Abstract:

This study adopts the deep learning-based “DiTing” single-station method for near-source earthquake detection and seismic phase picking. At the technical level of single-station localization, a novel multi-feature fusion approach is proposed to accurately estimate the epicentral distance. This method integrates the virtual wave velocity technique, the travel-time table method, and a multi-orbit surface wave approach. To determine the back azimuth, the methodology combines principal component analysis(PCA), the single-value moving average method(SV), the radial amplitude maximum method, and the surface wave polarization technique. Furthermore, it introduces a dynamic weighting mechanism that adjusts the contribution of each method based on their respective uncertainties and the reliability of the phase picking. This approach effectively addresses two longstanding challenges in traditional single-station localization: the 180° azimuthal ambiguity and high uncertainty in back azimuth estimation.

Using continuous waveform data from the two nearest broadband seismic stations to the 2025 Dingri M6.8 earthquake in Tibet, the study conducted rapid analysis for the period from January 7 to 15, 2025. The analysis identified 2, 255 and 1, 730 aftershocks at each respective station, revealing the spatiotemporal characteristics of the aftershock sequence. The aftershocks display a predominantly north-south spatial distribution that aligns closely with the strike direction of the causative fault. The aftershock zone extends approximately 70 kilometers along the fault, primarily concentrated to the west of the Dengmecuo fault. Moreover, the spatial distribution of aftershocks shows a strong correlation with the pattern of mainshock co-seismic slip, with aftershocks clustering in regions of relatively low slip. This correspondence supports the hypothesis that stress redistribution following the mainshock governs aftershock occurrence. Comparison with aftershock catalogs produced by other researchers further confirms the consistency and reliability of the results obtained in this study.

Extensive experimental results demonstrate that the integration of multiple localization algorithms significantly enhances the stability and accuracy of single-station solutions. The study establishes a cross-validation framework whereby results from different algorithms are compared, enabling the identification and elimination of erroneous data affected by noise or local anomalies. This approach substantially improves the robustness of single-station localization, particularly in complex seismic environments.

The single-station localization technique allows for the rapid inversion of key earthquake parameters within seconds of an event, offering a substantial reduction in response time compared to traditional multi-station localization systems. This improvement translates into critical additional seconds for issuing warnings and initiating evacuations, thereby mitigating casualties and property loss in the affected regions. As seismic data processing techniques and single-station localization algorithms continue to evolve, further improvements in localization accuracy are anticipated. With the accumulation of large-scale seismic datasets and the integration of advanced intelligent computing technologies, the potential of single-station localization for near-field strong earthquake early warning is expected to be significantly expanded in the future.

Key words: 2025 Dingri M6.8 earthquake, DiTing, single-station location, aftershocks

摘要： 文中基于深度学习“谛听”单台法检测近震和震相拾取, 采用多种算法联合求解地震震中距和后方位角, 快速解析2025年西藏定日M6.8地震发生后距震中最近的2个固定宽频带台站2025年1月7—15日期间记录的连续波形资料, 分别获得了2 255和1 730个余震的时空分布特征。结果表明: 多种算法结果交叉验证了文中单台定位结果的可靠性。从余震分布来看, 其大体呈现出SN向展布, 与发震断层走向吻合较好。余震沿着断层走向展布, 长约70km, 主要分布在登么错断裂以西。余震分布与主震破裂滑动量分布联系密切, 主要位于同震滑动量较小的区域。基于文中所述方法, 可快速获取此次西藏定日M6.8地震破裂过程和余震分布结果, 服务于探究地震致灾范围和机理, 为地震灾害评估提供基础资料。

关键词: 定日M6.8地震, “谛听”, 单台定位, 余震