Landslides (2012) 9:179–187
DOI 10.1007/s10346-011-0294-4
Received: 8 April 2010
Accepted: 24 August 2011
Published online: 10 September 2011
© Springer-Verlag 2011
Louis Bugnion I Brian W. McArdell I Perry Bartelt I Corinna Wendeler
Measurements of hillslope debris flow impact pressure
on obstacles
Abstract We present mea sure ments of hil lslope debris flow
impact pressures on small obstacles. Two impact sensors have
been installed in a real-scale experimental site where 50m
3
of
water-saturated s oil material are released from rest. Impact
velocities vary between 2 and 13m/s; flow heights between 0.3
and 1.0m. The maximum impact pressures measured over 15
events represent between 2 and 50 times the equivalent static
pressures. The measurement s reveal that quadratic velocity-
dependent formulas can be used to estimate impact pressures.
Impact coefficients C are constant from front to tail and range
between 0.4<C<0.8 according to the individual events. The
pressure fluctuations to depend on the sensor size and are
between 20% and 60% of the mean pressure values. Our results
suggest that hazard guidelines for hillslope debris flows should be
based on quadratic velocity-dependent formulas.
Keywords Hillslope debris flow
.
Field scale tests
.
Impact
pressure
.
Impact coefficient
Introduction
A long-standing problem in the study o f landslides is to
quantitatively understand the pressures they can exert o n
obstacles as a function of impact speed, flow height and debris
mixture properties. Simple and reliable formulae are required in
mitigation studies to delimit hazard zones and strengthen
buildings or dimension structures such as masts and pylons in
debris flow torrents and runout zones or where hillslope debris
flows (or open-slope debris flows) are expected. Ring-net barriers
(Wendeler 2008), a new method of debris flow mitigation, also
require estimates of impact pressure to adequately dimension
structural elements and anchors. Pressure formulae are also
helpful to analyse damage—both on man-made structures (Egli
2005) and trees (Stoffel and Bollschweiler 2009)—in historical
case studies and therefore can be used to establish magnitude–
frequency relations for landslide or debris flow activity in a
particular region. The problem is especially relevant because new
simulation tools are available that can predict flow speed and
height in three-dimensional terrain (e.g., Christen et al. 2010;
McDougall and Hungr 2004). Without accurate estimates of the
corresponding pressures, the potential of these new tools cannot
be fully realized.
Impact pressures generated by geophysical flows are compli-
cated because they depend both on the flow mixture (which
contains mud, rocks and large boulders and sometimes woody
debris and air) that varies from point to point in the surge
(Iverson et al. 2010) and with time and on the geometry and size
of the flow as well as of the obstacle. Because of the material
inhomogeneity, impact forces can fluctuate strongly over time.
The rheology of the mixture is important because it influences the
ability of the flow to redistribute stresses in the region of the
obstacle. The presence of a solid phase can increase local impact
pressure due to hard contact and material locking, a phenomenon
that is observed in dry granular flows (Levy and Sayed 2008) and
at the head of some debris flows.
In
this
report, we present measurements of flow height, front
and surface velocity and impact pressure of field-scale hillslope
debris flows (following the terminology of Hutchinson 1988 to
refer to unchannelized debris flows on a hillslope). The volume of
the material involved is 50 m
3
, the maximum flow height at the
location of the obstacles (30 m downstream) ranges between 0.3
and 1.0 m and the front speed at the location of the obstacles
ranges between 2 and 13 m/s. These observations are made in a
wide trapezoidal channel excavated into a hillslope (Fig. 1). The
hillslope debris flows we generate are roughly equivalent to small
debris flows on weakly channelized surfaces. Exampl es ar e
described by Imura and Shimojo (2007) in Japan (Table 1) and
Rickli and Bücher (2005) in Switzerland (Table 2). These field
studies typically involve post-failure documentation of the initial
landslide failure zone as well as the runout distance for natural
events—unfortunately, data on flow velocity and impact pressure
are rarely available. In some cases, rough estimates of flow
properties are available (e.g., Egli and Vanomsen 2005). The flows
described herein are roughly equivalent to the median volume,
flow height and slope angle value of Tables 1 and 2.
Our results indicate that impact pressures on large sub-
merged obstacles (flow heights of same order as the obstacles)
scale with the square of the flow velocity. This dependency is
interpreted in terms of the fraction of material that is stopped
during the impact process. Although not described herein, the
overall objective of this work was to investigate the performance
of flexible barriers against hillslope debris flows and small shallow
landslides, so considerable effort, involving evaluations by Swiss
landslide experts (including practitioners), was made to produce
realistic flows with a realistic sediment mixture and flow
behaviour.
Methods
A 41-m-long, 8-m-wide channel is constructed on the side of a
rock quarry near Veltheim, Canton Aargau, Switzerland (Fig. 1a).
Sediment deposits on the hillslope are excavated down to the
bedrock surface, which is a natural bedding plane parallel to the
land surface, with an average slope inclination of 30°. The channel
sidewalls are 1 m high and consist of soil material (which is
generally not entrained by the flows). At the upper channel end, a
wall is constructed out of wood beams and steel columns to create
a sloping reservoir (Fig. 1b) with a maximum capacity of 50 m
3
of
debris (approximately 100 metric tons). The retaining wall is 1.8 m
high; the lower 0.8 m operates as a trap door, hinged at the base,
that is held closed by cables. The cables are released to spill the
debris material down the slope. Because the reservoir is located
near a road, it can be filled by trucks which transport the material
from the mixing place. The material is obtained by mixing soil
and bedrock material from the quarry with water. Systematic and
reproducible variation of the flow mixture composition and water
Landslides 9 & (2012)
179
Original Paper
content is not possible. On one hand, the composition of the raw
materials from the quarry (bedrock and soil material) is expected
to vary over large volumes and from one test to another. On the
other hand, water loss during transport is inevitable and difficult
to quantify. For these reasons, the proportion of the raw materials
varies from one test to another and samples of material (see
below) are taken from the reservoir to assess mixture composition
and water content. The filling of the reservoir and the release take
place within 2 h to minimize settling and consolidation. Plastic
sheets are placed at the bottom of the reservoir to prevent water
loss and to ensure evacuation of all material. At the lower end of
the channel, a flexible steel mesh barrier stops the flow.
Upon release the material accelerates rapidly out of the
reservoir and flows down the slope. Deposition on the slope
begins immediately—fast flows deposit little material and accel-
erate over the entire channel length while slower flows deposit
more material and decelerate along the second half of the channel
(see Fig. 2). The volume stopped by the flexible barrier at the
Fig. 1 Channel excavated down to bedrock in the hillslope of a disused quarry
face (a) and material reservoir comprising the release zone (b). Note that the
downstream end of the channel ends at a flexible barrier, which artificially limited
the runout of the flows
Table 1 Summary of 669 events dated between 2001 and 2005 in Japan from
Imura and Shimojo (2007)
Min. Max. Median
Slope (°) 15 80 44
Failure depth (m) 0.2 5 1.2
Volume (m
3
) 25 500 66
Slope refers to the slope angle in the failure zone
Fig. 2 Lateral view of release 15.1 just after release (a) and at the time of impact
on the pressure plates (b)
Table 2 Summary of 133 events from year 2002 in Switzerland from Rickli and
Bücher (2005)
Min. Max. Median
Slope (°) 23 50 34
Failure depth (m) 0.2 1.5 0.8
Volume (m
3
) 30 1,100 72
Runout distance (m) 4.5 71 16.5
Original Paper
Landslides 9 & (2012)180
lower end of the channel ranges between 15 and 40 m
3
. Vertical
(slope-normal) accelerations are visible in the vicinity of the trap
door, but observations of the flow surface from video recordings
indicate that these rapidly disappear after several meters, leading
to flows in which the primary velocity component is parallel to
the slope. Video observations of the front reveal that the flow
velocities across the channel width are quite constant at least until
position 2; material at the slope sidewalls lags slightly behind,
indicating some sidewall friction. After position 2 a small
secondary surge often develops on the right side of the channel.
In places, micro gullies are present on the bedrock surface,
principally on the sides and in the lower part of the channel.
Minor amounts of material transported by runoff due to rainfall
between the tests lie on the bed surface and can be entrained by
the flow. At the end of a test day, the channel is cleared out with a
pressurized water hose. On tests with successive releases (between
2 and 4), deposited material (between 5 and 30 m
3
) from the first
releases is still present on the channel bed during the consecutive
releases.
The channel is instrumented with vertically oriented laser
distance sensors located 14 m (position 1) and 26 m (position 2)
downstream from the starting reservoir (distances measured
parallel to the slope). The sensors hang from cables and are used
to determine the flow heights at the middle of the channel. The
flow heights reported here correspond to the height perpendicular
to the bed surface. At position 2, two laser distance sensors are
spaced 30 cm with their beams parallel and directed at the flow
surface. The velocity of the upper flow surface is derived using the
discrete cross correlation function h
1
h
2
(t, Δt) of the two height
signals
h
1
h
2
t; DtðÞ¼
X
tþbin size
t
0
¼tbin size
h
1
t
0
ðÞE
h
1
h
1
h
2
t
0
þ DtðÞE
h
2
h
2
ð1Þ
where h
1
(t) and h
2
(t) are the two flow height time series. The
bin_size parameter is the time interval over which the cross
correlation function is calculated. It was set to 0.125 or 0.25 s. E
h
1
and E
h
2
are the mean values of the flow heights over the bin_size
while
h
1
and
h
2
are the standard deviations of the flow heights
over the bin_size.
Four metres downstream of position 2 at 30 m downslope
(position 3), two pressure plates are mounted on cube-shaped
wedges to measure impact forces (see Fig. 3). The smaller wedge
has dimensions 160 mm (width) × 225 mm (height); the larger
wedge 240 mm (width) and 295 mm (height). The heights of the
wedges are measured from the concrete plate that is flush with the
channel bed. Thus, both wedges are completely submerged when
the flow heights are greater than 295 mm, which is the case most
of the time (see Table 3 and Fig. 5). Cylindrical strain-gauge
sensors with diameter 120 mm and measuring range up to 20 kN
are built in to the wedges. They are each surmounted by two
square steel plates with side lengths 120 and 200 mm (A=0.0144
and 0.04 m
2
), respectively. The steel plates are separated by an
elastomer layer of 20 mm thickness for overload protection (SBR
elastomer with hardness 65 shore A). The screws holding the steel
plates and the elastomer layer together only press on the outer
plate so that the impulses are damped by the elastomer layer. The
centres of the steel plates are located 0.14 m and 0.17 m above the
ground, and the two sensors are separated laterally by 1 m.
Between tests 11.2 and 13.1, the locations of the large and small
pressure plates were exchanged.
The strain-gauge sensors deliver a 2-kHz signal. They are
filtered to remove oscillations due to hard contacts between
solid grains and the pressure plates. The filtering consists of
replacing each signal value by the mean value over 0.05-s time
intervals.Thechoiceofthetimeintervalisbasedon
calibration tests of the pressure plates and corresponds to the
duration of the pressure oscillation consecutive to a hard
contact. A s teel sphere (mass=190 g) attached to a string (see
Fig. 4) is released from di fferent heights on the pressure pl ate
(impact velocities between 0.5 and 1 m/s, rebound velocities
between 0.1 and 0.3 m/s). Because the pressure plate is not
perfectly rigid (due t o the elas tomer layer), it vibrates under
the effect of the hard contact and the recorded pressure
oscillates around zero (see Fig. 4b). The duration of the
oscillation is of the order of 0.02 to 0.03 s. The mean pressure
value taken over the duration of the oscillation scales with the
momentum exchanged between the steel ball and the pressure
platewhilethemedianvaluetakenoverthedurationofthe
oscillation is close to zero. The filtering of the pressure signal
does not modify the integral of the pressure signal over time
that is equal to the momentum exchanged between the flow
Fig. 3 a Small and large wedges and b the design drawing of the large wedge.
The pressure sensors are located on the upslope-oriented face of the wedges
Landslides 9 & (2012) 181
and the plate. The pressure oscillations that are characteristic
of the pressure plate vibrations are replaced by mean pressure
values that are related to the solid grain properties. The mean
values, median values and standard deviation of the pressure
signals are comp uted over the same 0.05-s time inter vals. Only
small deviations are noticed between the signals filtered with
the mean or with the median value indicating that the
contribution of hard contacts to the pressure signals are short
and p unctual. The deviations coincide with peaks in the
standard deviation time series.
Table 3 Summary of Veltheim tests
Release
no.
Mean front
velocity (m/s)
Max. flow
height at
position 2 (m)
Wet density
(kg/m
3
)
Water mass
fraction (%)
Fines mass
fraction (%)
Gravel mass
fraction (%)
Liquid
limit (%)
Plastic
limit (%)
Plasticity
index (%)
4 5.3 0.59 1,850 22 46 28 30 17 13
5 8.4 0.41 1,920 21 36 40 28 17 11
6 10.3 0.51 1,950 24 21 59 27 15 12
7.1 9.5 0.32 1,760 20 34 33 29 16 13
7.2 10.4 0.66 –––––––
8.1 8.0 0.38 1,840 25 38 35 32 15 17
8.2 9.2 0.79 1,880 23 31 44 31 15 16
9.1 10.2 0.29 1,790 28 48 16 27 13 14
9.2 9.6 0.9 –––––––
9.3 9.9 0.6 –––––––
10 8.2 0.4 1,900 18 21 46 24 15 9
11.1 9 0.38 2,060 16 27 48 33 17 16
11.2 9.4 0.42 –––––––
13.1 8.4 0.33 1,880 22 28 37 26 17 9
13.2 9.1 0.54 –––––––
14.1 9.1 0.4 1,990 17 25 48 31 18 13
14.2 9.6 0.88 2,030 14 – – –––
14.3 9.1 0.8 1,930 19 – – –––
14.4 8.6 0.8 –––––––
15.1 8.9 0.37 1,830 23 25 41 29 16 13
15.2 9.1 0.99 –––––––
16.1 6.4 0.37 2,110 14 41 26 33 17 16
16.2 9.1 0.74 –––––––
On tests 1 to 3 and 12, the pressure plates are not installed or are defective
The mean front velocity is the mean front velocity calculated between position 1 and position 2. “Fines” refer to both clay and silt content (d<0.063 mm). The liquid and plastic
limits are the consistency limits for the fines content and the plasticity index is the difference between the liquid and plastic limits
Fig. 4 a Setup for calibration tests of
the pressure plates. b Force oscillation
measured during one hard contact
between the steel ball and the small
pressure plate
Original Paper
Landslides 9 & (2012)182
Results
Between September 2008 and September 2010, we have performed
16 tests (see Table 3). Most tests consist of one single release of
50 m
3
debris material. On tests 7, 8, 11, 13, 15 and 16, two successive
releases are carried out while on test 9 three releases and on test
14 four releases take place successively (the volume released is
always 50 m
3
). Between consecutive releases (between 2 and 3 h)
deposited material remains on the channel bed surface. Material
samples are collected from the top of the reservoir before every
first release (and before second release 8.2 and in the deposit after
release 7.1) and are analysed in the laboratory for density, water
content, grain-size distribution and consistency limits of the fines
content (see Table 3 and Fig. 5). The volume of the samples (about
30 kg) is large eno ugh to assess the grain-size distribution
accurately for grain sizes up to 60 mm. For grain sizes up to
100 mm samples of 250 kg would be necessary (in order to have a
volume of the sample hundred times larger than the volume of
the grain). The mass fraction of cobbles larger than 60 mm
(ranging between 0% and 5%) is thus removed from the results
and the other mass fractions are adjusted accordingly. On release
7.1, two material samples are collected; one in the reservoir and
one in the deposit at the flexible barrier. The properties of the two
samples are very close (maximum deviation of 8% for the clay
mass fraction), indicating that the sample taken at the surface of
the reservoir is representative for the whole material contained in
the reservoir.
The density of the debris mixtures ranged between 1,760
and 2,110 kg/m
3
and the water content between 14% and 28%
(percentage of mass). The mass fraction of fines (grain size d<
0.063 mm) and gravel (2 mm<d<60 mm) vary between 21%
and 48% and 16% and 59%, respectively. The mean front
velocity calculated between position 1 and 2 var ies between 5.3
and 10.4 m/s, and the maximal flow heights at po sition 2 ra nge
between 0.29 and 0.99 m. The maximum impact pressure
ranges between 15 and 200 kPa (see Table 4). The passing time
of the front is defined as the time when the height is equal to
0.05 m. From the pressure signals, it is defined when the
pressure reaches 10 kPa.
Flow height
Two main types of flow front can be recognized. Flows on the
slope free of sediment deposits have a moderately steep front
followed by a body with smoothly varying flow height. The
maximum flow heights range between 0.3 and 0.5 m and are
measured between 0.5 and 1.5 s after the passing of the front.
Flows on the slope with previously deposited material show
steeper fronts and a shorter body. The maximum flow depths
range between 0.6 and 1 m and are measured between 0 and 0.5 s
after the passing of the flow front. The peak heights correspond to
the aerated (and unsaturated) upper part of the front region of
the flow. A prominent feature of all the flows is the presence of a
tail, indicated by decreasing flow heights with time. A constant
surface elevation at the end of the measurement indicates that the
material has stopped moving. The deposition heights at positions
1 and 2 range between 0.03 and 0.28 m. The deposition heights
correlate negatively with the mean front velocity from Table 3.
The mean of deposition heights at positions 1 and 2 after first
releases are considered (because the deposition heights after
consecutive releases are affected by the deposition heights from
the preceding releases) (Fig. 6).
Front and surface velocities
Mean front velocities between positions 1 and 2 are computed
from the front passing times at the laser sensors (see Table 3).
They range between 5.3 and 10.4 m/s. For the analysis of the
impact pressure data, the front velocities between positions 2 and
3 are determined from the front passing times at the laser sensor
and from the impact times on the pressure plates (see Table 4).
This results in different front velocities at the locations of the
large and small pressure plates due to different impact times.
Surface velocities at position 2 are presented in Fig. 7 for releases
14.1 to 14.2.
Comparing first releases (on the bedrock flow surface
without sediment deposits from subsequent releases), we can
distinguish a difference in flow behaviour among releases with a
large fraction of coarse particles (gravel fraction between 45% and
50%, releases 6, 10, 11.1, 14.1 and 15.1). They all are fast flows
independently of water content or liquidity index. The liquidity
index LI is defined on the base of the plastic limit PL and the
liquid limit LL of the fines content and situates the water content
W in connection to the consistency limits PL and LL:
LI ¼
W PL
LL PL
ð2Þ
A second category includes releases with large fines content,
i.e., clay and silt fraction larger than 40% (releases 4, 9 and 16).
The front velocities of these releases are lower on average and
display scatter but correlate with the liquidity index.
Impact pressure
The maximum impact pressures range between 15 and 200 kPa, i.e.,
over 1 order of magnitude. Alternatively, the impact pressures are 2
to 50 times larger than the static pressure values. In Fig. 8,theraw
and filtered time series of the impact pressure signals measured
during release 15.1 by the large pressure plate are plotted. The
maximum impact pressures are measured in the front of the flow
Fig. 5 Grain-size distributions for releases 4 to 16.2, the data points represent the
silt, clay, sand and gravel mass fraction. For release 7.1, the grain-size distributions
of samples from the reservoir and the deposit are plotted
Landslides 9 & (2012) 183
