Enhancing Seismic Monitoring Through the SCSN’s Earthworm/AQMS Real-Time System

The Southern California Seismic Network (SCSN) has long relied on the Earthworm/AQMS real-time monitoring system, a robust framework developed in the late 1990s that continues to serve various regional seismic networks today. This system employs a short-term average/long-term average (STA/LTA) phase picker, known as pickEW, alongside an event post-processing module called hypomag. Its real-time capabilities enable rapid processing of seismic data, ensuring timely detection and analysis of earthquakes.

Real-Time Monitoring Process

The real-time monitoring system operates by making phase picks almost instantaneously as data is received, typically every second. These picks are stored in what is known as a "pick ring." An associator module reviews the picks, assembling them into cohesive events. After 70 seconds from the origin time, an event is deemed complete, marking the closure of additional picks and initiating subsequent steps, such as magnitude calculation and relocation.

The pre-existing workflow for event post-processing at SCSN involves sending events with magnitudes below 3 to the hypomag module. The system processes the waveforms associated with these events, which is typically a 1–2 minute window preceding and following each event. By utilizing pickEW, the system generates a new set of picks focused on specific timeframes relative to expected phase arrivals. Initially identified picks can either be replaced by new ones or retained, facilitating a refined analysis.

In an innovative update to this workflow, SCSN has recently integrated a sophisticated deep-neural-network picker, known as PhaseNet, into the hypomag module. This transition aligns with modern advancements in seismic network processing, enhancing the system’s efficacy while maintaining existing server configurations.

Development and Evaluation of hypoPN

The introduction of hypoPN, the refined phase picker based on PhaseNet, involved meticulous testing. The goal was to overcome a few identified challenges that could hinder the transition from theory to practical application. Initial evaluations showed promise, with PhaseNet demonstrating superior performance in terms of the number and quality of picks compared to the traditional hypomag system.

One significant challenge tackled during the development was the PhaseNet’s inability to determine polarities, crucial for focal mechanisms. To address this, components of the pickEW algorithm were integrated into PhaseNet to enhance its functionality. Another hurdle involved the disparity between PhaseNet’s phase scores and the quality ratings expected by Earthworm/AQMS. These scores required careful mapping to maintain consistency in analysis and ensure accurate event relocation.

Subsequent testing involved historical backlog events, providing insights into the operational performance of hypoPN compared to its predecessor. Analysts reviewed these events, allowing for expert evaluation of the functionality and efficiency of the new system.

Near Real-Time Production Tests

Once hypoPN was deemed operationally ready, it was subjected to trials within the near real-time production environment. Initial technical tests focused on determining its reliability in a live setting, which included monitoring various seismic events. This phase of testing led to additional improvements in phase pick prioritization and quality mapping, further augmenting hypoPN’s capabilities.

During these assessments, trends began to emerge regarding the effectiveness of hypoPN in improving seismic event location accuracy. Reports indicated hypoPN consistently demonstrated proficiency in retaining higher-quality picks, which expedited the review process by analysts, particularly during periods of heightened seismic activity.

Performance Results and Insights

Analysis of performance data revealed that hypoPN nearly doubled the number of picks relative to hypomag, significantly enhancing the network’s operational efficiency. However, it was noted that a higher proportion of PhaseNet picks were either deleted or required adjustment by analysts. Although this sounds concerning, it is essential to consider the overall number of missed picks was considerably lower with hypoPN, showcasing its effectiveness in capturing seismic events.

A direct evaluation of location accuracy confirmed that hypoPN outperformed hypomag in determining epicenters and depths of seismic events. While both modules produced generally accurate results, hypoPN emerged as the superior choice, particularly in challenging scenarios that bespoke the need for refined analysis.

Interestingly, the workload for analysts showed promising trends, with hypoPN allowing quicker reviews and enabling analysts to finalize events more efficiently than with the older hypomag system. This innovation has the potential to alleviate the backlog that often accumulates during intense seismic activity, thereby streamlining the entire review process.

Forward-Looking Considerations

While the integration of hypoPN represents a leap forward, the SCSN recognizes the importance of continuous improvement and adaptation. Further testing remains essential, especially during periods of extreme seismicity, to fully assess hypoPN’s capabilities across diverse scenarios. The ongoing evolution of this system underscores SCSN’s commitment to leveraging cutting-edge technology for improved seismic monitoring outcomes, ensuring public safety and scientific advancement in understanding seismic activity.

As the field of seismology continues to evolve, the SCSN’s dedication to refining and advancing its methodologies will undoubtedly play a pivotal role in enhancing earthquake response and understanding.