Scholarlink

<p>This study evaluates the SAR model, combining RNN, GRU, and multi-head attention mechanisms, for microseismic phase picking using 87,050 three-component waveforms from the Three Gorges Reservoir area (2017–2020). Data preprocessing involved bandpass filtering (2–40 Hz) and phase association via the PAL method. Parameter optimization yielded a learning rate of 1×10<sup>−4</sup>, 15 epochs, and a single attention head. The optimized model achieved P- and S-wave arrival-time errors of 0.68 s and 1.28 s, respectively, with greater than 92% valid picks within 2σ. Compared to the PAL method, it reduced the mean residual to 0.12 s. The SAR-derived catalog detected 3,022 events (3 times more than the network catalog) while maintaining a consistent completeness magnitude (Mc0.8) and identifying previously undetected shallow events (0–5 km) aligned with fault zones.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Sensitivity Analysis of a SAR Model for Microseismic Monitoring in the Three Gorges Reservoir Area

  • 期刊论文

  • 发表日期：

    2026年3月11日

查看全文
  • Xiao-tong Shen,
  • Li-fen Zhang,
  • Jing-gang Li,
  • Yan-nan Zhao,
  • Sheng-qi Dai,
  • Jing-xuan Huang

摘要

This study evaluates the SAR model, combining RNN, GRU, and multi-head attention mechanisms, for microseismic phase picking using 87,050 three-component waveforms from the Three Gorges Reservoir area (2017–2020). Data preprocessing involved bandpass filtering (2–40 Hz) and phase association via the PAL method. Parameter optimization yielded a learning rate of 1×10−4, 15 epochs, and a single attention head. The optimized model achieved P- and S-wave arrival-time errors of 0.68 s and 1.28 s, respectively, with greater than 92% valid picks within 2σ. Compared to the PAL method, it reduced the mean residual to 0.12 s. The SAR-derived catalog detected 3,022 events (3 times more than the network catalog) while maintaining a consistent completeness magnitude (Mc0.8) and identifying previously undetected shallow events (0–5 km) aligned with fault zones.