Natural disasters such as earthquakes, tsunamis, cyclones, and typhoons are a not-so-friendly reminder of how powerful our planet actually is. In a matter of minutes, thousands of hundreds of homes and buildings can be swept away, incurring a high number of casualties.
With their power to destroy entire coastal communities lead to unimaginable victims, tsunamis are particularly destructive. What’s even worse, this catastrophic colossal wave traveling at the speed of a jet plane cannot be prevented or stopped.
To avoid being killed, people can only get out of its way. To lessen the death and tragedy tsunamis cause, people inhabiting coastal areas need all the time in the world they can get to evacuate. Unfortunately, despite the latest technology and innovative tools, predicting a tsunami is not an easy feat.
Source: Do.Something.org
Oftentimes, people don’t have enough time to leave, which may result in thousands and thousands of victims. Just remember possibly the most devastating tsunami hitting Indonesia on December 26, 2004, following an earthquake of 9.1 magnitude. The tsunami lasted for seven hours, inundating over a dozen countries, involving, India, Sri Lanka, Indonesia, Myanmar, Maldives, Thailand, Somalia, and Seychelles, and killing a minimum of 225,000 people in its wake.
Owing to artificial intelligence, scientists are devising novel early warning systems that will give coastal residents more time to evacuate. The two researchers, Usama Kadri from Cardiff University and Bernabe Gomez Perez from UCLA seem to have produced an ingenious solution published in the journal Physics of Fluids.
The new solution combines the most advanced acoustic technology with artificial intelligence to promptly classify earthquakes and identify potential tsunami risk.
Old vs. new system
Tsunamis are typically triggered by underwater earthquakes of a magnitude of 7.5 or above when a huge amount of water gets displaced. However, not all underwater earthquakes will generate seismic waves. Thus, determining what type of quake will trigger a tsunami is critical to evaluating the tsunami risk and avoiding false alarms.
The existing warning systems rely on seismic waves produced by undersea earthquakes. Data obtained from seismographs and buoys is sent to control centers which issue a tsunami warning. These current tsunami monitors also use ocean buoys that mark the continents’ coastlines to confirm an approaching wave.
In the open ocean, tsunamis travel at an average speed of 800 kmh, equal to that of a jet plane. They drop down to a car-like pace of 50 do 80 kmh as they get close to a coastline. There is hardly enough time for escape after the buoys have been triggered and the tsunami warning issued. By the time waves approach buoys, people have a maximum of few hours to evacuate. There were cases when that wasn’t enough.
The new system that the duo devised pairs up two algorithms to evaluate tsunamis. An artificial intelligence model evaluates the magnitude and type of an earthquake. An analytical model analyzes the size and direction of a tsunami the earthquake generates. As soon as the software obtains the required data, it can forecast the source and size of a tsunami, as well as coasts of impact, in approximately 17 seconds.
The key differences
Relying on buoys to measure the level of water doesn’t give enough time for evacuation. Due to this, the two researchers suggest measuring the acoustic radiation, i.e., sound, that the earthquake produces. The acoustic radiation or sound, which travels considerably faster than tsunami waves, carries information about tectonic movements.
Acoustic radiation travels through the water column much faster than tsunami waves. It carries information about the originating source and its pressure field can be recorded at distant locations, even thousands of kilometers away from the source. The derivation of analytical solutions for the pressure field is a key factor in the real-time analysis.
Usama Kadri, co-author of the study
The acoustic waves are recorded by underwater microphones referred to as hydrophones. Besides recording, these devices also monitor tectonic activity in real time.
The hydrophones and computational model triangulate the earthquake’s source, and AI algorithms categorize the slip kind and magnitude. The scale of the tsunami is then determined by factors such as effective length and width, uplift speed, and duration.
The AI software can also distinguish between different kinds of earthquakes and their propensity to trigger tsunamis, a typical issue with present systems. Tsunamis are significantly more likely to be caused by vertical earthquakes that elevate or lower the ocean bottom, rather than by horizontal tectonic slips. Still, both can produce similar seismic activity and cause false alarms.
Knowing the slip type at the early stages of the assessment can reduce false alarms and complement and enhance the reliability of the warning systems through independent cross-validation.
Bernabe Gomez Perez, the co-author of the study
Ability to predict all sorts of tsunamis
Tsunamis are typically associated with earthquakes. This is not so surprising, given that more than 80% of them are generated by earthquakes. However, tsunamis can be also triggered by landslides (typically from earthquakes), extreme weather, and volcanic eruptions. In rare instances, they can be a result of a meteorite impact.
The new AI-powered system that the two researchers have developed can predict tsunamis that are not triggered by earthquakes. This is achieved by monitoring vertical movement of the water.
The AI model was trained with historical data from more than 200 earthquakes. Seismic waves and acoustic-gravity waves were employed to identify the size and the scale of a tsunami. Acoustic-gravity waves refer to sound waves moving through oceans at much higher speeds compared to ocean waves, thus providing a quicker prediction method.
The researchers tested their AI model employing available hydrophone data. They found out that it almost immediately and successfully depicted the earthquake parameters with low computational demand. The model is currently being improved by including more information to improve the accuracy of tsunami characterization.
According to a statement from Cardiff University, this recent study on tsunami risk prediction is a component of a long-running endeavor to improve natural hazard warning systems all across the world. Though far from perfect, their most recent innovation is included in user-friendly software that will later this year be made available to national tsunami warning centers.
