Earthquake early warning (EEW) systems in Japan use data from as little as a single observation point, even at the stage in which only the P wave is detected at one point, to estimate the epicentral distance and magnitude to rapidly issue a warning. This approach aims to maximize lead time but has the drawback of significant estimation errors (epicentral distance: half to double, back azimuth: ±30°). Using distributed acoustic sensing (DAS) with a dense network of observation points allows for rapid and precise earthquake hypocenter determination, significantly reducing estimation errors compared to single-station methods in EEW system. However, no prior studies have investigated the immediacy of using DAS data—which is indispensable for early earthquake warnings. Therefore, we examined the use of DAS for rapid and precise hypocenter determination with the goal of integrating it into the EEW system. We used fiber-optic cables along the Kyushu Shinkansen to analyze strain-rate waveforms, detect P waves, and determine earthquake hypocenter locations. We detected P waves using short-term average /long-term average simultaneously across multiple channels. We then determined the epicenter using Hypomh. For rapid determination of the hypocenter location after P-wave detection, we assessed the hypocenter location each time the number of P-wave detection points increased by 20. For earthquakes with an epicentral distance of less than 50 km, we successfully determined the hypocenter within 12 km of the Japan Meteorological Agency-reported location. This was achieved by identifying the point where the sum of the maximum-likelihood estimation errors was minimized within 2 s of the initial P-wave detection in any channel. The findings suggest that an EEW system using DAS holds potential as a powerful tool to provide rapid and accurate hypocenter information, especially for near-field earthquakes that occur directly beneath populated areas, where lead times are short, and the potential damage is substantial.
4 July 2017 in Bulletin of the Seismological Society of America
To enhance the performance of earthquake early warning (EEW) systems that aim to issue alerts as quickly as possible, it is crucial to improve the accuracy of the epicentral distance estimated via the single‐station method. Although the conventional method estimates from the slope of the initial P‐wave envelope, this study applies deep learning techniques that can extract a variety of information from the waveform data. By analyzing ∼20,000 records observed at Kyoshin Network stations in Japan, the convolutional neural network (CNN) method achieved higher accuracy than the conventional method. Increasing the data length or the number of iterations of convolution, activation, and pooling layers in the typical CNN model did not significantly improve the accuracy of estimation. An automatic structure search (AutoSS) technique, in which model structure and hyperparameters are randomly varied, was employed to identify models that yield higher accuracy. A typical CNN model was used as the initial structure. The models obtained through this technique showed improved accuracy with increased data length or computational cost. The models that delivered the highest accuracy among those generated using the AutoSS technique outperformed the typical CNN model in terms of accuracy, although their computational costs were comparable. The AutoSS technique offers a significant advantage in that it allows the selection of a model that matches the computational capabilities of the hardware used for its implementation, thereby ensuring optimal computational efficiency. This exhibits that deep learning technologies can be used to improve the performance of EEW systems.
To improve the performance of earthquake early warning (EEW) systems, we propose an approach that utilizes the time-dependence of P-wave displacements to estimate the earthquake magnitude (M) based on the relationship between M and the displacement. The traditional seismological understanding posits that this relationship achieves statistical significance when the displacement reaches its final peak value, resulting in the adoption of time-constant coefficients. However, considering the potential for earlier establishment of the relationship’s significance than conventionally assumed, we analyze waveforms observed in Japan and determine the intercept in the relationship as a function of time from the P-wave onset. We demonstrate that our approach reduces the underestimation of M in the initial P-wave stages compared to the conventional technique. Consequently, we find a significant rise in the number of earlier warnings in the Japanese railway EEW system. Due to the inherent trade-off between the immediacy and accuracy of alarm outputs, the proposed method unavoidably leads to an increase in the frequency of alerts. Nonetheless, if deemed acceptable by system users, our approach can contribute to EEW performance improvement.