Item

Abstract

Detection and localization of microseismicity is an inevitable task to monitor fluid injections into subsurface rocks during hydraulic stimulations. Traditionally, such observations are done with downhole geophones, but in the last years, the usage of distributed acoustic sensing (DAS) using fiber-optic cables placed into boreholes has become a standard technique. However, DAS registrations still have lower signal-to-noise ratios than geophones, i.e., they are not able to detect small-magnitude events. In this work, we develop a convolutional neural network (CNN) that is capable to detect and distinguish P- and S-arrivals in continuous DAS recordings. This network is trained on several hours of DAS data recorded at the Utah FORGE EGS project in 2019. By also incorporating arrival-time information from geophones placed in the same borehole during the training, we are able to shift the detection threshold towards smaller magnitude events significantly. Although the number of microseismic events (≈150) used for training is small, the tested network performance is high and provides a complete event catalog down to magnitude Mw=-1.6. This means, that after a short training phase, the network can be used for long-term real-time observation. Although the network currently only detects seismicity, event localization is feasible with small network adaptations. Furthermore, the methodology and network architecture can be easily adjusted for similar case studies, where long-term seismic monitoring is required (e.g. EGS or CO₂ sequestration).