Independent Recovery of Vanishing Sources on POSS-I Photographic Plates Using Automated Source Detection and Cross-Epoch Matching

Zachary Hayes

arxiv: 2604.04810 · v1 · submitted 2026-04-06 · 🌌 astro-ph.IM · astro-ph.SR

Independent Recovery of Vanishing Sources on POSS-I Photographic Plates Using Automated Source Detection and Cross-Epoch Matching

Zachary Hayes This is my paper

Pith reviewed 2026-05-10 18:55 UTC · model grok-4.3

classification 🌌 astro-ph.IM astro-ph.SR

keywords vanished sourcesPOSS-I platessource detectioncross-epoch matchingphotographic platesastronomical transientssky surveysreplication

0 comments

The pith

An independent pipeline recovers most reported vanished sources from POSS-I plates but detects no significant temporal clustering.

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

The paper builds an automated pipeline to detect candidate vanished sources on digitized POSS-I Red photographic plates from the early 1950s. It combines PSF filtering of detected sources, local astrometric refinement, and cross-epoch matching against later blue plates and modern catalogs to filter out persistent objects. Benchmark tests on known cases recover eight of nine sources in one field and all three in another, with a low false-positive rate on control fields. A full scan across the survey footprint produces a catalog of 2.85 million candidates after additional cuts. Statistical tests for day-specific clustering in the 1949-1957 observation window return null results, leaving the temporal pattern inconclusive while the catalog-level replication holds.

Core claim

The pipeline recovers 8/9 sources in the April 1950 benchmark field and 3/3 in the July 1952 field. It matches 3450 of the 5399 entries in the published Solano et al. catalog (63.9 percent) at a median separation of 0.94 arcsec. A full-footprint search over POSS-I coverage yields 2.85 million candidates after PSF cuts, deduplication, and Pan-STARRS DR1 rejection. Application of day-window tests across the 368 observation nights produces a post-test relative risk of 1.35 that is not statistically significant, and a negative binomial model of nightly counts likewise shows no effect.

What carries the argument

The automated detection pipeline that performs PSF-filtered source detection on POSS-I Red DSS cutouts, applies local astrometric registration refinement, performs cross-epoch matching to POSS-I Blue and POSS-II Red plates, and rejects candidates using Pan-STARRS DR1.

If this is right

The pipeline reproduces 63.9 percent of the prior catalog entries with sub-arcsecond positional agreement.
A large, filtered catalog of 2.85 million candidates is generated for further investigation after post-processing.
No statistically significant excess of candidates appears on any particular calendar day within the 1949-1957 interval.
The false-positive rate remains low at approximately 0.2 per 10-arcmin field in non-crowded control regions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The recovered candidates could be followed up with targeted high-resolution imaging to check for reappearance or other properties.
Larger samples or alternative statistical models might be needed to detect any subtle day-of-year pattern if one exists.
The same cross-epoch rejection approach could be extended to additional historical plate archives to expand the search for vanished sources.

Load-bearing premise

That PSF filtering, local astrometric refinement, and Pan-STARRS DR1 rejection correctly separate true vanished sources from artifacts or misdetections without systematic bias across the full survey footprint and varying plate conditions.

What would settle it

Detection of faint counterparts to most of the 2.85 million candidates in deeper modern surveys such as Gaia or LSST within 3 arcsec would indicate they are persistent rather than vanished.

Figures

Figures reproduced from arXiv: 2604.04810 by Zachary Hayes.

**Figure 3.** Figure 3: Distribution of angular separations between independently detected Vela sources and their nearest Solano et al. (2022) counterpart. The median separation of 0.94 arcsec reflects the combined astrometric precision of both catalogs. • Approximately 54% (∼1,010) are detected on the plate in the raw catalog but rejected during PSF morphology filtering—they appear as real sources but do not meet our point-so… view at source ↗

**Figure 1.** Figure 1: April 1950 benchmark field. Left: POSS-I Red with 8 detected transient candidates circled (of 9 known). Center: POSS-I Blue (same date)—no candidates detected. Right: POSS-II Red (∼1996)—no counterparts at candidate positions [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** Figure 2: July 1952 benchmark field. All 3 known transients from Solano et al. (2024) recovered. Same panel layout as [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 4.** Figure 4: Classification of all 5,399 Solano et al. (2022) vanishing-source candidates by independent detection status. 3,450 are independently recovered; ∼1,010 of the unrecovered sources are detected in our raw catalog but rejected by PSF filtering; ∼870 show no detection above 5σ; 66 fall outside survey coverage. None of the unrecovered sources have a Pan-STARRS DR1 counterpart within 3 arcsec. dian, ellipticity … view at source ↗

**Figure 5.** Figure 5: Left: Bruehl-style calendar-day summary of study days with at least one candidate vanished source, binned by relation to nuclear test dates. The [+1] window is elevated relative to the baseline (17.9% vs 13.3%; RR = 1.35), but the effect is not statistically significant and is schedule-sensitive because all 368 study-window observation nights contain at least one candidate. Right: published Bruehl & Villar… view at source ↗

read the original abstract

We present an independent pipeline for detecting candidate vanished sources on digitized first-epoch Palomar Observatory Sky Survey (POSS-I) photographic plates. The pipeline detects and PSF-filters sources on POSS-I Red DSS cutouts, applies local astrometric registration refinement, and identifies candidates by cross-epoch matching against POSS-I Blue and POSS-II Red with Pan-STARRS DR1 rejection. On a 20-case benchmark harness, the pipeline recovers 8/9 sources in the April 1950 field and 3/3 in the July 1952 field, with a false positive rate of about 0.2 per 10 arcmin field on random non-crowded controls. A full-footprint sweep over the POSS-I coverage using 30 arcmin patches yields a filtered catalog of 2.85 million candidate vanished sources after post-processing PSF cuts, deduplication, and Pan-STARRS DR1 rejection. Cross-matching against the published 5,399-source Solano et al. (2022) catalog yields 3,450 matches (63.9%) with median separation 0.94 arcsec; among unrecovered catalog entries within our footprint, we find no Pan-STARRS DR1 counterpart within 3 arcsec. Applying Bruehl and Villarroel (2025)-style temporal windows to this catalog over the 368 POSS-I observation nights in the 1949-1957 interval gives a post-test calendar-day relative risk of 1.35 for the +1 day window, but the effect is not statistically significant (95% CI 0.91-2.00; two-sided Fisher p = 0.17) and is sensitive to coding unobserved days as zero-transient days. A negative binomial model of nightly candidate counts with nightly patch coverage as exposure is likewise null (IRR = 1.03, 95% CI 0.89-1.18, p = 0.71). The catalog-level replication is strong; the temporal association remains inconclusive.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper delivers a new 2.85-million-candidate catalog from a full-footprint POSS-I pipeline with decent replication of prior work, but the purity claim rests on thin validation.

read the letter

The main takeaway is a new automated pipeline that processes the entire POSS-I coverage and produces a catalog of 2.85 million candidate vanished sources, along with a null result on temporal clustering using the same style of test as earlier papers. The catalog overlaps 63.9% with the Solano et al. list at sub-arcsecond separations, which is a concrete replication point. Recovery on the 20-case benchmark hits 8/9 and 3/3 on the two test fields, and the false-positive rate is quantified at roughly 0.2 per 10 arcmin on non-crowded controls. The statistical checks include confidence intervals and p-values on both the calendar-day risk and the negative binomial model of nightly counts, so the null findings are reported with some rigor. That is the useful output: a large independent list plus explicit numbers on how it lines up with previous efforts. The methods are described at a level that lets you see the steps—PSF filtering, local astrometric refinement, Pan-STARRS rejection, and deduplication. The work is therefore straightforward to reproduce or extend if someone wants to apply similar cuts elsewhere. The soft spot is the validation scope. The benchmark and false-positive measurements stay limited to non-crowded random fields and only 20 known cases across two dates. When the same cuts run over the full survey, plate quality, crowding, and artifact rates vary, yet no injection-recovery test or density-stratified control is shown. If the effective false-positive rate climbs even a little in typical regions, the 2.85 million entries could contain a much higher fraction of artifacts while still producing the observed overlap with the smaller prior catalog. Post-processing cut details and the full error budget are also not fully visible from the abstract. This paper is for researchers who mine historical plates or build cross-epoch catalogs for transient searches. Readers who need a large, independently generated list with basic replication stats will find it usable as a starting point or comparison set. It deserves a serious referee because the quantitative replication and statistical tests give something concrete to evaluate, even if the purity across the whole footprint needs more work. I would send it to peer review and ask the authors to add crowding-stratified controls or injection tests before final acceptance.

Referee Report

2 major / 3 minor

Summary. The paper presents an independent automated pipeline for detecting candidate vanished sources on digitized POSS-I Red photographic plates. The method performs source detection and PSF filtering on DSS cutouts, applies local astrometric registration refinement, and identifies candidates via cross-epoch matching to POSS-I Blue and POSS-II Red plates followed by Pan-STARRS DR1 rejection. On a 20-case benchmark, it recovers 8/9 sources in the April 1950 field and 3/3 in the July 1952 field with a reported false-positive rate of ~0.2 per 10 arcmin on non-crowded controls. A full-footprint run over 30-arcmin patches produces a filtered catalog of 2.85 million candidates. This catalog overlaps 63.9% (3,450 matches, median separation 0.94 arcsec) with the Solano et al. (2022) catalog of 5,399 sources. Temporal association tests using Bruehl & Villarroel-style windows over 368 nights yield a non-significant relative risk of 1.35 (p=0.17) and a null negative binomial result (IRR=1.03, p=0.71).

Significance. If the pipeline's purity and completeness hold across the full survey, the 2.85M-candidate catalog would constitute a substantial independent resource for archival transient studies, enabling statistical analyses of source disappearance rates in the 1949-1957 era. The 63.9% replication of the prior Solano catalog is a clear strength, demonstrating consistency between independent detection methods. However, the temporal clustering test remains inconclusive, limiting immediate astrophysical conclusions. The work advances automated archival searches but its impact depends on robust validation of the large catalog.

major comments (2)

[Benchmark harness and full-footprint sweep] Benchmark harness and false-positive controls: The recovery rates (8/9 and 3/3) and FP rate (~0.2 per 10 arcmin) are derived from only 20 cases on two specific dates using non-crowded random controls. The full 2.85M catalog is generated across the heterogeneous POSS-I footprint (varying plate quality, stellar density, and artifact rates) without reported injection-recovery tests, crowding-stratified FP measurements, or plate-quality stratification. This gap directly affects whether the reported purity can be extrapolated to the full catalog.
[Catalog generation and cross-matching] Pan-STARRS DR1 rejection and post-processing: The pipeline relies on Pan-STARRS DR1 rejection plus PSF cuts and deduplication to produce the 2.85M catalog, yet no quantitative assessment of completeness, false-negative rate, or systematic bias in rejection across the survey is provided. The note that unrecovered Solano entries lack Pan-STARRS counterparts within 3 arcsec is useful but insufficient without an error budget or simulation to confirm the rejection step does not preferentially remove real transients.

minor comments (3)

[Methods and abstract] The abstract and methods would benefit from explicit listing of the exact PSF cut thresholds, local astrometric refinement tolerances, and deduplication criteria rather than referring to them generically as 'post-processing cuts'.
[Benchmark results] Consider adding a table or figure summarizing the 20-case benchmark recoveries, including individual source properties, separations, and reasons for the single non-recovery.
[Temporal association test] The temporal analysis correctly reports non-significance and sensitivity to day coding, but a brief power calculation for the negative binomial model would help readers interpret the null result.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback, which highlights both the potential utility of our 2.85M-candidate catalog and the need for clearer discussion of validation limits. We address each major comment below with clarifications on our approach and scope. Revisions to the manuscript will incorporate expanded caveats and a dedicated limitations section to better contextualize the benchmark and rejection steps.

read point-by-point responses

Referee: Benchmark harness and false-positive controls: The recovery rates (8/9 and 3/3) and FP rate (~0.2 per 10 arcmin) are derived from only 20 cases on two specific dates using non-crowded random controls. The full 2.85M catalog is generated across the heterogeneous POSS-I footprint (varying plate quality, stellar density, and artifact rates) without reported injection-recovery tests, crowding-stratified FP measurements, or plate-quality stratification. This gap directly affects whether the reported purity can be extrapolated to the full catalog.

Authors: We acknowledge that the 20-case benchmark on two fields with non-crowded controls provides only a baseline demonstration rather than a comprehensive purity assessment across the full heterogeneous footprint. Our primary validation metric remains the 63.9% overlap with the independent Solano et al. (2022) catalog, which offers external consistency without relying solely on our internal FP estimates. The reported FP rate serves as a conservative lower bound for non-crowded regions. In the revised manuscript we will add a 'Limitations and Future Work' section explicitly discussing the absence of injection-recovery tests and the potential impact of crowding and plate quality variations, while noting that the catalog is presented as a resource for further statistical studies rather than a purity-guaranteed list. revision: partial
Referee: Pan-STARRS DR1 rejection and post-processing: The pipeline relies on Pan-STARRS DR1 rejection plus PSF cuts and deduplication to produce the 2.85M catalog, yet no quantitative assessment of completeness, false-negative rate, or systematic bias in rejection across the survey is provided. The note that unrecovered Solano entries lack Pan-STARRS counterparts within 3 arcsec is useful but insufficient without an error budget or simulation to confirm the rejection step does not preferentially remove real transients.

Authors: The 3-arcsec Pan-STARRS DR1 rejection radius is chosen to accommodate known astrometric uncertainties in the digitized POSS-I plates while removing sources that remain detectable in modern surveys. The observation that unrecovered Solano entries also lack PS counterparts within this radius supports that the step is not selectively discarding real transients from the prior catalog. We agree that without dedicated injection simulations we cannot provide a full error budget or false-negative quantification. The revised manuscript will expand the methods and discussion sections to justify the radius choice, report the fraction of Solano sources rejected by this criterion, and include a paragraph on possible biases, while emphasizing that the approach is intentionally conservative to minimize contamination from persistent sources. revision: partial

Circularity Check

0 steps flagged

No significant circularity; pipeline and catalog rely on external data and independent processing

full rationale

The derivation chain consists of source detection on POSS-I plates, PSF filtering, local astrometric refinement, cross-epoch matching, and rejection against Pan-STARRS DR1 plus the external Solano et al. (2022) catalog. Benchmark recovery rates and the 2.85M catalog are generated from these steps without fitting parameters to the target outputs or redefining quantities in terms of themselves. The temporal window analysis applies a cited external style to newly produced data and yields a null result; no self-citation chains, ansatz smuggling, or fitted-input-as-prediction reductions are present. The central claims remain self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work rests on standard domain assumptions in astronomical image processing and catalog cross-matching with no new postulated entities or heavily fitted parameters visible in the abstract.

axioms (1)

domain assumption Sources absent from Pan-STARRS DR1 within the matching radius are either truly vanished or artifacts from the older plates
Used as the rejection criterion for candidate selection.

pith-pipeline@v0.9.0 · 5681 in / 1448 out tokens · 51327 ms · 2026-05-10T18:55:53.951099+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

9 extracted references · 9 canonical work pages

[1]

2025, Scientific Reports, 15, 34125, doi: 10.1038/s41598-025-21620-3

Bruehl, J., & Villarroel, B. 2025, Scientific Reports, 15, 34125, doi: 10.1038/s41598-025-21620-3

work page doi:10.1038/s41598-025-21620-3 2025
[2]

C., Magnier, E

Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, arXiv e-prints

work page 2016
[3]

M., Lattanzi, M

Lasker, B. M., Lattanzi, M. G., McLean, B. J., et al. 2008, The Astronomical Journal, 136, 735, doi: 10.1088/0004-6256/136/2/735

work page doi:10.1088/0004-6256/136/2/735 2008
[4]

2022, Monthly Notices of the Royal Astronomical Society, 515, 1380, doi: 10.1093/mnras/stac1880

Solano, E., Garc´ ıa-Alvarez, D., Marcos-Arenal, P., et al. 2022, Monthly Notices of the Royal Astronomical Society, 515, 1380, doi: 10.1093/mnras/stac1880

work page doi:10.1093/mnras/stac1880 2022
[5]

2024, Monthly Notices of the Royal Astronomical Society, 527, 6312, doi: 10.1093/mnras/stad3726

Solano, E., et al. 2024, Monthly Notices of the Royal Astronomical Society, 527, 6312, doi: 10.1093/mnras/stad3726

work page doi:10.1093/mnras/stad3726 2024
[6]

https: //ui.adsabs.harvard.edu/abs/2021NatSR..1112794V

Villarroel, B., Imaz, I., Joshi, J., et al. 2021, Scientific Reports, 11, 12794, doi: 10.1038/s41598-021-92162-7

work page doi:10.1038/s41598-021-92162-7 2021
[7]

Villarroel, B., Solano, E., & Marcy, G. W. 2025, arXiv e-prints

work page 2025
[8]

A., et al

Villarroel, B., Solano, E., Mattsson, L., et al. 2020, The Astronomical Journal, 159, 8, doi: 10.3847/1538-3881/ab570f

work page doi:10.3847/1538-3881/ab570f 2020
[9]

2026, arXiv e-prints

Villarroel, B., Streblyanska, A., Bruehl, S., & Geier, S. 2026, arXiv e-prints

work page 2026

[1] [1]

2025, Scientific Reports, 15, 34125, doi: 10.1038/s41598-025-21620-3

Bruehl, J., & Villarroel, B. 2025, Scientific Reports, 15, 34125, doi: 10.1038/s41598-025-21620-3

work page doi:10.1038/s41598-025-21620-3 2025

[2] [2]

C., Magnier, E

Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, arXiv e-prints

work page 2016

[3] [3]

M., Lattanzi, M

Lasker, B. M., Lattanzi, M. G., McLean, B. J., et al. 2008, The Astronomical Journal, 136, 735, doi: 10.1088/0004-6256/136/2/735

work page doi:10.1088/0004-6256/136/2/735 2008

[4] [4]

2022, Monthly Notices of the Royal Astronomical Society, 515, 1380, doi: 10.1093/mnras/stac1880

Solano, E., Garc´ ıa-Alvarez, D., Marcos-Arenal, P., et al. 2022, Monthly Notices of the Royal Astronomical Society, 515, 1380, doi: 10.1093/mnras/stac1880

work page doi:10.1093/mnras/stac1880 2022

[5] [5]

2024, Monthly Notices of the Royal Astronomical Society, 527, 6312, doi: 10.1093/mnras/stad3726

Solano, E., et al. 2024, Monthly Notices of the Royal Astronomical Society, 527, 6312, doi: 10.1093/mnras/stad3726

work page doi:10.1093/mnras/stad3726 2024

[6] [6]

https: //ui.adsabs.harvard.edu/abs/2021NatSR..1112794V

Villarroel, B., Imaz, I., Joshi, J., et al. 2021, Scientific Reports, 11, 12794, doi: 10.1038/s41598-021-92162-7

work page doi:10.1038/s41598-021-92162-7 2021

[7] [7]

Villarroel, B., Solano, E., & Marcy, G. W. 2025, arXiv e-prints

work page 2025

[8] [8]

A., et al

Villarroel, B., Solano, E., Mattsson, L., et al. 2020, The Astronomical Journal, 159, 8, doi: 10.3847/1538-3881/ab570f

work page doi:10.3847/1538-3881/ab570f 2020

[9] [9]

2026, arXiv e-prints

Villarroel, B., Streblyanska, A., Bruehl, S., & Geier, S. 2026, arXiv e-prints

work page 2026