Enhancing the speed and accuracy of earthquake source- location estimation is crucial for developing effective Earthquake Early Warning (EEW) systems. In this study, we aim to design an innovative machine learning–based approach that will utilize P- wave arrival times from initial seismic stations and compute differential arrival times relative to a reference station for epicenter estimation. We plan to train the model using an extensive earthquake catalog to evaluate its accuracy, robustness, and adaptability under limited data conditions and with fewer recording stations. This work will address key limitations of traditional seismological methods, such as latency and accuracy issues, by providing faster and more reliable location estimates. The proposed model is expected to offer improved scalability across different geographic regions and will be capable of learning effectively from minimal data. Among the four machine learning algorithms to be tested, we anticipate that the Random Forest (RF) classifier will demonstrate the best performance. Integrating the developed model into EEW systems is expected to significantly enhance earthquake monitoring, enabling timely and precise alerts to improve preparedness, reduce risks, and support rapid response during seismic events.
Abhishek et al. (Mon,) studied this question.