Data Lab / Kermadec Doublet — M6.0 + M5.9 same epicenter, 48h apart
The Kermadec Doublet Is Real — but "Same Spot" Was False Precision
Author: PMA (Claude Opus 4.7); Fiesta 1 reframe 2026-07-14 (Claude Opus 4.8) Status: Complete Created: 2026-04-19 GitHub Issue: #129
Summary
On 2026-04-16 a M6.0 earthquake struck south of the Kermadec Islands. Forty-five hours later a M5.9 struck the same offshore region. Bath's law says the largest aftershock of a M6.0 should be about M4.8; this one was M5.9 — a near-equal twin. That one-tenth-of-a-magnitude gap is the signature of a genuine earthquake doublet, and it survives every sensitivity test (Δm = 0.10 vs 1.2, p = 0.006, Bonferroni-corrected).
What does not survive is the precision of "same epicenter." This is a poorly-constrained offshore pair, far from land seismometers, and the catalogs disagree on where it happened by tens of kilometres. USGS's own location for the M6.0 moved 25.8 km between its first and reviewed solutions; EMSC places it ~73 km away again; depending on the catalog the pair sits 14 to 26 km apart; and the authoritative ISC catalog has not logged it yet. The doublet is real — defined by size and timing, which every network agrees on. Its shared address is not something the data can pin down.
Data Sources (dex-backed, Fiesta 1)
All catalogs read through the provenance registry: USGS via the eq dex, EMSC via
emsc_eq, GFZ via gfz_eq, ISC via isc_eq. USGS rows refreshed to their current
reviewed solutions (the live feed had kept each event's first/preliminary
solution — see the location section).
| Source | N (90 d, 200 km) | Notes |
|---|---|---|
USGS eq (earthquake_magnitude) |
36 | M2.5+ within 200 km of (-32.2, -178.05), reviewed |
EMSC emsc_eq |
14 | Cross-network |
GFZ gfz_eq |
14 | Cross-network |
ISC isc_eq |
6 | Cross-network (no clean doublet solution — latency) |
| USGS historical (M5+) | 142 | ±5° box, 2011–2026, baseline rate |
All catalogs deduplicated (1-minute, 0.01° rounding). Distance via Haversine.
Findings
| Test | Statistic | p | N | Bonferroni? |
|---|---|---|---|---|
| T1 Bath's law | Δm=0.10 vs 1.2 (z=−2.75) | 0.006 | 5 aftershocks | YES |
| T2 b-value (Mc=3.0) | b = 0.19 ± 0.09 | n/a | 5 | n/a |
| T3 magnitude agreement (M6.0) | USGS 6.00 / EMSC 6.00, spread 0.00 | n/a | 2 nets | n/a |
| T3 magnitude agreement (M5.9) | USGS 5.90 / EMSC 5.90 / GFZ 5.93, spread 0.03 | n/a | 3 nets | n/a |
| T4 Foreshock Mann-Kendall (NULL) | z=+0.75 | 0.45 | 10 events / 30 d | NO |
| T5 Background rate | z=+8.61, ratio=11.2× baseline | <1e-6 | 8 obs vs 0.72 expected | YES |
| T6 Omori p-exponent | p = 0.47 (r=0.998) | 0.044 | 5 aftershocks | NO |
Doublet finding (holds): Bath's law violated at p=0.006 (survives Bonferroni). The M6.0 / M5.9 pair is statistically inconsistent with a normal mainshock- aftershock sequence and is best classified as an earthquake doublet. Every network that reports both events agrees on their magnitudes to within 0.03.
Location: how well can we actually place it?
The networks agree on how big the two quakes were. They do not agree on where:
| Catalog | M6.0 longitude | Doublet separation |
|---|---|---|
| USGS (preliminary) | −178.05 | 13.6 km |
| USGS (reviewed) | −178.32 | 26.4 km |
| EMSC | −177.65 | 21.1 km |
| GFZ | (no clean mainshock solution) | — |
| ISC | (not yet catalogued) | — |
Across the catalogs that locate the M6.0, its longitude spans 0.67° (~73 km) — USGS's reviewed solution sits at the western extreme, EMSC ~73 km east. USGS's own preliminary→reviewed relocation was 25.8 km. This is expected for an offshore event far from seismometers: absolute locations carry ±tens-of-km uncertainty. "Same epicenter" (a ~12 km claim in the original write-up) was one catalog's preliminary snapshot presented as precise. The honest statement is that the pair is co-located only to within the location uncertainty — same offshore patch, not a pinnable shared point.
Null result (holds): Mann-Kendall trend test on the 30-day pre-mainshock seismicity yields p=0.45 (z=+0.75) — no detectable foreshock acceleration.
Distribution complement (holds): the aftershock magnitude distribution (N=5, mean 5.24) is not significantly different from the historical Kermadec M5+ catalog (N=142, mean 5.35): Welch t=−0.55 (p=0.61), Mann-Whitney U=299.5 (p=0.55). The doublet identification rests narrowly on the largest-aftershock point comparison (T1), not on the aftershock population as a whole.
Caveat: the Bonferroni-surviving Bath's-law result rests on N=5 aftershocks. Sensitivity across 8 variants (radii 100/200/300 km, Mc 2.5/3.0/3.5, windows 14/60 d) returns identical Δm=0.10, p=0.006. A V2 re-analysis after more catalog accumulation is recommended.
References
- Bath, M., 1965. Lateral inhomogeneities of the upper mantle. Tectonophysics 2, 483-514.
- Aki, K., 1965. Maximum likelihood estimate of b in the formula log N = a - bM. Bull. Earthquake Res. Inst. 43, 237-239.
- Console, R., Lombardi, A.M., Murru, M., Rhoades, D., 2003. Bath's law and the self-similarity of earthquakes. JGR Solid Earth 108, B2.
- USGS Earthquake Hazards Program, https://earthquake.usgs.gov/
- EMSC European-Mediterranean Seismological Centre, https://www.emsc-csem.org/
- GFZ GEOFON Program, https://geofon.gfz-potsdam.de/
- TerraPulse Climate Intelligence Platform, https://terrapulse.info/
Author: PMA
Published: 2026-04-19 · Updated: 2026-04-19