You are on page 1of 13

Original Paper

Landslides R. C. H. Koo I J. S. H. Kwan I C. W. W. Ng I C. Lam I C. E. Choi I D. Song I W. K. Pun


DOI 10.1007/s10346-016-0715-5
Received: 3 February 2016
Accepted: 22 April 2016 Velocity attenuation of debris flows and a new
Springer-Verlag Berlin Heidelberg 2016
momentum-based load model for rigid barriers

Abstract Effective design of mitigation measures against debris in engineering practice worldwide for landslide risk assessment
flow hazards remains a challenging geotechnical problem. At pres- and barrier design. However, debris-barrier interaction is usually
ent, a pseudo-static approach is commonly used for the calcula- ignored.
tion of impact load acting on a rigid debris-resisting barrier. The Recently, Kwan and Koo (2015) proposed a step-change func-
impact load is normally calculated based on the maximum velocity tion to approximate the debris velocity-time relationship (debris
observed in the transportation zone under free-field conditions velocity hydrograph) at the location of a barrier. In this approach,
without considering debris-barrier interaction. In reality, the im- the debris velocity hydrograph is generated for the free-field situ-
pact load acting on a barrier varies with the change of debris ation, i.e. flow impedance by the barrier wall is not considered.
momentum flux but this is seldom considered in barrier design. While this approach advances the design practice by considering
To provide a scientific basis for assessing debris momentum flux the variation of debris velocity, it is also worthwhile to consider the
during impact, this paper presents results from a study of debris- change in debris velocity due to the obstruction of the barrier,
barrier interaction using physical flume modelling. This study which is the focus of this paper.
showed that, following the first stage of impact, the accumulated In the present study, physical flume tests were conducted to
debris behind a barrier formed a stationary zone and caused the demonstrate the attenuation of debris velocity and change of
remaining debris to slow down in a run-up process. In the exper- momentum flux due to debris-barrier interaction in a dry sand
iments, the peak debris momentum was 30 % lower compared to flow. A literature search revealed that no detailed flume experi-
that observed under free-field conditions. A new momentum- ments have been conducted to investigate the change of momen-
based model was developed to take into account attenuation of tum flux due to debris-barrier interaction even for dry sand flows.
momentum flux for predicting debris impact load on rigid bar- Although it is acknowledged that landslide debris is a multi-phase
riers. The new rationalised model was assessed using data from the material and the flow mechanisms are complex, it is necessary to
notable Yu Tung Road debris flow in Hong Kong. The assessment start from a simple case such as a dry granular flow, in order to
showed that the design bending moment at the base of the barrier understand the fundamental mechanisms during the impacting
wall could be reduced more than 30 % using the proposed model, process. From the results of the flume experiments, a new
compared with the current design approach. The adoption of the momentum-based impact load model has been developed to pre-
proposed model could offer a new opportunity for practitioners to dict debris impact loading with consideration of debris-barrier
optimise the design of rigid barriers. interaction for the run-up process. The performance of the pro-
posed model was evaluated using data from of the Yu Tung Road
Keywords Velocity attenuation . Momentum flux . Rigid debris flow in Hong Kong.
barrier . Impact . Flume modelling
Review of debris impact load models on rigid barriers

Hydrostatic approach
Introduction Two approaches, namely the hydrostatic and the hydrodynamic
A debris flow typically appears as an elongated continuum travel- approaches, are commonly adopted for calculating the dynamic
ling along natural drainage lines. When a debris-resisting rigid impact load on rigid, debris-resisting barriers. In the hydrostatic
barrier is subjected to the impact of a debris flow, the debris will approach, the dynamic impact force (Fd) acting on a barrier wall
continuously fill-up the barrier in several stages. The buildup of due to debris impact is estimated using Eq. (1):
debris behind barriers formed at different stages of a single debris
flow, and it involves debris-barrier interaction. Guidelines on the F d w g h2 =2 1
design of rigid barriers, such as ASI (2013), Kwan (2012), NILIM
(2007), MLR (2006) and VanDine (1996), recommend a pseudo- where is a dimensionless empirical factor to account for the
static approach for calculating debris impact loads, which are dynamic effect of flow impact, is debris density (in Mg/m3), g is
assumed to be constant for all phases of debris impact. In reality, gravitational acceleration (in m/s2), w is the barrier width and h is
both the velocity and the thickness of debris vary with time at the debris thickness (in m). It can be seen that Eq. (1) is a form of the
barrier location. Therefore, the load acting on a barrier at different hydrostatic pressure calculation; an obvious limitation is that it
stages of impact is not the same. Debris velocity and thickness does not consider the debris velocity. Armanini (1997) carried out
under free-field conditions can be estimated using debris mobility small-scale tests and found values of ranging between 7 and 11.
models such as DAN-W (Hungr 1995) and 2d-DMM (Kwan and
Sun 2006). These models simulate the post-failure motion of rapid Hydrodynamic approach
landslides by discretising the debris mass into interconnected The second approach is based on the widely adopted hydrody-
slices and solving shallow water depth-averaged equations using namic model. Design guidelines published in Canada (VanDine
a Lagrangian formulation. Both models have been widely adopted 1996), China (MLR 2006) and Japan (NILIM 2007) are based on

Landslides
Original Paper
this approach where the dynamic impact pressure is assumed to be taken the peak value. Proske et al. (2011) reviewed the recent
proportional to the square of debris flow velocity. Since debris literature and reported values of between 1.0 and 2.5 for
flows arrive at a barrier in multiple stages, a multi-stage hydrody- passable obstacles. According to Wendeler (2008) and Kwan
namic model was developed by Wendeler (2008) and modified by (2012), the value of can range between 2.5 and 5.0 for rigid
Kwan (2012) (Fig. 1). This multiple-stage model assumes that the barriers.
debris involved in the first impact deposits behind the barrier and Although VanDine (1996) reports reductions in debris impact
the remaining debris rides over the deposit to give a second load once deposition of the debris begins, the change in debris
impact. The process continues until the barrier is filled to the impact velocity during the different stages of impact is not con-
design retaining height (hret). As the barrier is designed to retain sidered in the prevailing design practice. The reduction phenom-
the total volume of the debris, no overflow is assumed. The total enon may be attributed to two reasons: (i) debris is faster at the
impact load acting on the barriers can be estimated by considering front and slower in the rear portion and (ii) interaction between
the dynamic impact pressure and the lateral earth pressure (K) of the debris deposited behind a barrier and the remaining run-up
the deposited debris. The dynamic impact load (Fd) acting on the debris. Recently, Hong et al. (2015) reported debris impact loads
barrier is calculated using the hydrodynamic equation: measured during 139 debris flows that occurred between 1961 and
2000 at the Jiangjia Ravine in Yunnan Province, China. The field
F d w v2 h 2 measurements indicated that the peak impact pressure caused by
the debris front was almost double the mean impact pressure of
where is empirical dynamic pressure coefficient (dimension- the remaining debris. This is probably due to the change in debris
less), is density of debris (in Mg/m3), v is debris velocity (in m/s), velocity during different stages of impact.
w is the barrier width (in m) and h is debris thickness (in m). The If hard inclusions such as large rock blocks or boulders are
value of has been established by back-analyses of instrumented present in debris flows, additional load due to the impact of the
field and laboratory tests, and the values of v and h are usually hard inclusions should be considered in the design. Kwan (2012)
recommended the use of Hertz equation modified by a load
reduction factor for establishing the dynamic loads of large parti-
cles impacting on rigid barriers.

An improved hydrodynamic approach considering variations of debris


Debris flow
velocity
Barrier
impact pressure Compared to the prevailing design practice in which a constant
peak impact pressure is assumed, consideration of debris velocity
h (h = design debris
impact depth) change during the impacting process could lead to a more realistic
assessment of the debris impact pressure and thus more efficient
(a) First impact when the barrier is empty design of barriers. Recently, Kwan and Koo (2015) proposed an
improved approach using a step function (Fig. 2) to represent the
attenuation of debris velocity under free-field conditions. The step
function considers the volume of debris deposited after each stage
of impact, explained as follows: first, for simplicity, the geometry
Debris flow impact pressure
of the deposited debris is regarded as inclined blocks as shown in
h
Fig. 3a. With the maximum height of the deposited debris (h1, h2
Barrier
hn) known, the volume of each stage of debris (V1, V2Vn)
Static Depth of debris stopped by the barrier can be calculated. It follows that the time
pressure accumulated in
of debris preceding stages intervals for each stage can be calculated using a cumulative
debris-volume hydrograph generated by a suitable debris mobility
(b) Subsequent stages hitting the barrier model as shown in Fig. 3b. Last, with the time interval for each

Peak velocity used in conventional


Debris flow impact pressure t1 t2 hydrodynamic impact model

h Debris Step function used in improved


Velocity hydrodynamic impact model
Static pressure Design (m/s)
Barrier of debris retaining
height of
debris, hret Debris-velocity
hydrograph

(c) Last stage filling up the barrier to the design debris retaining height
Time (s)
Fig. 1 Design scenarios of multiple stages of debris impact. a First impact when
the barrier is empty. b Subsequent stages hitting the barrier. c Last stage filling up Fig. 2 Velocity hydrographs considered in establishing the design debris impact
the barrier to the design debris retaining height load

Landslides
(a) Instrumentation included a laser sensor to measure the up-
stream debris thickness and two high-speed cameras (one posi-
tioned at the top and one on a side) to capture images of the debris
hn and the barrier. The high-speed camera captured 200 images
n th stage
(frames) per second at a resolution of 1632 864 pixels. These
images were analysed using the particle image velocimetry (PIV)
program developed by White et al. (2003) to determine the debris
h2 velocity. Photo-sensors installed at a spacing of 0.5 m along the
2 nd stage
channel bed also measured the frontal velocity of the debris.
h1 1 st stage Dry Leighton Buzzard (fraction C) sand was used for the tests.
The sand has an internal friction angle () of 30 (Choi et al. 2014)
(b) and a density of 1.68 Mg/m3. The weight of the source material was
100 kg, delivered from a rectangular soil storage tank with a
Cumu lativ e v o lume o f
d eb ris p assin g th rou gh

volume of 0.06 m3 mounted at the upstream end of the flume.


barrier location

Vn To start a test, the flip door of the storage tank was opened to allow
n th stage the sand to flow out and run along the channel. The basal
V2 (interface) friction angle () between the sand and the channel
2 nd stage bed was 22.6, measured following the test procedure in Hungr
V1
1 st stage (2008).
Froude number (Fr), the ratio of flow inertia to the external
gravitational field, i.e. Fr = v/(gh)0.5 (where v is flow velocity, g is
(c) gravity and h is flow thickness), was used to characterise the flow
to aid the determination of an appropriate channel slope. Authors
Velo city o f d eb ris p assing

vp v1
th ro u gh b arrier lo cation

such as Armanini et al. (2011) and Choi et al. (2014) show that the
v2 Froude number controls the impact behaviour of granular flows.
The Froude numbers of channelised debris flows typically range
vn
between 3.0 and 5.0 (Hubl et al. 2009; Peng et al. 2015). A Froude
number of around 3 was targeted in the present flume experiments
and was used to calculate the targeted frontal velocity and the
maximum flow thickness. Several trials were carried out to estab-
t1 t2 tn lish an appropriate flume inclination. It was subsequently decided
that the flume should be inclined at 26 to generate sand flows
Time
corresponding to a Froude number of 3.0.
Fig. 3 The suggested debris impact scenario for design of barriers: a number of
stages, b design cumulative volume and c design envelope of velocity profile Test under free-field conditions
A reference test was first carried out to study the flow character-
istics of the debris (dry sand) under free-field conditions, i.e.
without a barrier. The velocity obtained from the PIV analysis
stage known, the debris velocity corresponding to each stage of was the average velocity across the thickness of the flowing debris.
impact (v1, v2vn) can be obtained from the velocity hydrograph The observation point was located 0.8 m away from the flip door
(as shown in Fig. 3c). In this approach however, only a maximum of the storage tank, that is, the location of the barrier in the other
reduction of 30 % in the peak debris velocity is permitted for tests. The thickness of the debris was measured by the laser
design purpose. The debris velocity determined in this way then distance sensor and confirmed from the high-speed images.
can be substituted into Eq. (2) for the calculation of debris impact Figure 5 shows the velocity and thickness hydrographs of the
force. A major limitation of this approach is that it does not debris. It can be seen that debris reached the observation point
consider the interaction between debris and the impact-resisting at about 0.3 s with a frontal velocity of 2.7 m/s. The velocity of the
barrier. remaining incoming debris then gradually decreased to 0.7 m/s at
5 s. It can also be seen that the thickness of the debris reached a
Laboratory flume tests peak value of 0.09 m at about 1 s and then reduced to 0.065 m after
2.5 s. Similar patterns of debris velocity and thickness are also
Test setup and materials used observed in experiments conducted by others. For example, Fitze
In the present study, physical flume tests were conducted to (2010), describing the large-scale flume experiments conducted by
investigate velocity attenuation and change of momentum flux Iverson et al. (2010), notes that the debris velocity and thickness
due to debris-barrier interaction using dry sand flows. Subpanels were significantly reduced by more than 50 % at downstream
a and b of Fig. 4 show, respectively, a photograph and a schematic locations.
cross section including instrumentation. The flume was 5 m long
and 0.2 m wide. The transparent side walls of acrylic glass were Test with a rigid barrier
0.5 m high and erected perpendicular to the channel bed. The Following the reference test, a second test was used to investigate
slope angle of the channel () was 26. the interaction between the dry sand flow and a rigid barrier. The

Landslides
Original Paper

Aluminum rigid
barrier

(a) Photograph of the laboratory flume


Top high-speed
camera
Photoconductive sensor
Laser sensor

High Speed Camera

Door Rigid
barrier

200 H
500
800
500

Side high-speed
700 500
camera

400

Note: All dimensions in mm Slope of channel bed

(b) Side view of the laboratory flume and instrumentation


Fig. 4 Laboratory flume and instrumentation. a Photograph of the laboratory flume. b Side view of the laboratory flume and instrumentation

flume setup was the same as the reference test, except that a between static and moving debris. It can be seen that the deposi-
vertical rigid barrier was installed at 0.8 m downstream from the tion zone enlarged during the flow process until debris completely
storage tank (i.e. the observation point chosen for Fig. 5). The filled the barrier. The development of a deposition zone and a
barrier was a 10-mm-thick, 0.26-m-tall aluminium plate and was debris run-up mechanism are also observed in both natural and
about three times taller than the average debris thickness mea- laboratory debris flow studies (e.g. ASI 2013; Armanini et al. 2011;
sured under free-field conditions (Fig. 5). Wendeler 2008). Tiberghien et al. (2007) note that the deposited
Figure 6 shows images taken during the model debris flow at materials behind a barrier play an important role in the dynamic
four different times: 4.64, 4.68, 4.75 and 4.85 s. A static (deposition) load transfer process, and Ashwood (2014) observes in flume tests
zone was present behind the barrier after the first stage of impact that debris deposited behind a barrier exert friction with the
between 4.68 and 4.75 s. The remaining debris then ran over the moving debris and thus result in energy dissipation of the
deposited material to hit the barrier at a higher elevation. The subsequent debris flow stages. A recent numerical study by
dotted line in Fig. 6c, d approximately delineates the boundary Kwan et al. (2015) revealed a debris-barrier interaction similar to

Landslides
4 0.2

Arrival me of
3.5 Velocity
debris at barrier

3
Thickness
0.15
Peak free-field velocity
v = 2.7 m/s
Rigid
Debris Velocity (m/s)

2.5
barrier

Thickness (m)
2 0.1 (260 mm)
1.5

1 0.05 Channel bed

0.5

0 0 (a) 4.64 s
0 1 2 3 4 5
Time (s)

Peak run-up velocity


Fig. 5 Free-field velocity and thickness hydrographs at 0.8 m from the soil storage
v = 1.1 m/s
tank
Rigid
barrier
the process observed in the flume test. However, in these studies, (260 mm)
there is no attempt to quantify the degree of impact load reduction
due to debris-barrier interaction.
To quantify the magnitude of impact load reduction, the pho- Stationary
tographic records obtained from the flume experiments were zone Channel bed
analysed using the PIV program to determine the debris velocity
and the extent of the deposited zone. Subpanels a and b of Fig. 7,
respectively, show the velocity vector plots for 4.64 and 4.75 s (b) 4.75 s
(corresponding to Fig. 6a, c). The dotted lines in Fig. 7b delineate
the region of flowing material. The debris momentum was trans- Fig. 7 Debris velocity vectors from PIV analysis. a 4.64 s. b 4.75 s
ferred to the barrier during the first impact, instantly forming a
deposited static zone immediately behind the barrier. The remain- that some of the materials bounced back after hitting the barrier
ing debris then rode over the deposited material to hit the barrier and collided with incoming materials, hence reducing their
at a higher elevation after a run-up process. Figure 7b also shows velocity.

100 m
mm

Staonary
zone
Vercal rigid barrier

(a) 4.64 s (c) 4.75 s

Staonary
zone

(b) 4.68 s (b) 4.85 s

Fig. 6 Mechanism of a dry sand flow impacting on a rigid barrier. a 4.64 s. b 4.75 s. c 4.68 s. d 4.85 s

Landslides
Original Paper
Effects of barrier on debris flow properties thickness hydrographs for two cases. For the case with a barrier,
Figure 8a compares the velocity hydrographs for the two cases only the thickness of the moving debris was plotted; the thickness
considered, i.e. with and without a barrier. Time zero (t = 0 s) of the stationary debris was not included. It can be seen that the
refers to the instance of the first impact. Unlike the free-field case debris had an initial thickness of 30 mm at the time of the first
where the observation point was fixed at the barrier location (i.e. impact. The thickness increased to 44 mm after 0.02 s and then
0.8 m away from the storage tank upstream), for the case with a dropped to 22 mm after 0.2 s (i.e. a reduction of 22 mm near the
barrier, the maximum velocity was determined for each time completion of run-up). This reduction in debris thickness was
instance recorded by the camera images, so that the observation comparable to the reduction observed in the free-field case, where
point moved slightly away from the barrier during the filling the thickness dropped by 25 mm (from 90 to 65 mm) from the
process. This approach allowed the tracking of the change in peak value. Compared to the free-field case, however, the peak
debris velocity during the fill-up. It can be seen from the dotted thickness for the case with a barrier was significantly reduced (44
line in Fig. 8a that the velocity of the debris front was 2.7 m/s at the vs. 90 mm) due to the formation of a static deposit zone after the
time of the first impact and dropped to 0.5 m/s in less than 0.5 s. first impact.
The same degree of reduction took 5 s to occur in the free-field To assist in quantifying the debris impact load, Fig. 9 shows the
case as shown by the solid line. Figure 8b shows the debris hydrographs of normalised mass and momentum fluxes for the

3 3
Barrier
(run-up)
Without barrier
2.5 2.5
2.4
Barrier (run-up)

2 2
Velocity (m/s)

1.5 1.5

1 1

0.5 0.5

0 0
0 2 4 6 8 0 0.02 0.1
Time (s) Time (s)
(a)
Time (s)
0 2 4 6 8
100
90 Without barrier

80 Barrier (run-up)
Thickness (mm)

70
60
50
44
40
30
20
10
0
0 0.05 0.1 0.15 0.2
Time (s)
(b)
Fig. 8 Hydrographs obtained from approaching debris. a Velocity. b Thickness

Landslides
2.4 instead of 1.7). The change in momentum flux could be
2.2 Without barrier
attributed to friction between the moving and stationary de-
2 Barrier (run-up) bris as well as the transformation of some of the debris
40%
Normalised mass ux

1.8 kinetic energy to potential energy in the run-up.


1.6 reducon
Yu Tung Road debris flow in Hong Kong
1.4
1.2
Overview
1 A notable debris flow in Hong Kong was used to provide further
0.8 evidence of the attenuation pattern of debris velocity in the field. This
0.6 was a sizable channelised granular debris flow involving an active
0.4 volume of about 3500 m3. It occurred in a natural hillside catchment
0.2 above Yu Tung Road on Lantau Island, Hong Kong, in 2008. The
total runout distance of this debris flow was 650 m. A post-landslide
0
investigation of the event is given in AECOM (2012). Figure 10a shows
0 0.2 0.4 0.6 0.8 1
the post-landslide debris trail and a 7-m-high single-terminal rigid
Normalised me barrier constructed in 2013 after the event. The velocity of the debris
(a) torrent was estimated from a video taken by an onlooker. Back
analyses were carried out using the Voellmy rheological model and
2 the computer program 2d-DMM (Kwan and Sun 2006). The two
Without barrier model parameters, namely the basal friction angle and the turbu-
1.8
Normalised momentum ux

Barrier (run-up) lence coefficient, were adjusted until a good agreement between the
1.6
observed and back-analysed frontal velocity and runout distance was
1.4 30%
obtained. Calibrated model parameters are summarised in Table 1.
reducon
1.2
Debris properties
1 Figure 10b shows a velocity hydrograph at chainage 520 m (80 % of
0.8 the total runout distance) which is the outlet of the channel. Two
observations can be made from this figure. First, the velocity curve
0.6
given by 2d-DMM fits the field data reasonably well. This confirms
0.4 the usefulness of the program to simulate debris mobility for study
0.2 of momentum flux. Second, the front of the debris flow reached
the channel outlet at a velocity of 8.6 m/s and that the remaining
0 flow gradually slowed after the peak to about 4.2 m/s at 60 s.
0 0.2 0.4 0.6 0.8 1 Figure 11 shows the normalised velocity, thickness and momen-
Normalised me tum flux hydrographs at chainage 520 m by 2d-DMM, where a single-
(b) terminal barrier was built after the debris flow event. To allow
comparison with the flume test results, the normalised time scale
Fig. 9 Normalised hydrographs. a Mass flux. b Momentum flux (i.e. is adopted normalised time = 1.0 denotes the instant when the
entire debris run-out process at the observation point completes) is
used. Time zero indicates the first arrival time at the observation
two impact cases, i.e. with and without a barrier. The mass flux of a point. A significant reduction in velocity (8.6 to 2.9 m/s, approx.
debris flow (hv, in Mg/s/m, i.e. rate of mass transport per unit width) 66 %) between the debris front and the rear portion can be seen in
represents the rate of mass passing through a unit area at the barrier Fig. 11a. The thickness of the debris also reduced from the maximum
location. The momentum flux M (=hv2, in kN/m) represents thickness of 2.5 m to less than 0.5 m near the end of the event.
the rate of transfer of momentum across per unit widththis The debris impact force (Fd) acting on a barrier is proportional to
parameter was adopted in the proposed momentum-based the momentum flux M (=hv2) (see Eq. (2)), and so the change in
model. Normalisation of these parameters with respect to momentum flux with time was calculated using the debris velocity and
their initial values allowed them to be compared for the two thickness given in Fig. 11a, b, with results shown in Fig. 11c. The flume
scenarios. In addtion, a normalised time scale of hydrographs test results (with and without a barrier) were also plotted in Fig. 11c for
is adopted for plotting hydrographs i.e. normalised time = 1.0 comparison. Two observations can be made from this combined plot.
denotes the instant when the entire debris run-out process at First, after showing initial peaks, the momentum fluxes reduced to
the observation point completes. Figure 9a shows that the almost zero towards the end of the flow events in all three cases.
peak normalised mass flux for the case with a barrier was Second, the normalised peak momentum fluxes were quite similar
about 40 % lower than that for the free-field case (1.3 instead between the Yu Tung Road debris flow and the laboratory flume test
of 2.3). This meant that the barrier only had to resist about without a barrier (1.70 vs. 1.75). This is probably because the Froude
40 % of the peak mass flux of the free-field flow. Figure 9b number and the gradient of the run-out path at the observation point
similarly shows that the peak normalised momentum flux was of both cases are similar. This suggested that attenuation of debris flow
reduced by about 30 % due to the presence of the barrier (1.2 momentum occur both in the laboratory and in the field.

Landslides
Original Paper
Completion of a 7 m high
Source
single terminal barrier (plan
area 24 m 37 m) in 2013
after the Yu Tung Road
debris flow event in 2008
Transport zone

Yu Tung Road Ch 520

12 June 2008
(a) Aerial photograph
12

Observed data from video


10 Error bar of video resoluon
Back-analysed by 2d-DMM
Debris Velocity (m/s)

0
20 25 30 35 40 45 50 55 60
Time (s)

(b) Velocity hydrograph at Chainage of 520 m

Fig. 10 Yu Tung Road debris flow in Hong Kong. a Aerial photograph. b Velocity hydrograph at chainage of 520 m

A new momentum-based impact load model field conditions until the moment of impact. Since the time histo-
ries of the debris velocity and thickness (i.e. velocity and thickness
Overview hydrographs) can be estimated using a suitable numerical model
Based on the findings reported in previous sections, a new impact load (e.g. DAN-W or 2d-DMM), the peak debris velocity and thickness
model was formulated to take into account the change in debris can be obtained and used to determine the design momentum flux
momentum flux during an impact. The proposed model is applicable M. The design impact force Fd of the first stage of impact is thus:
to granular debris flows showing a staged run-up process upon impact
on a rigid barrier. Figure 12 shows the idealised run-up behaviour of the F d w M Rm 4
landslide debris. To take into account the effect of changing momen-
tum with time on the impact force, Eq. (2) is rewritten as follows: where Rm is used to take into account the reduced debris
momentum flux due to debris-barrier interaction. A value of 0.7
F d t w M t 3 may be used based on the results shown in Fig. 9b.

where Fd (t) is dynamic impact force, w is the barrier width and Theoretical evaluation
M (t) is momentum flux h(t)v(t)2 with respect to time. Initially, Following the first impact, the remaining debris with initial veloc-
the barrier is impacted by the frontal debris flowing under free- ity v and thickness h are assumed to ride over the top of the

Table 1 Calibrated model parameters for the Yu Tung Road debris flow
Volume (m3) Observed runout distance (m) Back-analysed Voellmy rheological parameters
Turbulence coefficient (m/s) Basal friction angle (deg)
3500 700 500 8

Landslides
10
9 Yu Tung Road debris ow
87.9

Velocity (m/s)
7
6
5
4
3
2
1
0
0 0.2 0.4 0.6 0.8 1
Normalized Time

(a) Normalised velocity hydrograph

10
9 Yu Tung Road debris ow
8
Thickness (m)

7
6
5
4
3
2.3
2
1
0
0 0.2 0.4 0.6 0.8 1
Normalized Time

(b) Normalised thickness hydrograph

2
Normalised Momentum Flux

No barrier (experiment)
1.5 With barrier (experiment)

No barrier (Yu Tung Road debris ow)


1

0.5

0
0 0.2 0.4 0.6 0.8 1
Normalised Time
(c) Normalised momentum flux hydrograph

Fig. 11 Normalised velocity, thickness and momentum hydrographs. a Normalised velocity hydrograph. b Normalised thickness hydrograph. c Normalised momentum
flux hydrograph

(a) First impact (b) Impact by run-up debris

Free-field flow
prior to run-up

hru vru
Fru v, h
hd2 R
vp hd1 d d
Fd Fs
hp

Fig. 12 Schematic diagram for the proposed impact model. a First impact. b Impact by run-up debris

Landslides
Original Paper
deposited materials and then hit the barrier. The values of v and h The reduction in terms of debris velocity squared (v12) due to
can again be estimated from velocity and thickness hydrographs frictional resistance energy loss can be calculated by equating the
generated using a suitable debris mobility model. The observation energy loss obtained from Eq. (9) (ER) with the loss of kinetic energy:
point for the hydrographs should now be located immediately at
1
the back of the deposited materials instead of the location of the ER NtanL mv21 11
2
barrier. As shown in Fig. 12b, the deposited materials are assumed
to form a wedge with height hd1 immediately behind the barrier.
Assuming N = mg and re-arranging thus gives:
The top surface of the wedge is assumed to form an angle d with
the original slope surface, which in turn is inclined at an angle to v21 2gtan L 12
the horizontal. The total force F acting on the barrier now com-
prises the run-up impact force Fru and a static force Fs due to the
During the run-up process, some of the debris kinetic energy is
lateral earth pressure of the deposited materials, as follows:
converted into potential energy. The associated reduction in terms
F F s F ru cosd 5 of debris velocity squared (v22 ) due to run-up can thus be
calculated using the following simple equation:
The static force Fs term in Eq. (5) can be calculated using v22 2ghd2 13
conventional earth pressure theory. Regarding the second term,
Law (2014) found that (d ) typically varies between 10 and 30
where hd2 = L sin (d )
depending on the initial flow velocity and the internal friction
Since only the frontal debris will experience full velocity reduc-
angle of the debris (). In the flume experiments discussed in
tion due to the two effects discussed above, the lumped reduction
the earlier section, the value of (d ) was 30 for dry sand flows.
factor (Rru), should be established based on 0.5 L and 0.5 hd2
The run-up impact force Fru can be calculated using the following
(i.e. v2 v2 (1 Rru)2 = g (tan L + hd2) and can be calculated
equation:
using the following form:
F ru w M t 6 r
g tanL hd2
Rru 1 1 14
where the time-dependent momentum flux M(t) is a function of v2
the run-up velocity (vru) and run-up thickness (hru). For design
purpose, hru can be taken as h, i.e. assuming no change in thick- This factor can be substituted into Eq. (7) to obtain the velocity
ness during the run-up process for simplification. The estimation of the moving debris after run-up but before impact.
of run-up velocity is discussed in the following section.
Application of the proposed model to flume test results
Estimation of debris run-up velocity The performance of the proposed load model was evaluated by using
The run-up velocity can be estimated by applying a lumped it to back-analyse the change in momentum flux M of the debris in
reduction factor (Rru) to the free-field velocity prior to run-up the laboratory flume test with a rigid barrier (see Figs. 6, 7, 8 and 9). It
(v). is worthy to emphasise that the proposed model focuses only on the
run-up process up until the moment of impact. Although in the flume
vru v1Rru 7 test some of debris was deflected vertically upwards after impact
(Fig. 7b), the proposed model is still applicable as the deflected debris
where the lumped factor (Rru) takes into account the reduction was not found to shield the barrier during the following run-up stage.
in debris velocity due to two distinct effects: (i) energy loss due to Table 2 summarises the values of the parameters used. Since the
frictional resistance between the moving and stationary debris and actual run-up thickness was already available from the experi-
(ii) conversion from kinetic to potential energy due to a change in ment, it was used in the calculations so that the momentum
the elevation (hd2). For the first effect, the frictional resistance (R)
between the moving and stationary debris and the resulting energy
loss (ER) can be calculated using the following two expressions:
Table 2 Summary of parameter values used to back-analyse momentum flux for
R N tan 8 the laboratory flume test with a rigid barrier
Parameter Value Determination method

ER N tan L 9 Internal friction angle 30 Measured from laboratory


of soil () tests (Choi et al. 2014)

where N is weight of debris acting on surface of the deposited Net deposition angle 30 Measured (Fig. 7)
(d )
material and is equal to mg, and L is the debris travel distance
during the run-up process. From geometry, the travel distance can Free-field velocity (v) 2.4 m/s Measured (Fig. 8a)
be calculated using the following expression: Free-field thickness (h) 44 mm Measured (Fig. 8b)
Run-up velocity (vru) Varies Calculated (Eqs. (7) to (14))
hd1 cos
L 10 Run-up thickness (hru) 44 mm Same as h
sind

Landslides
0.45
Momentum ux (kN/m ) Experiment conventional hydrodynamic model was used to predict the debris
0.4
0.35
Back-analysis using constant thickness impact force during the design (Eq. (2)). As discussed, this ap-
Back-analysis using varied thickness
0.3 proach does not take into account debris-barrier interaction. To
0.25 demonstrate the effect of the new momentum-based model, it was
0.2 used to re-calculate the impact load (Fd) and bending moment
0.15
(BM) (per unit width) at the wall base of the said rigid barrier. This
0.1
0.05
case was also re-analysed using the improved hydrodynamic mod-
0 el incorporating velocity reduction (Figs. 2 and 3), so a total of
0 0.05 0.1 0.15 three calculation approaches are compared. Table 3 summarises
Time since impact (s) the values of the parameters used for these approaches, and
Table 4 summarises the calculation results for three impact stages,
Fig. 13 Comparison of experimental and back-analysed momentum fluxes where stage 1 is for initial impact and stages 2 and 3 are for run-up
impacts.
Both the conventional and improved hydrodynamic models
reduction factor was unity in this case. Figure 13 shows the back- give the same impact load for the first stage of impact (970 kN/
calculated and measured results. The observed maximum momen- m), but when debris-barrier interaction is considered, the peak
tum flux (hv 2 = 1.68 Mg/m 3 0.044 m 2.4 m/s 2.4 m/ load given by the momentum-based model is reduced by 43 %
s = 0.42 kN/m) occurred at 0.02 s. Two curves are provided, one (691 kN/m). This is due to the use of the peak momentum flux and
generated using a constant debris thickness (hru) and one using a a momentum reduction factor (Rm) of 0.7 as observed in the flume
variable hru. It can be seen that if the variation in the flow thick- experiments reported in this study; any other justifiable values
ness is considered (varied from 44 to 24 mm, see Fig. 8b), a better could also have been used. Due to the force reduction, the BM at
agreement is obtained although for forward prediction this infor- the wall base is also reduced by 43 %. Note that for the calculation
mation is not easily available so a constant thickness should be of the lever arm (L), the point of impact is taken at the centre of
assumed for design purposes. Some energy may also be lost due to the debris thickness (h).
inter-particle friction and turbulence within the moving debris For stages 2 and 3, the improved hydrodynamic model gives
during the run-up process, but this is not considered in the model. lower run-up impact forces and wall bending moments due to the
reduction in debris velocity over time (Figs. 2 and 3). However,
Application of the proposed model to Yu Tung Road debris flow these reduced velocities still correspond to the free-field condi-
After the 2008 Yu Tung Road debris flow event, a rigid barrier was tion. The new momentum-based model, which considers debris-
constructed near the channel outlet as shown in Fig. 10a. The barrier interaction, allows the predicted forces and bending

Table 3 Summary of parameter values used to back-analyse the Yu Tung Road debris flow case
Impact load model Parameter Value Determination method
a
Conventional hydrodynamic model (Kwan 2012) Dynamic impact coefficient () 2.5 Empirical (Kwan 2012)
Density of debris () 2.2 Mg/ Conservative estimate (Kwan 2012)
m3
Peak free-field debris velocity (vp) 8.6 m/s Back-calculated from video records
Peak free-field debris thickness (hp) 2.6 m Back-calculated from debris mobility
model (2d-DMM)
Hydrodynamic model incorporating Free-field debris velocity for phases Varies Based on free-field velocity hydrograph
free-field velocity reduction (Kwan other than the first phase of (Fig. 11a)reduction according to a
and Koo 2015)a, b impact (v) step function and up to 0.7 of the
peak value (Fig. 2)
Momentum-based model (this study)a Internal friction angle of soil () 10 Back-analysed equivalent value (Kwan
et al. 2015)
Net deposition angle (d ) 10 Same as (assumed)
Slope inclination angle () 20 From site geometry
Free-field debris velocity (v) Varies From velocity hydrograph
Free-field debris thickness (h) Varies From thickness hydrograph
Run-up debris velocity (vru) Varies Calculated (Eqs. (7) to (14))
Run-up debris thickness (hru) Varies Same as ha (assumed)
Momentum reduction factor (Rm) 0.7 Estimated from Figs. 9b and 11c
a
Dynamic impact coefficient () and debris density () are the same as the conventional approach
b
For the first phase of impact, the peak free-field velocity (vp) and thickness (hp) are used as in the conventional approach

Landslides
Original Paper
Table 4 Summary of the results from different impact load models for rigid barrier designYu Tung Road debris flow case
Design impact stage Conventional Hydrodynamic model incorporating Momentum-based model
hydrodynamic model free-field velocity reduction (this study)
(Kwan 2012) (Kwan and Koo 2015)
Stage 11st impact Fd = hp vp2 Fd = hp vp2 Fd = M Rm where M = h v2
= 2.5 2.2 2.5 8.42 = 2.5 2.2 2.5 8.42 = 2.5 2.2 2.5 7.92 0.7
= 970 kN/m = 970 kN/m = 601 kN/m
Lru = 2.5/2 m = 1.25 m Lru = 2.5/2 m = 1.25 m Lru = 2.5/2 m = 1.25 m
BM = Fru Lru = 1213 kNm/m BM = Fru Lru = 1213 kNm/m BM = Fru Lru = 751 kNm/m
Stage 22nd impact (run-up Fru = 970 kN/m Fru = 2.5 2.2 2.5 7.02 = 674 kN/m Fru = M cos(d ) Where
to 5 m high) Lru = 2.5 + 2.5/2 = 3.75 m Where v = 7.0 m/s from free-field M = hru vru2
BM = 3638 kNm/m Lru = 2.5 + 2.5/2 = 3.75 m Fru = 2.5 (2.2 2.5 5.72) cos
BM = Fru Lru = 2528 kNm/m 10
= 440 kN/m
Where hru is taken to be 2.5 m for
direct comparison
Lru = 2.5+2.5/2 = 3.75 m
BM = 1650 kNm/m
Stage 32nd impact (run-up to Fru = 2.5 2.2 2.0 8.42 Fru = 2.5 2.2 2.0 (0.7 8.4)2 = 380 Fru = 2.5 2.2 2.0 1.62 cos 10
7 m high, i.e. top of barrier) = 776 kN/m kN/m = 28 kN/m
Lru = 5 + 2/2 = 6.0 m Where v = 0.7 8.4 = 5.9 m/s from Where hru is taken to be 2.0 m for
BM = 4656 kNm/m free-field direct comparison
Lru = 5 + 2/2 = 6.0 m Lru = 5 + 2/2 = 6.0 m
BM = Fru Lru = 2280 kNm/m BM = Fru Lru = 166 kNm/m
Lru is the level arm along the barrier height. For all three models, debris thicknesses for the first three impact stages are 2.5, 2.5 and 2.0 m for direct comparison between them

moments to be further reduced more than 30 % as shown in the the support of the HKUST Jockey Club Institute of Advanced
table. Study.

Concluding remarks
The results of laboratory flume tests demonstrate the transient
nature of debris flows under both free-field (no barrier) and
obstructed conditions. Both the velocity and thickness of a debris References
flow vary during the flow process; the impact load acting on a AECOM (2012). Detailed study of the 7 June 2008 landslides on the hillside
barrier thus also varies during the impact. The peak momentum of above Yu Tung Road, Tung Chung. GEO Report No. 271, Geotechnical
debris can thus be 30 % lower than the case without debris-barrier Engineering Office, Hong Kong, 124 p.
interaction. A new momentum-based model is thereby developed Armanini A (1997) On the dynamic impact of debris flows. In: Recent developments on
debris flows, vol 64, Lecture notes in earth sciences. Springer, Berlin, pp 208226
to allow prediction of debris impact load with consideration of
Armanini, A., Larcher, M. & Odorizzi, M. (2011) Dynamic impact of a debris flow against a
debris-barrier interaction. Back analysis of the Yu Tung Road vertical wall. Proceedings of the 5th International Conference on Debris-Flow Hazards
(granular) debris flow in Hong Kong was carried out to assess Mitigations: Mechanics, Prediction, and Assessment, Padua, Italy, 1417 June 2011,
the performance of the proposed model. The assessment shows pp. 10411049
that, if the new model were to be used, the design bending moment Ashwood, W. (2014) Numerical model for the prediction of total dynamic landslide forces
on flexible barriers. Masters Thesis of Applied Science, Geological Engineering, The
at the barrier wall base could be further reduced more than 30 %
University of British Columbia, 162 p
from the current design approach of Kwan and Koo (2015). The ASI (2013) ONR 24801 protection works for torrent controlactions on structures
new model may provide a potential means to optimise the design (draft). Austrian Standard Institute, Austria, p 25
of rigid debris-resisting barriers. However, further investigation Choi CE, Ng CWW, Song D, Kwan JSH, Shiu HYK, Ho KKS, Koo RCH (2014) Flume
into its applicability to different types and magnitude of debris investigation of landslide debris-resisting baffles. Can Geotech J 51(5):540553
Fitze, P. (2010) Runout analysis of rapid, flow-like landslides. Masters Thesis, Hochschule
flows as well as the effects of barrier geometry and channel
fr Technik Rapperswil, 101 p
geometry would be needed. Hong Y, Wang JP, Li DQ, Cao ZJ, Ng CWW, Cui P (2015) Statistical and probabilistic
analyses of impact pressure and discharge of debris flow from 139 events during
Acknowledgments 1961 and 2000 at Jiangjia Ravine, China. Eng Geol 187:122134
This paper is published with the permission of the Head of the Hubl, J., Suda, J., Proske, D., Kaitna, R., Scheidl, C. (2009) Debris flow impact estimation.
Geotechnical Engineering Office and the Director of Civil Engi- Proceedings of the 11th International Symposium on Water Management and
Hydraulic Engineering, Ohrid, Macedonia, 15 September 2009, pp 137148
neering and Development, Hong Kong SAR Government. The
Hungr O (1995) A model for the runout analysis of rapid flow slides, debris flows and
work described in this paper was supported by a grant from the avalanches. Can Geotech J 32(4):610623
Research Grants Council of the Hong Kong SAR (HKUST 06/CRF/ Hungr O (2008) Simplified models of spreading flow of dry granular material. Can
12R and T22-603/15N). The authors would also like to acknowledge Geotech J 45(8):11561168

Landslides
Iverson R, Logan M, LaHusen R, Berti M (2010) The perfect debris flow? Aggregated Tiberghien, D., Laigle, D., Naaim, M., Thibert, E. & Ousset, F. (2007) Experimental
results from 28 large scale experiments. J Geophys Res 115:F03005 investigations of interaction between mudflow and an obstacle. Proceedings
Kwan, J.S.H. (2012). Supplementary technical guidance on design of rigid of the 4th International Conference on Debris-Flow Hazards Mitigation: Me-
debris-resisting barriers. GEO Report No. 270, Geotechnical Engineering Office, chanics, Prediction, and Assessment, Chengdu, China, 1013 September 2007,
Hong Kong, 88 p. pp 281292
Kwan, J.S.H. & Koo, R.C.H. (2015). Enhanced technical guidelines for design of VanDine DF (1996) Debris flow control structures for forest engineering. Ministry of
debris-resisting barriers. Technical Note No. TN 2/2015, Geotechnical Engi- Forests, British Columbia, p 68
neering Office, Hong Kong, 33 p. Wendeler, C. (2008) Murgangrckhalt in Wildbchen Grundlagen zu Planung
Kwan JSH, Koo RCH, Ng CWW (2015) Landslide mobility analysis for design of multiple und Berechnung von flexible Barrieren. PhD thesis, ETH Zrich, Switzerland (in
debris-resisting barriers. Can Geotech J 52(9):13451359 German)
Kwan JSH, Sun HW (2006) An improved landslide mobility model. Can Geotech J White DJ, Take WA, Bolton MD (2003) Soil deformation measurement using particle
43(5):531539 image velocimetry (PIV) and photogrammetry. Geotechnique 53(7):619631
Law, P.H. (2014) Computational study of granular debris flow impact on rigid barriers
and baffles. PhD thesis, The Hong Kong University of Science and Technology, 252 p
MLR (2006). Specification of geological investigation for debris flow stabiliza-
tion. DZ/T 02202006, National Land Resources Department, China, 32 p (in R. C. H. Koo ()) : J. S. H. Kwan : C. Lam : W. K. Pun
Chinese). Geotechnical Engineering Office, Civil Engineering and Development Department,
NILIM (2007). Manual of technical standard for establishing Sabo master plan Hong Kong SAR Government,
for debris flow and driftwood. Technical Note of NILIM No. 364, Natural Hong Kong, Peoples Republic of China
Institute for Land and Infrastructure Management, Ministry of Land, Infrastructure e-mail: raymondchkoo@cedd.gov.hk
and Transport, Japan, 18 p. (in Japanese)
Peng C, Chao Z, Yu L (2015) Experimental analysis on the impact force of viscous debris C. W. W. Ng : C. E. Choi : D. Song
flow. Earth Surf Process Landf 40(12):16441655 Department of Civil and Environmental Engineering,
Proske D, Suda J, Hbl J (2011) Debris flow impact estimation for breakers. Georisk 5(no. Hong Kong University of Science and Technology,
2):143155 Clear Water Bay, Kowloon, Hong Kong, Peoples Republic of China

Landslides

You might also like