Seismic Corridor Activity as a Global Precursor to M ≥ 7 Earthquakes: A Retrospective Validation on 1,418 Events

We test whether the geometric organization of background seismicity (M ≥ 5) into elongated spatial clusters — seismic corridors — constitutes a statistically significant precursor to large (M ≥ 7) earthquakes. A deterministic corridor score is evaluated retrospectively across all 1,418 M ≥ 7 earthquakes recorded globally between 1930 and 2026 (8,096,024 events), using a fixed 500 km radius and 5-year observation window without post-hoc parameter tuning. Among 1,232 evaluable events, the chain-filtered V2 algorithm detects 722 (58.6%; 95% Wilson CI [55.8%, 61.3%]; z = +6.0 vs. p0 = 0.5). In the digital broadband era (1990–2026), detection rises to 86.2% (462/536), stable across four consecutive decades (87.4%, 88.5%, 88.9%, 76.3%). ROC analysis of the continuous score against 325 null windows yields AUC = 0.690 for modern-era targets (Mann-Whitney z = +7.0), with an era-matched contrast of 87.9% vs. 41.5% in the 1980s (z = +6.2). Detection scales monotonically with observational data density, from 13.7% in the sparsest quartile of M ≥ 5 history to 90.9% in the densest. The signal survives target declustering (42.9% on independent mainshocks, z = +8.0 vs. p0 = 0.30), threshold sweeps, and deduplication perturbations. The raw null false-positive rate (64.0%, 208/325) exceeds the aggregate detection rate; the discriminating signal resides in the era-matched contrast and the data-density gradient, both documented in full. This study is the geometric companion of a separate OMP-based forecasting validation ('Precursor Signal Independent of Clustering', separate record): the two observe the same loading process through independent observables — static spatial geometry vs. dynamic stress-front convergence — with mutually reinforcing findings. Data: ISC-GEM v9.1; validation datasets and figure scripts available from the corresponding author.