A M6.6 Hit Sanriku.
No Foreshocks Came First.

The Question

The Japan Trench is one of the world's most heavily instrumented subduction zones. When a M6.6 strikes 150 km from a recent M7.4, two questions follow immediately. First: was there a detectable foreshock cluster — a swarm of smaller events in the days or weeks before — that could have flagged this rupture in advance? Second: is the M6.6 statistically part of the M7.4's aftershock sequence, or an independent event that happens to be nearby in time and space?

Both questions have standard tests. Both tests return null results.

What Three Networks Recorded

The USGS, EMSC, and GFZ catalogs all report the mainshock within 6 seconds and 30 km of the USGS epicenter. ISC has not yet ingested the event — typical 30-day latency.

Both epicenters lie on the Pacific-Plate subduction interface that ruptured catastrophically in the 2011 Tohoku M9.1.

Zero Foreshocks — A Clean Null

A 30-day, 50-km pre-window centered on the M6.6 epicenter contains zero events at magnitude 4 or above in the USGS catalog (the regional completeness floor offshore Japan).

To know whether zero is anomalously few or expected, we built a 16-year historical baseline: 199 non-overlapping 30-day windows of M5+ activity within 50 km of the M6.6 epicenter, 2010 through 2026. In that baseline, only 3 of 199 windows contained any M5+ events — a base rate of 1.5%. Under a binomial test with that rate, the probability of observing zero events in a single window is 0.985. The pre-M6.6 quiet is exactly what the historical baseline predicts.

Sensitivity: zero M4+ events at every radius from 10 to 50 km. At wider radii (100 km / 60 days yields N=4; 150 km / 30 days yields N=12) the count rises, but those events sit ~150 km away in the M7.4 cluster, not in the M6.6 source volume. The 50 km cap isolates the M6.6's own source from M7.4 contamination.

Statistically Independent of the M7.4

The Gardner & Knopoff (1974) declustering algorithm defines an empirical space-time window for any mainshock magnitude inside which subsequent events are classified as aftershocks. For a M7.4, the window is r ≤ 79.3 km and t ≤ 945.6 days.

The M6.6 sits 151.7 km from the M7.4 epicenter — roughly twice the GK aftershock radius. Under standard declustering, the M6.6 is classified as an independent event, not as a delayed aftershock of the M7.4.

A Coulomb-stress-transfer calculation could still find that the M7.4 perturbed the M6.6 nucleation patch above ambient noise — those calculations are physics-of-rupture work, not catalog-level statistics, and are outside the scope of this analysis.

A Suggestive Finding About the M7.4 Sequence

While running the test suite we noticed something about the M7.4: its aftershock sequence appears to be lighter than Bath's law would predict.

The largest aftershock of the M7.4 in the 28-day post-event catalog is M5.6 — giving an observed Δm = 1.8. The canonical Bath's-law value is Δm ≈ 1.2, with historical scatter σ ≈ 0.4. The M7.4's gap is about 1.5 standard deviations above the expected value, in the direction of weaker aftershocks.

We are not claiming a confirmed anomaly. A 1.5σ effect on a single sequence is suggestive, not significant. The bootstrap interval we computed on the aftershock catalog ([1.8, 2.2]) is a sampling-distribution bound on the observed Δm, not a hypothesis test against Bath's law itself. A proper test requires comparing this sequence to the empirical Δm distribution from matched subduction-zone M≥7 mainshocks, which is V2 work scheduled for late June 2026 when the post-M7.4 catalog matures.

If the deficit holds up under formal comparison, the pattern would be consistent with a subduction-zone rupture that exhausted most of the available slip budget on a single asperity — leaving little stress left over to feed large aftershocks.

Figure: Sixty Days of Seismicity

What This Result Does and Does Not Say

It says: the M6.6 was not preceded by a detectable foreshock cluster, and it cannot be statistically classified as a delayed aftershock of the M7.4 under standard declustering. Both nulls are clean and robust to choice of analysis window.

It does not say: that foreshocks were physically absent at scales below the regional completeness floor (Mc ≈ 4 offshore Japan); a high-resolution JMA-quality catalog could in principle resolve M3 events that USGS cannot. It also does not say that the M7.4 had no physical effect on the M6.6 nucleation — Coulomb-stress and post-seismic-deformation effects can operate at distances well beyond the empirical Gardner-Knopoff window, but those are physics-of-rupture questions that catalog statistics cannot adjudicate.

Deferred to V2

Three follow-on tests will run once 30+ days of post-M6.6 catalog accumulate (expected late June 2026):

• Bath's law and Omori decay for the M6.6 sequence itself — currently only 3 days of post-event data, with zero detected aftershocks inside 100 km.
• Formal comparison of the M7.4's Δm = 1.8 against the empirical distribution of matched M≥7 subduction events, to convert the 1.5σ suggestion into a hypothesis test.
• Fourth-network confirmation via ISC, which has not yet ingested either event.

Reproducibility

All code, data extracts, and the full statistical pipeline live in the m6-6-sanriku-2026-05-15-foreshock-signat workspace. Every number on this page comes from data/results.json. Pipeline: USGS FDSN + EMSC + GFZ + ISC catalog pulls, earthquake deduplication, multi-network agreement check, Bath's-law point test with bootstrap, Mann-Whitney foreshock test against 199 historical windows, binomial test against the historical base rate, Aki MLE b-value, Gardner-Knopoff aftershock declustering for the M7.4.