sciscape Publications
← All publications
astro-ph.IM · Instrumentation and Methods for Astrophysicsastro-ph.EP · Earth and Planetary Astrophysics

Plate-matched tests of public VASCO POSS-I data find no significant Earth-shadow deficit at 42,164 km for 107,875 transients or the top ML decile; only 2 to 4 plates inform the decile test

Question: Earlier studies report that VASCO transients on POSS-I red plates (1949-1956) avoid the Earth shadow at geosynchronous distance. Bruehl et al. (2026) report the largest deficit in the top machine-learning decile D9: 0.31 % in shadow against a 0.692 % control rate (p = 2e-7). They also report D9 against D0-D8 z = -2.78. We test if this deficit stays when the expected count comes from the same plates and source positions. Method: We use only public data. We assign the 107,875 catalogue sources to 636 POSS-I E plates with the STScI GSSS plate list and the VizieR VI/25 plate catalogue. We rebuild the plate exposure times. The GSSS date is one day early for 708 of 937 plates. For each plate, 8 dihedral reflections of the source positions give the null distribution. We use 7 time and observer conventions and 4 shadow radii (8.42 to 8.96 deg). For D9 against D0-D8, we also calculate a plate-stratified Mantel-Haenszel odds ratio and a within-plate label permutation test. Results: With the published rule, the public data give the published total of 495 shadow sources in the strict-date sample. Decile counts differ by up to 10 (D9: 22 against 31 published), because the public scores differ. Over all conventions and radii, observed/expected for all sources is 0.88 to 1.01 (z from -1.17 to +0.07; smallest one-sided p = 0.13). For D9, observed/expected is 0.56 to 0.93 (smallest p = 0.060). The stratified odds ratio of D9 against D0-D8 is 0.36 to 0.98. All 95 % intervals include 1 and also include the published pooled value of 0.60. Only 2 to 4 plates have shadow sources in both groups. The within-plate label test gives p = 0.10 to 0.33. The published significance comes from the plate-uniform control rate. In the published geometry at r = 8.50 deg, the expected rate at the actual source positions is 0.52 % for all sources and 0.29 % for D9. Only 5 to 13 plates touch the shadow, and the sources are not uniform on the plates. Limits: The 626-plate published control is not public. The public scores are from a different model version. D9 has low power, so these data cannot exclude a D9 deficit of the published size. The package has all code, the two CC BY 4.0 VASCO control files, a download script with SHA-256 checks for the other inputs, and all results. A second run from a clean folder gave the same result tables, byte for byte.

Earth shadow · POSS-I · VASCO · vanishing sources · photographic plates · geosynchronous orbit · Palomar Observatory Sky Survey · plate exposure times · Digitized Sky Survey · machine learning scores · control sample · permutation test · Mantel-Haenszel odds ratio · replication · null result · transients

Admitted 0 citing publications

References

8
  1. challengesdoi:10.48550/arXiv.2604.18799

    Bruehl et al. 2026 (Sci. Rep., in press). It reports that the top ML decile D9 has 0.31 % of sources in the Earth shadow against a 0.692 % control (p = 2e-7) and D9 against D0-D8 z = -2.78. With its rule and the public data, we get its total of 495 strict-sample shadow sources; decile counts differ by up to 10 because the public scores differ. With a plate-matched expected count, the D9 deficit is not significant (smallest one-sided p = 0.060 over 7 conventions and 4 radii). All plate-stratified odds ratio intervals include 1.

  2. challengesdoi:10.1088/1538-3873/ae0afe

    Villarroel et al. 2025 (PASP 137). It reports a 22 sigma deficit of POSS-I transients in the Earth shadow against hemispheric coverage, and 7.6 sigma against plate-based coverage. For the 107,875-source VASCO catalogue, our expected count at the actual source positions gives observed/expected 0.88 to 1.01 over 7 conventions and 4 radii, with no significant deficit. We did not rerun its exact sample or code.

  3. challengesdoi:10.48550/arXiv.2604.00056

    Doherty 2026 replication. It counts a source as in shadow if it is less than 8.50 deg from the antisolar point. It reports 499 of 107,875 transients (0.46 %) in shadow, against 0.78 % from plate coverage and about 1.4 % from the hemisphere. With the same 8.50 deg rule, we count 501 (0.46 %), and uniform points on the plates give 0.77 %. The plate-coverage value assumes sources that are uniform on each plate. At the actual source positions on the same plates, the expectation is 0.52 %, and 0.27 % to 0.52 % over 7 time conventions. Observed/expected is 0.89 to 0.99, with no significant deficit.

  4. supportsdoi:10.48550/arXiv.2601.21946

    Watters et al. 2026 find no significant shadow deficit and show that the uniform background assumption is false. Our independent plate-matched test with exposure times rebuilt from GSSS and VI/25 gives the same result for all sources and for the top ML decile.

  5. doi:10.48550/arXiv.2602.15171

    Villarroel et al. 2026 response to Watters et al. It notes that the critique has no uncertainty estimates. We give null distributions, p-values and confidence intervals for each test.

  6. doi:10.48550/arXiv.2609.05105

    Villarroel et al. 2026 model the transients as reflections from objects at 20,000 to 35,000 km. We test the shadow only at 42,164 km geocentric distance.

  7. usesdoi:10.1093/mnras/stac1552

    Solano et al. 2022 give the VASCO POSS-I vanishing-source sample from which the 107,875-source catalogue comes.

  8. doi:10.1038/s41598-025-21620-3

    Bruehl and Villarroel 2025 report a link between transients and nuclear test dates. We do not test that link.

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.