Abstract
Real-time accurate passive seismic event detection is a critical safety measure across a range of monitoring applications, from reservoir stability to carbon storage to volcanic tremor detection. The most common detection procedure remains the short-term average to long-term average (STA/LTA) trigger developed in the 1970s, in part due to its easy implementation and real-time processing capability. However, it has several well-documented limitations, such as requiring a signal-to-noise ratio greater than one and being highly sensitive to trigger parameters. Although numerous alternatives have been proposed, they often are tailored to a specific monitoring setting and therefore cannot be widely applied, or they are too computationally expensive and therefore cannot be run in real time. This work introduces a deep learning approach to event detection that is an alternative to the STA/LTA trigger. A bidirectional, long short-term memory, neural network (NN) is trained solely on synthetic traces. Evaluated on synthetic and field data, the NN approach significantly outperforms the STA/LTA trigger on the number of correctly detected arrivals as well as on reducing the number of falsely detected events. Its applicability is proven with 600 traces processed in real time on a single processing unit.