Developing on-site earthquake early warning systems has been a challenging problem because of time limitations and the amount of information that can be collected before the warning needs to be issued. A potential solution that could prevent severe disasters is to predict the potential strong motion using the initial P-wave signal and provide warnings before serious ground shaking starts. In practice, the accuracy of prediction is the most critical issue for earthquake early warning systems. Traditional methods use certain criteria, selected through intuition or experience, to make the prediction. However, the criteria thresholds are difficult to select and may significantly affect the prediction accuracy. This paper investigates methods based on artificial intelligence for predicting the greatest earthquake ground motion early, when the P-wave arrives at seismograph stations. A neural network model is built to make the predictions using a small window of the initial P-wave acceleration signal. The model is trained by seismic waves collected from 1991 to 2019 in Taiwan and is evaluated by events in 2020 and 2021. From these evaluations, the proposed scheme significantly outperforms the threshold-based method in terms of its accuracy and average leading time.

The extended integrated particle filter (IPFx) method is an automatic source determination algorithm designed for the Japanese earthquake early warning (EEW) system. The method improved earthquake source determination during active seismicity by incorporating the smart phase association scheme. We applied this method to the 2018 Hualien earthquake sequence and evaluated its performance by comparing it to the manual catalog. We used 1-month continuous waveforms from February 2018 at 170 stations. Owing to the higher noise level, we improved the phase association algorithm to avoid noise contamination. Out of 127 earthquakes with a seismic intensity ≥ 4, 105 were successfully detected in one month, of which 103 had good accuracy with a location error of < 30 km. The detectability of earthquakes decreased immediately following large events. The IPFx method showed good performance in detecting earthquakes with seismic intensity ≥ 4 during the 2018 Hualien earthquake sequence. The method was also applied to the 1-day continuous data on April 18, 2021, and detected 14 earthquakes with a magnitude ~ 2 that were not on the manual catalog. Currently, the Central Weather Bureau in Taiwan uses the effective epicenter method to locate earthquakes for the EEW system. However, source determination for offshore events is difficult as most of the stations are on land. We expect the IPFx method to provide better location estimates for offshore earthquakes and during the period of active seismicity. It also provides an earlier warning as it sends the first message when three stations are triggered. This new method can potentially improve the speed and accuracy of the Taiwanese EEW system.