FAST Recursive Cross-Correlation Detection

Harvard EPS55

Fingerprint And Similarity Thresholding - detecting similar earthquakes by cross-correlating all events with each other

0.60
2.0s

Cross-Correlation Matrix

Events →
Events →
1.0
0.5
0.0
0
Total Events
0
Unique Templates
0
Event Clusters
0.00
Avg Similarity

How FAST Works

FAST automatically discovers similar earthquakes without pre-defined templates:

1. Fingerprinting: Create compact representations of all detected events
2. All-vs-All Correlation: Compare every event with every other event
3. Similarity Matrix: Build N×N matrix of correlation values
4. Clustering: Group events above similarity threshold
5. Template Selection: Choose best representative from each cluster
Similarity(i,j) = max(xcorr(event_i, event_j))

Advantages of FAST

  • No prior templates needed - discovers patterns automatically
  • Finds repeating sources - identifies earthquake families
  • Robust clustering - handles noise and variations
  • Scalable - efficient fingerprinting reduces computation
  • Objective - no manual template selection bias

Key Innovation: Discovers earthquake families without knowing what to look for!

Applications & Performance

Ideal for:

  • Discovering repeating earthquake sources
  • Identifying aftershock families
  • Finding induced seismicity patterns
  • Volcanic earthquake classification
  • Building template catalogs automatically

Typical Performance:

Metric Value
Processing Speed ~1000 events/hour
Minimum Similarity 0.6-0.8
Cluster Accuracy >90%