3:30 PM - 3:45 PM
[MTT38-01] Direction of arrival and traveling speed estimation of infrasound observed after 2024 Noto Peninsula Earthquake
★Invited Papers
When 2024 Noto Peninsula Earthquake happened, infrasound signals suspectedly related to tsunami were observed at some observation points of "Infrasound Monitoring Network" [1] operated by Japan Weather Association. Hence, in order to scrutinize potential relationships between the observed infrasound signals and the tsunamis occurred after the earthquake, we analyzed direction of arrival (DOA) and traveling speed of the infrasound signals, using an estimation algorithm based on time delay of arrival (TDOA).
The signals used for analysis were those observed in Toyohashi city of Aichi prefecture, which is a site consisting of three observation points and yet closest to Noto peninsula. Infrasound signals with amplitude of approximately 2 Pa were observed at these three observation points at about 16:31 (JST). We employed a method proposed by P Annibale et al. [2] to estimate their DOA and traveling speed. This method can derive a closed-form solution, based on the geometrical relationships of the observation points on a plane and the TDOA information among them. It needs three observation points at least and be basically applicable to a sensor array of an arbitrary shape. The signals were first divided into time frames of 1-minute duration with every 1-second shift, and then for each frame TDOA was calculated as the time lag with which the cross-correlation function between the signals of different observation points takes its maximum value. The resolution of TDOA was 10 ms because the sampling frequency was set to 100 Hz.
For the portion where the infrasound signal with amplitude of approximately 2 Pa was observed at about 16:31, the estimate of DOA was 349.4 degree in azimuth. Therefore, we drew an azimuth line from the center of the array using a tool on the Japanese GSI website operated by Geospatial Information Authority of Japan [3] and found that the azimuth line crosses over Wajima city of Ishikawa prefecture. The traveling speed of the infrasound was estimated simultaneously as 391 m/s. In addition, for another infrasound signal observed during a time period from 16:33:30 to 16:33:40, the estimates of DOA and traveling speed were approximately 356 degree in azimuth and 387 m/s, respectively. It was found by the similar analysis using the Japanese GSI website that the azimuth line crosses over Suzu city of Ishikawa prefecture.
These two infrasound signals were estimated of different DOAs, but it is unknown whether they had different origins or they had the same origin but different DOAs through different traveling paths. Moreover, the traveling speed estimated seems considerably faster than the general sound speed. Possible reasons are the traveling direction was not horizontal in disagreement with the assumption of the algorithm, or small noise affected the estimation results because the number of observation points (i.e. independent variables) was the minimum for the employed estimation algorithm. The distance from the observation site to Wajima city is 291 km, and the infrasound signals were observed 21 minutes later than the occurrence of earthquake. Therefore, assumed that the infrasound was generated soon after the earthquake, the traveling speed is calculated as 231 m/s. There is considerable discrepancy between this speed and the estimated one. One possible reason could be that the infrasound was generated sometime after the occurrence of earthquake. Another reason could be that they traveled through the upper atmosphere, resulting in a longer traveling path than the horizontal one. More number of observation points will be required for us to investigate these possibilities further in detail.
[1] https://micos-sc.jwa.or.jp/infrasound-net/observed/
[2] P. Annibale et al., "TDOA-Based Speed of Sound Estimation for Air Temperature and Room Geometry Inference," IEEE Trans. on Audio, Speech, and Language Processing, vol. 21, no. 2, pp. 234-246, 2013.
[3] https://maps.gsi.go.jp/
The signals used for analysis were those observed in Toyohashi city of Aichi prefecture, which is a site consisting of three observation points and yet closest to Noto peninsula. Infrasound signals with amplitude of approximately 2 Pa were observed at these three observation points at about 16:31 (JST). We employed a method proposed by P Annibale et al. [2] to estimate their DOA and traveling speed. This method can derive a closed-form solution, based on the geometrical relationships of the observation points on a plane and the TDOA information among them. It needs three observation points at least and be basically applicable to a sensor array of an arbitrary shape. The signals were first divided into time frames of 1-minute duration with every 1-second shift, and then for each frame TDOA was calculated as the time lag with which the cross-correlation function between the signals of different observation points takes its maximum value. The resolution of TDOA was 10 ms because the sampling frequency was set to 100 Hz.
For the portion where the infrasound signal with amplitude of approximately 2 Pa was observed at about 16:31, the estimate of DOA was 349.4 degree in azimuth. Therefore, we drew an azimuth line from the center of the array using a tool on the Japanese GSI website operated by Geospatial Information Authority of Japan [3] and found that the azimuth line crosses over Wajima city of Ishikawa prefecture. The traveling speed of the infrasound was estimated simultaneously as 391 m/s. In addition, for another infrasound signal observed during a time period from 16:33:30 to 16:33:40, the estimates of DOA and traveling speed were approximately 356 degree in azimuth and 387 m/s, respectively. It was found by the similar analysis using the Japanese GSI website that the azimuth line crosses over Suzu city of Ishikawa prefecture.
These two infrasound signals were estimated of different DOAs, but it is unknown whether they had different origins or they had the same origin but different DOAs through different traveling paths. Moreover, the traveling speed estimated seems considerably faster than the general sound speed. Possible reasons are the traveling direction was not horizontal in disagreement with the assumption of the algorithm, or small noise affected the estimation results because the number of observation points (i.e. independent variables) was the minimum for the employed estimation algorithm. The distance from the observation site to Wajima city is 291 km, and the infrasound signals were observed 21 minutes later than the occurrence of earthquake. Therefore, assumed that the infrasound was generated soon after the earthquake, the traveling speed is calculated as 231 m/s. There is considerable discrepancy between this speed and the estimated one. One possible reason could be that the infrasound was generated sometime after the occurrence of earthquake. Another reason could be that they traveled through the upper atmosphere, resulting in a longer traveling path than the horizontal one. More number of observation points will be required for us to investigate these possibilities further in detail.
[1] https://micos-sc.jwa.or.jp/infrasound-net/observed/
[2] P. Annibale et al., "TDOA-Based Speed of Sound Estimation for Air Temperature and Room Geometry Inference," IEEE Trans. on Audio, Speech, and Language Processing, vol. 21, no. 2, pp. 234-246, 2013.
[3] https://maps.gsi.go.jp/