<strong>Towards End-to-End Earthquake Monitoring Using a Multitask Deep Learning Model</strong>

Weiqiang Zhu

Presentation information

[E] Oral

S (Solid Earth Sciences ) » S-CG Complex & General

[S-CG40] Science of slow-to-fast earthquakes

Tue. May 28, 2024 9:00 AM - 10:15 AM Convention Hall (CH-B) (International Conference Hall, Makuhari Messe)

convener:Aitaro Kato(Earthquake Research Institute, the University of Tokyo), Asuka Yamaguchi(Atomosphere and Ocean Research Institute, The University of Tokyo), Yohei Hamada(Japan Agency for Marine-Earth Science and Technology), Akemi Noda(Meteorological Research Institute, Japan Meteorological Agency), Chairperson:Aitaro Kato(Earthquake Research Institute, the University of Tokyo), Ryuta Arai(Japan Agency for Marine-Earth Science and Technology)

9:00 AM - 9:15 AM

[SCG40-01] Towards End-to-End Earthquake Monitoring Using a Multitask Deep Learning Model

★Invited Papers

*Weiqiang Zhu¹ (1.University of California Berkeley)

Advancements in seismic data processing provide crucial insights into earthquake characteristics. Conventional methods used for earthquake monitoring tasks, such as earthquake detection and phase picking, are being enhanced by the rapid advancements of deep learning. However, most of the current research focuses on developing separate models for each specific task, leaving the potential of an end-to-end framework relatively unexplored. To address this gap, we extend the PhaseNet model to introduce a multitask framework. This enhanced model, PhaseNet+, can simultaneously perform tasks of phase arrival time picking, first motion polarity determination, and earthquake source parameter prediction. The outputs of these perception-based models can then be processed by specialized physics-based algorithms to accurately determine earthquake locations and focal mechanisms. Our approach aims to enhance seismic monitoring by adopting a more unified and efficient framework.