Localization and Tracking of Urban Anthropogenic Vibration Sources Using Dense Seismic Arrays and ConvLSTM-Based Spatiotemporal Modeling
Mingcai Chai, Miao YU, Hao Meng — Journal of Applied Geophysics, 2025
Under Revision
Abstract
We present a framework for localizing and tracking urban anthropogenic vibration sources — including moving vehicles, construction activities, and industrial machinery — using dense seismic sensor arrays combined with a ConvLSTM-based spatiotemporal deep learning model. The approach integrates beamforming-based source localization with recurrent convolutional architectures to capture both spatial and temporal dependencies in the vibration wavefield. Experiments on real urban datasets demonstrate superior tracking continuity and robustness compared to conventional array-processing methods, enabling practical deployment for smart city infrastructure monitoring.
Status: Under revision for Journal of Applied Geophysics (2025)
Contribution: Co-author; contributed seismic signal processing methodology and urban monitoring system architecture design.
Significance for EB-1A: This work bridges geophysical array processing with state-of-the-art deep learning architectures for urban intelligence applications — directly demonstrating the translational impact from fundamental geophysical methods to city-scale monitoring solutions.