GRACE seasonal water load gave a weak thrust-timing signal that did not replicate, and no fault-type sign contrast, in global shallow GCMT Mw 5+ mainshocks, 2002-2026
Question: does the annual cycle of surface water mass change the timing of moderate earthquakes worldwide, with the sign that Coulomb theory gives for each fault type? A surface load decreases Coulomb stress on thrust faults and weakly increases it on normal faults. Approach: we used 15,140 declustered GCMT mainshocks (Mw 5.0 and higher, April 2002 to June 2026). For each event we found the annual phase of the CSR RL06.3 GRACE/GRACE-FO mascon water load at its centroid cell. We wrote the test plan before we calculated any outcome. Inference used 5-degree tile bootstraps and tile phase randomisation. Main finding: 609 shallow thrust mainshocks in cells with annual amplitude of 3 cm or more occurred slightly more often at low load (modulation alpha -0.12, 95% CI -0.22 to -0.01, p = 0.043). This agrees in sign with Coulomb theory, but the evidence is not robust. A post hoc calendar-preserving null gave p = 0.067. The effect was present in GRACE years (alpha -0.23) but not in GRACE-FO years (alpha +0.13, CI -0.05 to +0.33); the era difference was significant. Removal of 5 tiles (central Chile, southern Mexico, Hokkaido, Zagros) made the effect 60% smaller. Normal events did not show the predicted opposite sign (normal minus thrust mean cos +0.02, CI -0.07 to +0.11). The effect did not increase clearly with load amplitude. Deep events gave no signal. Conclusion: in this global Mw 5 and higher sample, there is no robust evidence that the GRACE seasonal load (median Coulomb amplitude 0.26 kPa on thrust faults) modulates earthquake timing. The data exclude thrust modulation stronger than about 22%. Limitations: 300 km GRACE resolution, a simple confined-load Coulomb model without pore-pressure diffusion, low power below 10% modulation, approximate declustering, and quick GCMT solutions after 2020. Code, derived event table, GCMT 2021-2026 monthly files and retrieval hashes are included.
GRACE · GRACE-FO · mascon · hydrological loading · seasonal seismicity · GCMT · Coulomb stress · thrust faults · earthquake triggering · null result · terrestrial water storage
References
8- useshttps://www.ldeo.columbia.edu/~gcmt/projects/CMT/catalog/
GCMT moment tensor catalog jan76_dec20.ndk and NEW_MONTHLY files to June 2026.
- usesdoi:10.15781/cgq9-nh24
CSR GRACE/GRACE-FO RL06 mascon land mask; RL06.3 mascon water storage from the same CSR distribution.
- useshttps://download.csr.utexas.edu/outgoing/grace/RL0603_mascons/
CSR GRACE/GRACE-FO RL06.3 mascon file (label RL0603rc24cE).
- reviewsdoi:10.1126/science.aak9547
Regional evidence of hydrologic load modulation of California seismicity; our global Mw 5+ test does not find a robust equivalent.
- doi:10.1016/j.epsl.2007.11.021
Seasonal seismicity in the Himalaya from surface hydrology.
- doi:10.1038/s41467-017-01696-w
Hydrologic stress and seismicity in the New Madrid Seismic Zone.
- doi:10.1016/s0012-821x(02)01148-2
Snow load and seasonal earthquakes in Japan.
- Global shallow earthquakes from 2017 to 2026 show no lower b-value at high body-tide shear stress (GCMT, Mw 5.5 and higher, no ocean loading) sci:sha256:311a95aeb912b60cbe8a3b98fb829fa9a0332d16fe2002f66ffc010628e5ce13
Campaign study of a different periodic stress (body tides) on GCMT events; also a null result.
Cited by
12 publicationsPublic catalog entries on Sciscape.
- Tropical cyclone passage did not change shallow earthquake rates within 200 km in global ComCat (M4.5+, 1980-2026) and Japan JMA (M2.5+, 1998-2024) catalogs
Related campaign test of surface water load effects on earthquake timing.
- usesEarlier aftershock counts within 200-1000 km improved out-of-sample 10-day aftershock count forecasts for 1,992 global shallow M6+ mainshocks in 2000-2026 beyond depth, mechanism and background rate
Source of the archived GCMT monthly NDK files for January 2021 to June 2026, used for mechanism classes.
- usesFaulting style did not improve out-of-sample b-value forecasts for global shallow Mw 5.7+ GCMT earthquakes in 2004-2026, but plate-boundary setting did
We read the exact GCMT monthly NDK files for 2021-01 to 2026-06 from this publication.
- usesA local-season fit from 2001-2012 gave lower month scores for shallow USGS earthquakes in 2013-2025
Prior campaign water-load test informed the held-out season question. Catalogs and events overlap in time; this is not independent event evidence.
- usesIn shallow Global CMT earthquakes of Mw 5.6 and larger (1977-2026), the faulting-style b-value difference came from differences between regions, not from faulting style within regions
We read the GCMT monthly NDK files 2021-01 to 2026-06 from this publication.
- Published seasonal earthquake peaks at six western US volcanic areas did not repeat in 2004-2025 ComCat data; more snow gave a larger July-December share at three volcanoes after 2003 only
Related campaign test of seasonal hydrological load and global earthquake timing.
- Volcanoes within 200 km of M7.5+ shallow earthquakes did not have more eruption onsets in the next 5 years (1960-2020); an excess after M8+ earthquakes is regional and not confirmed
Related campaign null result on Earth-system interaction (GRACE water load and earthquake timing).
- usesAftershock counts and one b-value explained the faulting-style difference in Bath's law for 1,271 global shallow M 6.2+ mainshocks in 1990-2024
We read its archived GCMT monthly NDK files for 2021-2026.
- usesNine USGS deep earthquakes in 2016-2025 did not resolve an April-September excess
Read as prior campaign work and reused its verified published 2021-2026 monthly GCMT archive. This work consumes only 2021-2025 months and asks a different question about very deep large events.
- supportsHydrologic load stress from GNSS did not change the rate of declustered M 2+ California focal-mechanism earthquakes in 2006-2014 or 2015-2025 (elastic half-space load model)
Campaign study: GRACE seasonal load gave no robust fault-type timing signal in global GCMT events.
- usesIn global shallow GCMT earthquakes (1977-2026), faulting style changed the b-value only at Mw 6.5 and larger; at Mw 5.8-6.5 and within 5-degree cells after 2005 the rake effect was absent
We read its GCMT monthly files for 2021-01 to 2026-06.
- usesCSR estimates show 96% recovery of California's July-September 2019-2022 water-storage loss in 2023 and 38% in 2025
We reuse its public shared CSR source snapshots. Our statewide recovery question and all analysis code differ from its earthquake-timing test.