ESurfEarth Surface DynamicsESurfEarth Surf. Dynam.2196-632XCopernicus PublicationsGöttingen, Germany10.5194/esurf-5-283-2017Single-block rockfall dynamics inferred from seismic signal analysisHibertClémenthibert@unistra.frhttps://orcid.org/0000-0003-3457-6617MaletJean-Philippehttps://orcid.org/0000-0003-0426-4911BourrierFranckProvostFlorianeBergerFrédéricBornemannPierrickTardifPascalMerminEricInstitut de Physique du Globe de Strasbourg, CNRS UMR 7516, University of Strasbourg/EOST, 5 rue Descartes, 67084 Strasbourg, FranceInstitut national de recherche en sciences et technologies pour l'environnement et l'agriculture (IRSTEA), 2 Rue de la Papeterie,
38402 Saint-Martin-d'Hères, FranceClément Hibert (hibert@unistra.fr)23May20175228329221December201610January201716April201718April2017This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://esurf.copernicus.org/articles/5/283/2017/esurf-5-283-2017.htmlThe full text article is available as a PDF file from https://esurf.copernicus.org/articles/5/283/2017/esurf-5-283-2017.pdf
Seismic monitoring of mass movements can significantly help to
mitigate the associated hazards; however, the link between event dynamics and
the seismic signals generated is not completely understood. To better
understand these relationships, we conducted controlled releases of single
blocks within a soft-rock (black marls) gully of the Rioux-Bourdoux torrent
(French Alps). A total of 28 blocks, with masses ranging from 76 to 472 kg, were used
for the experiment. An instrumentation combining video cameras and
seismometers was deployed along the travelled path. The video cameras allow
reconstructing the trajectories of the blocks and estimating their velocities
at the time of the different impacts with the slope. These data are compared
to the recorded seismic signals. As the distance between the falling block
and the seismic sensors at the time of each impact is known, we were able to
determine the associated seismic signal amplitude corrected for propagation
and attenuation effects. We compared the velocity, the potential energy lost,
the kinetic energy and the momentum of the block at each impact to the true
amplitude and the radiated seismic energy. Our results suggest that the
amplitude of the seismic signal is correlated to the momentum of the block at
the impact. We also found relationships between the potential energy lost,
the kinetic energy and the seismic energy radiated by the impacts. Thanks to
these relationships, we were able to retrieve the mass and the velocity
before impact of each block directly from the seismic signal. Despite high
uncertainties, the values found are close to the true values of the masses
and the velocities of the blocks. These relationships allow for gaining a better
understanding of the physical processes that control the source of high-frequency
seismic signals generated by rockfalls.
Introduction
Understanding the dynamics of rockfalls and other mass movements is critical
to mitigate the associated hazards but is very difficult because of the
limited number of observations of natural events. With the densification of
the global, regional and local seismometer networks, seismic detection of
gravitational movements is now possible. The continuous recording ability of
seismic networks allows a reconstruction of the gravitational activity on an
unprecedented timescale and the monitoring of unstable slopes
e.g..
More than the detection of these events, recent advances allow determining
the dynamics of the largest landslides on Earth from the very low-frequency
seismic waves they generate. Inversion and modelling of the long-period
seismic waves permits to infer the force imparted by these catastrophic
events on Earth and to deduce dynamic parameters (acceleration, velocity,
trajectory) as well as their mass
.
However, these approaches are limited by the size of the events. Only the
largest landslides will generate the long-period seismic waves used in the
inversion and the modelling methods. Moreover, these events constitute only a
small proportion of the landslides that occur worldwide.
In recent years, a new approach based on the analysis of the high-frequency
seismic signal has been proposed. High-frequency seismic waves are generated
independently of the size of the event, and can be recorded if seismometers
are close enough to the source. Hence, this allows a seismic detection of the
events that do not generate long-period seismic waves
e.g..
The limitation of this approach is that high-frequency seismic waves are more
prone to be influenced by propagation effects (attenuation, dispersion,
scattering) and, more importantly, that the source of the high-frequency
seismic waves associated with gravitational instabilities is not yet well
understood.
Studies have shown that several landslide properties can be linked to
features of the high-frequency seismic signals. In some cases, it has been
observed that the landslide volume is correlated to the amplitude
or to the radiated seismic energy of the
high-frequency signals . Other studies have
shown that the high-frequency seismic signals can also carry information on
landslide dynamics. determined with numerical
modelling that a good correlation exists between the short-period
seismic-signal envelope, the modelled friction work rate and the momentum
(product of the mass and the velocity) for two rock-ice avalanches. The
model-based approach proposed by predicts that a correlation
can be found between the modelled force and the power of the short-period
seismic signal for rockfalls that occurred at the Soufrière Hills volcano
on the island of Montserrat. have demonstrated that, for 11 large
landslides that occurred worldwide, the bulk momentum controls, in first
order, the amplitude of the envelope of the generated seismic signals filtered
between 3 and 10 Hz. These authors also demonstrated that the maximum
amplitude of the seismic signal, corrected for propagation effects, is
quantitatively correlated with the bulk momentum. These results are important
as they open the perspective to quantify landslide dynamics, independently of
their size, and directly from the seismic signals they generate (i.e. without
inversion or modelling). Being capable of quantifying landslide properties
directly from the seismic signals they generate is critical for the
development of future methods aimed at their real-time detection and
characterization using high-frequency seismic signals. However, before
considering an operational implementation of such methods, we need to better
understand the source of the generated high-frequency radiation and its link
with landslide dynamics
One of the assumptions that emerge from these studies to explain the link
between the landslide dynamics and the high-frequency seismic signal features
is that this relationship can potentially originate from small-scale
processes within the landslide mass, and between the landslide mass and the
substrate. The dynamic properties of a bouncing particle within a granular
flow might control the impulse imparted to the solid Earth at each impact,
and the amplitude of the seismic wave generated might be proportional to the
magnitude of the impulse. However, this assumption raises an important issue:
what is the link between the dynamics of a single bouncing particle (a rock
for example) and the seismic signal generated?
Theoretical developments as well as laboratory and field experiments were conducted by
to address this issue. These authors have shown that the
mass and the speed of an impactor can be related to the radiated elastic
energy and to the spectrum of the signal, following analytic developments
based on the Hertz theory of impact . However, the field
experiment conducted showed that, in this case, these simple relationships
did not perform well to quantify the velocity and the mass of single rocks
from the seismic signal it generates. Difficulties to synchronize the seismic
signals with direct observations and the use of a seismometer that was not
capable to record the high-frequency energy of the generated seismic waves
might explain why the analytic relationships were not confirmed by this
experiment.
View from (a) the first and (b) the second video cameras deployed at
the bottom of the slope. The ground control points are indicated by blue
points. (c) Trajectory reconstruction for block 4 on the DEM, built
from lidar acquisition, superimposed on an orthophoto of the Rioux-Bourdoux slopes. Each
point indicates the position of an impact and the colour gradient represents
the chronology of these impacts (blue for the first impact and red for the
last one). K2 is a three-component short-period seismometer and K1, K3 and K3
are vertical-only seismometers. CMG1 is a broad-band seismometer.
In this study we propose a new field experiment of controlled releases of
single blocks to investigate the relationships between block properties and
dynamics, and the features of the seismic signals generated by impacts with
the slope. We deployed several short-period and broadband seismic stations to
record the high-frequency seismic signal generated at each impact. The
trajectory of each block is reconstructed with video cameras that were
synchronized with the seismometers. The seismic signal processing allowed us
to infer the amplitude of the seismic signal at the source, corrected for
propagation effects, and the seismic energy radiated by the impacts. We then
compare the features of the seismic signal of each impact to the dynamics and
the properties of the released block.
The Rioux-Bourdoux experiment
The focus of the Rioux-Bourdoux controlled-releases experiment was to study the
seismic signal of single-block rockfalls on unconsolidated soft rock, which
is highly attenuating for seismic waves. The Rioux-Bourdoux is a torrent
located in the French Alps, approximately 4 km north of the town of
Barcelonnette (France). The slopes surrounding the torrent consist of
Callovo-Oxfordian black marls and are representative of the slope morphology
of marly facies observed in south-east France. Due to the high erosion
susceptibility of black marls numerous, steep gullies have formed on these
slopes.
We conducted the releases within one of these gullies (Fig. a and b). The advantage of launching the blocks in a
gully is that for every block the travelled path is roughly the same.
Moreover, the steepness of the gullies that developed in black marls allows
the block to rapidly reach a high velocity. The travel path had a length of
approximately 200 m and slope angles ranging from ∼45∘ on the
upper part of the slope to ∼20∘ on the terminal debris cone. A total of 28 blocks with masses ranging from 76 to 472 kg were manually launched.
Two video cameras (Sony α7 – 25 frames per second) were deployed at the
base of the gully, close to the torrent. Ground-control points were marked
for visual recognition on the videos and their 3-D coordinates were measured
using the Global Navigation Satellite System (GNSS). A reference digital elevation
model (DEM) at a spatial resolution of 0.5 m was built from terrestrial
lidar
acquisitions (Fig. c). The time of the cameras was
set to be synchronous with the seismic sensors' time (GPS). The seismic
network was composed of one broadband seismometer (CMG40T – sampling frequency
100 Hz) located north of the gully, and an antenna of four short-period
seismometers (one with three vertical components and three with one vertical component – sampling
frequency 1000 Hz) located south of the gully (Fig. c).
MethodsTrajectory reconstruction and dynamic parameters estimation
To reconstruct the trajectory, the impacts of each block were manually picked
on the frames of the videos. Thanks to the control points, the frames of the
videos were projected on the DEM. Hence, once an impact was identified on the
frame, the position of the pixel was reported on the DEM, which gave the true
position of the impact. This processing was repeated for the two cameras,
which gave an estimate of the uncertainties on the determination of the
position and the time of the impact. The velocity just before impact was
derived from the block trajectory and the duration of block flight before
impact. The kinetic energy was computed as
Ek=12mV2,
with m the mass of the block and V the velocity before impact. We also
determined the potential energy lost during the block flight before impact
from the difference of altitude of the block between two impacts, inferred
from the reconstructed trajectory, as
Ep=mg(ht1-ht2),
with g the gravitational acceleration and ht1 and ht2 the
altitudes of the block at the impacts that occurred at the two successive
times t1 and t2. Unfortunately, the resolution of the cameras and the
complex dynamics of the blocks during the first seconds of propagation did
not allow us to identify clearly the impacts on the upper part of the slope.
However, the trajectories of the blocks on the lower part of the slope were
well constrained, with an average uncertainty in the inferred velocity of the
blocks before impacts of 0.95 ms-1, for velocities with values between 6 and 17 ms-1.
Seismic signal processing
Several authors have shown that the seismic waves generated by gravitational
instabilities are dominated by surface waves
e.g.. These
high-frequency seismic surface waves are subjected to strong propagation
effects, especially in a highly attenuating medium such as black marls.
Figure shows the seismic signals recorded for the launch
of the block number 4. The attenuation is visible when comparing peaks in the
seismic signal recorded at the station located on the upper part of the slope
(Fig. a) to the ones recorded at the station on the
lower part of the slope (Fig. c), for the same time.
The amplitude of the peaks is clearly dependent on the distance between the
impact and the seismic station. Moreover, Fig. b shows
the attenuation of the highest frequency with the distance of the source to
the seismic station. To compare seismic signal features to the dynamic
parameter of the rockfall, we have to correct these attenuation effects.
proposed a simple attenuation law giving the amplitude A(r)
of a seismic surface wave recorded at a distance r as
A(r)=1rA0×e-Br.
If the distance between the station and the source is known, the computation
of the amplitude at the source A0 is straightforward. However, we have to
determine the frequency dependent parameter B that accounts for the
anelastic attenuation of seismic waves. If we consider ri the distance
between the source and station i and rj the distance to station j, the
apparent anelastic attenuation parameter Bij is then
Bij=log(A(ri)ri)-log(A(rj)rj)rj-ri.
By combining Eqs. () and (), we can compute the amplitude at the source
A0 for each pair of stations. This value is then averaged over all the
pairs of stations, and the standard deviation gives an estimate of the
uncertainty.
(a) Signal recorded at the short-period station located on the upper
part of the slope and (b) corresponding spectrogram, generated by block number
4 (mass of 209 kg). (c) Signal recorded at the broadband station located on
the lower part of the slope and (d) corresponding spectrogram, generated by
block 4.
Another quantity that we want to compare to the dynamics of the block is the
radiated seismic energy. The energy of a seismic surface wave can be computed
as
Es=∫titf2πrDhcuenv(t)2eBrdt,
with
uenv(t)=u(t)2+Ht(u(t))2,
where Ht is the Hilbert transform of the seismic signal u(t) used to
compute the envelope uenv(t), ti and tf the times of the
beginning and the end of the seismic signal respectively, h the thickness
and D the density of the layer through which the generated surface waves
propagate, and c is their phase velocity. The average velocity of surface
waves in black-marl formations observed in the area of the Rioux-Bourdoux
torrent is approximately 300 ms-1,
which, for seismic signal with central frequencies around f=20 Hz as
observed in Fig. , gives a propagation depth h,
computed as h=c/f, of ∼15 m. The density D of dry black marls is
approximately 1450 kgm-3.
Before computing the amplitude at the source and the energy of the seismic
signals generated by impacts, we first selected the seismic signals with the
following criteria. We excluded the seismic signals generated when (i) sliding
of the blocks occurred, (ii) the blocks stopped mid-slope and (iii) more
generally when the signal-to-noise ratio was too weak on the seismic stations
to perform the computation of the apparent anelastic attenuation parameter
Bij. Bij is dependent on the frequency of the seismic waves.
Therefore the seismic signals were band-pass filtered between 1 and 50 Hz.
This frequency band is chosen because most of the seismic wave energy is not
attenuated in this band within the span of the seismic network (Fig. b and d). For each seismic record selected, we manually
picked the peaks corresponding to the impacts on each station. This
processing results in a data set of 37 impact seismic signals, coming from 9
out of the 28 launches.
ResultsCorrelation between dynamic parameters and seismic signal features
From the reconstructed trajectories we inferred the velocity, the momentum
and the kinetic energy of the block before each impact (Eq. ), and the potential energy lost during the block
trajectory before impact (Eq. ). The velocities exhibit a
low variability, with values ranging from 6 to 17 ms-1
(Fig. ). We did not find significant correlation
between the mass and the impact velocity.
Histogram of the observed absolute velocities before impact.
The seismic signal processing yielded the maximum amplitude at the source
A0max and the radiated seismic energy Es at each impact. The average
uncertainty in the computation of the maximum amplitude A0max, inferred
from the standard deviation, and expressed as a percentage of the computed
values (i.e. A0max±x%A0max), ranges from 7 to 129 %, and is
58 % on average. Regarding the computation of the radiated seismic energy
Es, the uncertainty, estimated following the same approach, ranges from
55 to 152 % of the computed values, and is 86 % in average.
Spearman correlation coefficients, coefficients of the regression lines for proportional and linear relationships and
corresponding coefficient of determination R2.
Spearman correlation Proportional Linear Parameters (X,Y)ρp valuesαβR2αβR2A0max=α|p|+β0.671.1210-72.3510-900.632.2610-92.5010-70.64Es=αEp+β0.686.7510-64.4010-600.615.0410-6-0.010.61Es=αEk+β0.703.0110-62.5910-600.593.0910-6-0.010.64Es=αm+β0.511.310-31.4810-400.232.8510-4-0.030.31Es=αmVz13/5+β0.694.1610-64.8610-700.625.8510-7-0.010.63Es=αmVz0.5+β0.627.6310-55.2410-500.331.0710-4-0.040.47A0max=αEs+β0.448.210-3––––––
(a) Maximum of the amplitude A0max, corrected from attenuation,
as a function of the average momentum |p| of the block before the impact.
Radiated seismic energy Es of the seismic signal generated at the impact
as a function of (b) the kinetic energy before the impact Ek, (c) the
masses m of the blocks, (d) the potential energy lost Ep, (e) the
parameter mVz0.5, and (f) the parameter mVz13/5. Errors bars resulting
from the computation of the momentum, the kinetic energy and the amplitude at
the source are indicated by black lines. For each pair of parameters the
light-grey line corresponds to the best regression line computed for a linear
relationship and the dark-grey one to the best regression line computed for a
proportional relationship.
We investigated the possible correlations between (1) the maximum amplitude
at the source A0max of the seismic signal and the absolute momentum
|p| before the impact, (2) the radiated seismic energy Es and the
potential energy lost Ep, (3) the radiated seismic energy Es and the
kinetic energy Ek before impact, and (4) the radiated seismic energy Es
and the mass m of the blocks. The analysis based on Hertz's theory of
impact conducted by yielded the parameter mVz13/5,
with m the mass of the block and Vz the vertical velocity before impact,
which should in theory scale with the radiated seismic energy Es of the
seismic signal generated at each impact. However, when investigating this
relationship for real single-block rockfalls, they did not found a
significant correlation with this parameter. The best correlation they found
was with the parameter mVz0.5. We also investigated these two cases
with our data set. We computed for each pair of parameters the Spearman rank
correlation coefficient ρ and the corresponding p values (Table ) as we assume that the
parameters should scale following monotonic laws.
The best correlation coefficient ρ has a value of 0.70 for the pair of
parameters Es and Ek. Slightly lower correlation coefficient values are
observed between the maximum amplitude A0max and the absolute momentum
|p| (ρ=0.67) and the radiated seismic energy Es and the potential
energy Ep (ρ=0.68). The correlation coefficient between the radiated
seismic energy Es and the best empiric parameter mVz0.5 found by
is poorer (ρ=0.62) than the one observed between the
radiated seismic energy and the parameter mVz13/5 they derived from
the Hertz theory of impact (ρ=0.69). Finally, our results show that
there is no correlation between the maximum amplitude A0max and the
radiated seismic energy Es (ρ=0.44) and between the radiated seismic
energy Es and the mass of the blocks m (ρ=0.51). We also
investigated other correlations between dynamic parameters and seismic signal
features, with the vertical momentum or the vertical kinetic energy for
example, but we were unable to improve on the correlations found with the
modulus of the dynamic quantities.
To characterize the relationships between the parameters that are correlated,
we computed the regression lines that best fit the data (Fig. and Table ). According to
the theoretical analysis conducted by , the dynamic
parameters should scale proportionally with the seismic features. However,
several studies have shown that linear relationships allow a better fitting
of the data gathered from the observation of natural events
e.g.. We computed the
regression coefficients of the best fitting lines for the two types of
relationships and assessed the quality of the fitting by computing the
coefficients of determination R2.
Overall the R2 coefficient values do not exceed 0.64 (Table ). This is caused by a high scattering of the data
which comes from the high uncertainties on the computation of the seismic
attenuation parameters and hence on the values of A0max and Es, as
shown by the large error bars in Fig. . The best
R2 coefficients are yielded by the linear regression between the maximum
amplitude A0max and the momentum |p|, and the radiated seismic energy
Es and the kinetic energy Ek (R2=0.64 for both cases). For the
parameter couples Es/Ep and Es/mVz13/5, R2 coefficients
are slightly lower, with values of 0.61 and 0.63 respectively. The regression
of each pair of parameters by proportional relationships gives lower values
for the coefficient R2. However, the β coefficients of the best
linear regressions are close to 0. We assume that linear regressions allow to
better accommodate for the scattering of the data than proportional
regressions, and that β coefficients are not physically significant.
Retrieving block properties and dynamics from the seismic signal
We have shown that correlations exist between several dynamic quantities and
features of the seismic signal generated at each impact. In this section we
investigate whether these relationships can provide accurate estimates of the mass
and the velocity of the blocks, directly from the features of the seismic
signals generated by the impacts.
Our results show that the maximum amplitude and the seismic energy are not
correlated (Table ). Hence, we can combine the linear
relationships inferred for the maximum amplitude and the momentum, and for
the radiated seismic energy and the kinetic energy, with the coefficients
α and β yielded by the linear regressions. We can express the
mass mi as a function of A0max and Es as
mi=5.9×1011(A0max-2.50×10-7)2(Es+0.01).
Using Eq. (), we computed mi for each impact
of each block for which we were able to compute A0max and Es, and
compared the average estimates of mi to the measured mass mr of each
block (Table ). Overall, the inferred masses
mi are close to the real masses mr of the block. However, the
uncertainty in the inferred values is high, especially for blocks for which
we have a few number of exploitable impacts and therefore few estimates of
A0max and Es. This may also come from the uncertainties related to
the computation of the seismic quantities.
Comparison between the real mass mr of the blocks and the average inferred masses mi
computed with Eq. ().
We can also estimate the velocity of the block before each impact using the
linear regression and the corresponding coefficients found between the
maximum amplitude A0max and the maximum momentum p, or between the
seismic energy Es and the kinetic energy Ek, and with the masses
inferred with Eq. (). We choose to use the linear
relationship between the amplitude and the momentum because the uncertainties
associated with determining the amplitude at the source are lower than those
associated with the radiated seismic energy. The inferred velocity Vi can
be computed as
Vi=A0max-2.50×10-72.26×10-9mi.
Figure a shows the distribution of the absolute
difference between the velocities inferred Vi and the velocities Vr
derived from the trajectory reconstruction. The values of the difference are between 0.1 and 13.7 ms-1, with a median value of 2.4 ms-1. We also computed the ratio of the velocity absolute
|Vi-Vr| difference over the velocity derived from the trajectory
reconstruction Vr (Fig. b). The majority of
the values of the ratio fall below 0.5 (i.e. the difference is less than
50 % of the value of the velocity derived from the trajectory
reconstruction), and the median ratio is 0.2 (i.e. 20 % of the value of the
velocity derived from the trajectory reconstruction).
(a) Histogram showing the distribution of the difference between the
velocity before impact Vi inferred using
Eqs. ()
and () and the velocity Vr estimated
from the video cameras. (b) Same as (a) but normalized by the value of the
velocity Vr estimated via the video cameras.
Discussion and conclusion
The Rioux-Bourdoux experiment of controlled single-block rockfalls produced
important results to better understand the links between the dynamics of
rockfalls and the seismic signal associated. Our results suggest that
correlations exist between the seismic signal features and the energy, the
velocity and the mass of single-block rockfalls. We observed that the maximum
amplitude of the seismic signal generated at each impact and the momentum
(product of the mass and the velocity) of the blocks are correlated. Our
results also suggest that the energy of the seismic radiation released at
each impact scales linearly with the potential energy lost and the kinetic
energy.
By combining the scaling laws found, we were able to infer realistic values
of the masses and the velocities before impact of the blocks from the
amplitude and the energy of the seismic signals generated at each impact. The
difference between the mass of the blocks determined from the seismic
quantities and the real values is 27 % in average. Our results also
demonstrate that when the number of impact seismic signals used to determine
the mass of the blocks increases, the error made on the inferred values
decreases. For the velocities, the average difference between the inferred
and the real values of the velocity is 20 %. These errors might come from the
uncertainties on the computation of the seismic quantities. We determined,
from the computation of the seismic quantities on multiple pairs of stations,
that the average uncertainties are 58 and 86 % on the computed values of
the amplitude at the source and of the radiated seismic energy respectively.
We suppose that these uncertainties are mainly caused by the simple seismic
attenuation model used.
We found that the relationship derived from the Hertz's theory of impact
proposed by that links the radiated seismic energy of the
signal generated to the parameter mVz13/5 is verified with our data.
However, the scaling between the seismic energy and the parameter
mVz13/5 did not yield significantly better quantitative correlation
than the one observed between the radiated seismic energy and the kinetic
energy, or between the amplitude at the source and the momentum of the block
before impact (ρ=0.69, 0.70 and 0.67 respectively). This confirms the
combined role of the mass and the velocity before impacts of the block in the
generation of seismic waves, but does not allow us to identify a unique
dynamic parameter that would control the seismic signal features. Further
analytical and theoretical developments are needed to understand the physical
processes that explain these correlations, and ultimately what the
physical parameters are that control the characteristics of the seismic signal
generated.
An issue that arose from studies on the link between the seismic signals and
the dynamics of mass movements is about the energy transfer and more
specifically the ratio Rs/p between the radiated seismic energy and the
potential energy lost. found for 10 rockfalls that
occurred in the French Alps that this Rs/p ratio is between
10-5 and 10-4. have found a Rs/p ratio
of 10-3 for an artificially triggered rockfalls in the Montserrat massif
(Spain). In volcanic contexts, and have
observed Rs/p ratios ranging from 10-5 to 10-3. In this study,
we found a Rs/p ratio between the radiated seismic energy and the
potential energy lost of approximately 10-6 (Table ). Interestingly, a ratio of the same order is
observed between the radiated seismic energy and the kinetic energy. The
value of the Rs/p ratio is lower than those observed in other contexts.
We assume that this might be explained by the nature of the substrate as in
our case the rockfalls propagated on soft rocks, which may absorb more
potential energy (by deformation for example) than igneous
or metamorphic hard rocks
. Investigating this assumption on the role
of the substrate on energy transfer by replicating the experiment of
controlled releases of single blocks in other contexts constitutes one of the
perspectives of this work.
We identified several limitations that have to be addressed before
considering an operational application of seismology to quantify rockfall
properties. First, our results show that better attenuation models are needed
to reduce the uncertainties on the computation of the seismic signal
features. This could be achieved by deploying denser seismic networks for
example. Second, the range of the mass of the blocks used in our experiment
spans only 1 order of magnitude. The behaviour of the relationships we found
has to be investigated for a larger range of volumes. Third, the
relationships found may be specific to a particular context and may depend on
the substrate onto which the rockfalls propagate. This again underlines the
relevance and the necessity of reproducing similar studies in new contexts.
Finally, our results give a new insight into the processes that generate
high-frequency seismic signals associated with rockfalls, landslides,
rock avalanches, and granular flows in general. We show that the maximum
amplitude of the seismic signal generated by the impact of a single particle
is proportional to its mass and velocity. In a granular flow, a very large
quantity of particles interact with themselves and with the substrate at a
given time. The magnitude of these impulses imparted on the Earth by each
particle might be controlled by the mass and the velocity of the particles
within the flow according to the correlations we observed. The issue is now
to understand what controls the dynamics of the particles within the flow and
how their complex interactions influence the generation of seismic waves.
This should be more thoroughly investigated, using numerical granular flow
models for example, and is probably the key to model the high-frequency
seismic signal associated with gravitational instabilities in the future.
The codes and the data used in this study are accessible upon request by
contacting C. Hibert (hibert@unistra.fr).
C. Hibert, J.-P. Malet, F. Bourrier and F. Provost participated in the acquisition and the processing of
the seismic and kinematic data. F. Berger, P. Tardif and E. Mermin helped to design and perform the Rioux-Bourdoux experiment,
and for the acquisition of the video and the reconstruction of the trajectories of the blocks. P. Bornemann performed the lidar
survey and the processing of the data that allowed reconstructing the DEM of the gully into which blocks were launched.
The authors declare that they have no conflict of
interest.
Acknowledgements
We are very grateful to Anne Mangeney for helpful discussions and
insightful suggestions, and to Georges Guiter (RTM) for organizing the
practical details and the security of the experimental launch site. This work
was carried with the support of the French National Research Agency (ANR)
through the projects HYDROSLIDE (Hydrogeophysical Monitoring of Clayey
Landslides) and SAMCO (Adaptation de la Société aux Risques Gravitaires
en Montagne dans un Contexte de Changement Global) and of the EUR-OPA Major Hazards Agreement of the Council of Europe through the project “Development
of cost-effective ground-based and remote monitoring systems for detecting
landslide initiation”. The data were acquired using instruments belonging to
the French national pool of portable seismic stations RESIF-SISMOB
(CNRS-INSU). The authors gratefully acknowledge Maxime Farin and Francesco Panzera for insightful
reviews. We also would like to thank the editor, Jens M. Turowski, for his substantial comments, which helped to improve this
paper.
Edited by: J. Turowski
Reviewed by: M. Farin and F. Panzera
References
Aki, K. and Chouet, B.: Origin of coda waves: source, attenuation, and
scattering effects, J. Geophys. Res., 80, 3322–3342, 1975.Allstadt, K.: Extracting source characteristics and dynamics of the August
2010 Mount Meager landslide from broadband seismograms, J. Geophys. Res., 118, 1472–1490, 10.1002/jgrf.20110, 2013.Amitrano, D., Grasso, J. R., and Senfaute, G.: Seismic precursory patterns
before a cliff collapse and critical point phenomena, Geophys. Res. Lett., 32, 10.1029/2004GL022270, 2005.
Burjánek, J., Moore, J. R., Molina, F. X. Y., and Fäh, D.: Instrumental
evidence of normal mode rock slope vibration, Geophys. J. Int., 188, 559–569, 2012.
Burtin, A., Hovius, N., Milodowski, D. T., Chen, Y.-G., Wu, Y.-M., Lin, C.-W.,
Chen, H., Emberson, R., and Leu, P.-L.: Continuous catchment-scale monitoring
of geomorphic processes with a 2-D seismological array, J. Geophys. Res.-Earth Surf., 118, 1956–1974, 2013.
Chen, C.-H., Chao, W.-A., Wu, Y.-M., Zhao, L., Chen, Y.-G., Ho, W.-Y., Lin,
T.-L., Kuo, K.-H., and Chang, J.-M.: A seismological study of landquakes
using a real-time broad-band seismic network, Geophys. J. Int., 194, 885–898, 2013.
Clouard, V., Athanase, J.-E., and Aubaud, C.: Physical characteristics and
triggering mechanisms of the 2009–2010 landslide crisis at Montagne
Pelée volcano, Martinique: implication for erosional processes and
debris-flow hazards, Bulletin de la Société Géologique de
France, 184, 155–164, 2013.
Crampin, S.: Higher modes of seismic surface waves: Second Rayleigh mode
energy, J. Geophys. Res., 70, 5135–5143, 1965.Dammeier, F., Moore, J. R., Haslinger, F., and Loew, S.: Characterization of
alpine rockslides using statistical analysis of seismic signals, J. Geophys. Res., 116, F04024, 10.1029/2011JF002037, 2011.
Dammeier, F., Moore, J. R., Hammer, C., Haslinger, F., and Loew, S.: Automatic
detection of alpine rockslides in continuous seismic data using Hidden Markov
Models, J. Geophys. Res.-Earth Surf., 121, 351–371, 2016.Deparis, J., Jongmans, D., Cotton, F., Baillet, L., Thouvenot, F., and Hantz,
D.: Analysis of rock-fall and rock-fall avalanche seismograms in the French
Alps, B. Seismol. Soc. Am., 98, 1781–1796,
10.1785/0120070082, 2008.Ekström, G. and Stark, C. P.: Simple scaling of catastrophic landslide
dynamics, Science, 339, 1416–1419, 10.1126/science.1232887, 2013.Farin, M., Mangeney, A., Toussaint, R., Rosny, J. D., Shapiro, N., Dewez, T.,
Hibert, C., Mathon, C., Sedan, O., and Berger, F.: Characterization of
rockfalls from seismic signal: Insights from laboratory experiments, J. Geophys. Res.-Sol. Ea., 120, 7102–7137,
10.1002/2015JB012331, 2015JB012331, 2015.Favreau, P., Mangeney, A., Lucas, A., Crosta, G., and Bouchut, F.: Numerical
modeling of landquakes, Geophys. Res. Lett., 37, L15305,
10.1029/2010GL043512, 2010.
Galea, P., D'Amico, S., and Farrugia, D.: Dynamic characteristics of an active
coastal spreading area using ambient noise measurements – Anchor Bay, Malta,
Geophys. J. Int., 199, 1166–1175, 2014.
Gance, J., Grandjean, G., Samyn, K., and Malet, J.-P.: Quasi-Newton inversion
of seismic first arrivals using source finite bandwidth assumption:
Application to subsurface characterization of landslides, J. Appl.
Geophys., 87, 94–106, 2012.Helmstetter, A. and Garambois, S.: Seismic monitoring of Séchilienne
rockslide (French Alps): Analysis of seismic signals and their correlation
with rainfalls, J. Geophys. Res., 115, F03016,
10.1029/2009JF001532, 2010.
Hertz, H.: Über die Berührung fester elastischer Körper, Journal
für die reine und angewandte Mathematik, 92, 156–171, 1882.Hibert, C., Mangeney, A., Grandjean, G., and Shapiro, N. M.: Slope
instabilities in Dolomieu crater, Réunion Island: From seismic signals to
rockfall characteristics, J. Geophys. Res., 116, F04032,
10.1029/2011JF002038, 2011.
Hibert, C., Grandjean, G., Bitri, A., Travelletti, J., and Malet, J.-P.:
Characterizing landslides through geophysical data fusion: Example of the La
Valette landslide (France), Eng. Geol., 128, 23–29, 2012.Hibert, C., Ekström, G., and Stark, C. P.: Dynamics of the Bingham Canyon
Mine landslides from seismic signal analysis, Geophys. Res. Lett.,
41, 4535–4541, 10.1002/2014GL060592, 2014a.Hibert, C., Mangeney, A., Grandjean, G., Baillard, C., Rivet, D., Shapiro,
N. M., Satriano, C., Maggi, A., Boissier, P., Ferrazzini, V., and Crawford,
W.: Automated identification, location, and volume estimation of rockfalls
at Piton de la Fournaise volcano, J. Geophys. Res.-Earth Surf., 119, 1082–1105, 10.1002/2013JF002970,
2014b.Hibert, C., Stark, C. P., and Ekström, G.: Dynamics of the Oso-Steelhead landslide from broadband seismic analysis,
Nat. Hazards Earth Syst. Sci., 15, 1265–1273, 10.5194/nhess-15-1265-2015,
2014c.
Hibert, C., Ekström, G., and Stark, C. P.: The relationship between
bulk-mass momentum and short-period seismic radiation in catastrophic
landslides, J. Geophys. Res.-Earth Surf., in press,
2017.
Levy, C., Jongmans, D., and Baillet, L.: Analysis of seismic signals recorded
on a prone-to-fall rock column (Vercors massif, French Alps), Geophys. J. Int., 186, 296–310, 2011.
Levy, C., Mangeney, A., Bonilla, F., Hibert, C., Calder, E. S., and Smith,
P. J.: Friction weakening in granular flows deduced from seismic records at
the Soufrière Hills Volcano, Montserrat, J. Geophys. Res.-Sol. Ea., 120, 7536–7557, 2015.
Maquaire, O., Malet, J.-P., Remaıtre, A., Locat, J., Klotz, S., and Guillon,
J.: Instability conditions of marly hillslopes: towards landsliding or
gullying? The case of the Barcelonnette Basin, South East France, Engineering
Geology, 70, 109–130, 2003.Moretti, L., Mangeney, A., Capdeville, Y., Stutzmann, E., Huggel, C.,
Schneider, D., and Bouchut, F.: Numerical modeling of the Mount Steller
landslide flow history and of the generated long period seismic waves,
Geophys. Res. Lett., 39, L16402, 10.1029/2012GL052511, 2012.
Norris, R. D.: Seismicity of rockfalls and avalanches at three Cascade Range
volcanoes: Implications for seismic detection of hazardous mass movements,
B. Seismol. Soc. Am., 84, 1925–1939, 1994.Panzera, F., D'Amico, S., Lotteri, A., Galea, P., and Lombardo, G.: Seismic site response of unstable steep slope using noise
measurements: the case study of Xemxija Bay area, Malta, Nat. Hazards Earth Syst. Sci., 12, 3421–3431, 10.5194/nhess-12-3421-2012, 2012.Schneider, D., Bartelt, P., Caplan-Auerbach, J., Christen, M., Huggel, C., and
McArdell, B. W.: Insights into rock-ice avalanche dynamics by combined
analysis of seismic recordings and a numerical avalanche model, J. Geophys. Res., 115, F04026, 10.1029/2010JF001734, 2010.
Spearman, C.: The proof and measurement of association between two things,
American J. Psychol., 15, 72–101, 1904.
Tripolitsiotis, A., Daskalakis, A., Mertikas, S., Hristopulos, D., Agioutantis,
Z., and Partsinevelos, P.: Detection of small-scale rockfall incidents using
their seismic signature, in: Third International Conference on Remote Sensing
and Geoinformation of the Environment, International
Society for Optics and Photonics, 953519–953519, 2015.Vilajosana, I., Suriñach, E., Abellán, A., Khazaradze, G., Garcia, D.,
and Llosa, J.: Rockfall induced seismic signals: case study in Montserrat, Catalonia,
Nat. Hazards Earth Syst. Sci., 8, 805–812, 10.5194/nhess-8-805-2008, 2008.Yamada, M., Matsushi, Y., Chigira, M., and Mori, J.: Seismic recordings of
landslides caused by Typhoon Talas (2011), Japan, Geophys. Res. Lett., 39, L13301, 10.1029/2012GL052174, 2012.Yamada, M., Kumagai, H., Matsushi, Y., and Matsuzawa, T.: Dynamic landslide
processes revealed by broadband seismic records, Geophys. Res. Lett., 40, 2998–3002, 10.1002/grl.50437, 2013.
Zimmer, V. L. and Sitar, N.: Detection and location of rock falls using seismic
and infrasound sensors, Eng. Geol., 193, 49–60, 2015.