Introduction
Studies focusing on landslides occurring directly in moraine deposits
or moraine ridges are scarce (Beló et al. 2006), even though
they have potential to trigger impact waves responsible for the
generation of glacial lake outburst floods (Fig. 1). Glacial lake
outburst floods have been responsible for large economic costs
and the loss of thousands of lives worldwide (Iribarren Anacona
et al. 2014). In Cordillera Blanca, Peru (Zapata 2002), different
types of slope movements (e.g., landslides, and rockslides) caused
35 % of documented glacial lake outburst floods with known cause
(Emmer and Cochachin 2013). These slope movements in previously
glaciated regions have often been related to the effects of
multiple glacier retreats that result in the debuttressing of bedrock,
as well as in debris-mantled slopes and the deposition of steeply
inclined glacier sediments (Holm et al. 2004; Hugenholtz et al.
2008). It has been documented that the removal of ice-buttress
support has affected landslide occurrence on several lateral moraines
worldwide. One example is a compound landslide on the
southeast lateral moraine of Palcacocha Lake, which caused a
minor glacial lake outburst flood in 2003 (Cordillera Blanca,
Peru, Vilímek et al. 2005). Debuttressing also contributed to the
activity of large landslides on the moraine of Athabasca glacier in
the Canadian Rocky Mts. (Hugenholtz et al. 2008) and resulted in
moraine landslides with volumes of up to 0.7 × 106
m3
in the Alps
(i.e., Lower Grindelwald Glacier, Oppikofer et al. 2008; Huggel
et al. 2013).
The mechanical and hydrological properties of the moraine
slopes are the key to their stability and the possible characteristics
of landslides occurring in their material. Only a limited number of
studies have dealt with this topic, partly due to methodological
constraints in investigating materials with a heterogeneous mix of
particles varying from silt to boulders. The significant effects of the
fine fraction of the moraine material on its overall strength have
been described by Novotný and Klimeš (2014), who argued that
suction occurring within this material affects the peak shear
strength of the unsaturated moraine material (Simeoni et al.
2003; Springman et al. 2003). Other studies underline the importance
of mineral composition, attributing lower strength to sand
and gravel soils containing greater proportions of quartz, while
increased feldspar composition leads to higher strength due to the
greater angularity of the particles (Bolton 1986; Bareither et al.
2008). Slope stability calculations for this environment are even
scarcer (Hubbard et al. 2005) and very often involve moraine
material situated in contrasting geological settings, e.g., forming
the sliding plane of a rock slide (Schneider-Muntau and Zangerl
2005) or moraine deposits overlaying marls (Simeoni et al. 2003).
Another way to better assess potential future landslide hazard is
to investigate rates of past movements (Viero et al. 2010; Strozzi
et al. 2013; Sun et al. 2015), which ideally cover prolonged time
periods and include monitoring of explanatory variables (e.g.,
ground water levels, pore pressures). In remote high mountain
regions where none of these data are available, remote sensing is a
suitable alternative for detecting past movements of different
landforms (e.g., rock glaciers—Emmer et al. 2015), including landslides
(Farina et al. 2006; Righini et al. 2011; Tamburini et al. 2011;
Cigna et al. 2012; Hölbling et al. 2012; Strozzi et al. 2013).
Spaceborne Synthetic Aperture Radar (SAR) data and related
interferometry techniques allow the detection of movement over
wide areas with millimeter precision, usually representing total
displacements over periods of years.
The initiation and propagation of glacial lake outburst floods
are a complex phenomenon that has rarely been studied considering
all the different processes involved (Westoby et al. 2014). The
complex chain modeling of documented, but unmeasured glacial
lake outburst floods is presented, for example, in Worni et al.
Original Paper
(2014) or Schneider et al. (2014). The latter is an example from the
Cordillera Blanca using Rapid Mas Movement Simulation
(RAMMS) for modeling a triggering ice avalanche; wave generation
and propagation are modeled using the Iber hydrodynamic
model, which provides outflow hydrographs that in turn are used
for glacial lake outburst flood modeling, also using RAMMS.
This study aims at an evaluation of landslides from moraine
slopes as potential triggers of glacial lake outburst floods. We
carried out a variety of research activities, including detailed field
data collection, satellite SAR interferometry (InSAR), stability
calculations, and hydrodynamic modeling at Palcacocha Lake,
Cordillera Blanca, Peru.
Study area
Palcacocha Lake is situated in the upper part of the Cojup glacial valley
(4562 m above sea level (a.s.l.), Fig. 2), which is carved into granites and
granodiorites of the Cordillera Blanca batholith (Morales 1967, Estudios
glaciologicos-geologicos efectuados en la Cordillera Blanca (in Spanish),
unpublished report). This lake produced a catastrophic glacial lake
outburst flood in 1941 (Oppenheim 1946; Zapata 2002) and another
minor outburst flood in 2003 (Vilímek et al. 2005). The lake’s dramatic
increase in volume (Somos-Valenzuela et al. 2014), unstable moraines,
and the increasing vulnerability of the growing city of Huaráz
located downstream made it a suitable site for study.
The inclination of the trough slopes varies between 60° and
80°close to the ridges, dropping to 30°–40° near the valley floor,
where the slopes are often covered by talus (Vilímek et al. 2005).
The Cordillera Blanca foothills and the Santa River Valley are
underlain by sedimentary and volcanic rocks of Mesozoic age. In
this area, the Cojup River flows mainly through glaciofluvial deposits,
where it has carved a deeply incised canyon-like valley with
numerous bank scours. The valley floor progressively flattens and
finally opens into an alluvial cone, where the city of Huaráz is
located and the Cojup River joints the Quilcay River, which in turn
flows into the Santa River at an elevation of 3000 m a.s.l.
Palcacocha Lake is one of almost 600 lakes in the Cordillera
Blanca mountain range dammed by moraines (Vilímek et al. 2015).
Its current moraine developed during the Little Ice Age, in which
the first phase culminated in the Cordillera Blanca between 1590
and 1720 AD (Rabatel et al. 2013). Its terminal moraine (Fig. 2) rises
up to146 m above the valley floor. It was filled by lake water until 13
December 1941, when it breached, producing a catastrophic outburst
flood that claimed thousands lives in the regional capital city
of Huaráz (Lliboutry et al. 1977a; Zapata 2002; Carey 2010). The
triggering factor is unknown, but no earthquake was reported
before the outburst. Thus, ice fall or glacier calving seems to be
possible explanations (Oppenheim 1946). Oppenheim (1946)
pointed out that another possibility is dam failure without any
external trigger (see also Costa and Schuster 1988), due to (i)
internal erosion (widening) of subsurface channels leading to the
degradation of dam body and to its rupture (e.g., Richardson and
Reynolds 2000) or (ii) spontaneous incision of a surficial outlet in
the dam, followed by increasing discharge and additional erosion
and incision, and thus progressive enlargement of the dam breach
(positive feedback; Yamada 1998). Since then, the stability of the
lake dam and possible processes for glacial lake outburst flood
initiation have been an important topic for Peruvian and international
scientists, as well as for the local community (Carey 2005).
Fig. 1 Schematic diagram showing theoretical types of slope processes which could triggered GLOFs in the Cordillera Blanca Mts., Peru, representing only fraction of all
possible GLOF initiation processes (Emmer and Cochachin 2013; Richardson and Reynolds 2000). Ice/rock avalanche (I/R Av) caused the 2010 GLOF from Lake 513 (Klimeš
et al. 2014; Schneider et al. 2014); rockslide (Rs) initiated the 2002 GLOF from Safuna Alta Lake (Hubbard et al. 2005); landslide from lateral moraine caused the 1953
GLOF from Tullparaju Lake (Lliboutry et al. 1977a), the 2003 GLOF from Palcacocha Lake (Vilímek et al. 2005), and the 2012 GLOF from Artizon Alto Lake (Emmer et al.
2014); a debris flow (Df) from a valley dammed temporal lake, which initiated a landslide from the inner slope of the Cangrajanca Lake lateral moraine (Cordillera
Huayhuash, Peru, Engel et al. 2011). a Workflow diagram showing the steps taken to achieve complex knowledge of processes and conditions affecting hazard imposed by
GLOFs triggered by landslides originating in the lateral moraines of the Palcacocha Lake. Asterisks indicate steps where review of available literature was extensively
used. Ch. chapter, Fs factor of safety
Original Paper
After the 1941 event, the lake was dammed by a wide moraine with
low relief resting on the bottom of the previous lake. It is among
about 40 lake dams in the Cordillera Blanca where remedial works
were carried out (Reynolds 2003; Portocarrero 2014; Emmer et al.
2016). The dam was reinforced in the early 1970s by two concrete
dams with an outflow channel that keeps the water elevation
constant at 4562 m a.s.l. (Ojeda 1974). Since 1970, the glacier
tongue has retreated by more than 1200 m and the lake volume
has increased to a recent volume of about 17 million m3
(Instituto
Nacional de Defensa Civil 2011).
A glacial lake outburst flood triggered by a landslide into the
lake occurred on March 19, 2003. The landslide was probably
caused by saturation of the moraine material by heavy rain. It
created a wave more than 8 m high that overtopped the lake dam
and resulted in a minor flood in the Cojup Valley, blocking a main
Huaráz drinking water treatment facility with sediments. As result
of this event, the inhabitants of the Huaráz were without a drinking
water supply for more than a week (URGH (2003): Informe
técnico: Estado situacional de la laguna Palcacocha (in Spanish).
ANA, p 19). Earlier works suggested that the retreat of the glacier
tongue, which intensified after 1980, contributed to the landslide
initiation (Emmer et al. 2014; Vilímek et al. 2005). Investigations of
mechanical and strength properties showed that the inner moraine
walls maintain stability despite their very steep slopes (very often
above 50°), which exceed the slope angle of tested strength parameters
(e.g., angle of repose), which reach values of around 40°
(Novotný and Klimeš 2014). This can be partly explained by
additional factors affecting the slope stability, including secondary
cementation of the sediment, wedged boulders, or the sedimentary
structure of the moraine, with additional strength imparted by
Fig. 2 Palcacocha Lake and its surrounding on the 5-m DEM (Horizons South America S.A.C., 2013, Informe Técnico del Proyecto, Consultoría Para El Levantamiento
Fotogramétrico Detallado De La Sub Cuenca Del Río Quillcay Y La Ciudad De Huaráz BID-MINAM (PE-T 1168)) and on a photograph looking NW, toward Ranrapalca Mt.
(6162 m a.s.l.) showing east-facing inner slopes of its terminal moraine; S.R. Santa River, Q.R. Quilcay River
large rock slabs oriented perpendicular to the direction of possible
shear planes, and interlocking soil strength has to be applied.
Methods
Field research, satellite SAR interferometry, and numerical analysis
were carried out (Fig. 1a) to define realistic conditions needed
for assessment of glacial lake outburst floods triggered by landslides
from the lateral moraines. Field geomorphological mapping
was supported by analyses and interpretation of satellite images
(SPOT image from 5.8.2006, Google Earth image from 20.8.2013)
and historical aerial photographs (15.7.1950, SAN Perú; August
1970, NASA) to prepare a geomorphological and landslide inventory
map of the Palcacocha Lake terminal and lateral moraines.
The geomorphological map defined landforms based on their
morphology and formative processes. It focused on identifying
landslides and describing their current activity during the field
research. Fresh-looking features with opened cracks were classified
as active, while those more obscured by vegetation and denudation
were considered as inactive. The map also shows basic
hydrological information, focusing on describing possible sources
of seepage water (e.g., streams, springs) in the moraine slopes,
during the dry season (May–September) when fieldwork was
conducted.
In contrast to geomorphological mapping that defines landslides
according to visible morphological features and the presence
of sliding planes, differential SAR interferometry (DInSAR; Bamler
and Hartl 1998; Rosen et al. 2000) and Persistent Scatterer Interferometry
(PSI; Wegmüller et al. 2003; Werner et al. 2003) were
used to determine the state of activity of landslides based on
displacement rates derived from the interferometric signals
(Farina et al. 2006; Righini et al. 2011; Cigna et al. 2012; Hölbling
et al. 2012; Strozzi et al. 2013). In this study, DInSAR and PSI have
been applied to stacks of high-resolution ERS-1/2 SAR, ENVISAT
ASAR, and ALOS PALSAR (ground resolution on the order of
20 m) and very-high-resolution TerraSAR-X images (ground resolution
on the order of 3 m) acquired between 1992 and 2015
(Table 1). Images from ascending and descending orbits were
analyzed for better spatial coverage around the lake. Snow cover
did not cause phase decorrelation in the interferograms, and
therefore, all scenes could be included in the analysis. PSI analysis
was conducted using only the stacks of ENVISAT ASAR images of
the descending orbit and of ALOS PALSAR images of the ascending
orbit, because for the other image stacks, the number of
acquisitions was insufficient for this kind of processing. PSI results
consist of linear deformation rates and displacement histories in
the satellite line-of-sight (LOS) direction. Reference points were
selected individually for each of the sensors in areas estimated as
stable. The error of the average ENVISAT ASAR displacement rates
ranges from 1.0 to 1.8 mm year−1
(Crosetto et al. 2009) that of the
ALOS PALSAR displacement rates is better than 9 mm year−1
(Ng
et al. 2012). In DInSAR analysis, displacement is derived from the
interpretation of the phase difference of the signals acquired by
two satellite SAR acquisitions after compensation for topographic
effects using an external digital elevation model (DEM) with 5 × 5m
pixel size derived from LiDAR measurements (unpublished
report of Horizons South America S.A.C., 2013 Informe Técnico
del Proyecto, Consultoría Para El Levantamiento Fotogramétrico
Detallado De La Sub Cuenca Del Río Quillcay Y La Ciudad De
Huaráz BID-MINAM (PE-T 1168), Ministerio Del Ambiente A
Travel Del Fonam, Lima, Peru). Zones with similar displacement
rates were identified by interpretation of wrapped interferometric
phase images, following recommendations given in Barboux et al.
(2014) for interpreting DInSAR signals in mountainous areas.
Final determination of whether such zones represent landslides
was based on field inventory mapping. As a rule of thumb, an
error estimate of one fourth of a phase cycle is considered for the
DInSAR images to be a consequence of atmospheric artifacts
(Strozzi et al. 2001).
Electric resistivity tomography (ERT) profiles were measured
using a 4point light hp earth resistivity meter (Fischer et al. 2013)
with electrode spacing of 4 or 2 m, applying Schlumberger,
Wenner, and dipole-dipole array settings. The system allowed a
maximum profile length of 96 m; therefore, it was necessary to roll
up sections to extend the profiles. Due to the sloping unstable
sediment, it was not possible to ground the electrodes on the steep
inner slopes of the moraine, so the profiles end at the crest of the
moraine rim. The maximum acquired length and depth reach by
the profiles were 120 and 12.5 m, respectively. The data were
processed in the specialized software Res2DInv (by Geotomo Software,
Inc.), and the inversion procedure included the effect of the
topography of the profiles. The topographic profiles were measured
using a hand-held laser inclinometer (LaserAce 1000) with a
precision of 0.05 m per 150 m, which allowed all important morphological
features to be captured.
Slope stability calculations were carried out for a topographic
profile (Fig. 3) across a potentially active compound landslide in
the position of profile no. 1 (ERT1 on Fig. 7). Smaller, secondary
landslides of the compound, potentially active landslide were
considered for the calculations. The calculations were done in
GeoSlope software, applying a limit equilibrium method
(Morgenstern and Price 1965). The potential sliding surface (e.g.,
fully specified slip surface) was determined using field mapping
and the results of resistivity tomography sounding, which clearly
showed its depth, shape, and down-slope limit on the inner moraine
slope. Shear strengths of the moraine material were derived
from standard and large shear box tests and laboratory tests of
angle of repose published in Novotný and Klimeš (2014). The
tested soil sample was taken from the landslide scarp. These tests
allowed application of peak strength and strength in a critical state
to the stability calculations (Fig. 3). Laboratory tests determining
minimum and maximum compaction of the material were used to
estimate a value of the bulk density of the in situ material, with the
value being close to the maximum compaction. It was estimated to
be γ = 20 kN/m3
. The moraine material was divided into homogenous
layers parallel with the surface of the profile, which were
characterized by specific stresses and respective peak shear
strengths determined using large shear box tests (Fig. 3b). The
stability calculations reflected variability in water content within
the moraine material (applied as pore pressure coefficient ru in the
calculations, where ru represents the portion of water above slip
surface) and the effects of earthquakes, introduced as a value of
seismic kh coefficient, which is the ratio of horizontal acceleration
to the gravitational acceleration constant. The role of ice support
at the slope toe was also tested, but the possible effects of waterlevel
fluctuations were not considered, as the outflow from the
lake is fixed by a control structure.
For the hydrodynamic wave modeling, we used the LiDARderived
5 × 5-m DEM and bathymetric data measured in the field
Original Paper
in October 2003, using Humminbird 100SX sonar (unpublished
report of URGH, 2003 Informe técnico: levantamiento batimetrico
y topográfico de la laguna Palcacocha (in Spanish), ANA). The Iber
hydrodynamic model (Iber 2010) allows simulations of turbulent
free-surface unsteady flow (Schneider et al. 2014) using input
information about the volume of the falling mass and its velocity,
with the volume as the critical parameter for the wave type and
energy (Panizzo et al. 2005). The landslide volumes were estimated
using results of detailed geomorphological mapping (defining
aerial extend of the landslides) including topographic profiles,
ERT measurements (providing information about possible depths
of the main sliding planes of the landslides), and evaluation of prefailure
aerial photographs (1970). The measured aerial extents of
the landslide bodies were multiplied by estimated depths of the
main sliding planes to provide an estimate of the landslide volume.
The modeling approach used information about the 2003
landslide (aerial extent, depth of the sliding plane) and triggered
impact wave height to calibrate the model parameters (e.g., impact
wave velocity), which were then used to model the possible impact
waves caused by the potentially active landslides identified on the
moraine. The model results provide maximum wave height and
flow volumes. The wave height was compared with a description of
the 2003 historical glacial lake outburst flood initiated by a landslide
into the lake (Vilímek et al. 2005). Comparison of the historical
event with modeled scenarios was used to assess possible
future glacial lake outburst flood magnitudes and the degree of
hazard caused by the specified outburst flood initiation process.
Results
Geomorphological and landslide mapping of the Palcacocha Lake
moraine
The general morphology of the lateral moraines and the lake basin
is shown on two topographic profiles (b on Fig. 4), which illustrate
the gentler inclination of the outer slopes (around 35°) compared
to the inner moraine slopes (over 50°), where the majority of the
debris flows and landslides occur. No active landforms were identified
on the outer slopes. This general observation holds also for
available historical information. The outer slopes show some elongated
and concave landforms. Some of them were identified on the
July 1950 and August 1970 aerial photographs, and some of them
were also identified on a photograph by Hanz Kinzl (1932). This
Table 1 Satellite SAR data considered in our study (A: ascending orbit, D: descending orbit)
Sensor Track Time period No. of images Time interval (days)
ERS-1/2 SAR D 440 2003–2010 12 35
ENVISAT ASAR D 440 2003–2010 21 35
ERS-1/2 SAR A 218 1996–1999 7 35
ENVISAT ASAR A 261 2006–2009 13 35
ALOS PALSAR A 111 2007–2011 19 46
TerraSAR-X D 066 2014–2015 2 11
PSI was conducted with stacks of ENVISAT ASAR images of the descending orbit and ALOS PALSAR images of the ascending orbit (italic)
Fig. 3 Shear strength in peak (φ) and critical (φcr) state and topographic profile used for slope stability calculation in GeoSlope software (s.s. slip surface)
suggests that at least some are more than 80 years old and originated
as debris flows or lake spillways formed during the lake
development (1 on Fig. 4). However, they could be even older and
could have formed during the initial moraine development, when
the glacier completely filled the space behind the moraines (2 on
Fig. 4). The historical aerial photographs confirm landslide activity
on the inner slopes. A shallow landslide on the northwest lateral
moraine failed after August 1970, and more recently, a smaller
landslide (3 on Fig. 4) was detected in roughly the same spot.
The southeast lateral moraine shows a clearly visible scarp, parallel
to its crest, in the 1950 image, which 20 years later developed into a
landslide about 400 m wide. Part of this landslide remained until
recent times (4 on Fig. 4), whereas the rest failed and was replaced
by several recently documented landslides described in detail later
(5 and 6 on Fig. 4).
The inner slope of the southeast lateral moraine is dominated
by a compound landslide formed by a temporarily inactive landslide
from March 2003 (6 on Fig. 4) and an active debris flow (6a
on Fig. 4). The 2003 landslide triggered an impact wave, which
resulted in a minor glacial lake outburst flood. It has a mostly
planar shear plane and reaches a total length of about 400 m, of
which at least 150 m represents an accumulation deposit on the
lake bottom, as can be identified from the October 2003 bathymetric
measurements (URGH (2003b): Informe técnico:
levantamiento batimetrico y topográfico de la laguna Palcacocha
(in Spanish). ANA, p 6). The activity of the debris flows persists
since 2004 (Emmer et al. 2014). It has a fresh-looking accumulation
and a very steep and incised source area with a permanent
spring. The spring is at the contact between moraine sediments
and a bedrock outcrop (Fig. 5a) and has been detected repeatedly
during our fieldwork in dry periods and thus is probably permanent.
Another compound landslide was identified just southwest
of the active debris flow (5 on Fig. 4). It consists of a smaller,
secondary landslide (5a on Fig. 4), which rests on a steeply inclined
slope formed by the head of a deeper and larger landslide. Both of
these landslides were identified as potentially active. The secondary
landslide has a fresh-looking 3-m-high scarp, suggesting relatively
recent vertical movement. There are also two open tension
cracks on its body (open to a depth of 0.8 m and up to 2 m wide).
The body of the larger landslide is formed by ridges, which are
elevated with respect to the area outside the landslide by up to 1 m.
There are also several tension cracks generally sub-parallel to the
margin of the slope (Fig. 5b). A deeper landslide (more than 30 m)
with a retrogressive character (7 on Fig. 4) was identified close to
the outburst channel of the lake. It is probably related to the slope
oversteepening due to the incision of the outflow channel which
formed, probably very quickly, during the catastrophic moraine
breach in 1941. Shallow landslides dominate the inner slope of the
northwest side moraine. Landslide no. 8 (Fig. 4) has a low scarp,
suggesting low mobility of the slide mass or an initial stage of
landslide development, which is the reason why it was classified as
a potentially active landslide. The large landslide (9 on Fig. 4) is
shallow, as shown by rock outcropping in its accumulation area.
Internal structure of the potentially active landslides
A series of ERT profiles (Figs. 6 and 7) were measured to describe
the subsurface environment of the potentially unstable rim of the
lateral moraine. The results (Fig. 7) document major features
identified during the geomorphological mapping. In particular,
tension cracks on the secondary landslide (5 on Fig. 4) are clearly
visible on profile no. 1 (Fig. 7), which also shows the geometry of
the possible shear plane of the secondary landslide. Some of the
tension cracks identified during field mapping that mark the scarp
of the large, potentially active landslide are also clearly visible
within its body (profiles 2 and 3 on Fig. 7). High and narrow
resistivity zones (with resistivity over 50,000 Ωm) on the profile
no. 5 (Bt?^ on Fig. 7) were not recognized during the field mapping,
and their interpretation is unclear. They may represent tension
Fig. 4 Geomorphological map (a) and cross profiles (b) of the Palcacocha Lake terminal and side moraines (s. slope). Topographic profiles are not exaggerated. Inset c
shows body and main scarp of the secondary landslide (2a) (for explanations of numbers 1–9, see text). Background image is August 2013 Google Earth satellite image
Original Paper
fissures formed below the surface, which may be a precursor of
possible landslide development in the future.
Extremely high resistivity readings on profile no. 1 and no. 2
(Bbb^ on Fig. 7) are attributed to a zone built by large boulders
(diameter over 1 m), which contains numerous air-filled spaces,
which, along with bedrock, have very low conductivity. A pronounced
talus deposit was identified at the toe of the slope on
profile no. 2. The extremely high resistivity zone identified at
depth in the middle of this profile (Bbb?^ on Fig. 7) may represent
either a buried boulder accumulation or fragments of the bedrock.
The zones with low resistivity (Bw^ on Fig. 7) may be interpreted
as relatively wet areas, where the moraine sediments are soaked by
water from an underground stream. Material forming these zones
may also contain a higher portion of fine particles, which would
contribute to longer preservation of the wetness due to their lower
hydraulic conductivity (Novotný and Klimeš 2014). These zones on
profile nos. 1 and 2 do not correspond well with the topographically
lowest point of the side valley. This suggests some lithological constraint
on the low resistivity layers and the presence of fine-grained
and presumably clay-rich materials, which exhibit quite low
resistivity (e.g., 1–100 Ωm; Reynolds 2011). Lower resistivity
zones on profile nos. 4 and 5 are probably not related to
underground water but rather represent properties of the
moraine material that better preserve moisture. This material
is sharply delimited from higher resistivity regions, which are
interpreted as dryer and coarser moraine material. These two
contrasting materials on profile nos. 4 and 5 may also represent
different generations of moraine sediments, representing
complex development of the studied area.
Results of the geomorphological mapping and ERT profiles
were used to estimate the volumes of the mapped landslides. The
estimated volume of the secondary landslide (5a on Fig. 4) is
30,600 m3
, assuming 8 m as an average depth of the potentially
sliding plane. Many uncertainties are involved in estimating the
volume of the larger potentially active landslide, due to its more
irregular shape and the possibility that it continues below the lake
surface. We estimate the volume of the potential landslide to be
about 630,000 m3
, assuming that the average potentially sliding
Fig. 5 a Permanent spring found in the source area of the active debris flow, which is fed by a small stream as well as underground water flows bound to the contact of
moraine sediment and bed rock outcrop (lower part of the image). b Yellow line shows tension fissure (t), which marks the potential scarp of the large active
landslide. Trace of ERT profile no. 2 is also shown
Fig. 6 Detailed view of the potentially active landslides and ERT survey profiles, which 0 m is always on their NW end at the edge of the outer moraine slope. Inset shows
respective portion of the geomorphological map on Fig. 4 (background image is an orthophotograph provided by ANA)
body depth was 20 m and that the landslide did not extend below
the lake water level.
Satellite SAR interferometry
The ENVISAT ASAR and ALOS PALSAR PSI displacement rates
were plotted on the maps with selected geomorphological features
on Fig. 8. Negative values (red colors) indicate an increase in the
distance from target to satellite, i.e., a lowering of the surface. On
the northwest-facing side of the valley, ENVISAT ASAR PSI results
(Fig. 8a) indicate stability (i.e., displacement values below
2 mm year−1
) for large sections of the slope above the lake and
movements with a rate of a few millimeter per year near the
moraine crest above the temporarily inactive landslide (4 on
Fig. 8a). Because, in the field, this area shows no cracks or scarps
related to possible future sliding, the PSI signal might be related to
the movement of superficial material. ALOS PALSAR PSI results
(Fig. 8b) have a larger noise than results based on ENVISAT ASAR,
because of the smaller number of acquisitions available and the
larger wavelength. Nevertheless, the coverage with valid information
is much larger, confirming past interferometric studies
(Strozzi et al. 2005) that found that more spatially complete information
can be generally found with L-band with respect to Cband.
In addition, ALOS PALSAR images were acquired with a
larger incidence angle than ENVISAT ASAR ones (∼35° vs. ∼23°),
and therefore, areas masked by layover and shadow are smaller,
resulting in a better coverage with valid points on both sides of the
valley. ALOS PALSAR PSI results (Fig. 8b) of the ascending orbit
indicate stability (i.e., displacement values below 5 mm year−1
) for
large sections of the slopes above the lake and a few sectors with
displacement rates approaching 10 mm year−1
. One of these sectors
on the southeast moraine corresponds to the potentially active
landslide detected on the geomorphological map (5 on Fig. 8b).
Other areas with displacement rates slightly below 10 mm per year
were identified on or close to a temporarily inactive landslide (4
on Fig. 8b), close to the terminus of the lake (10 on Fig. 8b), and
between two temporarily inactive landslides on the southeastFig.
7 Electric resistivity tomography profiles constructed in direction along slope (1–3) and across slope (4 and 5), sp possible sliding plane of secondary landslide, t
tension fissures, c relatively drier and coarse moraine material, w moraine sediment with higher water content and also probably with finer particles, bb large boulders
with opened spaces among them and bedrock; question mark indicates features, which were not verified in a field and are only hypothesized
Original Paper
facing moraine slopes (9a on Fig. 8b). In the cases 4, 9a, and 10
(Fig. 8b), high-activity areas are on places where mainly movement
of superficial material can occur, often related to the pronounced
rills. The PSI results do not indicate significant movements along
any of the cracks or scarps found in the field related to possible
future sliding (5 on Fig. 8). The confirmation of no activity along
geomorphologically identified scarps and potential scarps of landslides
is very useful information with respect to their hazard
assessment.
Areas without PSI points are identified not only for water,
snow/ice, and layover/shadows, as expected, but also on many
parts of the moraine. For instance, on the scarp of the recently
active landslide (5 and 6 on Fig. 8a), no ENVISAT ASAR PSI points
were detected, and on large sections of the moraine on the northwest
side of the lake, there are no ALOS PALSAR PSI points (3 and
8 on Fig. 8). Because the presence of snow cover or vegetation can
be excluded, decorrelation on PSI results might indicate the presence
of movements more rapid than a few centimeters per year. In
these cases, interpretation of single SAR interferograms can be
useful for a more complete picture of the displacements. Using
differential InSAR (DInSAR) analysis, we identified zones with
displacement rates of 100 to 500 mm year−1
classified according
to Barboux et al. (2014, see Online Resource 1). Three such zones
with major coherent phase signals were detected between 2007 and
Fig. 8 Results of ENVISAT ASAR (a) and ALOS PALSAR (b) PSI displacement rates are shown along landslides from geomorphological map. Negative values (red colors)
indicate an increase in the distance from target to satellite, i.e., a lowering of the surface (for explanations of numbers 1–10, see the text). Background image is August
2013 Google Earth satellite image
2010 over the northwest side of the lake (yellow zones on Fig. 9).
Field investigation suggested that, in some cases, displacements
reflected mainly processes that do not generate landslide morphology
(zone 11 on Fig. 9) and are represented by near continuous
down-slope transport of superficial sediments and their accumulation
on the talus of the inner slope. In other places (zone 12 on
Fig. 9), it corresponded well to a shallow landslide identified
during geomorphological mapping, and thus, it reflected its movement
activity (compare with 8 on Fig. 4). Another major signal was
detected in 1998 on the northwest-facing slope of the lake (13 on
Fig. 9) with ERS SAR data over the potentially active landslides
and landslide from 2003. It could be considered as a possible
movement precursor signal for the 2003 landslide. After 2003,
ENVISAT ASAR interferograms taken over 35 days or more show
decorrelation in this area, which was probably caused by major
activity of this landslide following its occurrence in March 2003.
Unfortunately, the 11-day TerraSAR-X interferograms, which are
characterized by a very high resolution of 3 m and could have been
very useful for our investigations (Notti et al. 2010), had a very
large baseline of more than 500 m and were not useful for our
study. Over the area of the 2003 landslide, the TerraSAR-X interferogram
seemed to be decorrelated as well. In yearly DInSAR
interferograms, decorrelation was observed for large areas of the
moraine (green zones on Fig. 9). Remaining patches of snow or
snow avalanches could have caused the decorrelation detected
close to the glacier terminus. But, in other areas (zones 14 and 15
on Fig. 9), mobility of superficial material and landslide activity
may be responsible for the signal decorrelation.
Slope stability calculations of potentially active landslides
A factor of safety was calculated using a fully specified slip surface,
as well as a slip surface defined by calculation as kinematically best
suited for the development of the studied landslide. These surfaces
were almost identical (Fig. 3). Calculations of the factor of safety
(Fs) considering the strength in a critical state for dry material
resulted in values below the limiting state—Fs = 0.829. This
was not surprising, as the values of the angle of internal
friction were lower than the actual slope dip of the modeled
profile (Novotný and Klimeš 2014). Therefore, we tested the
slope stability by considering the peak strength, which probably
governed the recent slope stability conditions. The calculated
Fs of the dry, cohesionless, and homogeneous moraine
material fraction of up to 22-mm grains was very close to 1
(Table 2), which was unrealistic considering the recent steep
topography. Therefore, we introduced an apparent cohesion
into the calculations, which accounted for the stabilizing effects
of large boulders and their sedimentation characteristics,
as suggested by previous work (Novotný and Klimeš 2014). To
obtain, a Fs = 1.1 (above the stability threshold) of dry material
required using an effective cohesion of 10 or 12 kPa
(Table 2). It showed for the dry material that the slope was
stable, but still very sensitive to any factors possibly lowering
its stability (e.g., water saturation). Sensitivity calculations for
varying degrees of water saturation were carried out. For
example, the factor of safety was close to the limiting value
1 in the model case when 10 % of material above the slip
surface was fully saturated. A further increase of water content
above the slip surface mathematically reduces the factor
of safety to below 1. For example, in the model case, when
50 % of material above the slip surface was fully saturated,
the factor of safety was reduced to 0.65, for which the slope
was very unstable (Table 2).
The effects of seismicity are an important destabilizing factor
for slopes. Since the study area could be subject to earthquakes, we
Fig. 9 Results of the DInSAR analysis interpretation (for additional information, see Online Resource 1). Yellow outlines define zones with detected movement rates of
10–50 cm year−1
with years when the major displacements were detected. Areas on the orographic right slopes of the Palcacocha Lake were detected with ALOS PALSAR
image pairs with a time interval of 46 days and ENVISAR ASAR image pairs with a time interval of 35 days. The area on the left moraine slope was identified using ERS SAR
image pairs with 35-day interval. Areas with decorrelation observed in yearly or monthly DInSAR interferograms are shown by green lines. The signal decorrelation could
be caused by snow patches near the glacier or by large mobility of superficial material or landslide activity. Extent of the Palcacocha Lake in 2003 is shown by red
outline. Glacier extents from Randolph Glacier Inventory Version 4.0 (Arendt et al. 2014) are shown by pink lines. Background image is August 2013 Google Earth
satellite image
Original Paper
performed a series of calculations showing the effects of seismicity
on slope stability under different water saturation conditions
(Fig. 10). They show the high sensitivity of the factor of safety to
the seismic kh coefficient. In the case of a dry slope with increased
cohesion (c = 10–12 kPa), Fs drops to 1 when a kh of only 0.05 is
applied. This value of kh corresponds to earthquake magnitudes
from Mw 5 to Mw 7 (Idriss 1985; STN 730036).
Ice support of the adjacent slopes is a well-recognized factor
contributing to slope stability, and increased slope movement is
often related to glacier retreat (Evans and Clague 1994; Holm et al.
2004; Hugenholtz et al. 2008; Oppikofer et al. 2008). Therefore, we
calculated a theoretical height above the recent water level of the
glacier within the terminal moraine, which would increase the
landslide stability to over a threshold value of Fs = 1. With increasing
ice thickness, slope height above the ice decreased, which has a
major effect on slope stability increase. From a variety of possible
slope stability conditions, we decided to present results of two
scenarios representing possible worst-case scenarios for the studied
slope. The apparent cohesion was always set to 12 kPa. The first
scenario was defined by ru = 0.05 (10 % of material above slip
surface was fully saturated) and kh = 0.1 (approximately Mw 6–7,
Idriss 1985; STN 730036). The second scenario had ru = 0.25 (50 %
of material above slip surface was fully saturated) and kh = 0 (no
seismicity). The first scenario with applied seismicity and less
water above the slip surface (scenario 1 on Fig. 11) showed that
the factor of safety reached the threshold value when the slope
height above the glacier surface was 35 m or less (corresponding to
a glacier thickness of 95 m above the recent lake water level at
4562 m a.s.l.). Calculations which applied a greater water portion
above the slip surface and no seismicity (scenario 2 on Fig. 11)
showed that the slope stability reached the Fs = 1 when the slope
height was only 25 m, corresponding to a glacier thickness of
110 m. A significant re-advance of the glacier at Palcacocha Lake
can be excluded; nevertheless, these sensitivity calculations indicated
that unstable conditions in the moraine slopes of Palcacocha
Lake probably existed already during periods of ice-surface lowering,
when the glacier was still present within the studied
moraines.
Hydrodynamic model of landslide-induced impact waves
Results of the Iber model provided a relative measure of the
potential of a landslide to trigger impact waves of a specific size
related to landslides of specific dimensions (Table 3). Landslide
volume and impact wave height triggered by the 2003 compound
landslide (1 on Fig. 4) were used for back calculations, enabling the
best-fit estimation of landslide velocity entering the lake to be
7 m s−1
. Subsequently, the potentially active compound landslides
were modeled (Table 3), showing that the impact wave caused by
the main potentially active landslide (5 on Fig. 4) could reach 8 m
in height, with a maximum flow of 29,400 m3
s−1
before reaching
the reinforced dam. Its magnitude would be very similar to the
wave triggered by the 2003 compound landslide. The calculated
impact wave caused by the smaller, secondary landslide of the
potentially active compound landslide (2a on Fig. 4) would not
overcome the existing artificial dams, as its maximum height was
only 1 m, while the dam freeboard is 7 m. This result reflected the
small estimated volume of the secondary landslide. The height
similarity between impact waves caused by the 2003 compound
Table 2 Factor of safety calculated using peak strength
Water content Factor of safety
c = 0 c = 10–12 kPa
Dry 0.998/0.972 1.089–1.107
10 % of material above slip
surface is fully saturatedb
– 0.998–1.015
50 % of material above slip
surface is fully saturateda
– 0.640–0.657
a
Water content of the material above sliding surface was characterized by pore pressure
coefficient ru = 0.25
b
Water content of the material above sliding surface was characterized by pore pressure
coefficient ru = 0.05
Fig. 10 Changes of the factor of safety (Fs, y-axis) due to increasing seismic loading (kh, x-axis), 1a—dry with c = 0 kPa, 1b—dry with c = 12 kPa, 2a—10 % of material
above slip surface is fully saturated with c = 0 kPa, 2b—10 % of material above slip surface is fully saturated with c = 12 kPa, 3a—50 % of material above slip surface is
fully saturated with c = 0 kPa, 3b—50 % of material above slip surface is fully saturated with c = 12 kPa
landslide and the potential failure of the newly described landslide
suggested that also a glacial lake outburst flood due to the landslide
failure would have a similar magnitude to the one in 2003. In
this case, the outburst flood effects were only minor and restricted
to minor damage to the drinking water treatment facilities located
very close to the Cojup River. The ongoing drainage works, using
siphons at Palcacocha Lake, might even be sufficient to prevent a
spillover of such a small-magnitude outburst flood, since the lake
level (field visit in April 2015) was about 3.4 m below the artificial
outlet originally designed to keep the lake water level at
4562 m a.s.l.
Discussion
Slope stability and landslide occurrence
Debuttressing of debris-mantled slopes (i.e., removal of their toe
support) significantly contributes to landslide occurrence
(Haeberli et al. 1997; Hugenholtz et al. 2008) but may be compensated
for by post-glacial slope development. In this case, formation
of talus has reduced the slope angle from about 55° in the upper
part of the inner moraine slope to about 35° at its toe at Palcacocha
Lake. Also, our slope stability calculations showed that significant
ice buildup would be necessary to push the slope stability above
the critical threshold of the factor of safety. In the scenario in
which only water infiltration was considered as the destabilizing
factor (i.e., 50 % of material above slip surface was fully saturated),
the stability of the modeled slope dropped below Fs = 1 (Fig. 11)
when the slope height above the glacier surface was more than
25 m. In other words, our calculation results showed that when the
glacier thickness dropped to below 110 m or less above the recent
water level, the slope started to be highly unstable under the
specified hydrological and recent morphological conditions. The
low slope stability was confirmed by the landslide mapping, which
identified repeated and localized landslide activity, mostly on the
northwest-facing moraine slope, as well as PSI SAR data analysis
showing high decorellation explained by superficial movements
exceeding a few centimeter per year. The later process does not
include landsliding defined by development of sliding planes and
landslide forms, but rather individual movements of large boulders
or removal of material by sheet and rill erosion on steep
slopes. On the other hand, the fact that landslide occurrence in the
steep part of the moraine slopes was not very frequent, at least
based on available landslide inventory data, and the recorded
recent activity was spatially limited suggests that landslide initiation
was strongly governed by specific slope stability conditions
(e.g., buried ice, concentrated water flow, site-specific properties of
moraine material) as well as by the occurrence of specific triggering
events (e.g., precipitation or earthquakes). Comparable results
for landslide distribution stimulated regional-scale investigation
of the role of post-Little Ice Age (LIA) Neoglacial retreat on
landslide activity in British Columbia (Holm et al. 2004). This
study suggests that complex landslide occurrence patterns are
determined by the type of material, relief morphology, and availability
of suitable sediments, which is affected by glacier aggradation
and removal processes.
The slope stability calculations were done with good knowledge
of the basic mechanical and strength properties of the modeled
material, which was sampled directly from the scarp area of the
potentially active landslide. The large shear box tests performed on
the landslide material allowed us to define peak shear strengths as
function of normal stresses, representing different depths below
landslide surface. This information was used in the slope stability
calculations. Despite such detailed input data, the calculation
results need to be carefully evaluated, since some additional parameters
still could only be estimated. Among them were the
effects of large block wedging (Novotný and Klimeš 2014) or
moraine sedimentation fabrics that tend to parallel the outer
moraine slope surface (Lucas and Sass 2011) and which were also
confirmed by field observations at the Palcacocha Lake moraine.
Shear planes on the inner slopes have to cut across this
Fig. 11 Changes of the factor of safety due to increasing height of the glacier,
scenario 1—c = 12 kPa, ru = 0.05 (10 % of material above slip surface is fully
saturated), and kh = 0.1; scenario 2—c = 12 kPa, ru = 0.25 (50 % of material
above slip surface is fully saturated), and kh = 0
Table 3 Characteristics of the main input parameters and results of the Iber hydrodynamic model
2003 Compound landslide Potentially active compound landslide
Main landslide Secondary landslide
Landslide volume (m3
) 756,000 630,000 30,600
Landslide depth (m) 17.2 20 8
Qmax (m3
s−1
) 23,000 29,400 2500
Wh (m) 8 8 1
The results of the Iber hydrodynamic model are in italic. Landslide velocity at entry to the lake was estimated to 7 m s−1
wh maximum height of the impact wave before reaching the dam crest, Qmax maximal flow of the highest modeled wave before reaching the dam crest
Original Paper
sedimentation structure, where rock blocks oriented parallel to the
outer slope may cause apparent cohesion of the moraine material.
Another factor which could increase the stability of the moraine
slopes under natural conditions was suction (compare with
Springman et al. 2003). Its action at the study site was witnessed
during field work on wet moraine material at shallow depths
(already 0.3 m under the surface) even during dry season. All these
factors were accounted in the calculations by introducing an
apparent cohesion of 10–12 kPa. Nevertheless, there is uncertainty
in how well these additional stability factors are reflected by these
values. In the case of the 2003 compound landslide, water pressure
generated by the spring in the crown area was probably the factor
that possibly lowered the slope stability (compare with URGH,
2003 Informe técnico: Estado situacional de la laguna Palcacocha
(in Spanish). ANA, p. 19) and was not reflected in the stability
calculations.
Historical and recent geomorphological mapping showed
that landslides in the last 60 years occurred mainly on the
southeast lateral moraine of Palcacocha Lake, close to the
glacier. The strong persistence of landsliding in this location
was probably caused by the availability of surface and subsurface
water, which was being supplied by a small stream
coming from the higher parts of rock slopes of the trough.
The high water availability at this site was due to the fact that
the moraine was underlain by a rock slope at relatively
shallow depths (evidenced by rock outcrops seen in the active
debris flow source area, Fig. 5a). Thus, this slope could be
classified as a Brock-controlled slope,^ which represented terrain
most associated with surficial landslides in areas affected
by post-Little Ice Age glacier retreat (Holm et al. 2004). The
importance of water saturation on triggering landsides in
moraines was further supported by examples of landslides
which occurred after periods of heavy precipitation (e.g.,
Santa Teresa, Cuzco in Klimeš et al. 2007 or Giráldez et al.
2013) or due to water seepage through lateral moraine
(Cangrajanca Lake, Cordillera Huayhuash in Engel et al.
2011). Some other works do not mention this factor (e.g.,
Hugenholtz et al. 2008), but local slope morphology inferred
from published aerial photos suggested that water concentration
from slopes adjacent to the moraine could play a role in
landslide mobilization.
SAR analysis and landslide identification
The PSI analyses of ALOS PALSAR images showed displacement
rates approaching 10 mm year−1
on the southeast lateral
moraine, corresponding to the area of potentially active
landslide detected on the geomorphological map (5 on
Fig. 8b). Moving PSI points were nevertheless grouped over
the landslide body, and no displacement was detected on the
PSI over the identified scarps of the landslide described by
field mapping as potentially active in the future. This could
be explained by recent activity of the landslide scarp area
being too small to be detected by the SAR technology (in
particular, considering the unfavorable orientation of the
ALOS PALSAR with respect to the slope orientation), or the
possible displacements along the scarp were not captured by
the available time series of ALOS PALSAR images (2007–
2011). These findings, along with repeated field observations
(2011–2014) which revealed no major activity, represented
positive information with respect to the probability of rapid
future reactivation, which was more likely to occur in cases
when nearly continual deformation occurred. Field evaluation
of other PSI signals detected on the moraine with both
ENVISAT ASAR and ALOS PALSAR analyses attributed the
deformation to movements of superficial material or to a
compaction of the moraine. Two zones of major coherent
phase signals detected by DInSAR interferograms on the
northwest side of the moraine, with displacement rates of
10 to 50 mm year−1
(yellow zones on Fig. 9), illustrated
advantages of mutual PSI and DInSAR data interpretation.
These signals were attributed to both movement of superficial
materials (11 on Fig. 9) and landsliding (12, 13, 14, and 15
on Fig. 9), which showed the importance of complementing
SAR interferometry with field geomorphological mapping, in
particular, when the proper mechanism behind the detected
displacements needed to be defined. This notion was further
supported by the fact that the movements detected by the
PSI as well as the DInSAR analyses generally coincided with
landslide occurrence on the southeast moraine. The Bprecursor^
role of movements detected by DInSAR analysis may be considered
only if it is detected at places with favorable conditions
for landslide initiation. So far, it seems that such additional
conditions were limited to the March 2003 landslide (6 on Fig. 4)
and its close surroundings and were not detected on the northwest
moraine.
Glacial lake outburst floods triggered by moraine landslides
Field investigation and calculations from Palcacocha Lake, as
well as experience from historical events, showed that landslides
on the moraine slopes with the possibility of entering
the lake could produce impact waves responsible for initiating
glacial lake outburst floods. The magnitudes of these outburst
floods seemed to be limited, and similar results obtained for
the 2003 event as well as the new landslides scenarios suggested
that floods with only a small extent would be produced
by the studied moraine landslides under the specified conditions
(note that the actual water level was 3.4 m lower than
the one used in the hydrodynamic model). These contain the
landslide volume and velocity at the entry to the lake, as well
as shape of the lake basin and location of the material entry
into the lake. The reliability of the resulting model depended
on how reliably determined these input variables were. The
shape of the lake basin and location of the material entering
into the lake were reliably described by available topographic
data and field mapping. Higher uncertainty was related to
estimation of the landslide volume determined from the landslide
depth, which was reliably known only along the ERT
profile. This, however, is a large improvement compared to
other impact wave models (Schneider et al. 2014), in which
the volume was estimated by experts using remotely sensed
data. The velocity at the entry into the lake was determined
by the best-fit back calculation from the 2003 landslide event
still providing reliable results.
Simulation of the impact of an ice avalanche (SomosValenzuela
et al. 2014), with a volume of 5 × 105
m3
comparable
to our scenarios (Table 3), with the comparable higher velocity of
20 m s−1
entering Palcacocha Lake at its very beginning and
travelling along its longitudinal axis, results in an estimated
maximum height of the impact wave of 9 m. In this case, the
volume of the material was set arbitrary, using literature values.
When the ice avalanche volume is increased to 1 × 106
m3
, the
impact wave more than doubles to 21 m. The different conditions
of the ice avalanche and landslide simulations, as well as the
significant increase of the impact wave height when increasing
the total volume of the ice avalanche, suggested that the relatively
small volume of the avalanche and landslide material was responsible
for similar impact wave heights of 8 and 9 m respectively.
While in the case of the ice and ice/rock avalanches coming from
the glaciated parts of the mountain ridges (Schneider et al. 2014), it
is very difficult to reliably estimate possible volumes of the past as
well as future avalanches, volumes of the landslides from the
moraine slopes can be constrained more objectively, as shown in
this study. Even without detailed field information, the possible
variability in the landslide volumes was highly constrained by the
moraine morphology and total volume of available material.
Therefore, we assumed that glacial lake outburst floods resulting
from landslides in moraines will be probably smaller than outburst
floods triggered by other mass movements (Fig. 1), which may
involve much larger volumes of material entering the lake at
higher velocities and at a different angle with respect to the
lake longitudinal axis. This assumption has direct implications
for glacial lake outburst flood hazard assessment. Considering
that the calculated wave heights correspond to about the
height of the dam above the artificial outlet channel, the
progress of ongoing measures to lower lake water level has
an immediate effect on the potential hazard. In April 2015, the
lake level was already lowered to about 3.4 m below the
original level, which, according to the hydrodynamic impact
wave modeling, would already be sufficient to prevent a
spillover of the dam for both the 2003 landslide and the
future landslides.
Examples of documented glacial lake outburst floods triggered
by landslides from inner moraine slopes listed in the
caption of Fig. 1 suggest that Palcacocha Lake is not the only
lake in the Cordillera Blanca that has favorable conditions to
trigger outburst floods by moraine landslides. So far, there
has been no detailed regional evaluation of moraine slope
stability of glacial lakes focusing on this phenomenon, but
there are several well-known examples of lakes susceptible to
similar glacial lake outburst floods within the Cordillera
Blanca region (e.g., Jancarurish—8° 51′ 32″ S, 77° 40′ 22″ W;
Arhueycocha—8° 53′ 17″ S, 77° 37′ 44″ W; Allicocha—9° 15′ 03″
S, 77° 27′ 33″ W; Pag Pag—9° 04′ 04″ S, 77° 33′ 20″ W;
Tullpacocha—9° 25′ 10″ S, 77° 20′ 27″ W). Landslides similar
to those described in this study may have occurred at these
other lakes, but if they triggered only small magnitude outburst
floods, these may not have been recorded. Despite the
presented evidence and assumptions about glacial lake outburst
flood magnitudes, it could be very dangerous to underestimate
future outburst floods triggered by landslides from
moraines. Large-volume or high-velocity landslides from moraines
on sites with different morphological and geological
settings could initiate waves capable of even breaching the
moraine dams, possibly causing large glacial lake outburst
floods. Climate-driven environmental changes may critically
affect stabilities of slopes above glacial lakes, possibly
triggering large moraine landslides. Thus, continuous remote
sensing and field-based monitoring and related hazard assessments
(e.g., Reynolds 2003; Klimeš 2012; Emmer and Vilímek
2014) are important steps in risk management and effective
mitigation of glacial lake outburst flood hazards. They should
also consider the impacts of ongoing environmental change
(e.g., glacier retreat, permafrost degradation, and associated
slope response, mainly water infiltration to the lateral moraines).
Structural measures can help to decreasing the susceptibility
of moraine-dammed lakes to produce outburst
floods caused by overtopping or even breaching of the moraine
dams. Effective mitigation strategies should increase
dam freeboard and reinforce the lake outlet, i.e., especially
creating artificial dams, in combination with open cuts or
digging outflow tunnels (Emmer et al. 2016). The majority of
large moraine-dammed lakes of the Cordillera Blanca have
some of these structural measures; nevertheless, the 2003
outburst flood from Palcacocha Lake showed that floods
resulting from moraine sliding into the lake may occur in
already remediated lakes, which thus should not be excluded
from continuous hazard re-assessment. Such mitigation works
typically significantly reduce the hazard potential but do not
reduce it to zero. As an alternative to construction works,
other measures can be applied to lower the risks of floods, by
decreasing the vulnerability of the potentially affected population,
i.e., improving their preparedness and response in the
event of an outburst. Examples for such so-called non-structural
measures are early warning systems (EWSs) to alert the
population in case of a flood. In the Cordillera Blanca, such a
EWS has recently been successfully implemented at Lake 513
(Frey et al. 2014), which also experienced a glacial lake outburst
flood in 2010, despite structural mitigation measures
(Klimeš et al. 2014; Schneider et al. 2014).
Conclusions
Occurrences of landslides on inner slopes of lateral moraines as
possible triggers of glacial lake outburst floods from Palcacocha
Lake have been investigated. Detailed field morphological and
geophysical data are required to reliably assess volumes of
landslides causing impact waves. Proper estimation of the
volume of material entering the lake is important to substantially
increase the reliability of results of hydrodynamic wave
modeling. Local topography is also important in the slope
stability calculations that show that slopes of the lateral moraines
are close to the stability threshold and are sensitive
mainly to the water content. Nevertheless, the calculations
leave open questions about stabilizing factors, which ensure
the temporal stability of the moraine slopes, even during
potential triggering events of high intensity (e.g., highmagnitude
earthquakes). The presence of glaciers is often
regarded as a stabilizing factor for adjacent slopes. Parametric
stability calculations showed that considerable thickness of
glacier support of the slope is required to increase its stability
above the threshold limit under recent morphological conditions.
This suggests that large portions of moraine slopes were
highly unstable even during earlier periods of glacier thinning.
This finding, along with the documented strong spatial persistence
of landslides on the Palcacocha lateral moraine,
Original Paper
shows that site-specific conditions related to water saturation
are important factors for landslide occurrence in the inner
slopes of exposed moraines.
Back calculation of the impact waves triggered by a historical
landslide and landslide scenarios based on field evidence
and slope stability calculations suggest similar magnitudes for
past and potential future impact waves. Subsequent comparison
with the 2003 flood suggests that floods of only small
extent may be produced by the studied landslides in moraine
under the specified conditions of Palcacocha Lake. Ongoing
efforts to lower the lake level and increase the dam freeboard
may already be sufficient to prevent a spillover at the dam in
case of a landslide impact. Findings suggest that floods triggered
by landslides in moraines are smaller than floods
caused by other slope processes (e.g., ice/rock avalanches)
due to the generally smaller entry velocities and volumes of
landslides from the lateral moraines. This assumption has to
be critically evaluated against site-specific conditions at a
given lake (e.g., possible dam breach) and any possible environmental
changes that may cause increased water content in
the moraine material. Additional water input into the moraines
may mobilize larger volumes, possibly causing higher
impact waves and thus posing a higher glacial lake outburst
flood hazard.
Acknowledgments
The authors wish to acknowledge the financial support provided
by the Czech Science Foundation (Grant No. P209/11/1000), Grant
Agency of Charles University (GAUK project no. 70 413; GAUK
project no. 730 216), the European Space Agency (S:GLA:MO
project), and the Swiss Agency for Development and Cooperation
(SDC) (Proyecto Glaciares). This work was carried out thanks to
the support of the long-term conceptual development research
organisation RVO: 67985891. ERS and ENVISAT SAR data courtesy
of C1F.6504, © ESA. ALOS PALSAR © JAXA. TERRASAR-X data
courtesy HYD0562, © DLR. We thank John M. Reynolds for a very
detailed and helpful review as well as Matt Rowberry and Christian
Huggel for their valuable comments.
References
Arendt A, Bliss A, Bolch T, Cogley JG, Gardner AS, Hagen JO, Hock R, Huss M,
Kaser G, Kienholz C, Pfeffer WT, Moholdt G, Paul F, Radić V, Andreassen L,
Bajracharya S, Barrand N, Beedle M, Berthier E, Bhambri R, Brown I, Burgess E,
Burgess D, Cawkwell F, Chinn T, Copland L, Davies B, De Angelis H, Dolgova E,
Filbert K, Forester R, Fountain A, Frey H, Giffen B, Glasser N, Gurney S, Hagg
W, Hall D, Haritashya UK, Hartmann G, Helm C, Herreid S, Howat I, Kapustin
G, Khromova T, König M, Kohler J, Kriegel D, Kutuzov S, Lavrentiev I, LeBris R,
Lund J, Manley W, Mayer C, Miles ES, Li X, Menounos B, Mercer A, Mölg N,
Mool P, Nosenko G, Negrete A, Nuth C, Pettersson R, Racoviteanu A, Ranzi R,
Rastner P, Rau F, Raup B, Rich J, Rott H, Schneider C, Seliverstov Y, Sharp M,
Sigurðsson O, Stokes C, Wheate R, Winsvold S, Wolken G, Wyatt F, Zheltyhina
N (2014) Randolph Glacier Inventory—a dataset of global glacier outlines:
version 4.0. Global Land Ice Measurements from Space, Boulder Colorado,
Digital Media
Bamler R, Hartl P (1998) Synthetic aperture radar interferometry. Inverse Probl 14:R1R54.
doi:10.1088/0266-5611/14/4/001
Barboux C, Delaloye R, Lambiel C (2014) Inventorying slope movements in an Alpine
environment using DinSAR. Earth Surf Process Landf 39:2087–2099. doi:10.1002/
esp.3603
Bareither CA, Edil TB, Mickelson DM (2008) Geological and physical factors
affecting the friction angle of compacted sands. J Geotech Geoenvironmental
Eng 134:1476–1489
Beló M, D’Agata C, Smiraglia C, Plefini M (2006) Ice core moraine collapse at Froni Glacier
(Italian Alps): a case of tourist risk. Geophys Res Abstr 8 (08361)
Bolton MD (1986) The strength and dilatancy of sands. Géotechnique 36:65–78
Carey M (2010) In the shadow of melting glaciers: climate change and Andean society.
Oxford University Press, p 288
Carey M (2005) Living and dying with glaciers: people’s historical vulnerability to
avalanches 524 and outburst floods in Peru. Global Planet Change 47(2–
4):122–134
Cigna F, Bianchini S, Casagli N (2012) How to assess landslide activity and intensity with
persistent scatterer interferometry (PSI): the PSI-based matrix approach. Landslides.
doi:10.1007/s10346-012-0335-7
Costa JE, Schuster RL (1988) The formation and failure of natural dams. Geological
Society of America Bulletin 100(7):1054–1068. doi:10.1130/0016-
7606(1988)100<1054:TFAFON>2.3.CO;2
Crosetto M, Monserrat O, Bremmer C, Hanssen R, Capes R, Marsh S (2009)
Ground motion monitoring using SAR interferometry: quality assessment. Eur
Geol 26:12–15
Emmer A, Vilímek V, Klimeš J, Cochachin A (2014) Glacier retreat, lakes development and
associated natural hazards in Cordillera Blanca, Peru. In: Shan W, Guo Y, Wang F,
Marui H, Strom A (eds) Landslides in cold regions in the context of climate change,
Springer, pp 231–252
Emmer A, Cochachin A (2013) The causes and mechanisms of moraine-dammed lake
failures in the Cordillera Blanca, North American Cordillera and Himalayas. AUC
Geograph 48:5–15
Emmer A, Vilímek V (2014) New method for assessing the susceptibility of glacial lakes
to the outburst floods in the Cordillera Blanca, Peru. Hydrol Earth Syst Sc 18:3461–
3479. doi:10.5194/hess-18-3461-2014
Emmer A, Loarte EC, Klimeš J, Vilímek V (2015) Recent evolution and degradation
of the bent Jatunraju glacier (Cordillera Blanca, Peru). Geomorphology
228:345–355
Emmer A, Vilímek V, Zapata ML (2016) Hazard mitigation of glacial lake
outburst floods in the Cordillera Blanca (Peru): the effectiveness of remedial
works. J Flood Risk Manage, not yet assigned to an issue. doi:10.1111/
jfr3.12241
Engel Z, Česák J, Rios Escobar V (2011) Rainfall-related debris flows in Carhuacocha
Valley, Cordillera Huayhuash, Peru. Landslides 8:269–278
Evans SG, Clague JJ (1994) Recent climatic change and catastrophic geomorphic
processes in mountain environments. Geomorphology 10:107–128
Farina P, Colombo D, Fumagalli A, Marks F, Moretti S (2006) Permanent
scatterers for landslide investigations: outcomes from the ESA-SLAM project.
Eng Geol 88:200–217
Fischer T, Lentschke J, Küfmann C, Haas F, Baume O, Becht M, Schröder H (2013) Highmountainous
permafrost under continental-climatic conditions: actual results of
different mapping methods and an empirical-statistical modeling approach for the
Northern Tien Shan. Geoph Res Abs 15:EGU2013–13074
Frey H, García-Hernández J, Huggel C, Schneider D, Rohrer M, Gonzales Alfaro C,
Muñoz Asmat R, Price Rios K, Meza Román L, Cochachin Rapre A, Masias
Chacon P (2014) An early warning system for lake outburst floods of the
Laguna 513, Cordillera Blanca, Peru. In: Proceedings of the International
Conference on the Analysis and Management of Changing Risks for Natural
Hazards. 18–19 November 2014, Padua, Italy
Giráldez C, Choquevilca W, Fernández F, Frey H, García J, Haeberli W, Huggel C, Ludena S,
Rohrer M, Suarez W (2013) Large mass movements related to deglaciation effects in
southern Peru (Cusco). Geoph Res Abs 15:EGU2013–8183
Haeberli W, Wegmann M, Vonder Muhll D (1997) Slope stability problems related to
glacier shrinkage and permafrost degradation in the Alps. Ecologae Geologicae
Helvetiae 90:407–414
Holm K, Bovis M, Jakob M (2004) The landslide response of alpine basins to post-Little
Ice Age glacial thinning and retreat in southwestern British Columbia. Geomorphology
57:201–216
Hölbling D, Füreder P, Antolini F, Cigna F, Casagli N, Lang S (2012) A semiautomated
object-based approach for landslide detection validated by persistent
scatterer interferometry measures and landslide inventories. Remote Sens
4:1310–1336
Hubbard B, Heald A, Reynolds JM, Quincey D, Richardson SD, Zapata-Luyo M, SantillanPortilla
N, Hambrey MJ (2005) Impact of a rock avalanche on a moraine-dammed
proglacial lake: Laguna Safuna Alta, Cordillera Blanca, Peru. Earth Surf Process Landf
30:1251–1264
Hugenholtz CH, Moorman BJ, Barlow J, Wainstein PA (2008) Large-scale moraine
deformation at the Athabasca Glacier, Jasper National Park, Alberta, Canada. Landslides
5:251–260
Huggel C, Gruber S, Korup O (2013) Landslide hazards and climate change in high
mountains. In: Shroder J, James LA, Harden CP, Clague JJ, Shroder J (eds) Treatise on
geomorphology. Academic Press, San Diego, pp 288–301
Iber (2010) Two-dimensional modeling of free surface shallow water flow, Hydraulic
reference manual, Iber v1.0. http://www.iberaula.es/web/index.php. Accessed 20
January 2014
Idris IM (1985): Earthquake ground motions. In: EERI Course on BStrong ground
motion—seismic analysis, Design and code issues,^ April 10th 1987, Pasadena,
California
Instituto Nacional de Defensa Civil (2011) Informe de peligro N° 003-12/05/2011/COENSINADECI/15:00
horas (Informe N° 01): Peligro por aluvión en el departamento de
Ancash, Huaraz-Peru: COEN-SINADECI, INDECI, Huaraz
Iribarren Anacona P, Mackintosh A, Norton KP (2014) Hazardous processes and events
from glacier and permafrost areas: lessons from the Chilean and Argentinean Andes.
Earth Surf Process Landf. doi:10.1002/esp.3524
Klimeš J, Vilímek V, Vlčko J (2007) Debris flows in the vicinity of the Machu Picchu
village, Peru. In: Sassa K, Fukuoka H, Wang F, Wang G (eds) Progress in landslide
science. Springer, pp 313–314
Klimeš J (2012) Geomorphology and natural hazards of the selected glacial valleys,
Cordillera Blanca, Peru. AUC Geograph 47:25–31
Klimeš J, Benešová M, Vilímek V, Bouška P, Rapre AC (2014) The reconstruction of a
glacial lake outburst flood using HEC-RAS and its significance for future hazard
assessments: an example from Lake 513 in the Cordillera Blanca, Peru. Nat Haz
71:1617–1638
Lliboutry L, Morales BA, Pautre A, Schneider B (1977) Glaciological problems set by the
control of dangerous lakes in Cordillera Blanca, Peru. I. Historical failures of morainic
dams, their causes and prevention. J Glaciol 18:239–254
Morgenstern NR, Price VE (1965) The analysis of the stability of general slip surfaces.
Géotechnique 15:70–93
Ng AH-M, Ge L, Li X, Abidin HZ, Andreas H, Zhang K (2012) Mapping land subsidence in
Jakarta, Indonesia using persistent scatterer interferometry (PSI) technique with ALOS
PALSAR. Int J Applied Earth Obser and Geoinform
Notti D, García-Davalillo JC, Herrera G, Cooksley G (2010) Assessment of the
performance of X-band satellite radar data for landslide mapping and
monitoring: Upper Tena Valley case study. Nat Hazards Earth Syst Sci
10:1865–1875
Novotný J, Klimeš J (2014) Grain size distribution of soils within the Cordillera Blanca,
Peru: an indicator of basic mechanical properties for slope stability evaluation. J Mt
Sci 11:563–577
Ojeda N (1974) Consolidacion laguna Palcacocha (in Spanish). Electroperu, Huarás
Oppenheim V (1946) Sobre las lagunas de Huaraz (in Spanish). Boletin Sociedad
Geologica Peru., pp 68–80
Oppikofer T, Jaboyedoff M, Keusen H-R (2008) Collapse at the eastern Eiger flank in the
Swiss Alps. Nature Geoscience 1:531–535. doi:10.1038/ngeo258
Panizzo A, De Girolamo P, Petaccia A (2005) Forecasting impulse waves
generated by subaerial landslides J Geophys Res Ocean 110:23.
doi:10.1029/2004jc002778
Portocarrero CAR (2014) The glacial lake handbook—reducing risk from dangerous
glacial lakes in the Cordillera Blanca, Peru. USAID technical report, p 80
Rabatel A, Francou B, Soruco A, Gomez J, C’aceres B, Ceballos JL, Basantes R, Vuille M,
Sicart JE, Huggel C, Scheel M, Lejeune Y, Arnaud Y, Collet M, Condom T, Consoli G,
Favier V, Jomelli V, Galarraga R, Ginot P, Maisincho L, Mendoza J, Menegoz M,
Ramirez E, Ribstein P, Suarez W, Villacis M, Wagnon P (2013) Current state of glaciers
in the tropical Andes: a multi-century perspective on glacier evolution and climate
change. The Cryosphere 7:81–102
Reynolds JM (2011) An introduction to applied and environmental geophysics. Wiley
Ltd., p 712
Reynolds JM (2003) Development of glacial hazard and risk minimisation
protocols in rural environments. Methods of glacial hazard assessment and
management in the Cordillera Blanca Peru. Reynolds Geo-Sciences Ltd.,
Flintshire, p 72
Richardson SD, Reynolds JM (2000) Degradation of ice-cored moraine dams: implications
for hazard development. IAHS Publications 264:187–197
Righini G, Pancioli V, Casagli N (2011) Updating landslide inventory maps using
Persistent Scatterer Interferometry (PSI). Int J Remote Sens 33:1–29
Rosen P, Hensely S, Joughin I, Li F, Madsen S, Rodriguez E (2000) Goldstein R (2000)
Synthetic aperture radar interferometry. Proc IEEE 88:333–382
Schneider-Muntau B, Zangerl C (2005) Numerical modelling of a slowly creeping
landslide in crystalline rock—a case study. In: Konečný P (ed) Impact of human
activity on the geological environment. Taylor & Francis Group plc, London, pp
535–540
Schneider D, Huggel C, Cochachin A, Guillén S, García J (2014) Mapping hazards from
glacier lake outburst floods based on modelling of process cascades at Lake 513,
Carhuaz, Peru. Adv Geosci 35:145–155
Simeoni L, Tarantino A, Mongiovi L (2003) Effects of unsaturation on the stability
of a moraine slope. In: Schanz T (ed) Unsaturated soils: experimental studies,
Proceedings of the International Conference, From Experimental Evidence
towards Numerical Modeling of Unsaturated Soils, Weimar, Germany., pp
497–508
Somos-Valenzuela M, Chisolm RE, McKinney DC, Rivas D (2014) Hazard mapping of a
potential glacial lake outburst flood in Huaraz, Peru. Center for research in water
resources, Online Report 14–01. http://www.crwr.utexas.edu/online.shtml. Accessed
21 November 2014
Springman SM, Jommi C, Teysseire P (2003) Instabilities on moraine slopes induced by
loss of suction: a case history. Géotechnique 53:3–10
Strozzi T, Wegmüller U, Tosi L, Bitelli G, Spreckels V (2001) Land subsidence
monitoring with differential SAR interferometry. Photogramm Eng Remote Sens
67:1261–1270
Strozzi T, Farina P, Corsini A, Ambrosi C, Thuring M, Zilger J, Wiesmann A,
Wegmüller U, Werner C (2005) Survey and monitoring of landslide displacements
by means of L-band satellite SAR interferometry. Landslides
2:193–201
Strozzi T, Ambrosi C, Raetzo H (2013) Interpretation of aerial photographs and satellite
SAR interferometry for the inventory of landslides. Remote Sens 5:2554–2570.
doi:10.3390/rs5052554
Sun Q, Zhang L, Ding Y, Hu J, Liang H (2015) Investigation of slow-moving landslides
from ALOS/PALSAR images with TCPInSAR: a case study of Oso, USA. Remote Sensing
7:72–88
Tamburini A, Del Conte S, Larini G, Lopardo L, Malaguti C, Vescovi P (2011)
Application of SqueeSARTM to the characterization of deep seated gravitational
slope deformations: the Berceto case study (Parma, Italy). In: Margottini C,
Canuti P, Sassa K (eds) Landslide science and practice, vol 2. Springer, New
York, pp 437–444, Early Warning, Instrumentation and Monitoring
Viero A, Teza G, Massironi M, Jaboyedoff M, Galgaro A (2010) Laser scanningbased
recognition of rotational movements on a deep seated gravitational
Original Paper
instability: the Cinque Torri case (North-Eastern Italian Alps). Geomorphology
122:191–204. doi:10.1016/j.geomorph.2010.06.014
Vilímek V, Zapata ML, Klimeš J, Patzelt Z, Santillán N (2005) Influence of glacial
retreat on natural hazards of the Palcacocha Lake area, Peru. Landslides
2:107–115
Vilímek V, Klimeš J, Červená L (2015) Glacier-related landforms and glacial lakes
in Huascarán National Park, Peru. J Maps doi. doi:10.1080/
17445647.2014.1000985
Wegmüller U, Werner C, Strozzi T, Wiesmann A (2003) Multi-temporal interferometric
point target analysis. In: Proceedings of the Multi-Temp Conference, Ispra, Italy, 16–
18 July 2003
Werner C, Wegmüller U, Strozzi T, Wiesmann A (2003) Interferometric point target
analysis for deformation mapping. In Proceedings of IGARSS, Toulouse, France, 21–25
July 2003
Westoby MJ, Glasser NF, Brasington J, Hambrey MJ, Quincey DJ, Reynolds JM (2014)
Modelling outburst floods from moraine-dammed glacial lakes. Earth-Sci Rev
134:137–159
Worni R, Huggel C, Clague JJ, Schaub Y, Stoffel M (2014) Coupling glacial lake impact,
dam breach, and flood processes: a modeling perspective. Geomorphology
224:161–176
Yamada T (1998) Glacier lake and its outburst flood in the Nepal Himalaya., Japanese
Society of Ice and Snow
Zapata ML (2002) La dinamica glaciar en lagunas de la Cordillera Blanca (in Spanish).
Acta Montana 19:37–60
Landslides 13 & (2016) 1477