← Back to Sessions

Building Local sP and PmP Phase Datasets for Precise Earthquake Location and Whole-crust Imaging Using AI-assisted Methods

Description: 

Later seismic phases, such as sP depth phases and Moho-reflected PmP waves, provide critical constraints on earthquake source parameters and crustal structure, yet they are rarely used in routine earthquake studies. For instance, accurate focal depth is often difficult to determine due to trade-offs with origin time, especially in regions with sparse near-source stations. The sP phase offers a powerful constraint on depth, but its systematic use has been limited by identification challenges.

We develop an integrated workflow that combines expert phase analysis with artificial intelligence tools to build high-quality local sP and PmP phase datasets. Key characteristics of sP and PmP waves are first identified through careful analyst review to generate reliable training datasets. Based on these expert-labeled data, deep-learning models, including PmPNet for PmP identification and sPNet for sP detection, are trained to automatically identify these later phases from large seismic datasets.

Our primary goal is to construct comprehensive local sP and PmP datasets in seismically active regions such as California and Sumatra, enabling broader use of these phases in earthquake studies. Initial applications in California demonstrate that incorporating sP phases reduces focal-depth uncertainties by a factor of five for most crustal events. These expanded datasets will provide a robust foundation for improved earthquake relocation and high-resolution imaging of Moho topography and lower-crustal structure.

Session: New Frontiers in Seismic Observations and Modeling with Innovative Methods and Emerging Data on Earth and Other Planets - III

Type: Oral

Date: 4/17/2026

Presentation Time: 04:45 PM (local time)

Presenting Author: Ping Tong

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors