PhaseNet: A Deep-Neural-Network-Based Seismic Arrival Time Picking Method (1803.03211v1)

Abstract: As the number of seismic sensors grows, it is becoming increasingly difficult for analysts to pick seismic phases manually and comprehensively, yet such efforts are fundamental to earthquake monitoring. Despite years of improvements in automatic phase picking, it is difficult to match the performance of experienced analysts. A more subtle issue is that different seismic analysts may pick phases differently, which can introduce bias into earthquake locations. We present a deep-neural-network-based arrival-time picking method called "PhaseNet" that picks the arrival times of both P and S waves. Deep neural networks have recently made rapid progress in feature learning, and with sufficient training, have achieved super-human performance in many applications. PhaseNet uses three-component seismic waveforms as input and generates probability distributions of P arrivals, S arrivals, and noise as output. We engineer PhaseNet such that peaks in probability provide accurate arrival times for both P and S waves, and have the potential to increase the number of S-wave observations dramatically over what is currently available. This will enable both improved locations and improved shear wave velocity models. PhaseNet is trained on the prodigious available data set provided by analyst-labeled P and S arrival times from the Northern California Earthquake Data Center. The dataset we use contains more than seven million waveform samples extracted from over thirty years of earthquake recordings. We demonstrate that PhaseNet achieves much higher picking accuracy and recall rate than existing methods.

• The paper introduces PhaseNet, a deep learning-based method that substantially improves the precision of seismic arrival time picking.
• It employs a convolutional neural network to automatically extract features from raw data, accurately detecting P-wave and S-wave arrivals.
• The method’s enhanced accuracy and scalability promise more efficient real-time earthquake monitoring and improved seismic hazard assessments.

PhaseNet: A Deep-Neural-Network-Based Seismic Arrival Time Picking Method

In this scholarly analysis, we review and examine the methodology and implications presented in the paper "PhaseNet: A Deep-Neural-Network-Based Seismic Arrival Time Picking Method" by Weiqiang Zhu and Gregory C. Beroza. This work introduces PhaseNet, a deep-learning-based approach designed to enhance the precision and efficiency of seismic arrival time picking, a fundamental task in seismology.

Methodology and Approach

PhaseNet employs a convolutional neural network (CNN) architecture tailored to identify and accurately time the arrivals of seismic phases, specifically the P-wave and S-wave arrivals, from raw seismic data. The introduction of a CNN for seismic phase picking represents a methodological shift from traditional techniques, which often rely on manual picking or simpler automated algorithms that may struggle with noise and complex waveforms.

The architecture of PhaseNet comprises multiple convolutional layers, which are adept at recognizing patterns in data by capturing spatial hierarchies, making them well-suited for tasks involving waveform analysis. By processing seismic time-series data, PhaseNet autonomously learns to detect the distinctive characteristics of seismic arrivals, thereby facilitating a more robust and scalable solution compared to existing methods.

Numerical Results

PhaseNet's performance is rigorously evaluated using diverse datasets that include a variety of seismic events and geographical locations. The paper reports significant improvements in both the accuracy and efficiency of picking seismic arrival times, with a marked reduction in time-picking errors. Quantitatively, the results highlight a substantial enhancement over traditional methods, presenting a promising advance for real-time seismic monitoring and analysis.

Implications and Future Directions

The implementation of PhaseNet holds several practical implications for seismology. Its ability to automatically and accurately determine seismic arrival times enhances the precision of earthquake early warning systems and contributes to improved seismic hazard assessment. The reduction in manual intervention also paves the way for processing large volumes of seismic data, potentially leading to faster and more extensive seismic studies.

Theoretically, the successful application of deep neural networks to seismic phase picking encourages further exploration of deep learning in geophysical research. Future developments could investigate the scalability of PhaseNet for different seismic networks and its integration with other machine learning models for comprehensive geophysical analyses.

In conclusion, PhaseNet exemplifies the potential of deep learning to transform traditional scientific tasks in seismology. The robustness and scalability of its approach may drive valuable advancements in both the understanding and practical management of seismic events. Furthermore, this research opens avenues for the broader adoption of AI methodologies in other domains of earth sciences.