Logo

New AI earthquake early warning system promises more accurate predictions

New AI earthquake early warning system promises more accurate predictions

2024-04-11

With earthquakes such a frequent occurrence in Taiwan, effective earthquake response systems are a big industry. A new AI earthquake early warning system is in development, which promises to provide more accurate predictions for quake magnitude. The new system has been developed at National Central University. The deputy head of its Earthquake Disaster and Risk Evaluation and Management Center, Chan Chung-han, explains how the new system could be used to reduce risks and casualties.

Chan Chung-han
Earthquake Disaster and Risk Evaluation and Management Center
We’re looking at the Hualien region during the recent earthquake. The 32 represents the Milun Fault. The 22 represents the Huadong Longitudinal Fault. We all know that it has a higher probability of earthquake occurrence in the future.

National Central University has a two-year project, funded by the Central Weather Administration, to create a next-generation AI earthquake early warning system, TT-SAM. Its aim is to win citizens a bit more time to prepare for an earthquake, and thereby reduce casualties. It will be able to pick up on tremors faster, and provide intensity predictions for each county with at least 90% accuracy. Developers hope it will be used eventually in the official earthquake warning system.

Chan Chung-han
Earthquake Disaster and Risk Evaluation and Management Center
When our station has picked it up, within about six seconds we’ll be able to assess which places are facing tremors of an intensity of 4 or more, for almost everywhere in Taiwan. Then we can use this method to issue warnings for most places.

The system uses artificial intelligence to monitor earthquake imagery, drawing on AI quake data training models built in Japan and Italy. It takes into account the ways that tremors are magnified in certain places in Taiwan, to optimize predictions. The system could predict that the recent quake of magnitude 7.2 in Hualien would have an intensity of 4 or over in most parts of Taiwan. Apart from making such predictions more accurate, the system will also issue its analyses faster than traditional warning systems.

Chan Chung-han
Earthquake Disaster and Risk Evaluation and Management Center
In terms of the needs of official institutions, I am confident they will be able to send information to people who need it. In six seconds, they can’t make a high-speed train slow down from 200 kilometers an hour to zero. But they can effectively reduce the speed, and maybe avoid some potential dangers.

Taiwan is in a high-frequency earthquake region. With advanced warning systems, response times are in the seconds, and every second can make a big difference in improving outcomes.

For more Taiwan news, tune in:
Sun to Fri at 9:30 pm on Channel 152
Tue to Sat at 1 am on Channel 53

比現有預警還快! 央大打造新一代AI地震預警系統

2024-04-11

台灣身處地震帶,如何更快預警,變成重要課題,中央大學獲氣象署經費支持,花兩年時間,打造新一代AI地震預警系統,透過AI擷取地震波,能在測站接受後六秒,就提供各縣市震度,正確率高達九成以上,比現有的傳統預警表現更好。

[[央大地震風險中心副主任 詹忠翰]]
“這次發生地震的花蓮地區來說的話,那32就是代表著米崙斷層,然後33代表著花東縱谷斷層,我們都知道說它在未來,它有著相對來說,比較高的地震發生機率”

為了替民眾爭取多一點應變地震的時間,把傷亡降的更低,中央大學獲得氣象署補助,花兩年多時間,打造出新一代AI地震預警系統(TT-SAM),能更快擷取地震波,提供各縣市震度,正確率高達9成以上,希望未來可以應用在地震警報。

[[央大地震風險中心副主任 詹忠翰]]
“當我們測站收到之後,大概六秒之後,我們就可以去評估說,全台灣絕大部分的地方,都將面臨震度四級以上的震度,進而我們就可以用這樣子的方式,然後去對全台灣,大部分的地方發布預警”

這套預警系統,主要是以人工智能來觀測地震波影像,參考日本和義大利地震資料訓練的AI模型架構,再加入台灣不同區域震波放大特性,以優化台灣地震預警表現.以此次花蓮規模7.2大地震為例,透過震波分析,即可評估台灣多數地區震度可達四級以上。除了預估震度更加準確,發布速度也比傳統地震預警系統更快。

[[央大地震風險中心副主任 詹忠翰]]
“公部門的一些需求上面的話,那我相信說,他們會把這些相關的一些資訊,然後去交付給一些需要的一些人。六秒其實雖然沒有辦法,讓高鐵從兩百公里的時速然後降成零,但是它其實是可以有效的去減緩它的速度,那避免一些,可能潛在的威脅”

台灣身處地震頻繁區域,每次地震一來,反應時間都是以秒為單位,多幾秒的預警時間,就多幾秒的機會。

更多新聞內容,請鎖定:
民視台灣台(152頻道)週日至週五晚上9:30
民視新聞台(53頻道)週二至週六凌晨1:00

Related News

Lai receives delegation from Project 2049 Institute

2024-04-11

Experts and lawmakers weigh in on Ma’s meeting with Xi

2024-04-11

Taiwanese business leader donates 140 emergency vehicles to Ukraine

2024-04-11

De facto Japanese ambassador tours FTV headquarters

2024-04-11

Puyuma train derails in Hualien after Car 1 strikes falling boulders at night on Apr. 10

2024-04-11

April 3 quake causes vertical ground displacement of 45.1 cm in Hualien

2024-04-11

NCHU breaks ground on NT$580 million veterinary building

2024-04-11

Spotlight on Lucien Hedquist, Hong Konger artist resident in Taiwan

2024-04-11