隨著地震儀的品質及數量增加，地震記錄的資料量也呈現十足的增長，傳統以人工挑選波相到時的工作已面臨人力不足的問題。本研究建立了一個基於深度學習的地震相位挑選模型，自動並穩定地進行P 波及S 波的挑選。品質良好的波相挑選結果對於研究工作至關重要，我們設計並執行了一套工作流程以展示其潛在應用。首先，我們以反投影法進行初步的地震偵測及定位，得到一個粗略的地震目錄及其反演之波相到時。接著利用波相挑選模型調整波相到時，並進行地震重新定位。最後再對重新定位之結果進行基於波型相似度的模板搜尋，在連續資料上找出更多重複發生的地震事件。本研究將此套流程應用於台灣大屯火山區域於2014/02/08 至 2014/02/18，士林地震(規模4.2)發生期間之連續地震記錄。我們使用2015 年至2019年間，台灣中央氣象局(CWB)提供的598,795 筆地震資料進行波相挑選模型的訓練及驗證，並使用2013 至2014 年間的資料進行測試；而所有地震波相的訊噪比皆大於5。在模型的測試結果中，P 波及S 波的檢出率分別為99.96%及99.5%；平均時間誤差分別為0.07 秒及0.16 秒，相對應的標準誤差為0.07 秒及0.18 秒。而最終由模板搜尋法找到的地震數目(2503 個)約為氣象局目錄(51 個)的50 倍。比較這兩個地震目錄，分別有50 個事件發生的時間相隔不到1 秒鐘(約佔了氣象局目錄的98%)，我們認為它們分別指向同樣的地震事件。

With the growing volume of seismic recordings, routine work of manual seismic phase picking would be faced with insufficient manpower. In this study, we build a deep learning model, attention U-Net, for automatic and stable seismic phase picking (both P and S phase). The well-constrained seismic phases could also benefit scientific researches. We designed a workflow to demonstrate its potential applications. We utilized phase picking model to constrain the preliminary earthquake catalog produced by back-projection. Earthquake re-location is then implemented using the constrained picks; and finally the relocation results and re-picked results are fed to execute template matching. The experiment is conducted using seismic recordings in Tatun area in Taiwan during the time period (2014/02/08 - 2014/02/18) before and after Shi-Lin earthquake (ML 4.2) occurred. The phase picking model is trained/validated on 598,795 data provided by the Central Weather Bureau (CWB, Taiwan) during the year from 2015 to 2019, and tested on 305,026 data during the year from 2013 to 2014. Training/validation data and testing data of P and S phase were all filtered by SNR threshold of 5. Picking rate of P and S phase are 99.96% and 99.5%; mean time residual of P and S are 0.07s and 0.16s with the standard deviation of 0.07s and 0.18s. The final detected events (2503) produced by template matching are about 50 times larger than that of CWB catalog (51). In these two catalogs, respective 50 events occurred less than 1 second apart, occupied 98% of CWB catalog. We considered that they indicate the identical events separately.

Allen, R. V., (1982). "Automatic phase pickers: Their present use and future prospects". Bull. Seismol. Soc. Am., 72, S225.
Bahdanau, D., Cho, K., and Bengio, Y. (2014). "Neural Machine Translation by Jointly Learning to Align and Translate". ArXiv14090473 Cs Stat.
Chen, Y. (2018). "Automatic microseismic event picking via unsupervised machine learning". Geophys. J. Int. 212, 88–102.
Chen, C., and Holland, A.A. (2016). "PhasePApy: A Robust Pure Python Package for Automatic Identification of Seismic Phases". Seismol. Res. Lett. 87, 1384–1396.
Gentili, S., and Michelini, A. (2006). "Automatic picking of P and S phases using a neural tree". J. Seismol. 10, 39–63.
Huang, B.-S. (2008). "Tracking the North Korean nuclear explosion of 2006, using seismic data from Japan and satellite data from Taiwan". Phys. Earth Planet. Inter. 167, 34–38.
Huang, H.-H., Wu, Y.-M., Song, X., Chang, C.-H., Lee, S.-J., Chang, T.-M., and Hsieh, H.-H. (2014). "Joint Vp and Vs tomography of Taiwan: Implications for subduction-collision orogeny". Earth Planet. Sci. Lett. 392, 177–191.
Igarashi, T., Matsuzawa, T., and Hasegawa, A. (2003). "Repeating earthquakes and interplate aseismic slip in the northeastern Japan subduction zone: REPEATING EARTHQUAKES AND ASEISMIC SLIP". J. Geophys. Res. Solid Earth 108.
Ioffe, S., and Szegedy, C. (2015). "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift". ArXiv150203167 Cs.
Ishii, M., Shearer, P.M., Houston, H., and Vidale, J.E. (2005). "Extent, duration and speed of the 2004 Sumatra–Andaman earthquake imaged by the Hi-Net array". Nature 435, 933–936.
Jurkevics, A, (1988). "Polarization Analysis of Three Component Array Data", Bull. Seismol. Soc. Am., 78, 1725-1743.
Kao, H., and Shan, S.-J. (2004). "The Source-Scanning Algorithm: mapping the distribution of seismic sources in time and space". Geophys. J. Int. 157, 589–594.
Kiefer, J., Wolfowitz, J., (1952). "Stochastic estimation of the maximum of a regression function", Ann. Math. Stat., vol. 23, pp. 462-466.

Kingma, D.P., and Ba, J. (2014). "Adam: A Method for Stochastic Optimization". ArXiv14126980 Cs.
Koper, K.D., Hutko, A.R., Lay, T., Ammon, C.J., and Kanamori, H. (2011). "Frequency-dependent rupture process of the 2011 M w 9.0 Tohoku Earthquake: Comparison of short-period P wave backprojection images and broadband seismic rupture models". Earth Planets Space 63, 599–602.
Krizhevsky, A., Sutskever, I., and Hinton, G.E. (2017). "ImageNet classification with deep convolutional neural networks". Commun. ACM 60, 84–90.
Langet, N., Maggi, A., Michelini, A., and Brenguier, F. (2014). "Continuous Kurtosis‐Based Migration for Seismic Event Detection and Location, with Application to Piton de la Fournaise Volcano, La RéunionContinuous Kurtosis‐Based Migration for Seismic Event Detection and Location". Bull. Seismol. Soc. Am. 104, 229–246.
Lecun, Y., Bottou, L., Bengio, Y., and Haffner, P. (1998). "Gradient-based learning applied to document recognition". Proc. IEEE 86, 2278–2324.
Lee, E.-J., Liao, W.-Y., Lin, G.-W., Chen, P., Mu, D., and Lin, C.-W. (2019). "Towards Automated Real-Time Detection and Location of Large-Scale Landslides through Seismic Waveform Back Projection". Geofluids 2019, 1–14.
Leonard, J., and Kramer, M.A. (1990). "Improvement of the backpropagation algorithm for training neural networks". Comput. Chem. Eng. 14, 337–341.
Lomax, A., Michelini, A., and Curtis, A. (2009). "Earthquake Location, Direct, Global-search Methods". In Encyclopedia of Complexity and Systems Science, R.A. Meyers, ed. (New York, NY: Springer New York), pp. 2449–2473.
Lewis, J.P. (1995). "Fast normalized cross-correlation". Proceedings of Vision Interface, pp. 120–123
Mu, D., Lee, E.-J., and Chen, P. (2017). "Rapid earthquake detection through GPU-Based template matching". Comput. Geosci. 109, 305–314.
Oktay, O., Schlemper, J., Folgoc, L.L., Lee, M., Heinrich, M., Misawa, K., Mori, K., McDonagh, S., Hammerla, N.Y., Kainz, B., et al. (2018). "Attention U-Net: Learning Where to Look for the Pancreas". ArXiv180403999 Cs.
Peng, Z., and Gomberg, J. (2010). "An integrated perspective of the continuum between earthquakes and slow-slip phenomena". Nat. Geosci. 3, 599–607.
Robbins, H., and Monro, S. (1951). "A Stochastic Approximation Method". Ann. Math. Statist., vol. 22, pp. 400-407.
Ronneberger, O., Fischer, P., and Brox, T. (2015). "U-Net: Convolutional Networks for Biomedical Image Segmentation". ArXiv150504597 Cs.
Rosenberger, A. (2010). "Real-Time Ground-Motion Analysis: Distinguishing P and S Arrivals in a Noisy Environment". Bull. Seismol. Soc. Am. 100, 1252–1262.
Ross, Z.E., and Ben-Zion, Y. (2014a). "An earthquake detection algorithm with pseudo-probabilities of multiple indicators". Geophys. J. Int. 197, 458–463.
Ross, Z.E., and Ben-Zion, Y. (2014b). "Automatic picking of direct P, S seismic phases and fault zone head waves". Geophys. J. Int. 199, 368–381.
Ross, Z.E., Trugman, D.T., Hauksson, E., and Shearer, P.M. (2019). "Searching for hidden earthquakes in Southern California". Science 364, 767–771.
Saragiotis, C.D., Hadjileontiadis, L.J., and Panas, S.M. (2002). "PAI-S/K: A robust automatic seismic P phase arrival identification scheme". IEEE Trans. Geosci. Remote Sens. 40, 1395–1404.
Schaff, D.P., and Beroza, G.C. (2004). "Coseismic and postseismic velocity changes measured by repeating earthquakes: COSEISMIC AND POSTSEISMIC VELOCITY CHANGES". J. Geophys. Res. Solid Earth 109.
Sleeman, R., and van Eck, T. (1999). "Robust automatic P-phase picking: an on-line implementation in the analysis of broadband seismogram recordings". Phys. Earth Planet. Inter. 113, 265–275.
Thurber, C., and Eberhart-Phillips, D. (1999). "Local earthquake tomography with flexible gridding". Comput. Geosci. 25, 809–818.
Trugman, D.T., and Shearer, P.M. (2017). "GrowClust: A Hierarchical Clustering Algorithm for Relative Earthquake Relocation, with Application to the Spanish Springs and Sheldon, Nevada, Earthquake Sequences". Seismol. Res. Lett. 88, 379–391.
Turin, G. (1960). "An introduction to matched filters". IEEE Trans. Inf. Theory 6, 311–329.
Uchida, A., Ito, Y., and Nakano, K. (2011). "Fast and Accurate Template Matching Using Pixel Rearrangement on the GPU. In 2011 Second International Conference on Networking and Computing, (Osaka, Japan: IEEE), pp. 153–159.
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, L., and Polosukhin, I. (2017). "Attention Is All You Need". ArXiv170603762 Cs.
Yamada, M., Matsushi, Y., Chigira, M., and Mori, J. (2012). "Seismic recordings of landslides caused by Typhoon Talas (2011), Japan: SEISMIC RECORDINGS OF LANDSLIDES". Geophys. Res. Lett. 39, n/a-n/a.
Yoon, C.E., O’Reilly, O., Bergen, K.J., and Beroza, G.C. (2015). "Earthquake detection through computationally efficient similarity search". Sci. Adv. 1, e1501057.
Zhang, H. (2003). "Automatic P-Wave Arrival Detection and Picking with Multiscale Wavelet Analysis for Single-Component Recordings". Bull. Seismol. Soc. Am. 93, 1904–1912.
Zhao, Y., and Takano, K. (1999). "An Artificial Neural Network Approach for Broadband Seismic Phase Picking". Bull. Seismol. Soc. Am., vol. 77 89 pp. 670-680
Zhou, Z., Siddiquee, M.M.R., Tajbakhsh, N., and Liang, J. (2018). "UNet++: A Nested U-Net Architecture for Medical Image Segmentation". ArXiv180710165 Cs Eess Stat.
Zhu, W., and Beroza, G.C. (2018). "PhaseNet: A Deep-Neural-Network-Based Seismic Arrival Time Picking Method". Geophys. J. Int.
Zhu, L., Peng, Z., McClellan, J., Li, C., Yao, D., Li, Z., and Fang, L. (2019). "Deep learning for seismic phase detection and picking in the aftershock zone of 2008 Mw7.9 Wenchuan Earthquake". Phys. Earth Planet. Inter. 293, 106261.