Transients in the Palomar Observatory Sky Survey

[flarkey]

Extending this line of thought, Villarreal specifically suggests these were objects in geosynchronous orbits. That should constrain the potential change in position for a given "transient" between the times of the red and blue plates making a search for an alien artifact in both sets of plates quite feasible.

I think the current hypothesis is that the transients are very bright and very short in duration so unless the two plates were used for imaging at the same time then it would be very unlikely for a specific 'transients' to appear on both.

 

[jarlrmai]

Did we/they ever do a minimum magnitude/second calculation to see the lower limit of the how bright and short they needed to be given the sensitivity/reciprocity of the film?

Is that even possible?

 

[flarkey]

@eburacom 's post #229 had some musings about the brightness of the glints in comparison to stars...

It does this by saying that star images on a 50 minute exposure would be fuzzy, but a transient lasting a fraction of a second would be sharp. Yes, but so would a plate anomaly. So Handley and Blair's criticism is not refuted; it remains the case that these anomalies are overwhelmingly more likely to be flaws in the plate rather than alien satellites. To quote the PASP paper:

Note as well that any transient longer than a second or so would show a line, not a dot - and all these lines would be precisely aligned with the rotation of the telescope as it compensates for Earth's rotation. But there Is no evidence for such extended lines. So somehow all these NTA satellites are specifically configured to only glint for a fraction of a second. These satellites must be very strange shapes, too; they have tiny reflective surfaces that are only visible for a fraction of a second, but rarely rotate so that the glints are visible at regularly spaced intervals. One imagines a large black sphere with a tiny hand mirror glued to the side.

Any glint from a real object that lasted 0.5 seconds (the figure suggested in the paper) would be reduced in apparent brightness by a factor of 6000 (0.5 seconds over 50 minutes). This would result in a decrease in the apparent brightness on the plate by a factor of 10.4 magnitudes, not 9 magnitudes as stated in the paper.

As well as these incredibly dim anomalies that last a fraction of a second, we should also see a selection of longer transients, 1 or more seconds in length, which would show an easily recognisable signature of elongation parallel to the Earth's rotation. But we don't.

Not to mention glints from satellites in much closer orbits - if you want to observe events on the surface of a planet, a network of spy satellites in low Earth orbit would give a much clearer view than a grid thirty-five thousand kilometres away.

 

[Eburacum]

Extending this line of thought, Villarreal specifically suggests these were objects in geosynchronous orbits. That should constrain the potential change in position for a given "transient" between the times of the red and blue plates making a search for an alien artifact in both sets of plates quite feasible.

Not really. Geosynchronous satellites aren't just constrained to the Earth's equator; some range quite far north and south. A cloud of alien geosynchronous satellites might range all over the place, so it would be difficult to identify transients caused by the same object.

More uncertainty.

 

[HoaxEye]

I've completed the Vasco60 pipeline.

Final release version #1 contains just a bunch of result files: https://github.com/jannefi/vasco60/tree/main/releases/final_release_v1/report

Survivors list (441 rows): https://github.com/jannefi/vasco60/blob/main/releases/final_release_v1/report/survivors.csv Using all veto-criteria, including the experimental ones inspired by the work of @Ivo Busko

So where are the hundred thousand or more survivors? Or 5,000? With this criteria, it's not possible to get that many survivors. But it's clear that even after extremely hard vetoes, a handful of survivors remain. They can be plate defects or other artefacts (e.g. aureoles, internal reflections, halation, ghost images). They can be transients. Note: I have not done any visual vetting.

Largest amount of survivor candidates from my pipeline is 15,303: https://github.com/jannefi/vasco60/blob/main/releases/final_release_v1/run/stage_S0.csv This is the so called Stage Zero file. It includes all candidates that survived the full pipeline. Rest of the survivors are reduced by post-pipeline steps in stages. All stage files are available here: https://github.com/jannefi/vasco60/tree/main/releases/final_release_v1/run/stages and everything is explained in README and some other docs available in the repository.

With the published criteria as I understand and implemented them, I do not see a path to a six‑figure survivor count. This suggests either (a) materially different selection/coverage/definitions, or (b) additional inputs/criteria not captured in the paper text. If the underlying list becomes available, I can run direct cross-checks and quantify the difference.

The paper says data may be shared on request, but I don't currently have permission to redistribute the unpublished 107K list, so I can't run a public cross-check. If it becomes shareable, I'll test it against the full POSS‑I coverage and publish the results.

Maybe I should write some sort of final report/analysis, too.

 

[Gary C]

Not really. Geosynchronous satellites aren't just constrained to the Earth's equator; some range quite far north and south. A cloud of alien geosynchronous satellites might range all over the place, so it would be difficult to identify transients caused by the same object. ...

[image removed]

More uncertainty.

Fair enough but a 'cloud' of alien satellites would indicate 24/7 coverage of the planet in which case the correlation with nuclear testing dates is irrelevant. The test sites would always be under observation. The papers authors really do seem to trying to have all things, all ways, circling back to NHI when an agnostic neutral observer would keep coming back to defects and artifacts of the human recording/transcribing technologies.

 

[jdog]

[image removed]


Fair enough but a 'cloud' of alien satellites would indicate 24/7 coverage of the planet in which case the correlation with nuclear testing dates is irrelevant. The test sites would always be under observation. The papers authors really do seem to trying to have all things, all ways, circling back to NHI when an agnostic neutral observer would keep coming back to defects and artifacts of the human recording/transcribing technologies.

Recall that the earliest paper's source for this "geosynchronous" idea was someone else's thought experiment about looking for technosignatures around the planets of other stars, suggesting that an ancient alien civilization might have left a sufficiently large "Clarke belt" of space debris in long-term stable orbits large enough that there might be detectable partial occultation of stars as they pass behind the planet.

 

[HoaxEye]

The paper says data may be shared on request, but I don't currently have permission to redistribute the unpublished 107K list, so I can't run a public cross-check. If it becomes shareable, I'll test it against the full POSS‑I coverage and publish the results.

Maybe I should write some sort of final report/analysis, too.

Here's a technical report of the third-party candidate list crossmatch results: https://github.com/jannefi/vasco60/...elease_v1/docs/vasco_candidate_crossmatch.pdf

Main Conclusion
The dataset as-obtained is not a strict instantiation of the stated gates (as evaluated under the VASCO60 implementation of those gates). It contains systematic contaminants that could bias ensemble inference unless explicitly controlled.

In other words: there is a recurring flaw in this dataset. If you just calculate the averages without specifically removing or accounting for that flaw, your final conclusion will be skewed and inaccurate.

 

[HoaxEye]

Here's a technical report of the third-party candidate list crossmatch results: https://github.com/jannefi/vasco60/...elease_v1/docs/vasco_candidate_crossmatch.pdf

Main Conclusion
The dataset as-obtained is not a strict instantiation of the stated gates (as evaluated under the VASCO60 implementation of those gates). It contains systematic contaminants that could bias ensemble inference unless explicitly controlled.

I continued the analysis with RA/DEC comparison plots. On the left, the third-party 107K list. In the middle: Vasco60 survivor candidates, and on right, the Vasco60 full tile footprint.

Observations here: https://github.com/jannefi/vasco60/...lease_v1/docs/mnras_footprint_observations.md

Despite the fact that MNRAS 2022 describes a similar 30" circular cut that has been implemented in Vasco60, it does not seem to apply to the 107K dataset. Also, Vasco60 tessellation plan avoids plate edges, but the third-party candidate list does not.

The third-party POSS-I candidate list does not appear to have been derived from a dataset restricted to the astrometrically reliable central region of each plate. Candidates are distributed across the full plate area, including the plate-edge zone where geometric distortions and vignetting are strongest. No evidence of a circular inclusion boundary was detected in the spatial distribution.

(source https://github.com/jannefi/vasco60/...lease_v1/docs/mnras_footprint_observations.md)

The "machine learning-paper" (Bruehl et al. 2026, arXiv:2604.18799) (pre-print) confirms these observations independently.

The ML ensemble assigns low candidate probability (below 0.66) to approximately 80% of the 107,875 candidates. The authors find that the majority of the dataset is more consistent with plate artefacts, blended sources, or extended objects than with genuine point-source transients. This quantitative outcome is consistent with the SPREAD_MODEL gate failure rate observed in the VASCO60 pipeline check (32.8% of in-footprint pairs fail the morphology gate), and with the 70.6% elimination rate at the shape and morphology pipeline stages (S0S, S0M).

...and:

Among the features with the highest SHAP importance in the ML model, red_dist_to_edge_px (the candidate's pixel distance from the plate or image boundary in the red-band image) ranks in the top twelve predictors of candidate probability. A feature that encodes distance to the plate edge would not carry predictive power if all candidates were drawn from a region safely interior to the plate boundary. Its inclusion among the highest-importance predictors is direct empirical evidence that plate-edge proximity is a systematic quality signal in this dataset — i.e., a measurable fraction of candidates sit near or at the plate boundary where image quality degrades.

(source: https://github.com/jannefi/vasco60/...lease_v1/docs/mnras_footprint_observations.md)

 

[john.phil]

Anton from @whatdamath YouTube channel published a video on the subject:

0:00 Vanishing stars mystery
1:00 How this all began - VASCO project
2:58 How this was produced - KODAK 103A
4:40 Emulsion issues?
5:15 Nuclear weapon tests
7:00 Could this be real object? Earth shadow test and other correlations
8:05 Critical analysis - is this just a bias?
10:28 Recent study focused on space weather - geomagnetic storm relation
12:50 Conclusions, so what is it?
14:00 Conclusions (Not aliens)


Source: https://www.youtube.com/watch?v=aWy3pDR6skI

References:

1. Searching for Fast Astronomical Transients in Archival Photographic Plates (https://arxiv.org/pdf/2603.20407)
2. Geomagnetic storm suppression of photographic plate transient detections in the POSS-I archive: an independent physical variable strengthening the nuclear test correlation (https://arxiv.org/pdf/2604.04950)
3. Independent Recovery of Vanishing Sources on POSS-I Photographic Plates Using Automated Source Detection and Cross-Epoch Matching (https://arxiv.org/pdf/2604.04810)
4. Independent Replication of Nuclear Test-Transient Correlations and Earth Shadow Deficit in POSS-I Photographic Plates (https://arxiv.org/pdf/2604.00056)
5. Critical Evaluation of Studies Alleging Evidence for Technosignatures in the POSS1-E Photographic Plates (https://arxiv.org/pdf/2601.21946)
6. Archives of Photographic Plates for Astronomical USE (APPLAUSE) - Digitisation of astronomical plates and their integration into the International
7. Virtual Observatory (https://www.aanda.org/articles/aa/pdf/2024/07/aa48793-23.pdf)
8. https://vascoproject.org/vanishing-stars/
9. Transients in the Palomar Observatory Sky Survey (POSS-I) may be associated with nuclear testing and reports of unidentified anomalous phenomena (https://www.nature.com/articles/s41598-025-21620-3)

 

[yoshy]

I have now downloaded and processed 30% of POSS-I plates using Vasco60 pipeline. I decided to build a script to understand the differences between my and MNRAS 2022 final survivor set R. (5,399 rows in MNRAS 2022). Findings were surprising so I published them:
https://github.com/jannefi/vasco60/...as_2022_candidate_audit/mnras_audit_report.md

Did you move or rename this file? I'm getting a 404 error from the link

 

[yoshy]

I think the headline here should be that this new paper seems to prove the null hypotheses.

This isn't meant to be pedantic, but the proper phrasing would be "fails to reject the null hypothesis". P-values aren't able to prove or disprove the null or alternative hypothesis, since p-values are the probability of observing a statistic at least as extreme as the observed results, assuming the null hypothesis is true. I only make this distinction because misunderstandings of p-values have been a major problem across every quantitative field in science.

Stated in another way, p-values are Pr( T(data) | H), where T is a statistic calculated from the data and H is the null hypothesis. Pr(H | data) can be computed if using Bayesian analysis, however, where it is called the "posterior probability". Frequentist analysis, which is the paradigm most commonly used and where p-values are the cornerstone of the analysis, does not provide Pr(H | data). Luckily, Bayesian analysis is becoming more common to use due to the computing power needed for it being widely available now. (The integrals needed to perform Bayesian analysis are generally intractable beyond very simple models, so numerical methods like Markov Chain Monte Carlo sampling are necessary.)

 

[yoshy]

@HoaxEye compliments to your code quality! Thoroughly documented, type annotations, clean structure. Great work!

 

[HoaxEye]

Did you move or rename this file? I'm getting a 404 error from the link

I replaced it with the final release: https://github.com/jannefi/vasco60/tree/main/releases/final_release_v1

 

[HoaxEye]

Status update on Vasco60 work:

I've spent some time building a controlled "surgical" test harness to compare individual MNRAS candidate coordinates against SExtractor/PSFEx outputs on the same tile centres, including parity logic such as the ≤30′ circle cut around tile centres. The harness reproduces my own test classifications consistently (e.g., FLAGS==0, SNR_WIN>30, morphology gates, and also purely geometric exclusions like circle_cut (dist_from_center=…')).

The remaining blocker is reproducibility: without the original MNRAS 2022 SExtractor/PSFEx configs (at minimum) and any selection rules used to export the 107K list, it's impossible to do an apples‑to‑apples replication/comparisons. I'm pausing further development work until those artefacts exist. I was told by one of my remaining "paper contacts" that they do not have the original configs. But at the same time they claim my configs are wrong.

For now, continued claims about what the 107K list "represents" are not testable in a reliable way.

 

[EightyEight8]

Avi Loeb released a Medium article titled "We Should Not Mistake Cosmic-Rays for UFOs!" where he says cosmic rays "might have also been confused with UFOs in orbit around Earth" in regards to 2 of Beatriz' papers. I have also posted links to both papers referenced by Avi. I wanted to make absolute sure he was talking about this threads paper, and he is, its the 2nd link.

We Should Not Mistake Cosmic-Rays for UFOs!

Similarly, cosmic-rays might have also been confused with UFOs in orbit around the Earth. Two recent papers (posted here and here) reported transient sources of light in historical photographic plates taken by the Palomar Sky Survey in the early 1950s, before the human Space Age. The Palomar photographic plates have a characteristic area of about a thousand squared centimeters. During the typical exposure of order an hour, the number of cosmic-ray impacts on this area is about 60,000. In order for a cosmic-ray impact to appear as a point source (star-like) rather than a streak, its incidence angle must be aligned to within a degree from the direction perpendicular to the plate surface. Given that a hemisphere covers 20,627 square degrees, the number of point sources associated with cosmic-ray impacts would be of order ten per photographic plate. This census could potentially account for the observed population of transient point sources in the photographic plates of the Palomar Sky Survey, rather than being UFOs.

Source: https://avi-loeb.medium.com/we-should-not-mistake-cosmic-rays-for-ufos-8d10b0085b9c

1 - https://arxiv.org/pdf/2604.18799
2 - https://www.nature.com/articles/s41598-025-21620-3

 

[Eburacum]

I suggested many pages ago that cosmic rays might be depleted in the direction of the Earth's shadow, due to the Earth's magnetotail;
https://www.metabunk.org/threads/tr...servatory-sky-survey.14362/page-2#post-349898
but I (tentatively) rejected this hypothesis because cosmic rays are generally too energetic to be affected by the magnetosphere (instead they are generally stopped by the atmosphere).

The deficit probably has other causes (sampling errors, for instance).

 

[jdog]

Avi Loeb released a Medium article titled "We Should Not Mistake Cosmic-Rays for UFOs!" where he says cosmic rays "might have also been confused with UFOs in orbit around Earth" in regards to 2 of Beatriz' papers. I have also posted links to both papers referenced by Avi. I wanted to make absolute sure he was talking about this threads paper, and he is, its the 2nd link.


Source: https://avi-loeb.medium.com/we-should-not-mistake-cosmic-rays-for-ufos-8d10b0085b9c

1 - https://arxiv.org/pdf/2604.18799
2 - https://www.nature.com/articles/s41598-025-21620-3

As skeptical as I am about the transient claims, I can't give much credit to Loeb on this.

As an astrophysicist with little background in planetary science, he was way out over his skis attacking other scientists about the interstellar comet 3I/Atlas and he doesn't seem to have any more insight into the nuances of 1950s-era glass plate astrophotography than the Villarroel team.

 

[HoaxEye]

As skeptical as I am about the transient claims, I can't give much credit to Loeb on this.

As an astrophysicist with little background in planetary science, he was way out over his skis attacking other scientists about the interstellar comet 3I/Atlas and he doesn't seem to have any more insight into the nuances of 1950s-era glass plate astrophotography than the Villarroel team.

I came across an older but highly relevant paper while digging into photographic plate artefacts:

Greiner et al. (1990), Astronomy & Astrophysics, 234, 251–261
"Discrimination between star-like defects on photographic plates and possible γ-ray burst optical counterparts" (PDF attached)

Key points from the paper:
- Star-like defects on photographic plates can be indistinguishable from real stellar images using standard methods like: shape comparison, density profiles, photometry
- The authors show that many candidate transient detections (in their case, possible optical counterparts to gamma-ray bursts) were later identified as plate defects, despite initially appearing astrophysical
- A key issue: standard analysis operates on 2D image properties, whereas real stars and artefacts differ in 3D structure within the emulsion
- They also note that even probability arguments (e.g. multiple detections, positional coincidence) can be misleading if the objects are not first proven to be real
- Possible causes for star-like defects include: 1) contamination 2) chemical processes 3) particle-related effects (e.g. radiation interactions with the emulsion)
Morphology alone (even when objects look perfectly star-like) is not always sufficient to distinguish real astrophysical sources from artefacts on photographic plates.

The paper doesn't settle what these transients are, but felt directly relevant.

 

[Gary C]

@HoaxEye Excellent find!

 

[HoaxEye]

@HoaxEye Excellent find!

Thanks! It's very difficult to find these relevant, older papers, but I'll keep digging and add them here: https://github.com/jannefi/vasco60/blob/main/docs/PLATE_FORENSICS_LIBRARY.md