UK markets open in 15 minutes
  • NIKKEI 225

    27,402.05
    +55.17 (+0.20%)
     
  • HANG SENG

    22,029.01
    -43.17 (-0.20%)
     
  • CRUDE OIL

    76.90
    +0.49 (+0.64%)
     
  • GOLD FUTURES

    1,971.60
    +28.80 (+1.48%)
     
  • DOW

    34,092.96
    +6.92 (+0.02%)
     
  • BTC-GBP

    19,218.21
    +476.72 (+2.54%)
     
  • CMC Crypto 200

    542.81
    +300.13 (+123.68%)
     
  • ^IXIC

    11,816.32
    +231.77 (+2.00%)
     
  • ^FTAS

    4,252.85
    -2.87 (-0.07%)
     

AI cuts tsunami impact prediction time to fractions of second in Japan

A new potentially life-saving technology uses artificial intelligence to cut the time for predicting how an approaching tsunami will impact the coastline to fractions of a second, Japanese scientists report in a new study.

Since northeast Japan was struck by a catastrophic tsunami in 2011 that claimed the lives of about 18,500 people, this coast has built on its early warning system, developing the world’s largest network of sensors with over 150 offshore stations for monitoring movements on the ocean floor.

Scientists from the Riken Prediction Science Laboratory in Japan said the data generated by these sensors need to be converted into tsunami heights and extents along the coastline to help save lives.

This requires applying the data to solve difficult mathematical equations – a process that could take about 30 minutes on a standard computer.

But in the new study, published recently in the journal Nature Communications, scientists used machine learning to cut the calculation time to less than one second.

“Conventional tsunami modeling provides predictions after 30 minutes, which is too late. But our model can make predictions within seconds,” study co-author Iyan Mulia from Riken said in a statement.

“The main advantage of our method is the speed of predictions, which is crucial for early warning,” Dr Mulia said.

In the study, researchers trained their machine-learning system using over 3,000 computer-generated tsunami events.

They then tested the system with 480 other tsunami scenarios and three actual tsunamis.

The new machine-learning model could achieve comparable accuracy at only 1 per cent of the computational effort, scientists said.

“The proposed machine learning-based model can achieve comparable accuracy to the physics-based model with about 99 per cent computational cost reduction, thus facilitates a rapid prediction and an efficient uncertainty quantification,” scientists wrote in the study,

However, they said the method is accurate only for large tsunamis higher than about 1.5 metres, adding that there is a need to improve the system’s accuracy for smaller tsunamis.

Researchers believe the new method could be adopted to make similar predictions for other disaster scenarios where time is of the essence.

“The sky’s the limit – you can apply this method to any kind of disaster predictions where the time constraint is very limited,” Dr Mulia said.

“We foresee that underpinned by the accelerating progress in computer science, substantial improvements of the proposed method are most likely expected in the near future, thus enhancing the tsunami mitigation capability globally,” scientists noted.