🔬 physics

Seismic noise suppression: array stations, waveform cross-correlation, and noise stochastization

This paper evaluates the effectiveness of seismic array beamforming, waveform cross-correlation (WCC), and a novel "noise stochastization" technique in suppressing ambient noise to improve the detection of low-amplitude seismic signals.

Original authors: Ivan Kitov

Published 2026-04-27

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Ivan Kitov

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The "Noisy Party" Problem: How Scientists Listen for Earthquakes in a Storm

Imagine you are at a massive, crowded music festival. You are trying to hear a friend whisper a specific secret to you from across the field.

The problem? There are two things making it impossible:

The Background Hum: The constant, low-level roar of thousands of people talking and the bass from the speakers.
The "Fake" Secret: Suddenly, a massive group of people nearby starts shouting the exact same secret your friend is trying to tell you. Because they are shouting it so loudly and in unison, your brain can’t distinguish your friend’s quiet whisper from the loud, rhythmic shouting of the crowd.

In seismology (the study of earthquakes), scientists face this exact problem. They are looking for tiny, important "whispers" (small earthquakes or underground tests) that are being drowned out by the "shouting" of massive earthquakes or the constant "hum" of the ocean.

This paper explains new ways to "clean up the audio" so scientists can hear those tiny whispers again.

The Three Tools in the Scientist's Toolkit

The author, Ivan Kitov, discusses three main ways to deal with this noise:

1. The "Array" (The Group of Microphones)

Instead of using just one microphone, scientists use an array—a group of many microphones spread out over a large area.

The Analogy: Imagine if you had 20 microphones. If a random noise (like a person walking by) hits them at different times, it sounds messy and cancels itself out. But if a real signal (your friend's whisper) hits them all in a synchronized way, the microphones can "team up" to boost that specific sound.
The Catch: This doesn't work if the noise is "coherent"—meaning the noise is also synchronized, like a giant drum beat that hits all the microphones at once.

2. Waveform Cross-Correlation (The "Pattern Matcher")

This is a mathematical way of saying, "I'm looking for this specific shape of sound."

The Analogy: Imagine you have a photo of a specific person's face. You scan a crowded room, looking for anyone who matches that exact pattern. Even if the room is noisy, if you find a perfect match for the pattern, you know you've found your person. Scientists use "templates" (patterns of known earthquakes) to scan through the noisy data to find matches.

3. Noise Stochastization (The "White Noise" Trick)

This is the most creative part of the paper. If the noise is too "organized" (like that loud shouting group), it tricks the Pattern Matcher. To fix this, scientists add even more random noise on purpose.

The Analogy: Imagine you are trying to listen to a rhythmic drum beat that is masking your friend's whisper. To break that rhythm, you turn on a "white noise" machine—a static sound like a radio between stations. This random static breaks up the organized pattern of the drum beat, making the "shouting" sound less organized and allowing the "whisper" to pop out.

The Big Discovery: The "Sweet Spot"

The paper shows that there is a "Goldilocks Zone" for adding this extra noise.

If you add too little random noise, the loud, organized "shouting" still wins. If you add too much, you drown out the whisper entirely. But if you add the perfect amount, you "scramble" the loud noise just enough that the Pattern Matcher can suddenly see the tiny earthquake signal clearly.

Why does this matter?

When a massive earthquake (like the 2011 Tohoku earthquake in Japan) happens, it creates a "blind spot" for scientists. For several minutes or even hours, the Earth is "shouting" so loudly that we can't see any other small earthquakes happening.

By using these mathematical tricks—scrambling the noise and using smart pattern matching—scientists can "see" through the chaos. This helps them keep better records of Earth's activity, which is vital for monitoring nuclear tests and understanding earthquake patterns to keep people safe.

Technical Summary: Seismic Noise Suppression via Array Stations, Waveform Cross-Correlation, and Noise Stochastization

Author: Ivan O. Kitov
Subject Area: Seismology, Signal Processing, International Monitoring System (IMS)

1. The Problem: Coherent Noise and Detection Thresholds

The primary challenge in seismological monitoring—particularly for the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO)—is detecting low-magnitude seismic events buried in ambient noise. While standard seismic noise is often quasi-random (stochastic), a critical problem arises when the noise is coherent with the sought signal.

Coherent noise occurs when high-amplitude seismic waves (e.g., from a major earthquake like Tohoku 2011) propagate in a similar manner to the target signal. In such cases:

Beamforming (the standard array processing method) fails because it relies on the destructive interference of random noise; coherent noise instead interferes constructively, masking the signal.
Detection Thresholds increase significantly, creating "blind spots" where aftershocks or small events cannot be detected for minutes or even hours following a catastrophic event.

2. Methodology

The paper evaluates and compares several advanced signal processing techniques designed to suppress coherent noise and improve the Signal-to-Noise Ratio (SNR) using Waveform Cross-Correlation (WCC) and seismic arrays.

Waveform Cross-Correlation (WCC): A matched filter method that uses templates (actual or simulated seismic signals) to identify similar waveforms in the data. It is more robust than beamforming in certain noise environments.
Seismic Array Exploitation (Turning Incident Azimuths): A technique where a multichannel signal is "turned" (by adjusting relative arrival times at sensors) to an angle orthogonal to the sought signal. This transforms coherent noise into incoherent/stochastic noise, facilitating destructive interference.
Digital Noise Stochastization (Whitening): The addition of computer-generated, scaled random noise to the actual data before calculating the cross-correlation coefficient. This is used to "break" the coherence of the ambient noise.
Joint Application: The study tests the combined effect of adding both a regular (but non-coherent) seismic wave and digital stochastic noise to optimize detection.

3. Key Contributions

Quantification of "Shake" and "Stochastization" effects: The paper provides a mathematical and empirical framework for using scaling factors (SeisN for regular seismic noise and StochN for digital stochastic noise) to tune detection sensitivity.
Analysis of Nonlinear Interactions: It demonstrates that the effectiveness of noise suppression is highly dependent on the incidence angle between the noise wave and the sought signal.
Optimization Framework: The research identifies that there is no "one-size-fits-all" parameter; rather, an optimal $StochN$ must be selected based on the specific geometry and amplitude of the transient noise.

4. Results

WCC vs. Beamforming: In environments with high coherent noise (e.g., $SeisN=5.0$), the WCC method provides a significant advantage, improving the $SNR_{cc}/SNR$ ratio by a factor of 10 to 15.
Stochastization Performance: Adding digital noise can actually increase the $SNR_{cc}$ peak. For example, at station ASAR, adding stochastic noise ($StochN=0.17$) increased the $SNR_{cc}$ peak from 42 to 55.
Angle Dependency: The paper shows that when the noise plane is collinear with the sought signal (e.g., at ~55° in certain tests), the $SNR_{cc}$ drops into a "trough." However, applying a higher $StochN$ can mitigate this trough, effectively "lifting" the signal out of the coherent noise.
Detection Thresholds: The methods demonstrated the ability to find signals even when the added stochastic noise is 10 times higher in amplitude than the signal itself.

5. Significance

This research is vital for the International Monitoring System (IMS) and the International Data Centre (IDC). By implementing these advanced WCC and stochastization techniques, seismologists can:

Reduce the "Blind Spot": Detect aftershocks and small seismic events much sooner after a major earthquake.
Improve Catalog Completeness: Enhance the accuracy and consistency of seismic bulletins (like the REB and SEL3).
Enhance Treaty Verification: Improve the capability to detect low-magnitude underground explosions that might otherwise be masked by natural seismic activity.