Earthquakes come without warning, but even a few seconds of rapid detection can reduce damage. Scientists around the world are developing technology to gain those vital few seconds needed to stop trains, close harbor gates, and halt nuclear power plant operations.
A representative event that highlighted the importance of earthquake detection technology was the Great East Japan Earthquake in March 2011. At that time, initial warnings were inaccurate, and some areas had only about 10 minutes of evacuation time. The earthquake’s magnitude was also predicted to be lower than it actually was, leading to a diminished tsunami warning. Furthermore, the incident led to the Fukushima nuclear power plant disaster, which caused major shock around the world.
Japan, having learned lessons from the Great East Japan Earthquake, became the first in the world to complete an ‘underwater earthquake early warning system.’ They installed over 6,000 km of optical fiber cables in surrounding waters and established 186 observation stations connected above the underwater faults. This ‘neural network’ for earthquake detection under the seabed can measure seismic waves and pressure changes caused by tsunamis in real time.
◇Establishment of an optical fiber network extending 6,000 km in Japanese waters
According to the National Research Institute for Earth Science and Disaster Prevention (NIED) in Japan, this system consists of networks such as the Seafloor Observation Network for Earthquakes and Tsunamis (S-net), the Dense Ocean-floor Network for Earthquakes and Tsunamis (DONET), and the Nankai Trough Seafloor Observation Network for Earthquakes and Tsunamis (N-net).
S-net was built mainly around the Japan Trench where the 2011 Great East Japan Earthquake occurred, and it consists of over 5,600 km of cables and 150 observation stations. It extends from the waters off Hokkaido to the waters off Chiba Prefecture in Honshu.
DONET monitors a region of the Nankai Trough in southwestern Japan. N-net supervises the entirety of the Nankai Trough. Located in the waters in front of Nankai Island in Honshu, the Nankai Trough is where the Philippine Sea Plate is subducting beneath southwestern Japan, and it is predicted that a massive earthquake could reach a magnitude of 9.
Starting in 2019, Japan began construction of the N-net, which stretches from the waters off Kochi Prefecture in Shikoku to the eastern region of Miyazaki City in Kyushu, and it was completed last month. Japan is a country composed of four major islands: Hokkaido, Honshu, Shikoku, and Kyushu, extending from north to south. The completion of the underwater observation network for earthquakes and tsunamis connecting all four islands occurred after the Great East Japan Earthquake.
Previously, one had to wait until seismic waves reached sensors installed on land, but now earthquakes can be detected directly above the underwater faults. As a result, earthquake warnings can be issued up to 20 seconds earlier and tsunami warnings up to 20 minutes earlier. This provides sufficient time to reroute aircraft landings or close harbor gates.
Additionally, the underwater earthquake early warning system can detect slow slip phenomena, which are minor crustal movements known as precursors to major earthquakes. This phenomenon, where stress is released without an earthquake occurring, is considered a precursor to large-scale earthquakes and can be a key clue for early warning systems.
◇Detecting earthquakes using existing cables and observation stations
In February, a new technology was developed that utilizes the optical fiber communication networks laid under the oceans worldwide as sensors without needing to bury new underwater cables like Japan. This method detects subtle vibrations transmitted along the optical fiber cables.
Researchers from the Swiss Federal Institute of Technology in Zurich introduced a new algorithm that integrates data from optical fibers and existing seismometers to detect earthquakes more accurately. The algorithm tracks the detected energy to identify the location of earthquakes.
This technology can also be used to detect tiny vibrations in places where existing seismometers are difficult to install, such as underwater, mountainous regions, and urban underground. Existing seismometers are expensive and challenging to install, but optical fiber cables are already densely spread worldwide, making them highly usable.
However, optical fiber cables can only detect vibrations in the length direction and struggle to measure the ground’s movements in three dimensions like existing seismometers. They are also easily affected by external vibrations such as urban traffic or industrial noise.
Researchers at Cardiff University in the UK developed a seismic detection system called ‘GREAT’ using four underwater acoustic marine observation stations designed to detect nuclear tests. It detects pressure waves that are 10 times faster than tsunamis and interprets signals using machine learning algorithms.
The researchers noted, ‘This system can predict the magnitude of an earthquake, the type of fault movement, and the possibility of tsunami occurrence within seconds and issue warnings,’ but they estimate that 24 underwater acoustic observation stations will be necessary to detect tsunamis globally.
◇AI-based detection technology in areas with few earthquakes
Last year, in Korea, AI-based earthquake detection technology was developed to ensure the safety of transport systems sensitive to vibrations, like high-speed trains.
A research team led by Professor Yoo Min-taek from Gachon University’s Department of Civil and Environmental Engineering trained AI on the vibration data experienced by high-speed trains during normal operations to create a model that can distinguish normal vibrations from abnormal vibrations caused by earthquakes. This is an ‘unsupervised learning approach’ that can capture anomalies based on normal state data without needing earthquake data.
The research team compared the performance of the existing earthquake monitoring methods with that of the AI model. Analyzing the accuracy in ground acceleration regions, where there is a heightened risk of train derailments during an earthquake, they found that the AI model detected earthquakes more quickly and accurately. It does not react when the train is running normally and only sounds an alarm in actual earthquake situations, making it more efficient than existing methods.
This technology can be easily applied in areas where earthquakes are rare. Previously, it was challenging to establish early warning systems in areas where it was difficult to secure various earthquake data, but this AI can operate with only normal data, allowing for expansion into more regions.
The research team explained, ‘Korea also heightened its awareness of earthquakes following the Pohang earthquake in 2016 and the Gyeongju earthquake in 2017, and an earthquake detection and warning system that can quickly slow down or stop trains is necessary to mitigate damage.’
References
Geophysical Journal International (2025), DOI: https://doi.org/10.1093/gji/ggae459
Scientific Reports (2024), DOI: https://doi.org/10.1038/s41598-024-51354-7