Machine-Learning Phase Pickers in a Regional Warning Network: Picking Latency, Association and Magnitude Budgets — A Computed Study
A regional network replaces its classical triggering algorithms with neural phase pickers trained on a decade of analyst-labelled waveforms; the pickers read each incoming trace in tens of milliseconds, associate picks across stations into events, and hand cleaner, earlier P onsets to the same magnitude machinery the network has always used. Its operating principle: neural-network phase pickers trained on analyst-labelled archives, shortening detection by picking emergent onsets earlier and more consistently. Headline values are ml p travel ml to nearest station ml s = 2.694; ml detection latency ml s = 7.197; ml total alert ml latency s = 8.797; ml blind zone ml radius km = 29.12; ml warning time ml at 50 km ml s = 5.772; ml warning time ml at 100 km ml s = 19.92. The learned component is confined to picking and association in these entries, so its gains can be compared line by line with the classical chain it replaces. SIMULATED, NOT MEASURED. Stated as a computed study for public safety and infrastructure protection, the learned picker is detection software for a warning network, with no targeting function and with no weapon application directed at anyone, any nation, or any earthly body.
Authors
- William Alexander Aldred (ORCID: https://orcid.org/0009-0000-1215-577X)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-16
- DOI
- https://doi.org/10.5281/zenodo.24265340
- Primary Topic
- Seismology and Earthquake Studies
- Type
- article
- Field-Weighted Citation Impact
- 0.00