Centroid full moment tensor analysis reveals fluid channels opened by induced seismicity at EGS, Helsinki region, southern Finland

Understanding fluid injection induced seismicity is key to safe and successful operations of deep geothermal systems. Efficient geothermal energy extraction by an enhanced geothermal system (EGS) requires increased fluid flow between geothermal wells. The experimental 6-km-deep EGS in the Helsinki capital region, southern Finland is an intriguing natural laboratory in a cool Precambrian shield setting that yields excellent seismic data quality. We investigate the source processes of the earthquakes induced by weeks-long EGS stimulations in 2018 and 2020 via a probabilistic waveform fitting method. Detailed resolution of full moment tensor solutions and their opening components can reveal crucial information on earthquake nucleation and fluid flow patterns.

We present results of a centroid full moment tensor analysis for ~250 events from 2018 and 16 events from 2020 in the moment magnitude range 0.5–1.9. We use three-component data of ~30 stations within a 9-km radius of the well-head site. We fit P- and S-phases by modeling synthetic waveforms using Green’s functions with a 20 m grid spacing based on a homogeneous velocity model. We employ automatic high signal-to-noise ratio waveform selection and automatically determined channel-wise correction coefficients for time shifts and amplitude scaling to represent small scale crustal variations not reflected in the velocity model. With the application of both waveform selection and channel corrections, the uncertainty of the moment tensor decreases on average by ~60 % and the location uncertainty by ~85 %. This results in a catalog of well-resolved moment tensors and centroid locations.

The obtained high-quality solutions are dominated by reverse faulting mechanisms with variable compensated linear vector dipole (CLVD) contribution and non-significant isotropic component. The 3D event distribution reveals largest positive CLVD contribution in seismic sources close to the injection well, which indicates localized fracture opening under constant volume with a simultanous adjacent shear event. Farther from the well, seismic sources have pure double-couple mechanisms or even negative CLVD contribution which may be indicative of fracture lengthening or closing under constant volume at later stages of the stimulation.

Identifying clusters with respect to source type and location within the 3D event distribution supports the interpretation of physical source processes and reveals fluid flow channels, and zones of weakness. Events with positive CLVD component occurring close to fluid-filled fractures are potentially nucleated by direct contact with the injected fluid and the associated pore pressure change. Events with zero or negative CLVD component on the outer parts of the seismicity distribution may have been nucleated by poroelastic stress transfers without a direct hydraulic contact to the injected fluid. Our findings suggest that the full extent of injection induced seismicity may not be indicative of fluid flow and thus it should not be used to estimate the extent of an artificially created connected fracture network of a geothermal reservoir.