Frequency-dependent quality factors of S-coda waves from active ultrasonic transmission measurements in the STIMTEC project, URL Reiche Zeche, Germany

Mean S-wave coda quality factors (mean-Qc) were estimated from active ultrasonic transmission (UT) measurements acquired during the STIMTEC project in the URL Reiche Zeche (Saxony, Germany). We used S-coda waves of 88 selected UT measurements carried out in 3 differently oriented boreholes (BH10, BH12, BH16) to estimate the spatial change of the coda quality factor in the targeted rock volume, an anisotropic metamorphic gneiss. We also analysed temporal variation in attenuation before and after hydraulic stimulations performed in two boreholes (BH10, BH17). We formed in total 8 UT groups (see data table "2022-004_Blanke-and-Boese_mean_UT_event_locations") from neighbouring UT measurements within different depths and from separated time intervals (see also Tab. 1 in Blanke et al. 2023), and compare mean-Qc estimates of centre frequencies ranging 3-21 kHz of octave-width frequency bands. Our results show a characteristic frequency-dependence and we find that mean-Qc estimates reveal temporal-variations of attenuation more significantly than those obtained from velocity measurements. The temporal variations are strongly connected to hydraulic stimulation activities resulting in a reduction of the coda quality factor where AE events occurred. Analysis of mean-Qc estimates after a temporal resting phase (with no activity in the rock volume) suggests that frequencies > 15 kHz indicate healing of small-scale fractures induced by injections. The study shows that coda analysis is a powerful tool for the detection of damage zones and for monitoring changes of the local fracture network within reservoirs important for exploitation or underground storage of gases and liquids.

We applied the S-coda wave analysis of Phillips (1985), which is based on the single isotropic scattering model, to estimate the frequency dependent coda quality factor Qc for each UT measurement at each sensor in the mine. The approach of Sato (1977) allows to start the analysis early in the S-wave coda as waveforms are corrected for geometrical scattering effects. The applied method comprises two parts: 1) Moving window analysis: We followed the results of the sensitivity analysis of Blanke et al. (2019) to select the analysis parameters. We use a moving window length of 1,024 samples, a lapse time of 1.1 x ts (S-onset time), a coda length of 9,000 samples (9 ms), and a minimum signal-to-noise ratio of 2. A reference noise window is selected from the end of the seismogram. Seismograms were filtered in octave-width frequency bands and the Power Spectral Density (PSD) was estimated for the pre-defined moving windows and each frequency band. 2) Regression analysis: A regression line was fitted through the coda amplitude measurements of each frequency band. Qc values were estimated from the slopes of regression lines and uncertainties (2σ standard deviation) were calculated from the slope coefficient estimates. In a final step, mean-Qc estimates per centre frequency were estimated at each sensor for each UT group (see data tables 3-10). Mean-Qc values were estimated from a minimum of 3 neighbouring UT events. Only for group UT1BH16-AFT, some mean-Qc values were estimated from less UT events due to the short borehole section beyond a previously defined damage zone that spatially separates the UT groups.

The STIMTEC hydraulic stimulation experiment (see Boese et al. 2022 for details) was conducted between 2018 and 2019 in the URL Reiche Zeche in Freiberg (Germany). The experiment aimed at investigating the role of stimulation processes in enhancing hydraulic properties of crystalline rocks. Active and passive seismic measurements were acquired in strongly foliated metamorphic gneiss during several phases of hydraulic stimulation-, testing-, and validation phases. Active measurements were conducted along two galleries (driftway and vein drift), and in several boreholes with different and mostly downward dipping orientations in the monitored rock volume (dimensions 40 m x 50 m x 30 m). The seismic network consisted of 12 Acoustic emission (AE) sensors (see data table 2), high-frequency accelerometers, and a broadband sensor installed in short and mainly upward trending boreholes above the monitored rock volume. Sensor and UT data configurations are provided by Boese et al. (2021). Hydraulic stimulations were conducted in boreholes BH10 (16-18 July, 2018) and BH17 (21-22 August, 2019) in different depth intervals and with different total injected volumes, resulting in the occurrence of AE events. This AE activity highlights activation and reactivation of fractures at the decimetre scale. The 88 analysed UT measurements (out of > 300) were acquired from boreholes BH10, BH12, and BH16. BH10 and BH16 run subparallel about 4.5 m apart and dip approx. 15° downwards. BH12 dips 36° from the driftway and crosses BH10 and BH16 below at approx. 33.9 and 18 m borehole depth, respectively.