Answers are grounded in this paper and cite their sources.
Study enhances earthquake detection in the Luxian region, suggesting fluid-driven triggering mechanisms.
Dense seismic arrays provide high spatial sampling of the seismic wavefield, substantially enhancing the detection and location of small-magnitude events in industrial environments. Waveform-based methods such as the MAchine Learning aided earthquake MIgration location (MALMI) algorithm are more robust than travel-time-based approaches for the automated detection and location of these small events. However, the standard MALMI implementation is computationally expensive and lacks quality control for assessing the convergence of migration-based location results. To address these challenges, we enhance the MALMI workflow by (1) generating continuous phase-probability time series using a pretrained Bidirectional Recurrent Neural Network (BRNN) model, (2) accelerating data loading via an input/output-parallel strategy, (3) incorporating graphic processing unit acceleration into the migration-based location module, (4) applying a convolutional neural network model to identify converged location results, and (5) performing final relative relocation with GrowClust. We first validate the end-to-end workflow using waveforms from methane-related sources with known locations and a manually compiled earthquake catalog. We then apply it to approximately one month of continuous data from a dense array in the Luxian area of the Sichuan basin. The relocated catalog produces travel-time curves that are more coherent and physically plausible than those derived directly from the manual catalog. Methane source locations determined by the new workflow are within 1 km of their true positions. The new catalog contains approximately 13 times more events than the manual catalog and reduces the magnitude of completeness from 1.1 to 0.0. We identify four clusters at depths of 1–4 km, above the hydraulic-fracturing interval. Spatiotemporal analysis, temporal variations in b value, and analytical pore-pressure modeling suggest that most earthquakes are triggered on pre-existing faults by stress changes, while one cluster appears to be directly driven by fluid diffusion.
No takes yet. Share an insight, caveat, or question.
Sun et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: