论文部分内容阅读
利用奇异值分解(SVD)分析实现了一种新的P波和S波震相的自动检测方法。该方法是基于Rosenberger(2010)用于三分量地震波形奇异值分解的实时迭代算法。该算法通过利用奇异值分解并将波形分为P波和S波分量来识别视入射角。我们将该算法应用于滤波后的波形,然后在经过滤波和奇异值分解分开的信道上设置入射角和奇异值检测器,或者应用信噪比(SNR)检测器来拾取P波和S波。安扎地震台网和最近在圣哈辛托断层带地区部署的便携式仪器为测试不同设置的检测算法提供了一个非常密集的地震台网,包括不同震源机制的事件、具有不同场地特性的台站和射线路径偏离奇异值分解算法中使用的逼近。2~30Hz巴特沃思带通滤波器给出了各种事件和台站的最佳性能。我们在许多事件上应用奇异值分解检测,并且从2005年6月M_W5.2地震复杂、强烈的余震序列中得到结果。这个序列经过几位分析人员的彻底复查,确定了主震后第一个小时的294个事件都围绕主震密集分布。我们使用这个数据集来微调自动奇异值分解检测、关联和定位,实现了37%事件的自动识别和定位。所有检测到的事件都落在此密集区内,并且没有虚假的事件。普通的信噪比检测器不会超过11%的成功,并且位置分布更广泛(不完全在复查的群集内)。由奇异值分解检测器检测到震相(P波或S波)的预先知识显著降低了由震相盲信噪比检测器产生的噪声。
A new method of automatic detection of P-wave and S-wave phases is realized by singular value decomposition (SVD) analysis. The method is based on the Rosenberger (2010) real-time iterative algorithm for singular value decomposition of three-component seismic waveforms. The algorithm identifies the apparent incident angle by using singular value decomposition and dividing the waveform into P-wave and S-wave components. We apply this algorithm to the filtered waveform and then set the angle of incidence and singularity detectors on the separate channels filtered and singularly-valued decomposed, or pick up the P and S waves using a signal-to-noise ratio (SNR) detector. The Anza Seismic Network and the recent portable instruments deployed in the San Jacinto Fault Zone provide a very dense seismograph network for testing different setup detection algorithms, including events of different focal mechanisms, stations with different site characteristics And the ray path deviates from the approximation used in the singular value decomposition algorithm. The Butterworth bandpass filter, 2 to 30Hz, gives the best performance for a variety of events and stations. We applied singular value decomposition detection on many events and obtained the results from the complex and intense aftershock sequence of the June 2005 M_W5.2 earthquake. After a thorough review by several analysts, this sequence confirmed that 294 events in the first hour after the main shock concentrated around the main shock. We used this dataset to fine-tune the detection, correlation and localization of automatic singular value decomposition to achieve the automatic recognition and localization of 37% of events. All detected events fall within this dense area with no false events. The average signal to noise ratio detector will not exceed 11% of the success, and the location of distribution more widely (not fully in the review of the cluster). The prior knowledge of the phase (P-wave or S-wave) detected by the singular value decomposition detector significantly reduces the noise generated by the phase blind blind signal-to-noise ratio detector.