A microseismicity precursor to M ≥ 5.5 earthquakes replicates across four independent zone sets in compressive tectonic regimes

Whether precursory seismicity carries usable information on forthcoming earthquakes has been debated for four decades, and a 2014 meta- analysis concluded that the question remained open because of a lack of replication. We evaluated P-HRI, a proprietary composite index of the 30-day temporal structure of daily seismicity that is independent of event rate, with fixed weights, on 1,951,464 catalogued events from four networks (USGS, INGV, GeoNet, EMSC) in 101 circular zones organised in four disjoint sets. Skill was measured with Molchan error-diagram areas (𝐴; 0.5 = chance) against a rate benchmark (30-day count) and a standard-parameter ETAS-type benchmark, under a fair-evaluation design that excludes ongoing sequences, with bootstrap confidence intervals and hash-certified preregistration. Pooled across regimes, P-HRI and the count were indistinguishable. In compressive regimes, P-HRI outperformed the count at M ≥ 5.5 in the exploratory set (Δ𝐴= −0.048, 95% CI −0.073 to −0.020; 20 zones, 225 targets) and in three further sets of never-analysed zones defined before data download: Δ𝐴= −0.079 (−0.094 to −0.064; 28 zones, 541 targets), −0.071 (−0.093 to −0.046; 18 zones, 229 targets) and −0.091 (−0.119 to −0.063; 18 zones, 202 targets). Across all 66 zones of the first three sets, P-HRI also outperformed the ETAS-type benchmark (𝐴 = 0.400 vs 0.471). Four probabilistic analyses, including a 30-day information-gain test whose code was first shown to detect a perfect synthetic predictor and reject a random one, found no measurable gain over a zone-fitted ETAS model outside sequences. A preregistered variant of the index did not improve on the original. The temporal structure of microseismicity therefore carries replicable, regime-dependent information on the relative likelihood of M ≥ 5.5 earthquakes beyond event rate; it supports ranking, not probabilistic forecasting or deterministic prediction. Keywords earthquake forecasting · microseismicity · foreshocks · Molchan diagram · tectonic regime · ETAS · replication · preregistration