11:00 〜 11:15
[SCG63-02] 台湾の強震ネットワークを利用したIPFx法のオンライン試験
キーワード：緊急地震速報、台湾、台湾中央気象署、地震活動、強震動
We developed an online testing system of the IPFx method for the strong motion network in Taiwan. The original IPFx program was designed for the WIN system, which handles 1s packet of the WIN format with client-server network configuration. The Central Weather Administration in Taiwan uses the EarthWorm system, which is a real-time data transmission system with MiniSEED format. In this study, we created a Python conversion program using the MQTT protocol to convert the MiniSEED format to the input of the IPFx program. One of the largest issues is that the packet length of the MiniSEED format is not constant, and each channel is transmitted randomly. The input of the IPFx method is a 1-s packet with the same timestamp for all channels. Therefore, the conversion program uses a data buffer to store several seconds of data and outputs a 1-s packet with a timestamp 3 seconds before.
We performed online testing since October 2024. Real-time testing is more challenging than offline testing with a limited length of data. We found 2 major issues during our test. One was a little oversight in the data conversion program which created a lot of false picks and false detections of earthquakes. The other issue was the magnitude overestimation. We double-checked the location was reasonably accurate for most of the events, but the magnitude was unreasonably large for most of them. The CWA acceleration data included the long-period noise, which accumulated when we performed the double integration to get the displacement magnitude. We added a recursive filter to remove the long-period noise before processing the data. After this correction of the IPFx program, the magnitude estimation becomes close to the catalog magnitude.
During this online testing, a magnitude 6.4 earthquake occurred on January 20, 16:17:27 (UTC). The real-time system detected this earthquake at 16:17:36, including the data transmission delay and data format conversion. The location error was less than 5 km from the beginning. The performance is quite good, and we can shorten the alert time by a few seconds if we improve the data format conversion. The official earthquake early warning system sends the first alert about 8 seconds after the occurrence of the event to almost all cities and counties of the entire Taiwan island. We will continue the monitoring and improvement of the real-time IPFx system.
We performed online testing since October 2024. Real-time testing is more challenging than offline testing with a limited length of data. We found 2 major issues during our test. One was a little oversight in the data conversion program which created a lot of false picks and false detections of earthquakes. The other issue was the magnitude overestimation. We double-checked the location was reasonably accurate for most of the events, but the magnitude was unreasonably large for most of them. The CWA acceleration data included the long-period noise, which accumulated when we performed the double integration to get the displacement magnitude. We added a recursive filter to remove the long-period noise before processing the data. After this correction of the IPFx program, the magnitude estimation becomes close to the catalog magnitude.
During this online testing, a magnitude 6.4 earthquake occurred on January 20, 16:17:27 (UTC). The real-time system detected this earthquake at 16:17:36, including the data transmission delay and data format conversion. The location error was less than 5 km from the beginning. The performance is quite good, and we can shorten the alert time by a few seconds if we improve the data format conversion. The official earthquake early warning system sends the first alert about 8 seconds after the occurrence of the event to almost all cities and counties of the entire Taiwan island. We will continue the monitoring and improvement of the real-time IPFx system.