International Data Centre: Reviewed Event Bulletin vs. Waveform Cross Correlation Bulletin

TL;DR

This study evaluates waveform cross-correlation for automatic aftershock detection of the 2012 Sumatera earthquake, comparing it with the Reviewed Event Bulletin, and demonstrates its effectiveness in identifying and reviewing seismic events.

Contribution

It introduces a method using waveform cross-correlation with selected stations to enhance automatic aftershock detection and compares results with the official Reviewed Event Bulletin.

Findings

Cross-correlation detected approximately 1,000,000 signals.

2763 event hypotheses were generated and reviewed.

The method improved detection sensitivity and event hypothesis generation.

Abstract

Our objective is to assess the performance of waveform cross-correlation technique, as applied to automatic and interactive processing of the aftershock sequence of the 2012 Sumatera earthquake relative to the Reviewed Event Bulletin (REB) issued by the International Data Centre. The REB includes 1200 aftershocks between April 11 and May 25 with body wave magnitudes from 3.05 to 6.19. To automatically recover the sequence, we selected sixteen aftershocks with mb between 4.5 and 5.0. These events evenly but sparsely cover the area of the most intensive aftershock activity as recorded during the first two days after the main shock. In our study, waveform templates from only seven IMS array stations with the largest SNRs estimated for the signals from the main shock were used to calculate cross-correlation coefficients over the entire period of 44 days. Approximately 1000000 detections obtained using cross-correlation were then used to build events according to the IDC definition. After conflict resolution between events with similar arrivals built by more than one master all qualified event hypotheses populated the automatic cross-correlation Standard Event List (aXSEL). The total number of distinct hypotheses was 2763. To evaluate the quality of new events in the aXSEL, we randomly selected a small portion of XSEL events and analysts reviewed them according to standard IDC rules and guidelines. After the interactive review of a small portion of the final product of cross correlation was obtained which we call the interactive XSEL. We have constructed relevant frequency and probability density distributions for all detections, all associated detection,and for those which were associated with the aXSEL and final XSEL events. These distributions are also station and master dependent.