How often does the static Coulomb stress sign at a later large earthquake depend on the finite fault model? A preregistered test on great earthquakes, 2010 to 2026
Year: 2026
Type: preprint
Abstract: Papers that attribute a large earthquake to static stress transfer from an earlier one usually compute the Coulomb stress change with a single finite fault model of the earlier event. The uncertainty of that sign has been measured within single inversions, but not between independent published models across many earthquakes. We ask how often the sign at the hypocenter of a later large earthquake changes with the choice of model, and whether later large earthquakes sit in stress increases more often than earlier earthquakes in the same area. The study was preregistered before any stress was computed. We used the 86 earthquakes of Mw 7.5 and above from 2010 to 2026. For each, we resolved the stress change of the USGS finite fault model onto both nodal planes of every event of magnitude 6.5 and above within 300 km in the five years after the mainshock. Such events in the five years before it serve as the control. The share of positive changes was 0.54 after and 0.51 before the mainshocks (difference +0.03, 95 percent interval -0.20 to +0.25, 59 mainshocks), so the two cannot be separated with this sample. On 13 mainshocks with independent published models, 51 later events had a sign that agreed on both nodal planes under the USGS model. Replacing that model with another published one gave the opposite sign in 0.19 of the comparisons on average (interval 0.08 to 0.25), and 0.43 of these events changed sign under at least one model. A static triggering sign computed from one finite fault model near the rupture therefore carries a model dependence that should be reported with it. Version 2 (8 November 2026) corrects version 1. After version 1 was released, the total moment tensor of every finite fault model was compared with the GCMT mechanism of its mainshock; this check was not in the preregistered rules for reading the model files. One second model, SRCMOD s2011TOHOKU01IDEx (Tohoku, Ide et al. 2011), has slip opposite to the earthquake (rake -81 degrees on every subfault of a thrust fault, 154 degrees from GCMT) and was removed; the preregistered statistics code was run again with no other change. The main test uses the USGS model only and does not change. For the second question, the mean sign change rate changes from 0.195 (0.081 to 0.254) to 0.186 (0.081 to 0.248), and the share of receivers that change sign under at least one model from 0.53 to 0.43. Version 1 remains available. The code and data package reproduces every number in the preprint, including the check and the recalculation; see README.md inside the archive. Second models of other authors are downloaded by fetchsecond.py and are not redistributed.
Author: Lyu, Wenjie
Institution: China University of Mining and Technology
Language: English
Subfield: Geophysics
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225397664