Detection of Immediate Foreshocks Using Dense Seismic Array: A Case Study of the 2021 Ms 6.4 Yangbi aftershock sequence

Advancing our understanding of earthquake nucleation process can shed lights on earthquake prediction, early warning, and hazard assessment. Foreshocks, which usually refer to smaller earthquakes that occur before an earthquake, exhibit good temporal and spatial correlations with the mainshock. Investigating the relationship between foreshocks and mainshocks can therefore provide valuable insights into earthquake nucleation mechanisms and contribute to the improvement of earthquake prediction and early warning capabilities. A recent study on the 2019 Mw 7.1 Ridgecrest earthquake sequence suggests that immediate foreshocks often share similar waveforms to the P-waves of subsequent earthquakes, differing only in amplitude. This similarity is believed to arise from the fractal nature of fault fracture processes. Consequently, there might be many immediate foreshocks with similar waveforms hidden in ambient noise that have gone undetected. Two methods have been proved to be effective in detecting small events: the Matched Filter Technique (MFT) and the Source-Scanning Algorithm (SSA).  The MFT relies on template events to detect small events by stacking cross-correlograms between the waveforms of the templates and potential events. The conventional MFT, however, requires that the small events be located in the vicinity of one of the template events and does not provide the accurate locations of detected events. On the other hand, SSA is a migration-based approach that involves stacking non-negative waveforms, envelopes, and their extended characteristic functions. However, due to their tendency to provide absolute locations, SSA are heavily influenced by the accuracy of the velocity model and struggle to accurately detect earthquakes that are obscured by noise. In our study, we prioritize the accuracy of relative event locations when studying the relationship between foreshocks and mainshocks. To address this concern, we have developed an advanced method that combines the strengths of cross-correlation and beamforming analyses. This method allows us to detect and relatively locate small seismic events simultaneously using dense array data. For the 2021 Ms 6.4 Yangbi  aftershock sequence, we first compute the cross-correlograms of the contentious records with the P-waves/S-waves of the target earthquake, respectively. We then grid searches around the hypocenter using N-th root stacking to detect and locate the immediate foreshocks. Upon detecting numerous immediate foreshocks, we proceed to statistically quantify the earthquake nucleation process or investigate the nucleation mechanism.