Abstract Single-station seismic localization technology enables accurate earthquake location using merely three-component waveform data from a single station. It has significant value for enhancing seismic monitoring in regions with sparse station coverage and for improving the location accuracy of small earthquakes. Conventional single-station methods face two critical limitations: (1) epicentral distance estimations depend on regional velocity models, and inaccuracies of these models directly transfer into location errors, and (2) back-azimuth estimations based on waveform polarization often suffer from large uncertainties and low reliability, especially under low signal-to-noise ratio (SNR) conditions. To address these challenges, this study proposes a deep neural network–based approach for single-station three-component seismic localization. The method employs two networks that take three-component waveforms as input to predict epicentral distance and back azimuth independently. Completely data driven, our approach required no prior information or manual intervention. We trained the model using a dataset containing 367k high-quality seismic events, followed by systematic evaluation across three dimensions: (1) performance comparison with mainstream localization models, (2) robustness testing under low-SNR (20) conditions, and (3) generalization assessment using Kyoshin Network (K-NET) data. Experimental results demonstrate that our method achieves high-precision localization in most scenarios, validating its effectiveness and practical utility. Furthermore, we conducted an in-depth comparison between regression and classification models for back-azimuth prediction, finding that the classification model has slightly higher errors than the regression model, but it can provide uncertainty estimates for predicted back azimuths. This research provides a novel technical solution for advancing seismic monitoring capabilities, particularly in network-sparse regions and for the monitoring of small events.
Guo et al. (Wed,) studied this question.