A capture gap, not a precursor, before the Pisa earthquake

A M4.3 earthquake struck near Pisa on 4 August 2026 at 08:15 UTC, about 7 km from the city and 275 km from the Cumiana VLF receiver. Checking the two weeks before it against the Cumiana anomaly record looked briefly promising: the two highest-scoring days in the entire record fall inside that window.

They are an artifact. The collector recorded nothing between 16 and 29 July, and the anomaly scorer uses a 24-frame lookback, so every window straddling that 13-day gap reconstructs across a discontinuity and scores near the maximum. Those 21 windows average 0.9113; the next 24 clean windows drop as a step function to 0.4074. The high scores mark the collector restarting, not the ionosphere doing anything.

What the clean data shows runs the other way. The observed pre-event stretch averages 0.4127 against a 0.5429 pre-gap baseline, so the period before the earthquake was quieter than usual, and the strongest sustained activity in the record is after the event rather than before it.

The coverage problem is on its own disqualifying: with no captures from 17 to 28 July, the first eight of the fourteen pre-event days have no data, and more than half the window cannot be checked either way.

The useful outcome is a defect rather than a result. Every collector restart currently injects a false top-ranked anomaly into the record, which is precisely the failure mode that would manufacture a precursor claim from an outage. A gap-aware guard that flags windows spanning a capture gap longer than the lookback is now queued before this scorer is used for event association again.

This remains a single-event retrospective with one station, no directional information, and no matched control windows. It is not evidence for or against a VLF-earthquake relationship.

See the full analysis and next actions.