Attenuation properties and source spectra from earthquakes induced by the 2018 geothermal reservoir stimulation in Espoo/Helsinki, southern Finland

Earthquake coda waveforms contain information about the scattering properties of the propagation medium and the source process that is complementary to observations based on first-arrival analysis in reservoir characterization contexts. We explore the effectiveness of the coda wave approach by inverting full S-wave envelopes including direct S-waves as well as coda of 200 small and micro-earthquakes induced by the 6.5 km deep 2018 geothermal reservoir stimulation in Espoo/Helsinki, Finland, for scattering and intrinsic attenuation and source spectra. The Qopen software package was used for the inversion of waveforms recorded by a network of stations consisting of 100 short period instruments organized in several arrays and about a dozen broadband sensors at epicentral distances between 0 and 35 km. The method optimizes the fit between Green's functions obtained with the acoustic, isotropic radiative transfer theory and observed energy densities of induced earthquakes. We use events with a local magnitude range between 0.0 and 1.8 and perform the inversion in frequency bands with central frequencies between 3 Hz and 200 Hz. Intrinsic attenuation and scattering strength are relatively low at high frequencies compared to previous studies as expected for the granitic basement rocks in the study area. The scattering transport mean free path is ~400 km at 10 Hz, ~50 km for higher frequencies and ~120 km for lower frequencies. The inverse of the intrinsic quality factor is between 10^-5 and 10^-4 for frequencies above 10 Hz and up to 10^-2 for low frequencies. Constraining these values locally is pivotal for improved imaging of temporal velocity variations based on ambient noise correlations. Source displacement spectra of the events are presented together with a comparison of estimated moment magnitudes and local magnitudes. The robust results require no tuning during the inversion process which supports the application of the Qopen software for real-time automatic estimates in future stimulation experiments.