Spatial-temporal evolution of injection-induced earthquakes in the Weiyuan Area determined by machine-learning phase picker and waveform cross-correlation

Abstract

Anthropogenic induced seismicity has been widely reported and investigated in many regions, including the shale gas fields in the Sichuan basin, where the frequency of earthquakes has increased substantially since the commencement of fracking in late 2014. However, the details of how earthquakes are induced remain poorly understood, partly due to lack of high-resolution spatial-temporal data documenting the evolution of such seismic events. Most previous studies have been based on a diffusive earthquake catalog constructed by routine methods. Here, however, we have constructed a high resolution catalog using a machine learning detector and waveform cross-correlation. Despite limited data, this new approach has detected one-third more earthquakes and improves the magnitude completeness of the catalog, illuminating the comprehensive spatial-temporal migration of the emerging seismicity in the target area. One of the clusters clearly delineates a potential unmapped fault trace that may have led to the Mw 5.2 in September 2019, by far the largest earthquake recorded in the region. The migration of the seismicity also demonstrates a pore-pressure diffusion front, suggesting additional constraints on the inducing mechanism of the region. The patterns of the highly clustered seismicity reconcile the causal link between the emerging seismicity and the activity of hydraulic fracturing in the region, facilitating continued investigation of the mechanisms of seismic induction and their associated risks.

Keywords

• hydraulic fracturing
• sichuan basin
• machine learning
• injection-induced earthquake
• weiyuan earthquake