A Machine-Learning Approach for Earthquake Magnitude Estimation

TL;DR

This paper introduces a deep learning model that estimates earthquake magnitudes from raw waveforms at a single station, achieving high accuracy and robustness without needing data normalization or instrument response correction.

Contribution

It presents a novel single-station deep learning method combining CNN and RNN for earthquake magnitude estimation directly from raw seismic data.

Findings

Average error close to zero in magnitude prediction

Standard deviation of ~0.2 in estimates

Effective for local and duration magnitude scales

Abstract

In this study we develop a single-station deep-learning approach for fast and reliable estimation of earthquake magnitude directly from raw waveforms. We design a regressor composed of convolutional and recurrent neural networks that is not sensitive to the data normalization, hence waveform amplitude information can be utilized during the training. Our network can predict earthquake magnitudes with an average error close to zero and standard deviation of ~0.2 based on single-station waveforms without instrument response correction. We test the network for both local and duration magnitude scales and show a station-based learning can be an effective approach for improving the performance. The proposed approach has a variety of potential applications from routine earthquake monitoring to early warning systems.