 940, 136, doi: 10.3847/1538-4357/ac997b

    2022 doi

  43. [51]

    B., Kaplinghat, M., Kirby, E., et al

    Pace, A. B., Kaplinghat, M., Kirby, E., et al. 2020, MNRAS, 495, 3022, doi: 10.1093/mnras/staa1419

    2020 doi

  44. [52]

    J., Hanke, M., et al

    Reichert, M., Hansen, C. J., Hanke, M., et al. 2020, A&A, 641, A127, doi: 10.1051/0004-6361/201936930

    2020 doi

  45. [53]

    F., Meisner, A

    Schlafly, E. F., Meisner, A. M., & Green, G. M. 2019, ApJS, 240, 30, doi: 10.3847/1538-4365/aafbea Sch¨olkopf, B., Smola, A., & M¨uller, K.-R. 1998, Neural Computation, 10, 1299, doi: 10.1162/089976698300017467

    2019 doi

  46. [54]

    A., et al

    Sestito, F., Zaremba, D., Venn, K. A., et al. 2023, MNRAS, 525, 2875, doi: 10.1093/mnras/stad2427

    2023 doi

  47. [55]

    D., Cˆot´e, P., & Sargent, W

    Shetrone, M. D., Cˆot´e, P., & Sargent, W. L. W. 2001, ApJ, 548, 592, doi: 10.1086/319022

    2001 doi

  48. [56]

    F., Cutri, R

    Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163, doi: 10.1086/498708 Sk´ulad´ottir, ´A., Vanni, I., Salvadori, S., & Lucchesi, R. 2024, A&A, 681, A44, doi: 10.1051/0004-6361/202346231 ESTIMATING PHOTOMETRICFE ANDMG OF DWARF GALAXIES17 Sk´ulad´ottir, ´A...

    2006 doi

  49. [57]

    2010, A&A, 513, A34, doi: 10.1051/0004-6361/200913759

    Starkenburg, E., Hill, V ., Tolstoy, E., et al. 2010, A&A, 513, A34, doi: 10.1051/0004-6361/200913759

    2010 doi

  50. [58]

    2025, Research in Astronomy and Astrophysics, 25, 035001, doi: 10.1088/1674-4527/adb55b

    Sun, B.-K., Chen, B.-Q., & Liu, X.-W. 2025, Research in Astronomy and Astrophysics, 25, 035001, doi: 10.1088/1674-4527/adb55b

    2025 doi

  51. [59]

    S., Chiba, M., et al

    Takada, M., Ellis, R. S., Chiba, M., et al. 2014, PASJ, 66, R1, doi: 10.1093/pasj/pst019

    2014 doi

  52. [60]

    2023, A&A, 669, A125, doi: 10.1051/0004-6361/202244052

    Tang, B., Zhang, J., Yan, Z., et al. 2023, A&A, 669, A125, doi: 10.1051/0004-6361/202244052

    2023 doi

  53. [61]

    Thackeray, A. D. 1939, MNRAS, 99, 492, doi: 10.1093/mnras/99.6.492

    1939 doi

  54. [62]

    2009, ARA&A, 47, 371, doi: 10.1146/annurev-astro-082708-101650

    Tolstoy, E., Hill, V ., & Tosi, M. 2009, ARA&A, 47, 371, doi: 10.1146/annurev-astro-082708-101650

    2009 doi

  55. [63]

    2023, A&A, 675, A49, doi: 10.1051/0004-6361/202245717

    Tolstoy, E., Sk´ulad´ottir, ´A., Battaglia, G., et al. 2023, A&A, 675, A49, doi: 10.1051/0004-6361/202245717

    2023 doi

  56. [64]

    2015, MNRAS, 449, 761, doi: 10.1093/mnras/stv294

    Ural, U., Cescutti, G., Koch, A., et al. 2015, MNRAS, 449, 761, doi: 10.1093/mnras/stv294

    2015 doi

  57. [65]

    Venn, K. A. 2014, Mem. Soc. Astron. Italiana, 85, 519

    2014

  58. [66]

    G., Mateo, M., & Olszewski, E

    Walker, M. G., Mateo, M., & Olszewski, E. W. 2009, AJ, 137, 3100, doi: 10.1088/0004-6256/137/2/3100

    2009 doi

  59. [67]

    G., Mateo, M., Olszewski, E

    Walker, M. G., Mateo, M., Olszewski, E. W., et al. 2007, ApJL, 667, L53, doi: 10.1086/521998

    2007 doi

  60. [68]

    R., Dolphin, A

    Weisz, D. R., Dolphin, A. E., Skillman, E. D., et al. 2014, ApJ, 789, 147, doi: 10.1088/0004-637X/789/2/147

    2014 doi

  61. [69]

    L., Eisenhardt, P

    Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010, AJ, 140, 1868, doi: 10.1088/0004-6256/140/6/1868

    2010 doi

  62. [70]

    2022, A&A, 659, A181, doi: 10.1051/0004-6361/202142724

    Yang, L., Yuan, H., Xiang, M., et al. 2022, A&A, 659, A181, doi: 10.1051/0004-6361/202142724

    2022 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.