Separation of seismic signal and ambient noise using deep neural network

Jiuxun Yin, Marine A. Denolle, & Bing He

Submitted August 12, 2021, SCEC Contribution #11310, 2021 SCEC Annual Meeting Poster #197

Earthquake signals in seismic data are inevitably contaminated with signals from unwanted sources. Separating noise from earthquake signals can greatly improve seismic data analysis, such as earthquake characterization and ambient noise analysis. This work develops a machine learning method to extract transient signals from ambient signals directly in the time domain for 3-component seismograms. We benchmark our architecture development and performance against a time-frequency counterpart (similar to the DeepDenoisier). We explore the generalization of our time-domain denoiser by training on various scales of seismic data. First, we train purely on observed seismograms of local ( < 350 km) events using the STandford EArthquake Dataset (STEAD) data set. Second, we generate a data set for regional earthquakes (350 km-2000 km), combining realistic noises and seismograms generated by deterministic synthetic waveforms. We explore the robustness of the denoiser on various noise structures. Finally, we explore the quality of extracted signals for earthquake characterization and ambient noise seismology.

Key Words
machine learning, seismic signals

Yin, J., Denolle, M. A., & He, B. (2021, 08). Separation of seismic signal and ambient noise using deep neural network. Poster Presentation at 2021 SCEC Annual Meeting.

Related Projects & Working Groups