Stress drops of hydraulic fracturing induced microseismicity in the Horn River basin: challenges at high frequencies recorded by borehole geophones

Abstract

SUMMARY The ground motions caused by seismicity associated with fluid injection can pose a significant hazard. Borehole geophone arrays can provide access to tiny seismic events, which can extend the investigated magnitude range. However, the high frequency phase arrivals (i.e., >100 Hz) also present challenges associated with high frequency cut-offs (fmax), stronger attenuation and resonances within geophones. These effects limit our ability to accurately constrain attenuation models and high frequency source parameters. We investigate 112 −0.6 $\le \, M_{\rm w}\, \le$ 0.7 seismic events and calculate corner frequencies and stress drops from 90 of these events recorded during hydraulic fracturing treatment in the Horn River basin, British Columbia. High frequency resonances (>250 Hz) caused by spurious frequency excitation and/or coupling issues can significantly distort the shape of phase arrival spectra and affect source parameter estimates. Critically, resonances vary in strength between (nearly) colocated events, which may compromise the validity of a spectral ratio approach. For stations showing the cleanest spectra, the Brune model provides a decent fit to the displacement spectra. However, bandwidth limitations, low signal-to-noise ratios, high frequency cut-offs and significant attenuation still hinder our ability to retrieve high frequency source parameters. We find that a frequency independent Qp = 180 ± 40 provides a reasonable model for crustal attenuation but the large uncertainty caused by resonances prevents a robust constraint. From those events that show the best fits, we find a mean Madariaga corner frequency of 210 Hz ± 30 from P-phase arrivals, which is in the range of expected values if self-similarity extends into negative magnitudes. We also calculate a mean stress drop of 1.6 MPa ± 1.2, which is within the tectonic range but slightly lower than other deeper regional studies, which can be explained by lower effective stresses and/or a lower crustal shear strength. We find no evidence for a change in stress drop with depth or distance from the point of injection. A plausible explanation is that effective stresses are lowered relatively quickly over the entire fault zone via direct hydraulic connections. However, the large uncertainties make it difficult to interpret source parameter variability in detail. For high resolution monitoring and source properties of microseismicity, there is an urgent need for high quality high frequency recordings unaffected by spurious frequencies.