Can AI help predict earthquakes?

Last Wednesday, a very strong 7.4 magnitude earthquake hit the popular tourist city of Hualien, located on the eastern shore of Taiwan. The event triggered large landslides and hundreds of aftershocks, many of which were over 6.0 in magnitude. It was the strongest earthquake to hit Taiwan in 25 years.

The earthquake damaged many buildings, and some had to be demolished because they couldn’t be saved. However, the situation could have been much worse if the residents of Taiwan had not prepared for this type of event. The last big earthquake in Taiwan occurred in 1999, when a 7.7 magnitude hit the middle of the island.

Taiwan is the world’s largest maker of computer chips, and the earthquake last week did briefly disrupted operations. The country produces about 90% of the world’s advanced semiconductors that are required for phones, computers, cars, and other devices that rely on these chips.

Two days later, on April 5, a rare 4.7 magnitude earthquake hit the northeastern states. The epicenter was located in Lebanon, N.J., which is about 50 miles west of New York City. The earthquake was felt across 14 states and New York City and Philadelphia were some of the major cities that were shaken. In fact, many live television programs in New York City and other locations were interrupted by the event. It was the third-strongest earthquake near New York City in 70 years. Fortunately, there were no injuries or collapsed buildings that were reported as of late Friday.

The recent earthquakes in Taiwan and the Northeast do not necessarily indicate that earthquake activity is increasing. However, the events of last week do show the importance of future earthquake predictions.

Scientists have worked for years to develop techniques for the early prediction of earthquakes. The progress of this challenge may have taken a big step forward with the use of AI (Artificial Intelligence).

At the University of Texas at Austin, a team of researchers released a study about using AI to increase the accuracy of earthquake prediction. It was a seven-month trial that was conducted in China. Over that period, it was determined that 70% of the earthquakes were predicted in China about a week before they occurred.

According to the University of Texas, the research team trained the AI computers using algorithms of five years of seismic recordings in the region. The experiment did accurately forecast 14 earthquakes within about 200 miles of the epicenter. However, there was one earthquake that AI missed and did predict eight others that did not occur.

With increasing urban sprawl and newly discovered fault systems, earthquake prediction has become even more essential. In the U.S., there are approximately 1,000 faults that are capable of producing earthquakes. More faults have been discovered since the last study in 2018 as there have been 350 additional ones across the country. Currently, scientists determine that approximately 75% of the U.S. is at risk of damaging earthquakes.

The regions with the highest levels of seismic activity include California, Alaska and Hawaii. Washington, Oregon and parts of the central U.S., just to name a few, also have a high risk of damaging earthquakes with at least a 95% chance of a disastrous earthquake in California during this century.

The country with the most fault systems is Japan. It’s currently estimated that over 2,000 faults exist, which also makes this country a very high risk.

In terms of our local weather, rain returned to North Idaho last week as most locations picked up around a half-inch of precipitation. The normal moisture for Coeur d’Alene to date is approximately 8.20 inches. The recent rainfall put our seasonal total over 9.50 inches.

The Pacific storm late last week was cold enough to produce some snow at the lower elevations. For the season, the additional 0.1 inches put our seasonal snowfall total in Coeur d’Alene at 51.1 inches. The heavier snowfalls from that system were reported much farther south of our region. The 51.1 inches of snow is below the seasonal average of 69.8 inches, but as about as close as you can get to the predicted 51.4 inches that Cliff made last October.

In the mountains, seasonal snowfall totals were good. At Silver Mountain, approximately 240 inches of snow was measured at the summit. At Lookout Pass, over 350 inches of snow fell for the season. At Schweitzer, over 230 inches of snow was reported for the season.

Thanks to the recent storm activity, plus the expected occasional rainfall through the rest of the month, Coeur d’Alene should be close to the April normal 1.77 inches of rain and melted snow at the end of the month.

With sea-surface temperatures cooling rapidly along the equatorial regions, the outlook for May is looking warmer with near-normal precipitation. The upcoming summer season still looks drier than normal across the Inland Northwest.

• • •

Contact Randy Mann at randy@longrangeweather.com.