Two-Stage Microseismic P-Wave Arrival Picking via STA/LTA-Guided Lightweight U-Net

Area of Science:

• Geophysics
• Seismology
• Machine Learning

Background:

• Accurate P-wave arrival time picking is crucial for analyzing mining-induced seismic events.
• Traditional STA/LTA detectors are noise-sensitive, while deep learning methods can be computationally intensive.
• Existing methods struggle to balance efficiency and accuracy in noisy microseismic data.

Purpose of the Study:

• To develop a robust and efficient two-stage P-wave picking framework for microseismic event monitoring.
• To improve the accuracy and real-time performance of seismic arrival time detection.
• To address the limitations of conventional STA/LTA and deep learning methods.

Main Methods:

• A two-stage framework integrating STA/LTA for preliminary detection and a lightweight U-Net for refinement.
• STA/LTA scans continuous waveforms for candidate P-wave arrival windows.
• Lightweight U-Net performs sample-level regression for precise arrival time estimation.

Main Results:

• The proposed method achieved a 63.21% hit rate within a ±0.01 s tolerance.
• Demonstrated significant improvements over CNN (25.42%) and STA/LTA (40.47%) methods.
• Reduced mean absolute error to 0.0130 s with strong generalization and noise robustness.

Conclusions:

• The coarse-to-fine framework effectively balances computational efficiency and P-wave picking accuracy.
• The integrated approach shows significant potential for real-time industrial applications in microseismic monitoring.
• The method offers a robust solution for accurate seismic event analysis in challenging environments.