4:15 PM - 4:30 PM
[MTT45-04] Infrasound estimation base on array observations in multiple areas
★Invited Papers
Keywords:MEMS, Direction of arrival, Sound speed, LMSE, Submarine eruption
When a big natural phenomenon that may cause a disaster happens, infrasound is often generated subsequently. Aiming at early detection of such event, we have been developing infrasound observation techniques for multiple observation points. We have designed an infrasound measurement device using MEMS sensors to make it small and available at low price, thereby it could help to increase observation points. The developed devices were deployed in three areas: Kagoshima, Tokyo and Miyagi prefectures for evaluation, where an array of five, four and five observation points were constructed respectively.
A large infrasound was generated by the submarine eruption of the Tonga islands on January 15, 2022. This was the first time that a single infrasound was observed at all the arrays. We estimated the direction of arrival and sound speed using the observed data. First, arrival time differences were calculated as the time lag providing the maximum value of the cross-correlation function between all possible combinations of the observation points in each area. The direction of arrival and sound speed were simultaneously estimated as an optimal solution in the LMSE criterion where the error function was defined as the power of distortion from the plane wave assumption. When the number of observation points were four or more, an optimal solution can be derived because the number of equations is larger than that of variables to be estimated. In the calculation, information of altitude of the observation points was ignored and only that of longitude and latitude were considered. As a result, we need to estimate two variables: direction of arrival and sound speed. However, to make the formulation simple, we assigned a variable to each of the sine and cosine components of the direction of arrival. These variables were combined using the trigonometry axiom and included as a constraint in the formulation. Cross-correlation functions were calculated using adjacent time frames weighted by a hamming window of 20 minutes duration and one minute shift. It is seen from past 20:00 to past 22:00 on January 15 in JST that the estimates became stable and error functions showed small values. The estimated sound speed was approximately 300 m/s regardless of the observation area. The estimate of the direction of arrival was 135 degree to 140 degree for Miyagi area, which roughly meets the direction of the Tonga islands from Miyagi prefecture according to an online service by the Geospatial Information Authority of Japan. On the other hand, that of Kagoshima area was 132 degree to 135 degree, which is slightly southward than the true direction. At the beginning of the sound wave arrival, they were directed relatively true directions and then gradually changed toward one direction among all the observation areas. One possible reason is that we employed sufficiently overlapped analysis windows. There would be other potential reasons but they need further investigations.
Meanwhile, all the observation devices used in Tokyo area were MBL1000-8, which simultaneously uses eight BOSCHE BME280 to improve signal to noise ratio. For the observation in Kagoshima area, three MBL1000-8 and two other devices with Infineon Technologies DPS310 were used. In Miyagi area, all the observation devices were those using DPS310. Sampling frequency were 4 Hz, 32 Hz, and 40 Hz, depending on observation points. It was confirmed that even such heterogeneous observation conditions, we can obtain considerably stable estimates if multiple observation points were included in the analysis and appropriate interpolation were applied. As described before, estimates of the direction of arrival gradually changed. The influence of wind and pressure fields remain for future study.
This work was partly supported by JSPS KAKENHI Grant Numbers JP19H02396.
A large infrasound was generated by the submarine eruption of the Tonga islands on January 15, 2022. This was the first time that a single infrasound was observed at all the arrays. We estimated the direction of arrival and sound speed using the observed data. First, arrival time differences were calculated as the time lag providing the maximum value of the cross-correlation function between all possible combinations of the observation points in each area. The direction of arrival and sound speed were simultaneously estimated as an optimal solution in the LMSE criterion where the error function was defined as the power of distortion from the plane wave assumption. When the number of observation points were four or more, an optimal solution can be derived because the number of equations is larger than that of variables to be estimated. In the calculation, information of altitude of the observation points was ignored and only that of longitude and latitude were considered. As a result, we need to estimate two variables: direction of arrival and sound speed. However, to make the formulation simple, we assigned a variable to each of the sine and cosine components of the direction of arrival. These variables were combined using the trigonometry axiom and included as a constraint in the formulation. Cross-correlation functions were calculated using adjacent time frames weighted by a hamming window of 20 minutes duration and one minute shift. It is seen from past 20:00 to past 22:00 on January 15 in JST that the estimates became stable and error functions showed small values. The estimated sound speed was approximately 300 m/s regardless of the observation area. The estimate of the direction of arrival was 135 degree to 140 degree for Miyagi area, which roughly meets the direction of the Tonga islands from Miyagi prefecture according to an online service by the Geospatial Information Authority of Japan. On the other hand, that of Kagoshima area was 132 degree to 135 degree, which is slightly southward than the true direction. At the beginning of the sound wave arrival, they were directed relatively true directions and then gradually changed toward one direction among all the observation areas. One possible reason is that we employed sufficiently overlapped analysis windows. There would be other potential reasons but they need further investigations.
Meanwhile, all the observation devices used in Tokyo area were MBL1000-8, which simultaneously uses eight BOSCHE BME280 to improve signal to noise ratio. For the observation in Kagoshima area, three MBL1000-8 and two other devices with Infineon Technologies DPS310 were used. In Miyagi area, all the observation devices were those using DPS310. Sampling frequency were 4 Hz, 32 Hz, and 40 Hz, depending on observation points. It was confirmed that even such heterogeneous observation conditions, we can obtain considerably stable estimates if multiple observation points were included in the analysis and appropriate interpolation were applied. As described before, estimates of the direction of arrival gradually changed. The influence of wind and pressure fields remain for future study.
This work was partly supported by JSPS KAKENHI Grant Numbers JP19H02396.