Automated P-wave arrival picking in microseismic monitoring: Integrating multi-feature clustering and enhanced AIC-STA/LTA

Accurate P-wave arrival time picking remains a major challenge in microseismic event analysis due to variable signal-to-noise ratio (SNR) conditions and complex waveform characteristics. We propose a unified and adaptive framework that effectively handles both. For high-SNR signals, we enhance the AIC-STA/LTA method by introducing a novel sequence segmentation strategy and precise extremum identification. For low-SNR scenarios, we design a robust three-domain feature fusion scheme—combining time-domain short-time energy, cepstral-domain MFCC, and statistical-domain kurtosis—followed by K-means clustering to achieve accurate waveform segmentation. Validation on real engineering datasets shows that our method bridges the SNR disparity with significantly improved picking accuracy. Specifically, it increases the proportion of small-error picks (≤5ms) by 20 % and reduces large-error picks (>20 ms) by 50 %, thereby minimizing manual correction, which is essential for preserving the accuracy of subsequent event location. These advancements greatly reduce human intervention and enhance the reliability and scalability of automated microseismic monitoring systems.