Applied Research & Methods

We apply machine learning and deep learning to advance earthquake early warning systems, predict ground motions, and uncover hidden earthquakes from noisy datasets.

📄 Paper: Deep RNN (Recurrent Neural Network) in magnitude prediction 📄 Paper: RNN in real-time rupture geometry tracking and shaking prediction

Seismology is inherently data-rich and data-driven. We analyze large-scale observational and simulated seismic data. A typical broadband three-component seismometer collects over 20–65 million data points per day, enabling deep analysis and pattern discovery at scale.

📄 Paper: Unet CNN (Convolutional Neural Network) in earthquake signal data mining

We simulate kinematic rupture models, seismic wave propagation, and inversion of source. These physics-based models provide insights into rupture processes and support hazard assessment.

📂 Dataset: Rupture simulation archived on Zenodo 📄 Paper: Joint inverion for solving kinematic rupture processes

Leveraging HPC resources, we simulate realistic large earthquake ruptures and their ground motions. These simulations are vital for understanding seismic risk in urban and tectonically active regions.