Graduate Student Spotlight: Improving nuclear, earthquake safety with infrasound
Kei deploying smartphones for explosion data collection at Idaho National Laboratory
The story below was written by Nina Shuping for the Hawai‘i Institute of Geophysics and Planetology.
Samuel “Kei” Takazawa, a Ph.D. student in the Hawai‘i Institute of Geophysics and Planetology in SOEST specializing in infrasound and smartphone sensors, draws inspiration from his childhood experiences in earthquake-prone Japan and an interest in disaster response. His journey has led him to research in the realm of explosion monitoring, machine learning, and the innovative use of the sensors in smartphones. Growing up amidst frequent earthquakes and natural disasters in Japan, Kei’s early life experiences sparked a curiosity and a commitment to addressing challenges related to these phenomena. Kei relocated to California during high school and later pursued a double major in physics and applied math at Wheaton College in Illinois.
During his undergraduate studies, Kei’s involvement with the Humanitarian Disaster Institute, through a professor’s connections, allowed him to apply his developing skills in applied math to enhance disaster prediction and response. Much of Kei’s work as a relief worker and researcher was in response to the magnitude 9.1 Tōhoku earthquake and tsunami in Japan in 2011, which also included the Fukushima nuclear accident. Simultaneously, Kei explored his interest in audio engineering through his media technician job working on school events and concerts, setting the foundation for his future endeavors in sound technology.
After taking a year off from school he started his master’s in applied mathematics while also working at a reinsurance company in Philadelphia. There, he engaged in machine learning and disaster modeling, aligning with his prior work at the Humanitarian Disaster Institute. He worked under the insurance company’s hurricane expert who ultimately encouraged Kei to pursue a doctoral degree, recognizing his interest in research, which his current position did not entail.
Entering the world of infrasound
In his Ph.D. journey, Kei found his way to the University of Hawaiʻi after he was recommended to HIGP researcher Milton A. Garcés. Kei’s particular work experience and interests aligned well with Garcés’s research in infrasound monitoring. Kei’s own research, and doctoral dissertation topic, which is funded by the National Nuclear Security Administration under the Department of Energy, focuses on leveraging smartphones as monitoring devices. His goal is to employ machine learning techniques to detect, classify, and localize explosions, contributing to nuclear non-proliferation efforts.
A global network of sensors exists to detect seismic waves and infrasound waves. These are used to detect earthquakes, volcanic explosions, and other loud noises or events. Seismic waves are created when there is a sudden movement of materials and the energy is then propagated through Earth allowing sensors to record this movement even from far away. Infrasound is the other side of the same coin, and Kei said that it can be defined differently depending on who you ask. Generally speaking, it is defined as acoustic oscillations below what a human can hear, and others restrict infrasound to delivery via air transmission.
One of the main reasons why both infrasound waves and seismic waves are able to be picked up by sensors far away is due to their long wavelength. Infrasound wavelengths are over 17 meters and seismic waves are often 40-250 meters long.This makes it easy for the waves to pass through solid materials such as Earth. Large events even produce waves that can circle Earth, sometimes more than once. Such was the case with the Hunga Tonga–Hunga Haʻapai volcanic explosion in Tonga in 2022. However, one disadvantage of this network of sensors is that the seismic/infrasound event has to be large enough or close enough to a sensor to be picked up, and as a result smaller explosions and activity are often not detected.
Kei’s work is figuring out how to use the sensors in cell phones as a distributed monitoring system that is able to pick up on, and locate, smaller explosions. By using the app his professor founded, called RedVox Infrasound Recorder, the user can allow their phone’s sensors to be recorded including audio below 20 hz. Those concerned with privacy can be assured that at the default sample rate of 800 Hz, the phone can detect the lower vowel tones of speaking but not the consonant sounds, making human speech unintelligible..
Kei also is using the data to help train a machine learning model to identify an explosion and notify the user when there is one that it picks up on. The applications of this are diverse, and the funding agency that Kei is working with is interested in using this system to monitor small scale nuclear explosions that might not be detected otherwise as part of a non-proliferation tactic since smaller nuclear explosions may not trigger one of the sensors in the global sensor network. Using time difference of arrival analysis, the location can be roughly determined and the frequency detected can help estimate size of explosion because larger explosions have lower frequencies. However, the use of the app and smartphone sensors are not limited to explosions. A prime example includes the MyShake app that is used for earthquake alerts and data collection in California, Oregon, and Washington. These applications are great for citizen science efforts, and Kei encourages people to go collect or hunt infrasound sources such as meteors, aircrafts, or maybe even Sasquach, which is the goal of one of the app’s users.
Future of the smartphone sensor network
Looking ahead, Kei envisions expanding his work into natural disaster applications, inspired by his upbringing in Japan and prior research experience. They contemplate expanding the use of smartphone sensors for broader geographical areas, particularly those prone to geological activity, such as the Pacific and the Ring of Fire.
Kei is also passionate about open science and emphasized the need for accessible data in scientific research. He shared an anecdote about how it took two years to make an acoustic explosion dataset publicly available due to different organizational policies that funded the experiments. Data accessibility can often be a large barrier to innovation especially for those who are not well funded or lack the connections to access data.
In reflecting on his interdisciplinary approach, Kei encourages a “jack of all trades” mindset, highlighting the value of diverse interests in his field. His unique blend of expertise in applied math, physics, and audio engineering exemplifies the potential for creating substantial impact in disaster-related research. Kei’s passion for mitigating the consequences of disasters, combined with his versatile skill set, positions him at the intersection of technology, mathematics, and humanitarian efforts.