Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluating Multi-station Phase Picking Algorithm Phase Neural Operator (PhaseNO) on Local Seismic Networks

Qingkai Kong, Avigyan Chatterjee, Chengping Chai, Alex Dzubay, Kayla A. Kroll, Josh C. Stachnik, Scott Fertig, Jeffrey Liefer, Paul Friberg

TL;DR

This work evaluates PhaseNO, a multi-station phase picker based on Fourier Neural Operators and Graph Neural Operators, on four local seismic networks and benchmarks it against PhaseNet and EQTransformer. It demonstrates that leveraging coherent signals across multiple stations improves phase detection and event association at local scales, particularly for low-SNR arrivals. An ablation analysis comparing PhaseNO with a single-station variant (PhaseNO1) shows multi-station input enhances timing accuracy and reduces false positives. The results support PhaseNO as a valuable addition for real-time monitoring and catalog completeness in local seismic networks.

Abstract

Reliable automatic phase picking is important for many seismic applications. With the development of machine learning approaches, many algorithms are proposed, evaluated and applied to different areas. Many of these algorithms are single station based, while recent proposed methods start to combine surrounding stations into consideration in the problem of phase picking. Among these algorithms, the Phase Neural Operator (PhaseNO) shows promising results on regional datasets comparing to existing algorithms. But there are many use cases for the local seismic networks in our community, therefore in this paper we evaluate the performance of PhaseNO on 4 different local datasets and compared the results to PhaseNet and EQTransformer. We used both individual phase picking metrics as well as association metrics to illustrate the performance of PhaseNO. With manually reviewing the newly detected events, we find the PhaseNO model outperformed the single station-based approaches in the local-scale use cases due to its consideration of coherent signals from multiple stations. We also explored PhaseNO's behaviors when only using one station, as well as gradually increase the number of stations in the seismic network to understand it better. Overall, using the off-the-shelf machine learning based phase pickers, PhaseNO demonstrated its good performance on local-scale seismic networks.

Evaluating Multi-station Phase Picking Algorithm Phase Neural Operator (PhaseNO) on Local Seismic Networks

TL;DR

This work evaluates PhaseNO, a multi-station phase picker based on Fourier Neural Operators and Graph Neural Operators, on four local seismic networks and benchmarks it against PhaseNet and EQTransformer. It demonstrates that leveraging coherent signals across multiple stations improves phase detection and event association at local scales, particularly for low-SNR arrivals. An ablation analysis comparing PhaseNO with a single-station variant (PhaseNO1) shows multi-station input enhances timing accuracy and reduces false positives. The results support PhaseNO as a valuable addition for real-time monitoring and catalog completeness in local seismic networks.

Abstract

Reliable automatic phase picking is important for many seismic applications. With the development of machine learning approaches, many algorithms are proposed, evaluated and applied to different areas. Many of these algorithms are single station based, while recent proposed methods start to combine surrounding stations into consideration in the problem of phase picking. Among these algorithms, the Phase Neural Operator (PhaseNO) shows promising results on regional datasets comparing to existing algorithms. But there are many use cases for the local seismic networks in our community, therefore in this paper we evaluate the performance of PhaseNO on 4 different local datasets and compared the results to PhaseNet and EQTransformer. We used both individual phase picking metrics as well as association metrics to illustrate the performance of PhaseNO. With manually reviewing the newly detected events, we find the PhaseNO model outperformed the single station-based approaches in the local-scale use cases due to its consideration of coherent signals from multiple stations. We also explored PhaseNO's behaviors when only using one station, as well as gradually increase the number of stations in the seismic network to understand it better. Overall, using the off-the-shelf machine learning based phase pickers, PhaseNO demonstrated its good performance on local-scale seismic networks.
Paper Structure (19 sections, 4 equations, 25 figures, 6 tables)

This paper contains 19 sections, 4 equations, 25 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Data and Networks
  3. Methodology
  4. PhaseNet
  5. EQTransformer
  6. PhaseNO
  7. Association algorithm
  8. Evaluation Metrics
  9. Phase Picking Accuracy
  10. Association and Event-Level Metrics
  11. False Positive and Novel Detection Validation
  12. Results
  13. Phase Picking Metrics
  14. Association metrics (vs. GT Catalogs)
  15. False Positive and Novel Detection Analysis
  16. ...and 4 more sections

Figures (25)

  • Figure 1: Seismic network configurations as well as the catalog events from 4 different regions.
  • Figure 2: The magnitude, depth and S-P time distribution for the 4 different tests. Each row represents a region.
  • Figure 3: Phase picking metrics, the left panels show the histogram comparisons for different algorithms, the greyed histograms are the matched phases for each algorithm. The right panels show the precision, recall and f1 metrics for both P and S pickings. The rows from top to bottom are test 1 to test 4. The detailed numbers are listed in \ref{['tab:A1']}.
  • Figure 4: P wave picking time differences comparing to manual pickings. Mean, standard deviation and number of phases are showing in each panel. Top row panels are test 1 and test 2 from left to right, while the bottom row has test 3 and test 4. Similar figure for S wave picking differences is shown in \ref{['fig:A1']}. Detailed numbers are listed in \ref{['tab:A2']}.
  • Figure 5: Test 1 phase true positives, false negatives, and false positives distributions for both P and S phases against signal noise ratio. For test 2-4, please refer to \ref{['fig:A2']} - \ref{['fig:A4']}.
  • ...and 20 more figures