PUBLICATIONS Geophysical Research Letters RESEARCH LETTER 10.1002/2017GL075039 Key Points: • Quantitative observation of seismic precursors to a large landslide • Time evolution of precursors suggests a nucleation governed gouge rheology evolution • Amplitude evolution of precursors highlights an accelerating creep with time Correspondence to: P. Poli, ppoli@mit.edu Citation: Poli, P. (2017), Creep and slip: Seismic precursors to the Nuugaatsiaq landslide (Greenland), Geophys. Res. Lett., 44, 8832–8836, doi:10.1002/2017GL075039. Received 21 JUL 2017 Accepted 24 AUG 2017 Accepted article online 29 AUG 2017 Published online 9 SEP 2017 Creep and slip: Seismic precursors to the Nuugaatsiaq landslide (Greenland) Piero Poli1 1 Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA Abstract Precursory signals to material’s failure are predicted by numerical models and observed in laboratory experiments or using field data. These precursory signals are a marker of slip acceleration on weak regions, such as crustal faults. Observation of these precursory signals of catastrophic natural events, such as earthquakes and landslides, is necessary for improving our knowledge about the physics of the nucleation process. Furthermore, observing such precursory signals may help to forecast these catastrophic events or reduce their hazard. I report here the observation of seismic precursors to the Nuugaatsiaq landslide in Greenland. Time evolution of the detected precursors implies that an aseismic slip event is taking place for hours before the landslide, with an exponential increase of slip velocity. Furthermore, time evolution of the precursory signals’ amplitude sheds light on the evolution of the fault physics during the nucleation process. 1. Introduction There is a growing body of experimental [e.g., Marone, 1998; Johnson et al., 2013; Scuderi, 2016] and numerical [e.g., Ferdowsi et al., 2013] evidence suggesting that material’s failure (e.g., earthquakes) is preceded by a nucleation phase, during which rearrangement of fault’s material can lead to creep, which accelerates as time is getting close to the rupture and generates precursory signals. Although observation of precursors is fundamental for earthquake forecasting [Jordan et al., 2011], foreshocks are reported mainly before large megathrust earthquakes [Bouchon et al., 2011, 2013; Ruiz et al., 2014; Kato et al., 2015]. In a similar way, geodetically observed slow slips are reported in a few cases, as before the 2014 Iquique earthquake in Chile [Ruiz et al., 2014], and the 2011 Tohoku-Oki earthquake [Hasegawa and Yoshida, 2015, and reference therein]. The lack of systematic precursors preceding small-magnitude events is likely due to our inability to detect signals from very small scale dynamic processes taking place during the nucleation [Mignan, 2014]. On 17 June 2017 a large landslide occurred near the fishermen village of Nuugaatsiaq in the northwestern region of Greenland (Figure 1). The mass in motion during this event slipped into a fiord and generated a tsunami responsible for four fatalities and widespread destruction of infrastructure in the Nuugaatsiaq village. Preliminary analysis of aerial photographs suggests a very deep-seated landslide, occurring over an area of 1000 m in length and 300 m in width [Bessette-Kirton et al., 2017]. The mass in motion estimated from differential Digital Elevation Model is between 35 and 51 million cubic meters [Bessette-Kirton et al., 2017]. Most of this material entered the fiord, generating the destructive tsunami. As reported in previous studies [Ekström et al., 2007], this landslide generated seismic waves recorded at several seismic stations around the world (Figures 2b and 2c). Preliminary analysis of these waveforms from the United States Geological Survey suggests that the slide released as much energy as a magnitude 4 earthquake. The presence of a seismic station very close to the landslide (~30 km) permitted a detailed inspection of the continuous seismic data (Figure 1). Preliminary analysis suggests that beyond the signal from the main event (Figure 2), a series of precursory signals were generated for several hours before the main sliding (Figure 2). These signals (Figure 2d) differ from the mainshock (Figure 2b) not only in moment but also in duration. In fact, the main shock waveform shows a complex signal lasting ~200 s, which probably reflects different stages of the fall [Gualtieri and Ekström, 2016]. ©2017. American Geophysical Union. All Rights Reserved. POLI Visual inspection of 24 h of data suggests that precursors share strong similarities, implying that the process generating these signals is occurring in approximately same position [Nadeau and Johnson, 1998; SEISMIC PRECURSORS TO A LANDSLIDE 8832 Geophysical Research Letters 10.1002/2017GL075039 Figure 1. Map of the study area, with the star representing the landslide, while the square shows the position of the used seismic station (NUUG). Figure 2. (a) The daylong east component seismogram filtered between 2 and 9 Hz. (b) The large amplitude signal from the landslide. (c) A zoom into the signal in Figure 2b. Before this large amplitude signal, (d) a series of precursors can be also observed. To quantify the number of precursor events, a reference waveform (Figure 2d) is correlated against the continuous data (Figure 2b). POLI SEISMIC PRECURSORS TO A LANDSLIDE 8833 Geophysical Research Letters 10.1002/2017GL075039 Figure 3. Correlation coefficient trace, the red line is the threshold to declare a detection. Poli et al., 2017]. From quantitative analysis of these data, I report time and amplitude evolution of the precursory signals, which permit to constrain the psychics governing the nucleation of landslides. 2. Data and Methods To quantify the evolution of precursory signals (Figure 2), I exploit their similarity and use a coherence-based method [Gibbons and Ringdal, 2006] aimed at counting how many events are occurring in the hours before landslide. To that end, I arbitrarily select a reference three-component waveform at station NUUG (Figure 2d) and correlate it against 24 h of seismic data. The result is a daylong correlation coefficient trace (Figure 3). When the correlation coefficient is above the threshold (8 times the median absolute deviation of the daylong correlation), a precursor is detected (Figure 3). The result of this processing provides 83 newly detected events. Given the similarities between the newly detected events, they can be stacked to improve the signalto-noise ratio [Brown et al., 2008] and to define a new reference trace. Using this new reference signal, I run the coherence-based method [Gibbons and Ringdal, 2006; Brown et al., 2008] for the second time. This second stage provides 95 detections (Figures 3 and 4). The stack of 95 precursors shows clear P and S waves, similar to regular earthquakes. The P-S time provides information regarding the distance to the station at which the process generating these waves is occurring. I measure P to S delay of 4.6 s from the data in Figure 4b. This delay suggests that the waves are generated at 32 km from the recording station, which is also the distance from the landslide. Having confirmed that the precursors are generated from the landslide area, I study their evolution in time. In Figure 4a I show the cumulative number of events as function of time. The first precursor event is observed at ~5 A.M. on 17 June 2017. After this first event, there is a clear exponential-like growth of the precursors up to Figure 4. Time evolution of precursory signals. (a) Cumulative number of events as function of time. (b) The 95 detected events ranged as function of time. The stack of these signals gives the (c) reference trace in which clear P and S waves are observed. (d) The amplitude time evolution is in clear agreement with the exponential increment of events seen in Figure 4a. POLI SEISMIC PRECURSORS TO A LANDSLIDE 8834 Geophysical Research Letters 10.1002/2017GL075039 the time of the main event, similar to numerical modeling and laboratory observations on sheared granular materials [Johnson et al., 2013; Ferdowsi et al., 2013]. Repetitive signals, like the ones detected before the Nuugaatsiaq landslide, are understood to be recursive ruptures of a brittle area, charged by an aseismic slip occurring around this zone [Nadeau and Johnson, 1998; Poli et al., 2017]. Relaxing this hypothesis, evolution of seismicity before the landslide implies an exponential acceleration on the landslide rupture plane. I suggest that the landslide mass is aseismically sliding for several hours before the main event, similar to observations before some large earthquakes [Bouchon et al., 2011, 2013; Ruiz et al., 2014; Kato et al., 2015] in laboratory experiments [Johnson et al., 2013; Ferdowsi et al., 2013], and theoretical models [Das and Scholz, 1981; Kanamori, 1981; Ohnaka, 1992]. Due to the nearly collocated source for the precursors, any propagation effect (e.g., attenuation and scattering) does not confound analysis [Abercrombie, 2015] and amplitude evolution for the detected waveforms can be studied, at least in a relative sense [Abercrombie, 2015]. Amplitude evolution in Figure 4d follows the same exponential-like growth observed in the temporally accumulated event count, except for the period between 20 and 22 h, where the amplitude drops and rises, arriving at the time of the landslide. The increase in amplitude is proportional to the increase in the seismic moment [Aki and Richards, 2002]. As precursory signals are from nearly collocated sources, I suggest that the moment increment is due to a growing asperity. This behavior is similar to what is observed in numerical studies and field observations, which show that under faster aseismic slip, repetitive stick slip events tend to rupture larger areas, and thus release a larger seismic moment [Chen et al., 2010]. Following this reasoning, it is likely that the evolution of slip velocity and amplitude reported here was due to a growing rupture, which gave rise to the resulting significant landslide. Thus, the main rupture plane likely underwent an evolution of its properties (e.g., friction, effective normal stress, and elastic stiffness), which controlled the transition from stable (aseismic) to unstable (landslide) [Leeman et al., 2016; Scuderi, 2016]. 3. Conclusions The behavior of the signals identified as precursors to the Nuugaatsiaq landslide agrees with the nucleation model, where foreshocks are caused by premonitory slow slips within the nucleation zone of the mainshock [Das and Scholz, 1981; Kanamori, 1981; Ohnaka, 1992], suggesting that earthquakes and landslides underlay the same nucleation phase physics. The acceleration of seismicity highlights a transition from quasi-static to quasi-dynamic (at later time) nucleation process preceding the dynamic event [Ohnaka, 1992]. The details of the observed seismicity permit to constrain the small-scale physics controlling this evolution. In fact, the observed exponential-like growth of precursors with time is in agreement with the results from numerical and laboratory experiments [Johnson et al., 2013; Ferdowsi et al., 2013], which predicts that creep preceding dynamic rupture is related to particle rearrangement in sheared granular material. Such grain rearrangement also agrees with laboratory experiments, exhibiting an evolution of material properties prior to the dynamic rupture responsible for the reduction of P wave velocity associated with the beginning of asperity failures prior to a macroscopic frictional failure [Scuderi, 2016]. The similarity between the results for the Nuugaatsiaq landslide and experiments in sheared granular media suggests that nucleation occurs in a mature layer with granular rheology, with an evolution of slip behavior influenced by the fabric development of gouge material [Scuderi et al., 2017]. This behavior is similar to the results from numerical simulation of avalanches in granular media, predicting a transition to metastability, with grain reorganization and the emergence of friction instabilities [e.g., Staron et al., 2006; Zaitsev et al., 2008; Michlmayr et al., 2012]. This study further indicates that near-field high-resolution data are needed to monitor the evolution of seismicity and the degree of criticality of a given natural system, and thus improve our preparedness for large hazards by forecasting incoming disasters. Acknowledgments The data used in this study are freely available on the Incorporated Research Institute for Seismology and were downloaded using ObspyDMT (Kasra Hosseini. [2017, June 1]. obspyDMT: v2.0.2. Zenodo. http://doi.org/10.5281/ zenodo.801778). POLI References Abercrombie, R. E. (2015), Investigating uncertainties in empirical Green’s function analysis of earthquake source parameters, J. Geophys. Res. Solid Earth, 120, 4263–4277, doi:10.1002/2015JB011984. Aki, K., and P. G. Richards (2002), Quantitative Seismology, vol. 1, Univ. Sci. Books. Bessette-Kirton, E., K. Allstadt, J. Pursley, and J. Godt (2017), Preliminary analysis of satellite imagery and seismic observations of the Nuugaatsiaq landslide and tsunami, Greenland. [Available at https://landslides.usgs.gov/research/featured/2017-nuugaatsiaq/.] SEISMIC PRECURSORS TO A LANDSLIDE 8835 Geophysical Research Letters 10.1002/2017GL075039 Bouchon, M., H. Karabulut, M. Aktar, S. Özalaybey, J. Schmittbuhl, and M.-P. Bouin (2011), Extended nucleation of the 1999 Mw 7.6 Izmit earthquake, Science, 331(6019), 877–880. Bouchon, M., V. Durand, D. Marsan, H. Karabulut, and J. Schmittbuhl (2013), The long precursory phase of most large interplate earthquakes, Nat. Geosci., 6(4), 299–302. Brown, J. R., G. C. Beroza, and D. R. Shelly (2008), An autocorrelation method to detect low frequency earthquakes within tremor, Geophys. Res. Lett., 35, L16305, doi:10.1029/2008GL034560. Chen, K. H., R. Bürgmann, R. M. Nadeau, T. Chen, and N. Lapusta (2010), Postseismic variations in seismic moment and recurrence interval of repeating earthquakes, Earth Planet. Sci. Lett., 299(1), 118–125. Das, S., and C. H. Scholz (1981), Theory of time-dependent rupture in the Earth, J. Geophys. Res., 86, 6039–6051, doi:10.1029/ JB086iB07p06039. Ekström, G., R. A. Hansen, G. L. Pavlis, and P. Lipovsky (2007), Seismological detection and analysis of recent landslides in Alaska and the Yukon, Abstract S52B-05 presented at 2007 Fall Meeting, AGU, San Francisco, Calif. [Available at https://earthquake.usgs.gov/earthquakes/eventpage/us20009nlg#dyfi.] Ferdowsi, B., M. Griffa, R. A. Guyer, P. A. Johnson, C. Marone, and J. Carmeliet (2013), Microslips as precursors of large slip events in the stick-slip dynamics of sheared granular layers: A discrete element model analysis, Geophys. Res. Lett., 40, 4194–4198, doi:10.1002/ grl.50813. Gibbons, S. J., and F. Ringdal (2006), The detection of low magnitude seismic events using array-based waveform correlation, Geophys. J. Int., 165(1), 149–166. Gualtieri, L., and G. Ekström (2016), Seismic reconstruction of the 2012 Palisades Rockfall using the analytical solution to Lamb’s problem, Bull. Seismol. Soc. Am., doi:10.1785/012016023. Hasegawa, A., and K. Yoshida (2015), Preceding seismic activity and slow slip events in the source area of the 2011 Mw 9.0 Tohoku-Oki earthquake: A review, Geosci. Lett., 2(1), 6. Johnson, P. A., B. Ferdowsi, B. M. Kaproth, M. Scuderi, M. Griffa, J. Carmeliet, R. A. Guyer, P.-Y. Le Bas, D. T. Trugman, and C. Marone (2013), Acoustic emission and microslip precursors to stick-slip failure in sheared granular material, Geophys. Res. Lett., 40, 5627–5631, doi:10.1002/2013GL057848. Jordan, T., Y.-T. Chen, P. Gasparini, R. Madariaga, I. Main, W. Marzocchi, G. Papadopoulos, G. Sobolev, K. Yamaoka, and J. Zschau (2011), Operational earthquake forecasting: State of knowledge and guidelines for implementation, Ann. Geophys., 54(4), 315–391. Kanamori, H. (1981), The nature of seismicity patterns before large earthquakes, in Earthquake Prediction: An International Review, edited by W. Simpson and G. Richards, pp. 1–19, Wiley. Kato, A., T. Terakawa, Y. Yamanaka, Y. Maeda, S. Horikawa, K. Matsuhiro, and T. Okuda (2015), Preparatory and precursory processes leading up to the 2014 phreatic eruption of Mount Ontake, Japan, Earth Planets Space, 67(1), 1–11. Leeman, J. R., D. M. Saffer, M. M. Scuderi, and C. Marone (2016), Laboratory observations of slow earthquakes and the spectrum of tectonic fault slip modes, Nat. Commun., 7, doi:10.1038/ncomms11104. Marone, C. (1998), Laboratory-derived friction laws and their application to seismic faulting, Annu. Rev. Earth Planet. Sci., 26(1), 643–696. Michlmayr, G., D. Cohen, and D. Or (2012), Sources and characteristics of acoustic emissions from mechanically stressed geologic granular media—A review, Earth Sci. Rev., 112(3), 97–114. Mignan, A. (2014), The debate on the prognostic value of earthquake foreshocks: A meta-analysis, Sci Rep, 4, 4099, doi:10.1038/srep04099. Nadeau, R. M., and L. R. Johnson (1998), Seismological studies at Parkfield VI: Moment release rates and estimates of source parameters for small repeating earthquakes, Bull. Seismol. Soc. Am., 88(3), 790–814. Ohnaka, M. (1992), Earthquake source nucleation: A physical model for short-term precursors, Tectonophysics, 211(1–4), 149–178. Poli, P., A. M. Jeria, and S. Ruiz (2017), The Mw 8.3 Illapel earthquake (Chile): Preseismic and postseismic activity associated with hydrated slab structures, Geology, G38522-1. Ruiz, S., M. Metois, A. Fuenzalida, J. Ruiz, F. Leyton, R. Grandin, C. Vigny, R. Madariaga, and J. Campos (2014), Intense foreshocks and a slow slip event preceded the 2014 Iquique Mw 8.1 earthquake, Science, 345(6201), 1165–1169. Scuderi, M. M., C. Marone, E. Tinti, G. Di Stefano, and C. Collettini (2016), Precursory changes in seismic velocity for the spectrum of earthquake failure modes, Nat. Geosci., 9(9), 695–700. Scuderi, M. M., C. Collettini, C. Viti, E. Tinti, and C. Marone (2017), Evolution of shear fabric in granular fault gouge from stable sliding to stick slip and implications for fault slip mode, Geology, G39033–1. Staron, L., F. Radjai, and J.-P. Vilotte (2006), Granular micro-structure and avalanche precursors, J. Stat. Mech.: Theory Exp., 2006(07), P07014. Zaitsev, V. Y., P. Richard, R. Delannay, V. Tournat, and V. E. Gusev (2008), Pre-avalanche structural rearrangements in the bulk of granular medium: Experimental evidence, EPL Europhys. Lett., 6, 83, 64003. POLI SEISMIC PRECURSORS TO A LANDSLIDE 8836