A synthetic benchmark dataset for induced seismicity monitoring

Monitoring induced seismicity is an indispensable part of risk management during the creation and operation of enhanced geothermal systems. Due to the relative scarcity of manually labeled, informative datasets of induced seismicity, it can be challenging to evaluate the performance of monitoring tools ahead of time. We have created continuous synthetic seismic waveform data for an induced seismicity sequence at the Utah Frontier Observatory for Research in Geothermal Energy (FORGE). The data are based on a synthetic catalog that mimicks an injection-induced sequence at Utah FORGE and contains approximately 20 000 events occurring during 24 hours with the majority of events during the simulated hydraulic stimulation. Taking into account site-specific geology, induced event waveforms are computed using a spectral-element visco-elastic wave propagation solver and source-receiver reciprocity. Two types of seismic noise are added to create two subsets of test data: low-amplitude Gaussian noise and site-specific correlated noise. We test the retrieval of the known events from the continuous synthetic data using one real-time and one post-processing monitoring workflow based on SeisComP and MALMI  (MAchine Learning aided earthquake MIgration location). The results illustrate reliability and shortcomings of the two monitoring tools. For example, depending on the monitoring tool, the noise conditions and the behaviour of the sequence (injection vs. post-injection), the rate of detected events ranges from approximately 20% to 100%. In addition to this benchmark, the dataset generation also serves as a rough feasibility study for a digital “twin” of wave propagation in an enhanced geothermal system. While uncertainties concerning the elastic medium and receiver coupling, as well as the time to availability of observed induced event data and interpretations are likely to pose challenges, the performance of the reciprocal wave propagation modeling strategy is satisfactory for incorporation into a “twin” if updates to the geologic structure are infrequent.