Event groups raised seasonal null spread by 23% over cell-year groups in shallow M5.5+ earthquakes from 2013-2025

We tested how group boundaries affect a seasonal earthquake test in a fixed USGS ComCat sample. We learned an annual phase from 2,568 shallow extratropical M5.5+ events in 2000-2012 and tested it on 2,558 events in 2013-2025. Connected event groups with pair links of at most 200 km and 30 days gave an exact rotation-null standard deviation of 7.65 percentage points. Fixed 20-degree cell-year groups gave 6.22 points. This is a 22.9% increase in standard deviation and a 51.1% increase in variance. Splitting the same event groups at cell-year boundaries reduced their variance by 24.5%. All nine distance-time checks gave larger spread than cell-years. The seasonal coefficient stayed at +1.25%, with two-sided conditional p = 0.874 in the primary event-group model. The earlier finding of no clear phase transfer remains. The result measures sensitivity to grouping assumptions. It does not show true uncertainty coverage or a physical load mechanism. Exact variance identities, aggregate results, source hashes, retrieval and analysis code, an independent audit and a figure are included. Raw and record-level external data remain outside the bundle.

seasonal seismicity · earthquake clustering · null variance · spatial boundaries · phase rotation · method sensitivity