A Deep Neural Network to Identify Foreshocks in Real Time

Date: 4/24/2019

Time: 10:45 AM

Room: Elliott Bay

Identifying foreshocks, prior to the occurrence of big earthquakes, is crucial and will make this process a valuable precursor. However, until now foreshock events are only identified retrospectively. The motive of this paper is to identify them in real-time.

I curated a new dataset that contains time series data of around 200 events of foreshock, mainshock and aftershock events each. I created a novel method, based on deep learning, that can classify seismic data into the aforementioned categories. The AI model achieves high accuracy while tested in the curated dataset. This is a big breakthrough as it can forecast the occurrence of massive earthquakes that are imminent.

Presenting Author: Vikraman Karunanidhi

Authors