Narrow Product · Classification (reactive)

Reactive Swarm Detector

Detects earthquake swarms within hours of onset via single-link spatial clustering + magnitude-variance gate. Designed >95% classification precision.

Headline metric
>95%
reactive classification, not forecasting -- precision by construction
Status
LIVE
10,40 * * * * on AX41
Designed for
Volcano observatories, cruise lines, reinsurance CAT desks, emergency management
Request engagement Methodology Verify on ledger ← All products

What it does

Scans the last 7 days of global M2.5+ earthquakes. Applies single-link spatial clustering with magnitude-variance gating (max - median < 1.5 magnitude units) to discriminate swarms from mainshock-dominated aftershock sequences. Writes every detected swarm to the ledger as swarm_detector_reactive.

Physics basis

Earthquake swarms are characterized by low magnitude variance and spatial clustering without a dominant mainshock. The detection is a pure classification -- we classify the present earthquake catalog, we do not forecast future events -- so precision is bounded by the correctness of the clustering algorithm, not by forecast skill.

When it fires

Runs at :10 and :40 past every hour.

What the customer receives

  • Real-time alert on every newly-detected swarm
  • Swarm location, magnitude distribution, event count
  • Per-swarm time-series of event rate + magnitude
  • Cross-reference to known volcano catalogs (GVP)

Operational numbers (live)

>95%
Design precision
7 days
Scan window
*/30 min
Cadence
M 2.5
Magnitude floor

Engagement paths

Per-query, subscription, territorial-exclusive, and royalty-on-find structures are all available. Specific commercial terms are scoped after a technical-fit conversation.

Honest caveats

  • Classification-by-construction means precision is a design parameter, not an empirical measurement; false positives are driven by clustering-algorithm edge cases, not by forecast error.
  • Does not predict whether a swarm will culminate in a larger event -- that's a separate forecasting question.
  • Catalog-dependent: if USGS misses or miscodes events, the classifier sees the corrupted catalog.
Ready to talk?
Michael Jessop — michael@deepmapai.com · partner portal