Two minutes after the world's largest tectonic plate shook off the coast of Japan, the Japan Meteorological Agency issued a final warning to about 50 million residents: the tsunami caused by the magnitude 8.1 earthquake is approaching the coast. But it was not until hours after the waves arrived that experts estimated the true scale of the earthquake in northeastern Japan on March 11, 2011. In the end, its magnitude reached 9, killing at least 18000 people, and some areas have never even received warnings.
Now scientists have found a way to get accurate data more quickly, using computer algorithms to identify the wake of gravitational waves emitted at the speed of light from faults.
"This is a new way to identify large earthquakes." Richard Allen, a seismologist at the University of California, Berkeley, who was not involved in the study, said, "if we implement this algorithm, we will be more confident in determining whether this is a real big earthquake and can send out alarms on a larger scale earlier."
Scientists usually use seismometers and other equipment to detect earthquakes by monitoring ground vibration or seismic waves. But the amount of early warning provided by seismometers depends on the distance between the earthquake and it and the speed of seismic waves that travel less than 6 kilometers per second. Seismic networks in Japan, Mexico and California provide seconds or even minutes of early warning, which works well in relatively small earthquakes. "But above magnitude 7, the seismic wave will saturate the seismograph, which makes it difficult to identify the most destructive earthquakes, such as the earthquake in Northeast Japan." Allen said.
Recently, researchers involved in the search for gravitational waves (ripples produced by the movement of large objects in space and time) realized that these gravity signals propagating at the speed of light can also be used to monitor earthquakes. "The idea is that once the mass moves anywhere, the gravitational field will change and everything can feel it." "Surprisingly, this signal appears even in seismometers," said Bernard whiting, a physicist at the University of Florida who works on the gravitational wave observatory of laser interferometers
Sure enough, in 2016, whiting and colleagues reported that conventional seismometers could detect these gravity signals. The earthquake caused great changes in quality; These displacements will produce gravitational effects and deform the existing gravitational field and the ground under the seismograph. By measuring the difference between the two, scientists concluded that they could create a new earthquake early warning system. The gravity signal will appear on the seismograph before the first seismic wave arrives, and this part of the signal is usually ignored. By superimposing signals from dozens of seismometers, scientists can identify patterns to explain the size and location of large events, whiting said.
Now, Andrea Licciardi, a postdoctoral fellow at Blue Coast University, and colleagues have established a machine learning algorithm for pattern recognition. They trained the model for hundreds of thousands of simulated earthquakes and then tested it on real data sets in Northeast Japan. Researchers reported in nature that the model accurately predicted the magnitude of an earthquake in about 50 seconds - faster than other state-of-the-art early warning systems.
The gravity signal is too weak to detect earthquakes below magnitude 8.3 with the existing technology, and the system is unlikely to provide more early warning in the seismic area covered by the seismograph. But Allen said it can provide more reliable estimates of the size of large earthquakes, which is particularly important for predicting tsunamis, which usually take an additional 10 or 15 minutes to arrive. Jean Paul ampuero, a seismologist at Blue Coast University and co-author of the paper, said that with this technology, Japanese seismologists can accurately determine the magnitude of the earthquake and issue appropriate alarms "one to two minutes after the earthquake".
However, this technology has not yet been put into use because it has not processed the data in real time. The model will be deployed in Japan, but only for earthquakes generated by specific fault zones that may produce "major earthquakes". Licciardi said the algorithm needs to be trained separately for use in different regions, and researchers are currently training in seismic networks in Peru and Chile.
"We have the first generation algorithm... This is a great progress." Allen, "now let's see if it really works."