| Abstract Submission No. | ABS-4.1-0195 |
| Title of Abstract | Hybrid Quantum-Deep Learning Framework for First-Arrival Picking in Local Seismic Data |
| Authors | Dr. Bijayananda Dalai*, Dr. Prakash Kumar |
| Organisation | CSIR-National Geophysical Research Institute, Hyderabad-500007 |
| Address | CSIR-NGRI, HYDERABAD Hyderabad, TELANGANA, India Pincode: 500007 E-mail: bijayanandadalai@gmail.com |
| Country | India |
| Presentation | Oral |
| Abstract | Accurate picking of seismic first arrivals is fundamental for earthquake monitoring, source characterization, crustal structure estimation and subsurface imaging, yet remains challenging under noisy conditions. In this study, we present a hybrid quantum-deep learning framework that integrates time-frequency mapping, unsupervised feature extraction, and quantum clustering for automated first-arrival picking from local earthquake data.The raw seismograms are first transformed using the Generalized S-Transform to capture spectral-temporal patterns, followed by a convolutional autoencoder that extracts noise-resilient encoded features without the need for labeled data. Then these encoded features are combined with statistical metrics from the original waveforms to form an enriched feature space. Subsequently, quantum clustering, implemented in the simulated quantum environment, is applied to isolate signal-bearing segments and enable reliable first-arrival picking.The effectiveness of the proposed approach is first validated on a suite of synthetic datasets contaminated with various levels and types of noise, and subsequently applied to observed data from the STEAD global database and seismic stations in the Jammu and Kashmir Himalaya. The method demonstrates stable first-arrival picking performance under noisy conditions when compared with STA/LTA, AIC Picker, and unsupervised deep learning using classic K-Means. It also exhibits a broadly similar trend to state-of-the-art supervised models such as PhaseNet and EQTransformer, while maintaining computational efficiency even in low signal-to-noise ratio environments. |
| Keywords | Computational Seismology, Quantum Clustering, Unsupervised deep learning, Signal detection, First-arrival picking |