Deep-Learning-Based Noise Suppression for Earthquake Monitoring in an Urban Setting with Long Beach Dense Array
Abstract
Earthquake monitoring in urban areas is crucial to seismic risk mitigation, but is challenging due to strong cultural noise sources that can overwhelm relatively weak earthquake signals. The deep-learning-based DeepDenoiser algorithm has shown remarkable performance for the denoising of seismic data recorded in relatively quiet places, where it is originally trained. In this research we extend the application of DeepDenoiser to an urban seismic environment, by training the DeepDenoiser with a large catalog of urban noise sources from the Long Beach dense nodal deployment. The Long Beach deployment comprised thousands of single-component (vertical) sensors were deployed in two phases in 2011 and 2012. These data represent one of the richest available sources of seismological urban noise for training an enhanced deep-learning-based denoising algorithm. Despite the noisy environment, the Long Beach nodal array data has been successfully used for structural imaging and for imaging/detecting weak earthquake sources. The extended version of DeepDenoiser should allow us to recover earthquake signals more reliably from noisy urban seismic settings. We will use it to explore whether recently reported deep seismic events under Long Beach, which were imaged using these same data, are real or artifacts. More generally, denoising should help improve the performance of existing earthquake monitoring networks in urban settings by improving signal-to-noise ratio and clear identification of earthquake signals.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.S43E0692Y
- Keywords:
-
- 0555 Neural networks;
- fuzzy logic;
- machine learning;
- COMPUTATIONAL GEOPHYSICS;
- 1910 Data assimilation;
- integration and fusion;
- INFORMATICS;
- 1914 Data mining;
- INFORMATICS;
- 1942 Machine learning;
- INFORMATICS