Researchers use a deep-learning model to identify earthquake waves from the DAS data from the offshore cable

Unused telecommunications fiber optic cables can provide three seconds of improved warning time for offshore earthquake early warning systems, as researchers have shown. The researchers used a deep-learning artificial intelligence model to identify earthquake waves from the DAS data obtained from the offshore cable. There are over 1500 cable landing stations across the globe, and this technology allows the use of operational cables and integration of DAS systems without disrupting telecommunications data transportation. This presents an exciting opportunity for further research.

Seismic stations located offshore of heavily populated coastlines are lacking, which poses a significant challenge for earthquake early warning systems (EEW). These areas are some of the world's most seismically active regions. A new study published in The Seismic Record shows how the conversion of unused telecommunications fiber optic cable can address this issue for offshore EEW.

Jiuxun Yin, a Caltech researcher now at SLB, and colleagues utilized a 50-kilometer submarine telecom cable that runs between the United States and Chile. They sampled seismic data at 8,960 channels along the cable for four days using the Distributed Acoustic Sensing (DAS) technique. This technique uses the tiny internal flaws in a long optical fiber as thousands of seismic sensors.

During the study period, Yin and colleagues used the cable data to determine earthquake locations and estimate earthquake magnitudes for one onshore (magnitude 3.7) and two offshore (magnitude 2.7 and 3.3) earthquakes.

Their results showed that using this single offshore DAS array offers an approximate three-second improvement in earthquake early warning compared to onshore DAS arrays. In a simulation conducted by the researchers, they found that by deploying multiple DAS arrays spaced 50 kilometers apart and working together in the area, they could improve EEW alert times in the subduction zone by five seconds.

Yin expressed that they had anticipated some improvements due to the offshore placement of the DAS array. However, the actual speed gains were even greater than their initial projections. The array's offshore location eliminates the wait time for seismic waves to reach land-based stations, which is the primary advantage.

Offshore Chile and the Cascadia region offshore Canada and the U.S. Pacific Northwest are alike. They both have an active subduction zone, where tectonic plates collide, and one plate plunges beneath another, causing some of history's largest and most destructive earthquakes. Even Southern California's offshore region has witnessed numerous faults that have hosted earthquakes of magnitude 6 or more. In all these densely populated coastal areas, offshore earthquake early warning could help protect lives and property.

Yin explained that Chile's elevated seismic risk was the primary reason for selecting this cable. The region experiences frequent offshore earthquakes and has been affected by several significant magnitude 8+ earthquakes in history, including the largest ever recorded in 1960. Considering the high seismic risk and potentially devastating impacts of a large earthquake, there is a pressing need for a reliable offshore earthquake early warning system in Chile.

The researchers utilized a deep learning artificial intelligence model, which had been trained and validated on previous seismic and DAS data, to identify the earthquake waves from the DAS data of this offshore cable. According to Yin, the volume of data collected for DAS is substantial and pre-trained deep learning models offer a highly efficient and reliable option for real-time applications like EEW. However, other traditional seismological methods of picking earthquakes can still be effective in processing DAS data with automation.

Yin also noted that researchers require more data, particularly from larger magnitude earthquakes, to develop and test EEW algorithms effectively, as well as more information on how DAS instruments respond before building a real-time EEW system that integrates with existing EEW frameworks. He stated that there are plenty of places around the world to continue this research.

As per Yin, "There are more than 1500 cable landing stations around the globe, and the progress in the technology permits the use of operational cables and adding DAS systems without affecting [telecommunications] data transportation. We believe that this opens up a host of exciting research opportunities, and we are keen to explore these in future studies. We are looking for close interactions with cable owners, environmental agencies, and policymakers to scale the DAS-EEW for the benefit of coastal communities."