Independent validation of core findings from Bruehl and Villarroel (2025), "Transients in the Palomar Observatory Sky Survey (POSS-I) may be associated with nuclear testing and reports of unidentified anomalous phenomena," published in Scientific Reports.
Author: Brian Doherty Contact: briandohertyresearch@gmail.com Date: January 2026
--
This package reproduces the primary statistical claims from the Bruehl and Villarroel paper using the original dataset provided by Dr. Stephen Bruehl. The goal is straightforward: take the same data, run independent analyses, and confirm (or fail to confirm) the reported results. Everything here is written from scratch rather than reusing the original authors' code.
Two analyses are included
- Nuclear test window correlation (chi-square, negative binomial regression, permutation testing)
- Earth shadow deficit analysis (geometric shadow classification of transient positions)
The data folder contains the original research datasets and is not included in the public repository per agreement with the data providers. If you need access to the datasets to reproduce the analysis, please contact Dr. Stephen Bruehl at Vanderbilt University Medical Center or Dr. Beatriz Villarroel at Stockholm University.
The scripts expect three files in the data folder:
Transient_Nuclear_Analyzed_Dataset_ScientificReports.xlsx Original transient dataset from Dr. Bruehl, containing daily transient counts from POSS-I plates (November 1949 through April 1957, 2,718 days)
SUPERVIKTIG_HELAVASCO.csv VASCO transient catalog with source positions and plate metadata
SUPERVIKTIG_HELAVASCO_validated_v4.csv Validated subset of the VASCO catalog
Python 3.10 or higher.
Dependencies
pandas numpy scipy statsmodels astropy openpyxl (for reading the Excel dataset)
Install with
pip install pandas numpy scipy statsmodels astropy openpyxlFrom this directory
python nuclear_transient_correlation.py
python earth_shadow_validation.pyBoth scripts print results to the console and save output files to the results/ subfolder. The random seed is fixed at 42 for reproducibility of permutation tests.
Loads the original Bruehl dataset and runs three independent tests of the nuclear window correlation:
Test 1: Chi-square contingency test. Builds a 2x2 table comparing transient detection rates inside versus outside nuclear test windows (plus or minus 1 day from detonation). Computes chi-square statistic, p-value, and relative risk.
Test 2: Negative binomial regression. Fits a generalized linear model with nuclear test window as the primary predictor, controlling for precipitation, cloud cover, moon illumination, and a media coverage proxy. Reports incidence rate ratios with 95% confidence intervals.
Test 3: Permutation test. Randomly shuffles nuclear test date assignments 10,000 times, recalculates relative risk each time, and compares the observed value to the null distribution. This confirms whether the specific test dates matter or whether any random set of dates would produce a similar effect.
Output files
results/nuclear_correlation_validation.csv -- Summary statistics
results/nb_model_summary.txt -- Full negative binomial model output
Classifies each VASCO transient by whether its sky position falls inside Earth's geometric shadow at geostationary orbit altitude. Objects inside the shadow cannot reflect sunlight, so a deficit of transients in the shadow region has implications for what the transients physically are.
The script calculates the Sun position for each observation time using standard astronomical algorithms, computes the anti-sun point (shadow center), and measures the angular distance from each transient to the shadow center. Transients within the shadow radius at GEO altitude (approximately 8.5 degrees) are flagged.
Output files
results/shadow_classification.csv -- Full classification of all transients
results/umbra_transients.csv -- Only the transients falling inside the shadow
| Metric | This Validation | Original Paper |
|---|---|---|
| Chi-square | 6.47 | 6.94 |
| p-value | 0.011 | 0.008 |
| Relative Risk | 1.45 | 1.45 |
The relative risk matches exactly. The small difference in chi-square likely reflects minor differences in how the contingency table handles edge cases.
| Variable | IRR | 95% CI | p-value |
|---|---|---|---|
| Nuclear Window | 3.527 | 2.799 - 4.446 | < 0.0001 |
| UAP Reports | 1.253 | 1.204 - 1.304 | < 0.0001 |
| Precipitation | 0.370 | 0.329 - 0.418 | < 0.0001 |
| Moon Phase | 0.202 | 0.174 - 0.234 | < 0.0001 |
The nuclear test effect strengthens after controlling for environmental confounders (IRR = 3.53 versus the crude relative risk of 1.45), indicating that weather and lunar conditions were partially masking the true effect size.
Observed RR: 1.447 Permutation mean: 1.009 Permutation p-value: 0.004 95% CI from null distribution: 0.708 - 1.319
Only 0.4% of random shuffles produced a relative risk as large as the observed value. The specific nuclear test dates drive the correlation, not temporal autocorrelation.
Restricting to transients within 2 degrees of plate center (eliminating edge artifacts): Transients retained: 22,309 (44% of original) Transient-positive days retained: 306 of 310 (98.7%) Nuclear window effect: p < 0.0001
The finding survives edge artifact removal with increased statistical significance.
| Dataset | Total | In Shadow | Rate |
|---|---|---|---|
| VASCO (Palomar) | 22,309 | 50 | 0.22% |
| Geometric expectation | -- | -- | ~1.4% |
Shadow transient rate (0.22%) is significantly below geometric expectation (~1.4%), consistent with the original findings. All 50 shadow transients predate artificial satellites (earliest: February 1949, latest: April 1956, all before Sputnik in October 1957).
See VALIDATION_METHODOLOGY.md in this folder for the complete methodology documentation, including detailed descriptions of the statistical tests, assumptions, shadow geometry calculations, and how results compare to the original paper at each step.
Nuclear test dates: DOE/NV-209 Rev 16, Johnston's Archive Transient data: Provided by Dr. Stephen Bruehl (Vanderbilt University Medical Center) VASCO catalog: VASCO project (Villarroel et al.) Solar position calculations: Astropy library using standard Meeus algorithms
If you use this validation package, please cite:
Bruehl, S. and Villarroel, B. (2025). "Transients in the Palomar Observatory Sky Survey (POSS-I) may be associated with nuclear testing and reports of unidentified anomalous phenomena." Scientific Reports, 15, 34125.
Brian Doherty briandohertyresearch@gmail.com