P Wave Arrival Picking and First‐Motion Polarity Determination With Deep Learning

By Zachary E. Ross, Men‐Andrin Meier, Egill Hauksson. Journal of Geophysical Research - Solid Earth, 2018

430 citations recorded by the source corpus.

Determining earthquake hypocenters and focal mechanisms requires precisely measured P wave arrival times and first‐motion polarities. Automated algorithms for estimating these quantities have been less accurate than estimates by human experts, which are problematic for processing large data volumes. Here we train convolutional neural networks to measure both quantities, which learn directly from seismograms without the need for feature extraction. The networks are trained on 18.2 million manually picked seismograms for the Southern California region. Through cross validation on 1.2 million independent seismograms, the differences between the automated and manual picks have a standard deviation of 0.023 s. The polarities determined by the classifier have a precision of 95% when compared with analyst‐determined polarities. We show that the classifier picks more polarities overall than the analysts, without sacrificing quality, resulting in almost double the number of focal mechanisms. The remarkable precision of the trained networks indicates that they can perform as well, or better, than expert seismologists.

Subfields: Artificial Intelligence, Geophysics

View Paper: https://www.semanticscholar.org/paper/85df8db729aac6f3a693b0cc4bcc3601fabfdede

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Geophysics

Citations in the corpus

Generalized Seismic Phase Detection with Deep Learning (2018, methodology): “In seismology, deep learning has recently been used to detect and roughly locate earthquakes in Oklahoma [20], and to determine P-wave arrival times and first-motion polarities [21].”
Seismic Signal Denoising and Decomposition Using Deep Neural Networks (2018, methodology): “Deep learning has been shown to be a powerful tool for learning the characteristics of seismic data [47, 48, 49, 50, 51, 52, 53, 54].”
Deep Learning for Geophysics: Current and Future Trends (2021, background): “With a sufficiently large training set, one can achieve remarkably picking and classification accuracies higher than STA/LTA (Zhao et al., 2019; Zhou et al., 2019), even close to or better than human experts (Ross et al., 2018, 4.5 million seismograms training set).”
Machine Learning in Seismology: Turning Data into Insights (2018, methodology): “…to simultaneously A convolutional feature extraction system is combined with a fully connected neural network (NN) to produce class probabilities for P waves, S waves, and noise (Ross, Meier, Hauksson, et al. , 2018). detect and locate earthquakes based on single-station waveform classification.”
Deep-learning seismology (2022, methodology): “The classification capability of DNNs makes them ideal for processing tasks such as event discrimination ( 6 – 8 ), event detection ( 9 , 10 ), phase picking ( 11 , 12 ), first-polarity determination ( 13 , 14 ), and phase association ( 15 , 16 ) that together form a seismicity monitoring workflow.”
CRED: A Deep Residual Network of Convolutional and Recurrent Units for Earthquake Signal Detection (2018, background): “There are, however, two main problems that occur in the training of very deep networks (where “deep” is usually taken to mean more than 15 layers): (1) vanishing/exploding gradients and (2) degradation (saturation and degradation of accuracy).”