Seismic waves from large earthquakes are easy to see – think of the classic image of a seismograph, where a pencil draws distinct waves on a spinning paper when an earthquake hits. Even to a trained eye, PEGS are just curves, indistinguishable from noise. It’s hard to prove they’re there. In 2017, Early logo PEGS in Northeast Seismic Data received push back from other seismologists.
But over time, researchers have collected more observations from earthquakes around the world. “I’ve managed to convince myself that the theory is correct,” said Maarten de Hoop, a computational seismologist at Rice University who was not involved in the study. Inspired in part by the controversy over his earlier discovery, he set out to mathematically demonstrate whether gravitational fluctuations should be observable. The key, he says, is to look at the data from the earliest moments of the earthquake, before the P-waves reach the sensors. At that point, the two forces “do not completely cancel each other out,” which means that a signal could theoretically be found in the noise. But the question of whether seismologists can truly separate the two remains.
The new study provides initial validation, de Hoop said. What’s clear is that current instruments can only distinguish the gravity signal from other noisy data during the largest earthquakes—those larger than magnitude 8.0, such as the massive large-thrust earthquakes that affect places like Japan, Alaska, and Chile. Because these large earthquakes are rare, Licciardi’s team created a dataset of hypothetical earthquakes interspersed with noise from real-world earthquakes observed across Japan. This is used to train a machine learning algorithm that will detect the onset of an earthquake and estimate its size based on the gravity signal.
When the researchers applied the model to real-time data from sensors during the Tohoku earthquake, it took about 50 seconds of data to make an accurate detection, outperforming recent state-of-the-art methods, including space-based GPS methods of measuring the ground after the earthquake. sports. The 8-second difference may sound small, but “it’s still significant in the context of an early warning,” Licciardi noted, especially in situations such as the Tohoku earthquake, where coastal residents only have a few minutes to evacuate for an impending tsunami.
In addition, the researchers noted that the model was more accurate in estimating the magnitude of earthquakes, which is crucial for predicting the size of tsunamis. In Japan in 2011, preliminary estimates of earthquakes below magnitude 8.0 indicated that the seismic waves were much smaller.
This method is still far from practical. Caltech seismologist Thomas Heaton described the continued search for gravitational perturbations as “a hammer for nails” as advances have been made in more traditional methods of seismic detection — including in Japan, where officials have passed Add more sensors along Japan to respond to earthquakes in the Tohoku region. Offshore subduction zone and extend its model to account for large earthquakes above magnitude 9.0. For him, the biggest task of an early warning system is to make warnings more practical: battle-test existing methods so that when warnings are issued, people will hear them and know how to react. “Our problem isn’t the sensor. It’s how it gets data from the system and tells people what to do,” he said.
But de Hoop, who describes himself as “enthusiastic” about the new work, noted that it provides a roadmap for improving the method using better data and machine learning techniques. The key to making this work applicable to more common, smaller earthquakes will be figuring out how to lower the magnitude threshold for detecting gravity signals — which may require sensors that directly detect changes in the gravity field. “I think there’s a lot of information out there and a lot of work to be done,” he said.