Work

Solar wind proton density and large earthquakes: a reanalysis and out-of-sample test of Marchitelli et al. (2020)

Year: 2026
Type: preprint
Abstract: Note on versions. Version 2 replaces version 1. Changes: The abstract is shorter. The analysis record moved to the code README. The descriptions of the original paper's Molchan-type test and of Chen et al. (2025) are corrected. Four method details are stated explicitly. No results changed. Two sensitivity values were added. What this adds. This is the first test of the reported correlation between solar wind proton density and large earthquakes (Marchitelli et al. 2020, Scientific Reports 10:11495) on earthquakes after 2016, together with a reanalysis of the original 1996 to 2016 data. It uses the original catalogue (ISC-GEM) and a null hypothesis that preserves earthquake clustering. Key finding. On 2017 to 2026 data the relative rate is R = 1.06 (p = 0.32), against the claimed 1.82. None of 10,413 simulations with the claimed effect planted into the real data falls as low as the observed value, so the claimed effect size is rejected. Effects of R of 1.3 or more are disfavoured at about 95% confidence, assuming unclustered extra events; a small effect (R of about 1.1 to 1.25) cannot be ruled out. In-sample reanalysis. The original values are reproduced: R = 1.43, 1.78 and 2.27. The headline significance rests mainly on one qualifying day, 11 March 2011. On that day the Tohoku sequence contributed 92 ISC-GEM events of M 5.6 or larger. This explanation was first suggested in a public reader comment on the original article in 2024; this note quantifies it. Under a clustering-preserving null, the full in-sample result has p = 0.003. Without the Tohoku day the excess is marginal: p = 0.13, or 0.03 with calibrated removal of the largest qualifying day. A residual excess remains in the 1996 to 2005 half: R = 1.26 after removing its largest day, p = 0.007. This is before any allowance for the original parameter search. Deep and declustered subsets show no significant excess. Calibration. SOHO proton densities drift by about 25% relative to OMNI between 1996 to 2015 and 2017 to 2026. The density threshold is therefore mapped through OMNI. Disclosures. A first test with the preprint definitions (M 5.8 or larger), fixed before the out-of-sample data were compared, gives R = 0.96. The published-definition analysis was specified after it. Size. This is a short reanalysis and replication note. It is relevant to work on solar-terrestrial triggering of earthquakes and to statistical tests for clustered catalogues. Files. The paper (PDF). code_and_protocol_v2.zip: all analysis code, a data download script for the public sources (SOHO CELIAS/PM, OMNI, USGS ComCat, ISC-GEM via ComCat) and a README recording the analysis order.
Author: Nguyen, Dat
Language: English
Subfield: Geophysics
Open Access status: green
OpenAlex ID: W7225989069
Cites: 2
Cited by:
Related to: 10