Matched Filtering for Earthquake Detection

Harvard EPS55

Interactive demonstration of template matching technique for detecting small earthquakes hidden in seismic noise

0.70

Matched Filtering Process

1. Template Event

Known earthquake waveform

→

2. Continuous Data

Seismic station recording

→

3. Cross-Correlation

Slide template across data

→

4. Detection

Peaks above threshold

Template Event
No Detection
Max Correlation: 0.00

How Matched Filtering Works

Template matching uses cross-correlation to find similar waveforms in continuous data:

CC(t) = Σ[template(i) × data(t+i)] / √[Σtemplate²(i) × Σdata²(t+i)]
  1. Select a template waveform from a known earthquake
  2. Slide template across continuous seismic data
  3. Calculate normalized cross-correlation at each position
  4. Identify peaks above detection threshold
  5. Peaks indicate similar earthquakes at those times
Correlation Values:
• Background noise: ~0.0-0.2
• Different event: ~0.3-0.5
• Similar event: ~0.8-1.0

Detection Parameters

Parameter Typical Value Purpose
Template Length 2-10 seconds Capture P & S waves
Correlation Threshold 0.6-0.8 Balance detections/false positives
Frequency Band 1-15 Hz Focus on earthquake signals
Network Channels 3-12 Improve detection robustness

Detection capability: Can find earthquakes 10-100× smaller than traditional methods

Advantages & Applications

✓ Advantages:
• Detects events below noise level
• Finds repeating earthquakes
• Automated processing
• Objective detection criteria
• Works in high-noise environments

Applications:

  • Aftershock sequences: Find small events following large earthquakes
  • Induced seismicity: Monitor fracking, geothermal sites
  • Volcanic monitoring: Detect repeating volcano-tectonic events
  • Nuclear monitoring: Identify small explosions
  • Fault studies: Map microseismicity patterns