NewsLocal News


New earthquake detection software being used in Oklahoma

Posted at 6:36 PM, Nov 15, 2020
and last updated 2020-11-15 19:36:50-05

When lock downs started earlier this year, Oklahoma’s State Seismologist, Dr. Jake Walter, used his time to develop a new earthquake detection software package.

“We actually implemented machine learning, which is similar to artificial intelligence, to help us detect the very smallest earthquakes that we can detect in Oklahoma,” Walter said.

The machine learning technique can aid by sorting through the mountain of data and reveal earthquakes that might’ve been missed by humans, especially the smaller ones. This method works so well because the computing software has been trained.

“It’s been trained on millions of seismographs, so millions of different earthquakes that are big and small, and it can detect those earthquakes within some of the noise levels.”

Using this cutting-edge system, they’ve been able to observe twice as many earthquakes compared to conventional sensing methods and small, previously undetectable earthquakes.

“This new technique allows us to get our noise levels down so that we can pick out those smallest events.”

The newly detected earthquakes are too small for people to feel, usually under a two magnitude, but the data is important to study.

“The smaller earthquakes that we add to our catalog by utilizing this technique in tandem with our traditional way of detecting earthquakes allows us to have great detail and richer data sets to analyze and understand where the faults are.”

Walter made sure to say the new software has not replaced the real-time earthquake detecting sensors across the state. Rather, both systems work simultaneously to help spot tremors in the data.

He’s already made the software public in hopes it’ll help other seismologists.

“So we’ve already gotten some indications from colleagues around the world that they’re using it to detect earthquakes in their region or their particular scientific study.”

Stay in touch with us anytime, anywhere --