Datensatz

Zusammenfassung

By using machine learning (ML) techniques we can increase the efficiency of the detection of clusters and mapping the rupture areas, which may be further analyzed in the context of the seismogenic process related to local tectonics and hydrogeological background. We are currently seeing an eruption in ML applications from seismology, however, reservoir-triggered seismicity (RTS) has some aspects that require an individual approach. It is mainly due to the following factors: a small number of mainly weak earthquakes limiting the training of deep learning, and seismicity often appears in unexpected areas and as swarms. Above all, such type of seismicity depends on anthropogenic and hydrological conditions. We compared deep learning methods of event detection and location, our own artificial neural network, and other similarity detection methods.We study also the possibilities of ML in searching for seismic clusters in the time and space domain, in the interpretation of the event locations and focal mechanisms of the repeaters with the tectonics, and in dynamic relationships between hydrodynamic changes and the RTS with the use of recurrent neural networks. This work was supported by research project no. 2021/41/B/ST10/02618, funded by the National Science Centre, Poland under agreement no. UMO-2021/41/B/ST10/02618.