The shutdown-and-resume evidence chain.
Industrial clients (smelters, semiconductor fabs, rail) need more than "risk is elevated." They need to hear the alarm 7 days out, watch it narrow at 36 hours, and get a final industrial-grade alert at 24 hours — per region, with an expected magnitude band and a hash-chained audit trail. Then, after the event, an aftershock forecast so they can safely resume.
Model: magnitude_range_predictor_v1 built on Phase 6 regressor (MAE 0.41 subduction / 0.42 intraplate). Band width starts at M0.5 per region and auto-narrows to M0.25 once the region accumulates ≥ 30 matched observations with robust residual σ ≤ 0.30. Endpoint: /api/v1/research/magnitude-cascade.
Per-region activity (top 10 by cascade volume)
Each region independently decides whether to fire T-7d, T-36h, and T-24h based on its own confidence score. "Tiers fired" shows which cascade levels the region has actually triggered in the current window — the progression itself is the evidence.
| Region | Total emitted | Matched obs | Region hit rate | Tiers fired | Band width |
|---|---|---|---|---|---|
| Reading ledger… | |||||
What this page is (and isn’t)
Is: a per-region, per-tier ledger of exactly what DeepMap AI emitted, when, and at what confidence — with a hash-chained audit trail so any buyer can verify the tightening happened in real time. Each tier carries an expected magnitude band (M0.5 by default), a point estimate, a 95% CI, and a SHA-256 link to the prior prediction in the chain.
Isn’t: a claim that every region fires all three tiers. The cascade only tightens when the Phase 6 classifier confidence crosses 0.75 (T-36h) and 0.85 (T-24h). Many regions will sit at T-7d for long windows and never promote — that’s the system refusing to over-commit. The graduation of a region to M0.25 band width only happens after the observed residual sigma proves the tighter band is supportable; we don’t narrow by fiat.
Reference — calibration: Wells & Coppersmith, 1994, New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement, BSSA 84(4). Region ceilings derived from region-specific maximum instrumental magnitudes (USGS Unified Catalog 1900–present).