20 seconds warning, saved 127 children! Can AI predict the 6.8 magnitude earthquake in Luding, Sichuan?

The latest news is that the 6.8-magnitude earthquake in Luding, Sichuan has killed 29 people in Ganzi Prefecture and 17 people in Ya'an City. Another 16 people are missing and 50 people are injured.

I feel extremely heartbroken when I see this news...

In this summer’s super heat wave, Sichuan, as the main force in the West-East electricity transmission The great province has resisted all grievances for the people of the country.

This heroic city chose to carry its own weight: in order to continue supplying power to other brother provinces, it extinguished the street lights, turned off the air conditioner, and plunged itself into darkness.

Subsequently, Sichuan encountered an epidemic again, and cities such as Chengdu fell into silence.

Yesterday, Luding suffered another earthquake, and the number of casualties is still being counted.

Source: China Firefighting

20 seconds, save 127 children

In Hailuogou, the epicenter of the earthquake, two hotel buildings collapsed.

Source: Lizhi News

Mr. Xu, the hotel owner, said that he I still have lingering fears, "I have never seen such a violent earthquake" and "my hands are shaking now." There were many touching scenes during this earthquake. According to Red Star News, after receiving news of the earthquake, the Sichuan Provincial People's Hospital immediately organized an emergency rescue team to go to the rescue, and the National (Sichuan) Traditional Chinese Medicine Emergency Medical Rescue Team also rushed to Luding for rescue. The Provincial Orthopedic Hospital immediately formed an emergency team and headed to the epicenter to carry out rescue operations.

​Who can go? I can! Really in tearsThe Office of the State Council Earthquake Relief Command and the Ministry of Emergency Management have launched the national earthquake emergency level three response and dispatched and deployed forward rescue.

Sichuan has urgently dispatched "large-scale high-altitude full-network emergency communication drones" to fly to the epicenter to perform public communication coverage tasks.

Source: Xiaoyang Video

Chengdu Fire Ground and Air Rescue Forces are also urgently gathering to rush to the epicenter to carry out joint reconnaissance and search and rescue missions. Search and rescue helicopters are constantly circling in the sky, never letting go of any possible opportunity!

Source: China Firefighting

At a moment when people across the country are worried, Jiupai News reported a gratifying news: After seeing the 20-second earthquake warning countdown message, teachers from a kindergarten in Ya'an immediately rushed to the class, woke up the sleeping children, and evacuated them one by one to a safe place. . The whole process took 1 minute and 57 seconds, and 127 children were successfully evacuated to a safe place.

Source: Jiupai News

Earthquake Early Warning This Key 20 In seconds, 127 children were saved.

earthquake early warning has made great contributions

As of press time, the earthquake has killed 65 people. In addition to being heartbroken, many people have raised this question: Why is it that even though technology is so advanced now, it is still impossible to prevent casualties through earthquake early warning? In fact, earthquake early warning has played a big role in this earthquake. At 12:52 on September 5, when a 6.8-magnitude earthquake occurred in Luding County, Ganzi Prefecture, Sichuan Province, many Chengdu citizens heard a countdown broadcast of "56, 55, 54..." on their mobile phones. Citizens immediately evacuated to safe areas.

This earthquake early warning countdown, known as the "big speaker", comes from the "earthquake early warning" system independently developed by our country. After the Luding earthquake in Sichuan Province, earthquake warnings were issued in Chengdu, Ganzi, Luzhou, Ya'an and other places in Sichuan Province, including many people's TVs, mobile phones and other terminal devices. In short, earthquake early warning is to use the characteristic that electromagnetic waves (about 300,000 kilometers/second) are faster than seismic waves (about 4 kilometers/second) after an earthquake occurs to send out warnings that seismic waves are about to arrive in areas where the seismic waves have not yet arrived.

Source: Sichuan Provincial People’s Government official website

What are the earthquake early warnings? How important? Research shows that: if the warning time is 10 seconds, casualties are reduced by 39%; if the warning time is 20 seconds, casualties are reduced by 63%. The Wenchuan earthquake that year killed nearly 70,000 people, which caused us too much pain... If we had such an early warning system at that time, the death toll could be reduced by at least 30%.

After the Wenchuan earthquake, my country increased its investment in disaster early warning in areas prone to earthquakes. As an earthquake-prone area, Sichuan Province began pilot installation of an early warning system in Chengdu in 2012. By the end of 2020, the early warning system can cover the entire province of Sichuan. The "earthquake early warning" system that played a big role this time is the mainland earthquake early warning network jointly built by the Chengdu High-tech Disaster Reduction Institute and the China Earthquake Administration. Seismic waves are divided into longitudinal waves and transverse waves. The former propagates fast and has little destructive power, while the latter propagates slowly and has high destructive power. The basic principle of earthquake early warning is to use the time difference between the two to issue an alarm.

After receiving the seismic longitudinal waves, the seismic observation instruments closer to the earthquake source will estimate the seismic parameters and their effects in real time and continuously, and send signals to the possible disaster areas where the seismic shear waves have not yet reached, a few seconds in advance. An alarm message is sent out within tens of seconds. According to an interview with Wang Dun, director of the institute, in Science and Technology Daily, after an earthquake early warning network is established using seismic sensors, once an earthquake occurs, the earthquake early warning system will use the characteristics of electromagnetic wave transmission speed to be much faster than seismic waves, and deploy it through reading and analysis. Recorded data from earthquake monitoring stations in various places are transmitted in real time to quickly determine the location and approximate magnitude of the earthquake, and then fully automatically issue early warnings to users who have not yet been affected, from seconds to tens of seconds in advance.

##Source: China Science and Technology Network

Due to earthquake waves affecting different areas The time is different, and the location of the terminal equipment that receives the earthquake early warning message is also inconsistent, so the early warning countdown time will also be different. Wang Tun said that after the 6.8-magnitude earthquake occurred in Luding County, the early warning network issued an early warning in 5 seconds when the 6.8-magnitude earthquake occurred in Luding, 7 seconds in advance to Kangding City, 20 seconds in advance to Ya'an City, and 20 seconds in advance to Chengdu City. 50 seconds advance warning.

my country’s earthquake early warning achievements, developed since the Wenchuan aftershock area, are already at the leading level in the world. Currently, the second-generation mainland earthquake early warning network is widely used in Sichuan. The second-generation earthquake early warning technology has comprehensively upgraded the front-end early warning stations, mid-end transmission response channels, and terminal processing systems based on distributed processing. It further optimizes the algorithm through the cloud computing operation center and can use 5G communication technology to improve system response. speed. However, all we can do now is to issue early warnings quickly after an earthquake occurs.

Predicting earthquakes that have not yet occurred is still an unsolvable problem for the whole world. Earthquake forecasting refers to predicting the time, location, and magnitude of possible future earthquakes based on monitoring data before an earthquake occurs. According to Director Wang Tun’s introduction:

What we do is earthquake early warning, not earthquake prediction. Imminent earthquake prediction is still a problem in the world. Earthquake early warning can only reduce casualties, but cannot prevent them.

Because it can only provide early warning but cannot predict, casualties cannot be avoided.

Earthquake prediction is still a world problem

Why is it so difficult to predict the next major earthquake? Let's start with a simple question: Where will the next big earthquake happen?

​

Research shows that larger faults often cause larger earthquakes. In theory, if all the faults were mapped, we should be able to put limits on the strongest earthquakes a given region could experience. However, estimating fault size and the corresponding energy released is not always straightforward. Faults often exhibit complex geometries, which complicates modeling of fault regions.

Additionally, faults can rupture simultaneously: During New Zealand's 2016 Kaikōura earthquake, 13 different faults ruptured simultaneously. Furthermore, the magnitude of an earthquake does not necessarily correlate positively with its destructive power. Depending on where they occur, moderate-magnitude earthquakes can be more damaging than "large" earthquakes. For example, the 1994 Northridge magnitude 6.7 earthquake in California caused significant property damage and loss of life, while the 2018 Fiji magnitude 8.2 earthquake, which was 178 times stronger, did not cause any damage.

Now, let’s look at this more complex question: When will the next big earthquake happen? Predicting time is the most difficult challenge in earthquake prediction. There are two theories that explain why predictions about time are flawed.

The first theory is called the elastic rebound theory, which states that the earth's crust bends and deforms under intense pressure until it eventually breaks under the strain. Slip along a fracture (i.e., an earthquake) causes the rocks on either side to rebound to a less deformed state and release stored energy, allowing the process of accumulated strain to begin again.

The second theory, known as signature earthquakes, describes the faults produced by the most studied earthquakes that appear to have distinct segments. In the intervals between earthquakes, these plates rupture repeatedly, accumulating the same amount of strain and thus producing earthquakes of similar magnitude.

Assuming that these two theories have always existed, we can based on

1) The position of the maximum unstressed strain,

2) The time since the last earthquake, and precise knowledge of fault zones to predict when the next earthquake will occur.

It can be seen that due to various complex factors in reality, it is difficult for us to predict the time when an earthquake will occur.

So, what will happen if artificial intelligence technology is used?

Theoretically, in earthquake prediction data, there are many patterns and signals that humans cannot see but that artificial intelligence can clearly detect.

Is it feasible to use AI to predict earthquakes?

In 2018, researchers from Google and Harvard University demonstrated in a paper published in Nature How to predict aftershock locations using deep learning. The model looked for patterns in a database of more than 131,000 "mainshock-aftershock" events, and then tested its predictions on a database of 30,000 pairs of similar events. The results showed that the deep learning network was more reliable than a model known at the time as "Coulomb rupture stress change."

##Paper address: https://www.nature.com/articles/s41586-018-0438-y

However, one year after the article was published, data scientist Rajiv Shah discovered when analyzing the results: the accuracy of the neural network was a bit ridiculously high. In subsequent recurrences, a very serious problem emerged - the data used to train and test the model overlapped (data leak). The short answer is that the model already knows the answers to the test during training. And this means that the test results have almost no practical significance. However, Nature ignored the researcher's questions and suggestions in the end... Despite this, scientists have not given up exploring this field, and various related studies are emerging one after another.

At the 2021 Annual Meeting of the Seismological Society of America (SSA), a research team from Stanford University proposed a new method for real-time analysis of seismic signals - DeepShake, and later published In Seismological Research Letters. The training data for the model comes from earthquake records in California in 2019. The test results show that DeepShake can issue an alarm between 7 and 13 seconds before the arrival of high-intensity ground shaking.

Compared with traditional earthquake early warning, DeepShake can directly issue early warning based on the characteristics of ground motion, eliminating some intermediate steps used by existing systems. The researchers say DeepShake demonstrates the potential of machine learning models to improve the speed and accuracy of earthquake warning systems.

Paper link: https://doi.org/10.1785/02202101412022, Researchers from the Institute of Geology and Geophysics of the Chinese Academy of Sciences and Stanford University proposed a new method in a paper published in Science Advances that greatly improves the detection capabilities of earthquake monitoring networks located in urban areas.

UrbanDenoiser Algorithm By filtering out background seismic noise, it improves overall signal quality and recovers signals that may have previously been too weak to record. Applications to Phalanx data and seismic sequences in urban areas show that UrbanDenoiser can improve signal quality and recover signals at signal-to-noise ratios as low as ~0dB.

Paper link: https://www.science.org/doi/10.1126/sciadv.abl3564

What can be expected is that with the continuous improvement of computing power and the continuous improvement of data, AI will also have further development in the field of earthquake prediction. Finally, let us pray silently that more people affected by the disaster can be rescued. God bless Sichuan and God bless China.

The above is the detailed content of 20 seconds warning, saved 127 children! Can AI predict the 6.8 magnitude earthquake in Luding, Sichuan?. For more information, please follow other related articles on the PHP Chinese website!