A Detailed Spatiotemporal Wavelet Study To Improve The P-Phase Picking Performance For The 2007-2010 Shallow Earthquake Swarms Near Matata, New Zealand
BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA(2018)
摘要
We report a 45% overall improvement by incorporating wavelet scale thresholding (WST) into the GeoNet P-phase picker. We study the spatiotemporal effect of the WST on performance of the automatic phase picking using 6471 waveforms with manual picks recorded by seven stations for 3312 Matata events over the 4-year period of 2007-2010. We identify whether the Haar wavelet or Mexican hat wavelet performs the best for each station by comparing seven quality control parameters, including total and net improvements. We show that the Mexican hat wavelet outperformed the Haar wavelet for recordings of Stations EDRZ, LIRZ, KARZ, URZ, and TGRZ, whereas the Haar wavelet was preferred for Stations OPRZ and MARZ.Having applied the Mexican hat WST, we obtained the largest average total improvement (64.8%) and onset retrieval (32%) for the EDRZ station and the largest average onset revision (37.8%) and picking quality enhancement (29%) for the KARZ station in 2007-2010. Having the highest rate of manually picked onsets at the EDRZ station confirms the merit of the results and the superiority of the WST over the current Butterworth filtering. We experienced the smallest rates of onset retrieval (average of 2.8%) and its associated temporal variation at the URZ station, which is likely due to the high-quality data recorded by the only deep-borehole-installed sensor at this station. Applying the Haar WST to the MARZ recordings retrieved missed onsets with a total average rate of 22.9%, which is significant because the MARZ station contributed waveforms for 93.3% of the chosen events. Overall, the improved time-frequency localization of the WST resulted in improving 2914 out of 6471 onsets (including 1323 retrieved and 1591 revised) by enhancing either initial detection capability or noise-signal modeling quality (or both) at each station.
更多查看译文
AI 理解论文
溯源树
样例
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要