Professional Documents
Culture Documents
s10346 021 01784 5
s10346 021 01784 5
s10346 021 01784 5
Landslides (2022) 19:747– 759 S. Daniel Andrade · Stefanie Almeida · Emilia Saltos · Daniel Pacheco ·
DOI 10.1007/s10346-021-01784-5
Received: 25 April 2021 Stephen Hernandez · Wilson Acero
Accepted: 14 October 2021
Published online: 9 January 2022
© Springer-Verlag GmbH Germany,
A simple and general methodology
part of Springer Nature 2021 to calibrate seismic instruments for debris
flow quantification: application to Cotopaxi
and Tungurahua volcanoes (Ecuador)
Abstract Debris flows are hazardous phenomena occurring at vol- (Brown et al. 2017; Thouret et al. 2020). They can occur directly asso-
canoes, and monitoring them has proved as challenging as impera- ciated to eruptive periods, and when volcanoes are dormant or even
tive in several cases. The use of seismic instruments to measure extinct (Cascini et al. 2014; Vallance and Iverson 2015; Thouret et al.
and study the physical properties of debris flows has witnessed 2020). Multiple water sources associated to volcanoes can give place to
significant progress in the last years, with the use of improved the formation of debris flows, among them rain (Lavigne and Thouret
sensors, innovative methodologies and high-resolution analysis. 2003; Santangelo et al. 2021), crater lakes (Manville and Cronin 2007;
However, the application of such studies to the practical task of Massey et al. 2010), glaciers (Pierson et al. 1990; Thouret 1990) and
providing early warnings remains limited by the significant amount ground-water bodies (Johnson et al. 2018). The size of such debris
of infrastructural and technological resources commonly required flows depends on the availability of water and sediments during their
for their deployment. In Ecuador, debris flows at volcanoes are formation and transport, with volumes ranging from 103 to 108 m3 and
detected by means of seismic instruments which are usually part peak discharges from 1 to 104 m3/s (Thouret et al. 2020). In the present
of broader monitoring networks, thus requiring calibration to study, the term debris flow implies a volumetric solid content > 60%
provide quantitative information about the flows and feed early- vol. of the water–sediment mixture, while the term hyperconcentrated
warning systems. In the present work, a theoretical approach based flow implies solid contents between 20 and 60% vol. (Jakob and Hungr
on the Buckingham Π-theorem is used to determine an expression 2005; Vallance and Iverson 2015).
that linearly correlates the seismic signal produced by a transiting Different types of monitoring instruments may be deployed
debris flow with its discharge rate, for instruments installed in dif- with the aim of identifying the potential or the actual occurrence
ferent substrata and at variable distances from the drainage. The of debris flows. Such instruments can be varied and their meas-
expression is experimentally tested with Acoustic Flow Monitors urements linked to different physical parameters associated with
and Broad-band seismometers installed in the vicinity of drain- debris flow occurrence, including rainfall (Bacchini and Zannoni
ages at Tungurahua and Cotopaxi volcanoes, where actual debris 2003; Osanai et al. 2010; Berenguer et al. 2015), ground and/or infra-
flows occurred in relation to eruptive activity. The experiments sound vibrations (Arattano 1999; Kumagai et al. 2009; Hübl et al.
consist in comparing the measured peak amplitude values of the 2013; Kogelnig et al. 2014; Abancó et al. 2014; Schimmel and Hübl
seismic signal envelopes with the estimated peak discharge rates 2016), flow stage/depth (Arattano and Marchi 2008; Jacquemart
of several events. The results confirm the validity of the theoreti- et al. 2017), surface or internal velocity (Arattano and Marchi 2005;
cal expression with linear correlations observed between the seis- Yang et al. 2011; Theule et al. 2018; Cui et al. 2018), basal or internal
mic amplitudes and the discharge rates, thus defining calibration forces (McArdell et al. 2007; Iverson et al. 2011; Berger et al. 2011)
expressions that can be generally applied to varied environments and visual aspect by video tracking which is usually complemen-
and instruments. The seismic instruments calibrated through this tary with other instruments (Berti et al. 1999; Genevois et al. 2001;
methodology can provide instantaneous and reliable predictions of Marchi et al. 2002; McCoy et al. 2010; Vázquez et al. 2016). When
debris flow discharge rates within less than an order of magnitude transmitted and analyzed in real time, the data produced by those
and only requiring limited data processing and storage. Such level instruments are expected to feed early-warning systems that could
of prediction could help to improve early warning systems based on trigger actions downstream in order to reduce the risk associated
seismic instruments installed in locations where more developed to populations or infrastructures in critical locations (Arattano and
instrumental arrays are unavailable or unpractical. Marchi 2008; Coviello et al. 2019). However, the deployment and
maintenance of such instrumental networks can prove prohibitive
Keywords Debris flow monitoring · Seismic instruments · Peak- because of either logistical, financial or security constraints related
discharge rate · Buckingham Π-theorem to both the instrumentation itself as well as the resources and infra-
structure required for the transmission, storage and analysis of the
Introduction data they produce (Hübl and Mikoš 2018; Hürlimann et al. 2019).
Debris flows are common phenomena occurring at volcanoes, where Research related to the use of seismometers for debris flow mon-
they are known as lahars, and often represent major hazards for com- itoring in volcanic and non-volcanic environments has increased
munities and infrastructure placed close to the drainages downstream in recent decades due to their adaptability and reliability as well as
Fig. 2 Seismic records of debris flows as observed at: (a) JUI-01 Acoustic Flow Monitor (AFM) at Tungurahua volcano; and (b) BNAS broad
band seismometer (BBS) at Cotopaxi volcano
Fig. 4 (a) Photograph of the Quebrada Cutzualo ravine (Cotopaxi matic diagram showing the geometric array employed to estimate
volcano) during fieldwork, after a debris flow recorded on 2016–01- one value of r from three C points that define one circle. Field meas-
13. Geometrical parameters for the application of Eqs. (10) and (11) urements employed between 4 and 6 C points, from which up to 10
are shown. Note the marks left by the flow during transit. (b) Sche- different circles and values of r were obtained in each curve
of 2000 kg*m−3 for all events will be used during the analysis of obtained for each event are available as supplementary material
the results (Iverson 1997; Manville et al. 1998). A discussion will be to this work.
presented regarding this assumption. When the values of Q are plotted against S, lineal correlations are
observed for both instruments, in agreement with Eq. (9) (Fig. 5).
The slope values of the linear regressions obtained in each station
Results
allow the calculation of the product M * δS/δL present in Eq. (9), by
extracting the value of x2 corresponding to either JUI-01 or JUI-02.
AFM instruments at Tungurahua volcano
Then, Eq. (9) can be written for the case of any AFM installed at a
The AFM instrument JUI-01 is installed x = 35 m away from the
distance x from the drainage as
Quebrada Palmaurcu ravine (Fig. 3a) in a location where the sub-
stratum corresponds to an andesitic lava flow with an assumed Q = 4.7 × 10−5 ∗ x 2 ∗ S + 3.8 (12)
density δS of 2400 kg/m3, while JUI-02 is installed x = 31 m away
from the same drainage in a location where the substratum is com- when the instrument is installed in a substratum composed of
posed of unconsolidated pyroclastic flow breccias with an assumed andesitic lava flow, and as
density δS of 1800 kg/m3 (Fig. 3a). In total between 2012 and 2014,
Q = 5.1 × 10−5 ∗ x 2 ∗ S + 4.8 (13)
JUI-01 recorded signals corresponding to six debris flows that could
be assessed in the field, but only four of those events were also when it is installed in a substratum composed of unconsolidated
recorded by JUI-02 due to temporary malfunctions of the instru- pyroclastic or epiclastic volcanic breccias.
ment. All the recorded events correspond to rain-triggered debris Confidence intervals for Q can be calculated for each AFM so
flows formed as excess sediments accumulated in the upper flanks Eqs. (12) and (13) can be used for discharge rate predictions based
of Tungurahua due to ongoing eruptions. on measurements of S (Fig. 5). However, given that the sample size
Table 2 summarizes the instrumental data and the field measure- of discharge rates (Q) in both instruments is small, the intervals
ments for each debris flow event. The peak seismic signals S varied have been calculated using a confidence percentage of only 75%. For
from 114 to 1140 AFM counts for JUI-01 and from 72 to 1251 for JUI-01 (Fig. 5a), where the sample size is 6 events, the confidence
JUI-02, while the peak discharges Q as estimated during fieldwork interval is ± 10.0 m3/s, while for JUI-02 (Fig. 5b), where the sample
varied from ~ 15 to ~ 73 m3/s. The detailed seismic and field data size is only 4, the interval is ± 15.5 m3/s.
Table 3 Data for debris Date Instrument S (µm/s) No. curves No. circles Q (m3/s) St-dv
flows studied at Cotopaxi Q
volcano. The detailed
seismic and geometrical field 2015–08-28 BNAS 6.26 5 56 6.6 1.7
measurements of each event
are provided as supplementary 2015–09-20 BNAS 40.28 5 48 33.5 4.1
material to this work 2015–11-16 BNAS 23.89 2 14 10.2 4.5
2015–11-28 BNAS 54.43 7 58 33.2 7.5
2016–01-13 BNAS 74.10 9 76 52.0 7.1
2016–04-09 BNAS 11.77 3 26 11.5 3.5
It is important to notice that the seismic signals provided by Q = 5.48 × 10−5 ∗ x 2 ∗ S + 1.26 (14)
AFMs are not direct measurements of the ground velocity, and
due to the way these instruments work, the actual seismic signal when the instrument is installed in a substratum composed of
is lost during data acquisition. Thus, Eqs. (12) and (13) are only unconsolidated volcanic breccias. Confidence intervals for Q have
valid for AFM instruments that use the same sensors and signals been calculated so Eq. (14) can be used for predictions based in
processed in the same way as the instruments installed in Quebrada measurements of S. Similarly to AFMs, by using a percentage of
Palmaurcu (Tungurahua volcano). Consequently, Eqs. (12) and (13) only 75%, the confidence interval for Q is ± 8.4 m3/s, with a sample
should allow the calibration of any of those AFM instruments for size of 6 events.
debris flow quantification, provided that x is measured and the It is important to note that the data provided by BNAS corre-
geological characteristics of the substrate are known. This is the sponds to actual seismic velocities, so that Eq. (14) should apply to
specific case of all AFM instruments installed at Tungurahua and any seismic instrument displaying a flat instrumental response for
Cotopaxi volcano. frequencies above 1/60 Hz and seismic envelope (RSAM) measure-
ments from signals obtained at 50 sps, filtered above 10 Hz in non-
BBS instrument at Cotopaxi volcano overlapping windows of 60 s, during the occurrence of a debris flow.
The BBS instrument BNAS is installed x = 110 m away from the Que- Validation of results
brada Cutzualo ravine (Fig. 3b), on a substratum composed of recent
unconsolidated pyroclastic flow and epiclastic breccias with an Validation of results was made by direct and indirect approaches.
assumed density δS of 1800 kg/m3. A total of 6 rain-triggered debris The direct validation was based on testing either Eqs. (12), (13) or
flows were recorded that could be surveyed in the field between (14) for past debris flow events recorded in drainages different
August 2015 and April 2016, as Cotopaxi volcano experienced an from Quebrada Palmaurcu (Fig. 3a) or Quebrada Cutzualo (Fig. 3b),
eruptive phase which produced a significant accumulation of loose and by comparing observed peak discharges with those predicted
ash in its upper western flank (Bernard et al. 2016a; Hidalgo et al. from the seismic records. Although abundant seismic records of
2018). Little is known about the specific conditions of the Quebrada debris flowss actually exist at IGEPN (Mothes and Vallance 2015),
Cutzualo upper catchment because its dangerous location has pre- the results obtained during this work could only be validated for
vented in situ assessments. Table 3 summarizes the instrumental a few events. The main challenge has been obtaining reliable data
data and the field measurements for each event. The peak seismic for the peak discharges Q of the debris flows events used for direct
velocities S varied from ~ 6.26 to ~ 74.1 μm/s, while the peak dis- validation.
charges Q as measured in the field varied from ~ 6.6 to ~ 52 m3/s. The The direct validation approach was possible for three instru-
detailed seismic and field data obtained for each event are available ments: VAZ-01 (AFM) and BRUN (BBS), installed respectively 24 m
as supplementary material to this work. and 955 m away from Rio Vazcum drainage at Tungurahua volcano
Similar to the case of AFM’s and in agreement with Eq. (9), a (Fig. 3a), and MIS-01 (AFM) installed 105 m away from Quebrada
linear correlation is observed when the values of Q and S are plotted Mishahuaycu drainage at Cotopaxi volcano, which has Quebrada
(Fig. 6). Here again, the slope value of linear regression enables the Chanchunga as its main tributary (Fig. 3b). The substrate around
calculation of the product k * δS/δL present in Eq. (9), by extracting Río Vazcum is composed by both lavas and consolidated volcanic
the value of x2 corresponding to BNAS. Then, Eq. (9) can be writ- breccias, while at Quebrada Mishahuaycu it is composed of par-
ten for the case of BBSs placed at a distance x from the drainage as tially consolidated volcanic breccias. Thus, Eq. (12) was used for the
validation of events recorded at AFM stations VAZ-01 and MIS-01, An indirect validation approach was used with the aim to tackle
and Eq. (14) for the case of BRUN. the lack of peak discharge measurement information. For this, a
Three debris flow events were recorded by VAZ-01 between comparison of data coming from one AFM and one BBS monitor-
2005 and 2012 at Rio Vazcum, Tungurahua volcano (Table 4). The ing the same drainage from contrasting distances was performed,
peak discharges Q of those events were estimated at those times by under the hypothesis that both instruments should provide similar
IGEPN staff using Eq. (10) by measuring accurate cross sections predictions for simultaneously recorded events. Such a compari-
A at Rio Vazcum but having to assume values of v in the range of son was possible at Quebrada Mapayacu (Tungurahua volcano)
5–10 m/s. All three were rain-triggered events of exceptional size where MAP-01 (AFM) and BMAS (BB) instruments are respec-
and were studied in some detail just after their occurrence because tively installed at 45 m and 840 m away from the drainage (Fig. 7).
they produced some level of damage downstream in Baños town The substratum underneath MAP-01 is composed of young and
(Fig. 3a). On August 25, 2008, BRUN recorded a big debris flow at poorly consolidated volcanic breccias, and thus Eq. (13) was used
Rio Vazcum originating from a landslide occurred downstream for predictions. Conversely, the substratum underneath BMAS is
from VAZ-01 which produced damage in Baños (Fig. 3a). Finally, composed by both lavas and partially consolidated volcanic brec-
one debris flow was recorded by MIS-01 on November 28, 2015 at cias, and thus Eq. (14) was used for predictions.
Quebrada Mishahuaycu, Cotopaxi volcano (Table 4, Fig. 3b). In Five debris flow events occurred at Quebrada Mapayacu between
this case, the peak discharge Q was estimated using the exact same 2011 and 2014 that could be simultaneously recorded by MAP-01
methodology and staff as all other events detailed in the present and BMAS (Table 5). The seismic amplitudes recorded by both
work. instruments are varied and span more than one order of magni-
Table 4 shows that predicted and observed peak discharges are tude. Table 5 shows that the predictions of peak discharges pro-
generally in agreement, especially in the case of the event recorded duced by either instrument are in general agreement, with most of
by MIS-01 at Quebrada Mishahuaycu drainage. In all cases, the peak them closely overlapping. As in the direct validation, this approach
discharges obtained from Eqs. (12) and (14) would have allowed also suggests that the results presented here can be used to provide
issuing reliable predictions within less than an order of magnitude. reliable predictions of debris flows peak discharges within less than
This information could have improved the early warnings provided an order of magnitude from seismic records.
by IGEPN.
Table 4 Data for debris flow Date Station x (m) S (counts) Q predicted (m3/s) Q estimated
events used for the direct (m3/s)
validation approach of results.
VAZ-01 and MIS-01 are AFM 2005–12-12 VAZ-01 24 3582 91–111 100–200
instruments, while BRUN is
a BBS. Values of estimated 2008–08-25 BRUN 955 6.2 (*) 285–302 300–600
Q for events recorded with 2011–11-11 VAZ-01 24 1650 39–59 50–100
AFMs were calculated using a
velocity range of 5–10 m/s 2012–12-01 VAZ-01 24 3069 77–97 90–180
2015–11-28 MIS-01 105 75 32–52 35–40
Conclusions
The use of the Buckingham Π-theorem to represent a simplified
physical system in which a debris flow transits close to a seismic
station has provided a general theoretical expression which cor-
relates the seismic signal (S) with the discharge rate of the flow
(Q) through parameters like the flow density (δL), the substratum
density (δS) and the distance (x) between the drainage and the
seismic instrument.
Experiments carried out at Cotopaxi and Tungurahua volcanoes
in which the peak discharges of actual debris flows were compared
with their peak seismic signals have confirmed the validity of the
theoretical expression. As a result, expressions have been obtained
to calibrate the specific cases of AFM instruments operating at
Cotopaxi and Tungurahua, but also in a general way for the case of
seismic envelopes of either 1, 10, 30 or 60-s time-windows obtained
from signals recorded at 50 sps by an instrument displaying a flat
response for frequencies above 1/60 Hz.
The simplicity of the parameters required for the calibration as
well as the characteristics of the seismic signal makes the results
applicable to different geological environments and to a large diver-
sity of seismic instruments.
Given the relatively limited number of observations, the 75%
confidence intervals for the predictions of peak discharges Q from
seismic records S are relatively large. However, various validations
performed under different circumstances suggest that the predic-
Fig. 8 Plots of Q vs. S displaying the observations made at Quebrada
tions of discharge rates are reliable within less than an order of
Cutzualo (Cotopaxi volcano) for BNAS BBS instrument. The envelope magnitude. This information could be really helpful for the issuing
values of S were calculated with windows of a) 30 s; b) 10 s; and c) of early warnings in drainages where only seismic instruments are
1 s. Linear correlations between Q and S are observed for all time- available for debris flows monitoring.
windows
Acknowledgements
The authors acknowledge the hard and tireless work by the technical
Q = V ∗ t −1 (19) staff at Instituto Geofisico—Escuela Politécnica Nacional in charge of
keeping in function the monitoring networks of the Ecuadorian volca-
where V is debris flow volume and t is unit time, and thus
noes, especially to Cristina Ramos, Darío García and Carlos Macías who
𝛿L ∗ Q = m ∗ t −1 (20) are in charge of the debris flow detection instruments. The authors wish
to thank two anonymous reviewers for their insightful and constructive
which defines a “mass per unit time” term to be introduced in comments which helped to improve the original manuscript.
Eq. (18) instead of ma, and thus Ek may actually be defined as the
kinetic energy per unit time and be written as