Characterizing and Locating Very Weak (−2.2 ≥ M L ≥ −3.4) Induced Seismicity in Unstable Sandstone Cliffs by Nanoseismic Monitoring

Unloaded natural rock masses are known to generate seismic signals (G reen et al. , 2006 ; H ainzl et al. , 2006 ; H usen et al. , 2007 ; K raft et al. , 2006 ). Following a 1,000 m 3 mass failure into the Mediterranean Sea, centimeter-wide tensile cracks were observed to have developed on top of an unstable segment of the coastal cliff. Nanoseismic monitoring techniques (W ust -B loch and J oswig , 2006 ; J oswig , 2008 ), which function as a seismic microscope for extremely weak seismic events, were applied to verify whether brittle failure is still generated within this unconsolidated sandstone mass and to determine whether it can be detected. Sixteen days after the initial mass failure, three small-aperture sparse arrays (Seismic Navigation Systems-SNS) were deployed on top of this 40-m high shoreline cliff. This paper analyzes dozens of spiky nanoseismic (−2.2 ≥ M L ≥ −3.4) signals recorded over one night in continuous mode (at 200 Hz) at very short slant distances (3–67 m). Waveform characterization by sonogram analysis (J oswig , 2008 ) shows that these spiky signals are all short in duration (>0.5 s). Most of their signal energy is concentrated in the 10–75 Hz frequency range and the waveforms display high signal similarity. The detection threshold of the data set reaches M L −3.4 at 15 m and M L −2.7 at 67 m. The spatial distribution of source signals shows 3-D clustering within 10 m from the cliff edge. The time distribution of M L magnitude does not display any decay pattern of M L over time. This corroborates an unusual event decay over time (modified Omori’s law), whereby an initial quiet period is followed by regained activity, which then fades again. The polarization of maximal waveform amplitude was used to estimate spatial stress distribution. The orientation of ellipses displaying maximal signal energy is consistent with that of tensile cracks observed in the field and agrees with rock mechanics predictions. The M L – surface rupture length relationship displayed by our data fits a constant-slope extrapolation of empirical data collected by W ells and C oppersmith ( 1994 ) for normal fault features at much larger scale. Signal characterization and location as well as the absence of direct anthropogenic noise sources near the monitoring site, all indicate that these nanoseismic signals are generated by brittle failure within the top section of the cliff. The atypical event decay over time that was observed suggests that the cliff material is undergoing post-collapse bulk strain accommodation. This feasibility study demonstrates the potential of nanoseismic monitoring in rapidly detecting, locating and analyzing brittle failure generated within unconsolidated material before total collapse occurs.