Seismic activity caused by shallow （a few hundred meters） and deep （thousands of meters or more） mining operations is collectively referred to as mining-induced earthquakes. In regions where such earthquakes occur frequently, the overall geological structure of the mining area can become unstable. This not only poses significant risks to underground production safety but also threatens surface infrastructure. Establishing a comprehensive and accurate catalog of mining-induced earthquakes provides an effective regulatory reference for the mining industry, ensuring the safe exploitation of mineral resources. It also offers a scientific basis for assessing the safety of mining regions. Furthermore, these data serve as important references for future seismic hazard assessments and the delineation of active faults. By systematically recording and analyzing mining-induced earthquakes, we can better understand and predict seismic activity, thereby providing a solid scientific foundation for the development of earthquake disaster prevention and mitigation strategies. This not only safeguards the lives and property of miners but also reduces secondary disasters caused by mining-induced earthquakes, therefore facilitating social stability and sustainable development.

The core methodology of this research is the Match and Locate （M&L） method, a type of template matching technique that utilizes waveform similarity to detect weak or missed seismic signals within continuous waveform data. M&L method has the advantages of mature principle, low false detection rate and less retrieval omission. This method can not only reduce the dependence on the velocity model, but also further detect earthquakes with smaller magnitude and similar earthquakes farther away from the template event, and give the location information of the detected event. Eight representative mining-induced earthquakes recorded by the Beijing Digital Telemetry Seismic Network and the China Earthquake Networks Center were selected as template events. These events are characterized by clear waveform and high signal-to-noise ratio. Using these templates, continuous seismic waveform data from January 2016 to December 2017 were scanned to identify events with similar waveform features that were potentially missing in the official earthquake catalog.

By setting the correlation coefficient threshold and ensuring that detected peaks exceed ten times the median absolute deviation （MAD） of the ambient noise, a total of 280 mining-induced earthquakes were successfully detected. This number is significantly larger than that of the officially cataloged events. The newly detected events were primarily located near the Da’anshan and Muchengjian coal mining areas, consistent with known areas of high mining activity. This spatial clustering further supports the validity of the detection results.

Firstly, the correlation between the detected waveform and the template is recalculated. It is found that the waveform is very similar to the template waveform, which proves the effectiveness of M&L method for detecting mining-induced earthquakes. Secondly, all detected events were of relatively small magnitude, all below M3.0, with the majority falling between M1.5 and M2.5. This magnitude distribution aligns well with expectations for mining-induced earthquakes, which tend to generate low-energy seismic signals. Secondly, the temporal distribution of events was found to be scattered without any apparent periodicity or clustering patterns. This irregular timing, combined with the lack of synchrony with mining schedules, suggests that the detected events were not the result of deliberate human activities, such as blasting.

Then, time-frequency analysis using spectrograms showed that the waveforms of these events exhibited slow attenuation, well-developed surface waves, and a dominance of low-frequency components. These characteristics are consistent with known properties of mining-induced earthquakes, which are often associated with shallow focal depths and complex wave propagation paths through fractured rock masses.

Additionally, a pairwise cross-correlation analysis was conducted on all detected events within each year to assess internal waveform consistency and potential source similarity. The results showed that most of the correlation coefficients exceeded 0.5, indicating that many events were not isolated but instead shared similar source characteristics or occurred within similar geologic and stress environments. Further, additional parameters （such as focal mechanism, depth estimation, etc） are required to verify whether these events are of the same type.

This study provides a robust case for applying the M&L method in mining regions, particularly under conditions where traditional seismic monitoring methods might miss low-magnitude events due to limited station coverage or weak signal amplitudes. The detection framework established in this study not only enhances our understanding of mining-induced earthquakes in the Beijing area but also demonstrates a scalable and adaptable approach for other mining regions with similar monitoring needs.

Furthermore, the study suggests that while the detection thresholds used in this research were effective in the current context, they can be dynamically adjusted in future applications to optimize detection performance under different environmental conditions. For instance, in areas with higher ambient noise or lower station density, more stringent thresholds may be necessary to reduce false positives, whereas in low-noise environments, thresholds could be relaxed to increase sensitivity. The adaptability of the method enhances its utility for broader applications, including earthquake hazard assessment, mine safety management, and subsurface stress monitoring.