基于两阶段混合模型的地震数据到时拾取研究

Seismic Data Arrival Pickup Based on a Two-Stage Hybrid Model

摘要: 相位拾取是地震资料处理中一项基础而重要的工作。为解决在连续波形中，地震事件及初至波拾取准确率不高的难题，基于深度学习提出了一种两阶段混合拾取模型，分两步进行震相拾取。首先构建CNGNet模型检测地震事件；其次构建GPkNet模型并引入自注意力机制，对检测到的地震事件进行P波到时的精确拾取。用于网络训练和验证测试的数据来源于2019年加利福尼亚州地震序列的20个台站的连续波形。在测试集上，地震事件检测的精确率和召回率都达到99%以上，P波拾取到时的估计误差为0.019秒。

Abstract: Phase-Picking is a basic and important task in seismic data processing. In order to solve the problem of low accuracy of seismic event and first-to-wave pickup in continuous waveforms, a two-phase hybrid pickup model is proposed based on deep learning to perform phase pickup in two steps. Firstly, a CNGNet model is constructed to detect seismic events; secondly, a GPkNet model is constructed to accurately pick up the P-arrivals time for the detected events by introducing the self-attention mechanism and the gated recurrent unit. The data used for network training and validation testing are derived from continuous waveforms from 20 stations of the 2019 California earthquake sequence. The precision and the recall ratio of seismic event detection by our method are more than 98% for the test sets, and detection reaches more than 99%, and the estimated error of P-arrival is 0.019 s.