Improving Seismic Activity Monitoring with AI-Driven Earthquake Parameters using Short-to-Long-Term Average Analysis

Abstract

Earthquake Early Warning Systems (EEWS) play a crucial role in mitigating earthquake impacts by providing timely alerts to affected regions. This study explores an AI-driven algorithm leveraging seismic trace migration and stacking to enhance the detection of seismic events. By analysing data from a temporary network in a volcanically active region, the algorithm identified optimal detection parameters, achieving a remarkable 94% detection rate, significantly outperforming conventional systems. Additionally, it identified 209 previously undetected events while maintaining a lower false detection rate of 25%, compared to the system's 40%. Key innovations include the application of Kurtosis functions and short-to-long-term average variations, enabling precise detection of seismic traces. The method demonstrated efficacy in analysing large swarms of low-magnitude events with short inter-event times, making it especially suitable for monitoring regions with complex seismic activities, such as fluid injection, drilling, and volcanic areas. Despite its computational intensity, the algorithm’s scalability and accuracy present a promising advancement in real-time seismic monitoring. Empirical tests affirm the utility of small, representative data subsets for fine-tuning detection parameters, reinforcing the system’s robustness. This approach underscores the potential of AI-driven methods in advancing seismology, improving early warning systems, and contributing to disaster risk reduction globally.