Content uploaded by Chia-Ming Lo
Author content
All content in this area was uploaded by Chia-Ming Lo on Aug 22, 2014
Content may be subject to copyright.
1 23
Landslides
Journal of the International Consortium
on Landslides
ISSN 1612-510X
Landslides
DOI 10.1007/s10346-013-0435-z
Landslide at Su-Hua Highway 115.9k
triggered by Typhoon Megi in Taiwan
Chia-Ming Lo, Ching-Fang Lee, Hsien-
Ter Chou & Ming-Lang Lin
1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag Berlin Heidelberg. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.
Landslides
DOI 10.1007/s10346-013-0435-z
Received: 8 May 2013
Accepted: 17 September 2013
© Springer-Verlag Berlin Heidelberg 2013
Chia-Ming Lo IChing-Fang Lee IHsien-Ter Chou IMing-Lang Lin
Landslide at Su-Hua Highway 115.9k triggered
by Typhoon Megi in Taiwan
Abstract This study focused on the landslide case at Su-Hua
Highway 115.9k, Taiwan. A preliminary investigation was
conducted on geomorphologic features change and landslide
mechanisms using digital elevation models, geographical maps,
and remote sensing images at different times in conjunction with
geological surveys and analysis results. Using the results of geo-
logical surveys and physical model experiments, we constructed a
discrete element method to simulate the process of landslide
movement. The results revealed deformation in the metamorphic
rock slopes upstream of 115.9k. The slopes around the erosion gully
upstream presented visible slope toes cutting and tension cracks at
the crest as well as unstable rock masses. According to the results
of numerical simulation for typhoon Megi event, intense rains
could induce slippage in the rock debris/masses in the source area,
initially at a speed of 5–20m/s. Subsequently, steeper terrain could
cause the rock debris/masses to accelerate to form a high-speed
(>30m/s) debris slide quickly moving downstream to form an
alluvial fan downstream by the sea.
Keywords Su-Hua Highway 115.9k .Landslide mechanism .
Geological surveys .Discrete element method .
The process of landslide movement
Introduction
The Su-Hua Highway is the only highway linking the northern and
eastern regions of Taiwan. Much of the terrain along the route is
steep, and many sections were carved from mountains bordering
the sea, where slopes are constantly eroded by weathering, earth-
quakes, typhoons, and heavy rain. From time to time, collapses
disrupt traffic and seriously threaten the safety of road users.
On October 13, 2010, Typhoon Megi developed into a super
typhoon, when the northeast monsoon had begun in Taiwan. The
combined effects produced uncommonly high rainfall. Rainfall at the
Suao monitoring station was 939.5 mm in a single day, and 181.5 mm
between 13:00 and 14:00 on October 21, both of which established
new records for the site. The rainfall also triggered large collapses on
upper slopes and the loss of the roadbed on the Su-Hua Highway
(104k–117k), claiming 26 lives. This was the most devastating natural
disaster on the highway since its opening.
During Typhoon Megi, a debris slide upstream from Dakeng
Bridge on the Su-Hua Highway (115.9k) was the most serious
disaster (Fig. 1). Based on aerial photos, the total area and volume
of the collapse were estimated at 10 ha and 2.1×10
6
m
3
, respectively.
The resulting debris flow struck the road to the southeast and was
deposited along the Pacific coastline, which created an alluvial fan
covering about 4 ha. The collapses were primarily debris from
colluvium and weathered rock (Chou et al. 2012; Lee 2010). Other
than Typhoon Megi, Su-Hua Highway 115.8k was also damaged in
Typhoon Parma (October 2009), Typhoon Nalgae (October 2011),
and Typhoon Tembin (December 2012).
According to a preliminary investigation (Chu et al. 2012), the
debris flow at Su-Hua Highway 115.9k during Typhoon Megi can be
attributed to three crucial points. (1) Significant headward erosion
upstream from this section of road formed erosion gullies
extending from the same source area, which caused a gradual
expansion of the landslide along the south side of the Tungaoling
ridgeline. (2) Material comprising the rock slope in the valley
upstream of Dakeng Bridge was fractured and foliated. Under
the gravity force, the rock strata gradually deformed, and gully
erosion destabilized the slope toes on the gully sides causing
collapses. (3) A substantial amount of colluvium had accumulated
in the gully, which facilitated the infiltration of surface water. This
caused the fragmented amphibolite and colluvial deposits above
the fresh rock strata to become saturated. In heavy rain, the gully-
bottom deposits quickly migrated downstream, thereby enhancing
bank erosion and increasing the probability of collapse. The
source-slope terrain, the geology, and the distribution of colluvium
were crucial factors in the collapse. However, few in-depth inves-
tigations have been conducted on the evolution of this landscape
or surrounding landslide history.
Large high-velocity landslides, especially rockslides that devel-
op into debris avalanches (sturzstroms), are remarkable geological
phenomena (Steven and Simon 2006). Their speed is important in
determining the destructive potential (Hungr 2007). Terrain-
caused features such as mass separation, collision interaction,
and kinematic change from mass sliding to particle flow are
important in estimating the sphere of influence in order to protect
property downstream (Lo et al. 2011a,b). This study combined the
analysis of geomorphologic history and remote sensing images to
explain geographical changes and summarize the characteristics of
the landslide at Su-Hua Highway 115.9k. The study also numeri-
cally simulated the collapse to estimate the landslide speed. Thus,
we were able to explain the process and characteristics of the
landslide. This paper also summarizes the process of landslide
hazard zonation of the area.
Method
Terrain development and geomorphological interpretation
The study area is located in Suao Township of Ilan County, Taiwan
(Fig. 1). Geographical data included topographic maps from 1936
(1/50,000) and 1996 (1/25,000), a 5 m×5 m digital elevation model
(DEM) created by the Aerial Survey Office of the Forestry Bureau
in 2004, 1 m×1 m Lidar data following the disaster in 2010, and
orthophotos taken before and after the disaster (1/10,000).
According to aerial photos and geographical data (Fig. 2), the
important landmarks from northwest to southeast include
Tungaoling (elevation 821 m), Northern Unnamed Creek (750–
280 m), Southern Unnamed Creek (elevation 750–50 m), and
Dakeng Bridge (elevation 260 m). The elevations of the source
and deposition areas differ by approximately 700 m; the slope of
the creek bed from upstream to downstream ranges between 15°
and 50°, which was a critical influence in the landslide. The
watershed of the Tungaoling source was primarily the Southern
Landslides
Technical Note
Author's personal copy
Unnamed Creek with the Northern Unnamed Creek as secondary.
The two tributaries played significant roles in the landslides here,
both in this event and in other historical landslides.
The source area at the two unnamed creeks and the two banks
of the valley form an obvious hummocky surface (Fig. 2), more
significantly developed in the Northern than in the Southern
Unnamed Creek. Sliding masses were observed to be compressing
the creek bed upstream.
The landslide zone primarily included the hummocky area in the
upstream portion of the Southern Unnamed Creek, which in earlier
times may have been the location of debris deposits or rock-slope
deformation, before heavy rains initiated their collapse or sliding.
Fig. 1 The location of study area and the disaster situations after Typhoon Megi
Fig. 2 Three-dimensional terrain (left) and post-collapse aerial photo (right) of study region
Technical Note
Landslides
Author's personal copy
The watershed upstream of the Northern Unnamed Creek reaches to
elevations of 500 to 560 m and contains substantial hummocky
surfaces, and compressive folding disrupts the creek flow. The creek
may have been buried and blocked in earlier periods.
The focus of analysis in the terrain development and geomorpho-
logical interpretation include the Northern Unnamed Creek and the
Southern Unnamed Creek. Analysis of historical data and images
helped to clarify the accumulation of colluvium in the source valley
and links between rock deformation and changes in the tributaries.
Geology
A geological survey was conducted onsite to help identify landslide
mechanisms. The base map was the 1/50,000 Suao geological map
(Central Geological Survey of the Ministry of Economic Affairs, 1997;
Fig. 3). The landslide geology was then simplified for numerical
modeling. Onsite surveys and geomorphological comparisons were
performed in June 2011, July 2012, and January and February of 2013.
Geomorphological analysis helped to reveal the processes involved
in changing the terrain and the landslide mechanisms in the valley.
Overlaying the 1/50,000 geological map of Suao on 2012 satellite
images (Fig. 3) revealed that the study site is primarily Tananao
schist. The strata protruding into Tungaoling and neighboring
areas include Tungao schist from the late Paleozoic, Fongshushan
amphibolite from the late Paleozoic to Mesozoic, Nasuao forma-
tions from the Eocene to Oligocene, Suao formations from the
Miocene, and modern alluvium (Taipei Association of Applied
Geology Engineer 2011).
The exposed Tungao schist consists of two amphibolite rocks
(Taa). The one in the north extends from the Wuyenchiao coast-
line to the southeast, past Tungaoling, and then west to the south-
ern foot of Hsimaoshan in a band-like distribution. The rock mass
is about 7 km long and ranges in width between 200 and 1,000 m.
Outcrops include green to dark green schistose masses with
extremely high strength and brittleness. Cored samples contain
kink bands. This rock formed flakes and blocks at Tungaoling
(Taipei Association of Applied Geology Engineer 2011). The schis-
tosity of the local strata is oriented in the northwest–southeast
direction, tilted towards the southwest at approximately 40° to 70°
with significant fracturing. The key strata distributed within the
study area include amphibolite, graphite schist, chlorite schist, and
quartz–mica schist (Ta). A fault passes through the area down-
stream from Dakeng Bridge (elevation 150–200 m) (approximate
location is shown as a white dotted line in Fig. 3).
Numerical simulation
The three-dimensional discrete element program PFC3D™(Itasca
2002) was used to simulate the movements of the landslide at Su-
Hua Highway 115.9k. The material strength parameters requiring
input in PFC3D include normal stiffness, shear stiffness, and the
stiffness and strength of parallel bond between elements and were
set as outlined by Potyondy and Cundall (2004) using the results of
uniaxial compression tests on field samples (metamorphic rock and
schist) as a reference. We compared the results of actual uniaxial
experiment with PFC simulations (Fig. 4) in order to calibrate the
conversion formula for macro- and micro-parameters and to derive
the micro-parameters related to materials in the simulation.
To simulate the energy dissipation in material collision, we
referred to the coefficient of restitution obtained in onsite tests
by Giani et al. (2004) to convert damping parameters (Table 1).
This provided parameter settings that were more consistent with
field observations. The coefficient of kinetic friction plays a key
role in the motion of catastrophic landslides; however, measuring
it is difficult. Thus, the study compared the deposit formations
resulting from simulations with those produced using various
coefficient settings and selected the setting that came closest to
the actual landslide deposit.
Fig. 3 Geological map (1/50,000) of Tungaoling (modified from Central Geological Survey data)
Landslides
Author's personal copy
Results
Results of terrain development analysis and geomorphological
interpretation
(a) 1936: The topographical map of 1936 shows significant headward
erosion at two sites in the upstream section of the Southern
Unnamed Creek, and the distribution of erosion collapse is
situated precisely where the primary collapse of the Typhoon
Megi landslide occurred (bright white location in the left image
of Fig. 5a). Moreover, the contours of the slopes between the
source tributaries protrude slightly, which is early evidence of
the mechanisms underlying the erosion of the valley and defor-
mation features that played a crucial role in the large landslide to
follow. In contrast, the Northern Unnamed Creek originated
from the east side of Tungaoling. The valley in the upstream
section is steeper than that of the Southern Unnamed Creek. On
the left bank of the valley at elevations between 300 and 500 m,
localized hummocky ground is visible, and the terrain at the
source is more V-shaped than that in the Southern Unnamed
Creek. Less geomorphological evidence of collapse was appar-
ent; therefore, the development of collapse on the two creek
banks was less pronounced.
(b) 1996: The right image in Fig. 5a shows five additional hum-
mocky surfaces at the source of the Southern Unnamed
Creek, which developed before 1996. In addition, erosion
gullies to the southwest of Tungaoling at an elevation of
approximately 800 m were more progressed. By overlapping
post-event satellite images from Formosat-2, we discovered
that the primary bodies associated with the collapse of the
Typhoon Megi landslide events (bright white location in the
right image of Fig. 5a) were located within the range of the
hummocky surfaces. Particularly in the period between Ty-
phoon Megi and 2013, the landslide zone expanded to the
hummocky surfaces on Tungaoling above elevations of
800 m. This indicates that in addition to the distribution of
erosion leading to the collapse within the study area, rock
slope deformation or the debris deposits may also have been
key factors in large-scale landslide following heavy rains.
(c) 2004 and 2010: Fig. 5b indicates that the primary landslide
zones in the upstream section of the Southern Unnamed
Creek in the Typhoon Megi event changed from hummocky
surfaces to gully terrain (the tips of the V-shaped contours
progressed towards higher elevations). This shows that much
of the deposit or locations of rock deformation with hummocky
Fig. 4 Comparison of test results from simulated uniaxial compression and actual test results
Table 1 Conversion of onsite damping parameters (modified from Giani et al. 2004)
Normal restitution
coefficient
Converted normal
damping ratio
Shear restitution
coefficient
Converted shear
damping ratio
Bedrock slope 0.50 0.21 0.95 0.02
Bedrock slope covered with broken rock 0.35 0.32 0.85 0.05
Slope covered with rock debris and soil 0.30 0.36 0.70 0.11
Soil slope covered with lush vegetation 0.25 0.40 0.55 0.20
Technical Note
Landslides
Author's personal copy
surfaces in the upstream portion of the Southern Unnamed
Creek slid toward midstream and downstream deposit areas
during this event. However, a portion of the hummocky sur-
faces in the source area (at elevations between 700 and 800 m)
still exist and their slope toes have collapsed and fallen down-
stream. For this reason, concern remains over the possibility of
landslides and cliff retreat on Tungaoling. Furthermore, despite
the occurrences of only minor debris slides in the Unnamed
North Creek during the event in question, six hummocky sur-
faces are still apparent near the valley and may be locations of
rock deformation or previous landslide deposits on the river
banks. They should therefore also be continually monitored in
the future.
Results of geological survey
The purpose of the geological survey was to clarify the key mech-
anisms underlying the large-scale debris slide at Su-Hua Highway
115.9k. The primary results are described below.
(a) Results from analyzing the geological survey in the upstream
section of the Southern Unnamed Creek (Fig. 6)showa
minor scarp produced by a down-sliding tension crack in
the landslide area on the southern side of Tungaoling
(Fig. 6a). Protruding rock within the landslide zone was
mostly fragmented amphibolite, and due to gravity or tecton-
ic stress in the metamorphic rock layer, considerable forced
deformation was apparent in the sliding surface and in rock
layers near the toe of the slope (Fig. 6a, b), which increased
fragmentation in the bedrock. An investigation of water
seepage at the slopes indicated that most seepage in the
slopes was located at elevations between 450 and 500 m near
the Southern Unnamed Creek, which was close to the junc-
tion between the amphibolite and schist. Based on previous
experience gained during the New Yungchuen Tunnel pro-
ject (Fu et al. 2004), we know that this occurs near contact
surfaces between amphibolite and schist. The areas near the
contact zone were extremely fragmented, particularly in
regions rich in quartz veins under stress. These zones often
form permeable strata. Moreover, fault gouge was distribut-
ed along a portion of the contact surface, which constituted
impermeable strata. The stratum at the Tungaoling source
area is amphibolite, which is harder than schist. This stra-
tum is generally in areas of concentrated tectonic stress,
such that rock joint structures and fractures are more prom-
inent than in the schist. This enabled surface water to
infiltrate into the less permeable schist layer at the source
of Northern Unnamed Creek during rain. The water traveled
along this weak surface to form springs near the contact
zone. This was the primary reason for the continuous ex-
pansion of the landslide zone at the Southern Unnamed
Creek (Fig. 6c, d).
(b) No significant collapses occurred during Typhoon Megi in
the upstream section of the Northern Unnamed Creek
(Fig. 7), but a considerable amount of material was deposited
in the creek valley (visual observations showed deposit
heights of approximately 7 to 10 m), most of which was
derived from amphibolite, the primary composition of the
valley banks near the landslide (Fig. 7a, b). The deposits
found in the areas with hummocky surfaces were much
greater in volume and area; however, protruding rock layers
were rare. In the areas with non-hummocky surfaces, we
discovered slightly deformed rock layers protruding from
both banks of the creek and severe erosion of the slope toes
(Fig. 7c). Tension cracks were also observed along the top of
the rock layer, which are likely the main causes of collapse on
the two valley sides. Areas of seepage on the slopes of the
Northern Unnamed Creek are at elevations of 350–400 m, at
the junction between amphibolite and schist. The terrain is
steeper on average than in the Southern Unnamed Creek and
could lead to large landslides; therefore, this area could be a
focus for future monitoring.
Fig. 5 aTopographic maps of Tungaoling in 1936 (left) and 1996 (right). Basemaps comprise post-event satellite images from Formosat-2 (2010). bTopographic maps
of Tungaoling in 2004 (left) and 2010 (right). Basemaps comprise post-event satellite images from Formosat-2 (2010)
Landslides
Author's personal copy
(c) The strike of the schistosity of the local rock layers in the
Southern Unnamed Creek (Fig. 8) is northwest–southeast,
dipping to the southwest at 40–70°. The rupture surface
was well developed with two or more sets of joints and
weak surfaces. We also found exfoliation (sheeting) be-
tween the amphibolite and schist, which was parallel to
the slope. This results in cave-ins and tension cracks, also
parallel to the slope surface, in the road surface at 115.8k
and in the slopes of the creek banks (Fig. 8). The bedrock
was planar, and not smoothly protruding, from the surface
in areas of collapse at the source of the Southern Unnamed
Creek (Fig. 6a). This was the key weakness in the surface,
which caused the collapse (exfoliation; sheeting) (Chu et al.
2013). It may also have affected the deep-seated sliding
surfaces that created the large landslide along the Su-Hua
Highway.
(d) There are several areas of fault gouge forming a localized
impermeable stratum on the north slope of Dakeng Bridge
and on the lower slopes of Su-Hua Highway 115.8k to 115.9k
(Fig. 9). The right bank of the valley has several fault zones
under 1 m in width traveling from east to west and sloping
steeply toward the south. The left bank of the valley has
Fig. 6 Onsite survey at upstream portion of Southern Unnamed Creek (using a post-event aerial photo as the basemap)
Fig. 7 Onsite survey scene at upstream portion of Northern Unnamed Creek (using a post-event aerial photo as the basemap)
Technical Note
Landslides
Author's personal copy
two to three fault lines approximately 4 m in width travel-
ing from north to south and sloping slightly toward the
east. These may explain the on-going collapses near 115.8k
in recent years and should be another focus for future
monitoring.
(e) The collapse in the upper slopes at Su-Hua Highway 115.9k
can be divided into three phases (Fig. 10).
&The first phase was development of fractures in the rock
slopes (Fig. 10, Step1). Due to the free surfaces of the valley
sides and release of tectonic stress, the slopes along the
valley have exfoliated sub-parallel to the slope. Earlier exfo-
liation accelerated headward erosion, and surface water
seeped into the slope along exfoliation joints before exiting
in springs where the amphibolite is more fractured.
Fig. 8 Results of onsite survey and the measurement of surface weaknesses in the midstream portion of the Southern Unnamed Creek (using a post-event aerial photo as
the basemap)
Fig. 9 Survey of lower slopes at Su-Hua Highway 115.8k–115.9k (using a post-event aerial photo as the basemap)
Landslides
Author's personal copy
&The second phase was toe erosion and rock-slope defor-
mation (Fig. 10, Step2). The extent of exfoliation and the
amphibolite being harder and more fractured than the
quartz–mica schist made it easier for surface water to
infiltrate into the slope. Groundwater then percolated
through the permeable stratum to exit above the imper-
meable stratum formed by localized fault gouge near the
slope toe. Under the double influence of slope erosion
from groundwater and debris flow in the valley, the slope
toe was easily damaged. A loss of support from beneath
caused the area over the toe to deform, which caused
slippage and creep between the strata in the slope and
intensified the fracturing (Fig. 6b).
&During the third phase, increasingly severe fracturing, and
opening of fractures in the body of the slope, enabled
substantial water infiltration into the permeable stratum
initiated the collapse (Fig. 10, Step3). The groundwater per-
colated through the fragmented amphibolite rock strata and
along the less permeable schist stratum beneath it to accu-
mulate on the impermeable stratum near the toe of the
slope, undermining the stability of the entire slope. The
torrential rains brought by Typhoon Megi caused the
groundwater levels within the slopes in the source area to
rise rapidly, thereby expediting erosion in the slope toe and
destabilizing the entire slope. From the stair-like appearance
of the sliding surface produced by the collapse in the South-
ern Unnamed Creek source, we deduced that the collapse
gradually progressed to the slope crest. Therefore, during
this landslide, infiltration of substantial quantities of rainfall
and increased slope erosion in the valley caused the section
between the slope toe and the slope face to collapse first. In
that instant, the near-saturated sliding mass behind became
unstable and slid downstream. The mechanisms underlying
this collapse should be considered in slope monitoring or
remediation projects along the Su-Hua Highway.
Landslide simulation
The numerical model
Due to limitations in computational capacity, we used a 5 m×5 m
DEM to construct the sliding surface and movement terrain. In the
simulation, we only considered the processes of sliding and deposition.
Due to software limitations (PFC3D 3.0), factors that commonly trigger
landslides (such as sliding caused by rain infiltration) were not included.
The sliding surface in the numerical model was based on a post-
event (2010) 1 m×1 m DEM, but in the deposit area, we used a pre-
event (2004) 5 m×5 m DEM. A total of 78,786 wall elements were used
to construct the terrain (Fig. 11) over a length of 2,000 m and a width of
450 m. The sliding mass included the primary sliding mass and the
midstream and downstream colluvium, comprising 15,764, 5,088, and
4,044 ball elements (radii ranging between 1 and 4 m), respectively. We
monitored the speed of the simulated primary sliding mass to help
explain the interactions observed during the movement of rock mass.
The landslide at Su-Hua Highway 115.9k was triggered by heavy rain,
and weathering and rust-staining patterns were found on the surfaces
of rock fragments after sliding. Therefore, we decreased the coefficient
of friction on the sliding face during analysis and weakened the
strength of bonds between rocks so that the conditions of the debris
slide simulation were closer to those observed in the field.
Selection of parameter settings
(a) Particle contact stiffness and bonding strength: For the prelimi-
nary conversion of macro- and micro-parameters, the study
referred to test results from Tungao core samples (Directorate
General of Highways 2010) and conversion formulas (Potyondy
and Cundall 2004). We compared actual uniaxial compression
tests and simulation experiments (Fig. 4) to revise the conversion
formula for macro- and micro-parameters and derive the micro-
parameters of element contact stiffness and bonding strength in
the environment of the numerical simulation (Table 2).
(b) Damping coefficients: The study referred to the damping pa-
rameter conversion method adopted by Giani et al. (2004).
Because the two banks and the source area at the Southern
Unnamed Creek were covered with lush vegetation prior to the
landslide event, this study set the normal and shear damping
coefficients at 0.4 and 0.2 as a reference.
(c) Coefficient of friction: We initially adopted the coefficient of
static friction measured between onsite deposits of rock and
between onsite rocks and the slope. The variables compared
in the deposit formation included run-out distance and the
width of deposits. The simulations show that when the coef-
ficient of friction between the elements equaled 0.1, the
resulting formation and deposition of terrain in the simula-
tion was closest to the results observed on post-event aerial
photos (Table 3and Fig. 12). The fastest sliding speed reached
Fig. 10 Diagram of collapse mechanism: Step1 fissure development stage. Step2 slope toe erosion and deformation stage. Step3 Gradual collapse stage
Technical Note
Landslides
Author's personal copy
52.2 m/s (approximately 188 km/h), enabling the entire sliding
body to move downstream to the coast. As a result, we
selected this parameter to establish a numerical model to
explain the kinematic movements involved in the landslide
at Su-Hua Highway 115.9k.
Simulation of landslide movement process
To explain the processes involved in the debris slide triggered by
Typhoon Megi in 2010 at Su-Hua Highway 115.9k, we monitored
the simulated movement speeds throughout the entire process.
This revealed variations in speed during different stages and
clarified the key processes involved in the overall movement.
The results are outlined as follows.
&Figure 13 presents the simulated variations in landslide speed.
The model volume distribution was based on subtracting the
pre-event from the post-event DEM. From upstream to down-
stream, we designated five primary landslide masses. Two of
these were situated in the upstream source area (masses A and
B), one in the midstream section (mass C), and two in the
curve of the valley at the location of the collapse (masses D and
E). The sliding masses A and B slumped directly downstream
Fig. 11 Numerical model of landslide movement at Su-Hua Highway 115.9k
Table 2 Parameters of PFC numerical simulation
Parameter Uniaxial experiment simulation Full-scale numerical model
Number of particles 28,250 24,896
Unit weight of ball elements (kg/m
3
) 2,300 2,300
Range of particle radius (m) 0.0025–0.003 1–4
Normal stiffness (N/m) 1.2e8–1.44e8 4.8e10–1.9e11
Shear stiffness (N/m) 6.0e7–7.2e7 2.4e10–9.5e10
Friction coefficient of ball elements 0.6 0.05–0.2
Friction coefficient of wall elements 0.6 0.6
Normal stiffness of parallel bond (N/m) 1.66e12–2.4e12 1.5e9–6.0e9
Shear stiffness of parallel bond (N/m) 8.3e11–1.2e12 7.5e8–3.0e9
Normal bond strength of parallel bond (MPa) 30 30
Shear bond strength of parallel bond (MPa) 15 15
Normal damping coefficient 0.4 0.4
Shear damping coefficient 0.2 0.2
Landslides
Author's personal copy
as a result of gravity, whereas masses C, D, and E maintained
higher self-supportability due to their larger coefficient of
friction (0.6) and greater bonding strength. When masses A
and B came into contact with masses C, D, and E downstream,
the lack of support from the slope toe, which had been
hollowed out by erosion, caused them to collapse. This se-
quence matches the landslide mechanisms derived from the
onsite survey. The simulation also used a PFC3D fluid option
to install a water body downstream, where the sea would be.
When particle elements encountered the water body, their
movement speed was reduced by fluid resistance, causing them
to accumulate in an alluvial fan.
&The simulated landslide progressed through seven phases from
sliding, accelerating, and decelerating to a final stop (Figs. 13 and 14);
&In the first phase, masses A and B began moving (at the 5-s
point in Fig. 13). The speed of the landslide in this phase
ranged between 5 and 20 m/s (∼18–72 km/h) with an
average speed of 9 m/s (∼30 km/h, as shown in Fig. 13).
&Second phase (15-s point in Fig. 13): Masses A and B moved
past the steepest terrain in the region, converged at the
Southern Unnamed Creek, and reached the toe of mass C.
At this point, the front portions of masses A and B were
moving at the highest speed of 52.2 m/s with an overall
average of 20.4 m/s (Fig. 14). The rear portions were also
beginning their accelerated descent.
&Third phase (22.5-s point in Fig. 13): The front portions of
masses A and B arrived at the highway where the Chuang-Yi
tour group bus was parked. Because the terrain in the valley
turns gradually from southeast to south, the front portion of
the sliding mass began slowing down (the average speed
reduced to 10.36 m/s, Fig. 14). Here, the landslide gradually
transformed from a debris slide to a debris flow. Due to toe
erosion, mass C became unstable and began collapsing.
&Fourth phase (40-s point in Fig. 13): By this time, mass C
had collapsed, and the front portions of masses A and B
had reached mass D, situated on an undercut slope at the
bend of the watercourse. The front portions of masses A
and B accelerated again (increasing to 27.5 m/s) speeding
the erosion of the toe of mass D.
&Fifth phase (60-s point in Fig. 13): Mass D collapsed, and
the front portions of masses A and B had passed another
bend in the creek, past the toe of mass E. Erosion by the
sliding masses from upstream destabilized mass E on the
lower slope of Su-Hua Highway 115.8k.
&Sixth phase (70-s point in Fig. 13): By this time, mass E had
collapsed, and a portion of the sliding masses was beginning
to pass the second bend before reaching the sea. Due to the
curve of the valley, the masses accelerated again, reaching
speeds of 12.40–12.57 m/s (Fig. 14)duringtheperiod70–77.5 s.
At about 77.5 s, the speeds began gradually decelerating. At
70 s, the front portions of masses A and B had reached the sea
and were slowing down in the water.
&In the seventh phase (157.5-s point in Fig. 13), an alluvial fan
formed, ending the landslide movement at 157.5 s.
Table 3 Comparison of friction coefficients
Coefficient of friction Landslide movement
distance (m)
Deposit
width (m)
0.05 1,375 427
0.1 1,268 384
0.15 1,192 363
0.2 1,136 325
Pre-event measurements
(DEM)
1,276 387
Fig. 12 Comparison of numerical simulation results and post-event images. The orange area on the lower slope of Su-Hua Highway 115.9k presents the final simulation
results. SM-A the slide mass A, SM-B the slide mass B, SM-C the slide mass C, SM-D the slide mass D, SM-E the slide mass E
Technical Note
Landslides
Author's personal copy
Application
The applications of the model are as follows:
(a) Zoning similar landslide areas: The area above Su-Hua High-
way 115.9k has three geological and topographical character-
istics: (1) the geological structure includes a junction between
amphibolite and schist, and combines both permeable strata
(fracture zone) and impermeable strata (fault gouge). (2)
Topographical characteristics include hummocky deposits
and rock deformation, and the toes of the deposits were
unstable due to erosion. (3) The valleys of the tributaries were
narrowed by sliding masses on the two banks. In some cases,
talus cones or alluvial fans were apparent downstream. The
terrain upstream of Su-Hua Highway 115.9k still presents
similar geological structures and topographical characteris-
tics; in heavy rains, significant large landslides may still
occur; therefore, this area should be closely monitored.
(b) Assessment of landslide potential and hazard zonation: Zon-
ing of areas with landslide potential and sensitivity to geo-
logical disasters has been performed throughout the
metropolitan and mountainous areas of Taiwan for nearly a
decade. However, due to immature landslide simulation tech-
niques, these assessments still require further study. This
study used techniques capable of overcoming previous inad-
equacies in landslide simulation.
(c) Pre-disaster scenario simulation: When coordinated with
slope monitoring and 3D stability analysis, the presented
numerical model could provide accurate pre-disaster scenar-
io simulations. The dynamic presentation of large landslide
movements can provide disaster prevention units with a
valuable reference, and the simulation results should increase
the accuracy of assessment and alerts.
(d) Early warning systems for downstream villages: Knowledge
of movement speeds and affected range could enable plan-
ning for early warning and establishment of refuges.
Fig. 13 Movement process and speed variations in the simulated landslide at Su-
Hua Highway 115.9k. SM-A the slide mass A, SM-B the slide mass B, SM-C the
slide mass C, SM-D the slide mass D, SM-E the slide mass E
Fig. 14 Average speed statistics of simulated landslide at Su-Hua Highway 115.9k
Landslides
Author's personal copy
Conclusion
This study investigated the geomorphic evolution, landslide mech-
anism, and kinematic characteristics of the landslide area sur-
rounding Su-Hua Highway 115.9k in Taiwan. A numerical model
revealed the variations in the speed of the moving material and the
area of influence of the large landslide.
Analysis of geomorphology revealed hummocky surfaces in several
upstream areas of the Tungaoling source area in 1936. Collapses and
continued rock deformation in the banks of the source creeks devel-
oped into five hummocky surfaces at the Southern Unnamed Creek
prior the large collapse, which occurred during Typhoon Megi. On the
banks of the Northern Unnamed Creek, six significant hummocky
areas remain and should be monitored in the future. A field survey
revealed distributions of fault gouge in the mid- to downstream
section of the valley, which forms an impermeable stratum within
the slope body. Erosion of the slope toe caused by multiple collapses
has made the entire slope unstable, such that debris slides and under-
cutting of the roadbed may still occur following heavy rains.
Numerical simulation of the landslide movement illustrated seven
phases involved in the collapse at Su-Hua Highway 115.9k. The initial
speed of the sliding masses at the source was 5–20 m/s. After sliding for
15 s, the mass descended steeper terrain at an elevation of 550 m
reaching a peak of 52.2 m/s. At 22.5 s, the mass struck Su-Hua Highway
and the vehicles on it, with devastating effect. The front portion of the
sliding mass subsequently decelerated due to the leveling of the
terrain, which transformed the landslide from a large-scale debris slide
to a debris flow. The lower slope of 115.8k was located at an undercut
slope at a bend in the creek. When the debris flow passed by this area,
the lower slope was subjected to severe lateral erosion, which triggered
collapses in the lower slope. At approximately 157.5 s, most of the
debris flow had been deposited near the sea as an alluvial fan.
Acknowledgments
The research is mainly supported by the National Science Council
of Taiwan, grant no. NSC 101-2218-E-270-001. The advice, com-
ments,andhelpprovidedbytheeditorandtwoanonymous
reviewers have significantly strengthened the scientific soundness
of this paper. Their kind assists are gratefully acknowledged.
References
Chu HK, Lo CM, Chang YL (2012) “Stability issues and preliminary analysis of Su-Hua
highway with Su-Hua highway 115.9k as an example”, 2012 Taiwan Rock Engineering
Symposium, Chungli, pp 579–588
Chou HT, Lee CF, Lo CM, Lin ZP (2012) “Landslide and alluvial fan caused by an extreme
rainfall in Suao, Taiwan.”11th International Symposium on Landslides (ISL) and the
2nd North American Symposium on Landslides. Banff, Alberta, Canada, pp 1–7
Chu HK, Lo CM, Chang YL (2013) Numerical analysis of slope stability at the 115.9k point
of the Su-Hua highway. J Chin Soil Water Conserv 44(2):97–104
Directorate General of Highways (2010) Feasibility study and priority construction
planning tasks for the improvement of the mountainous sections on highway
9, the Su-Hua highway, Geotechnical Engineering Survey and Documentary
Report
Fu R, Cha NT, Hsueh WC (2004) The measures adopt to solve the NYC Tunnel high
pressure water. Chin J Rock Mech Eng 23(2):1–7
Giani GP, Migliazza M, Segalini A (2004) Experimental and theoretical studies to improve
rock fall analysis and protection work design. Rock Mech Rock Eng 37:369–389
Hungr O (2007) Dynamics of rapid landslides. Progress in landslide science, chapter 4.
Springer, Berlin, pp. 47–56
Itasca Consulting Group Inc. (2002) PFC3D particle flow code in 3 dimensions. User’s
guide. Itasca Consulting Group, Minneapolis
Lee YL (2010) Aerial survey of geological disasters along Su-Hua highway. Geology
Quarterly 29(4):21–25
Lo CM, Lin YH, Jeng TY, Wei LW, Chi CC, Lin HH (2011) “Kinematic process of the
catastrophic landslide through investigation and numerical simulation”, 14th Confer-
ence on Current Researches in Geotechnical Engineering in Taiwan, pp 1–12
Lo CM, Lin ML, Tang CL, Hu JC (2011b) A kinematic model of the Hsiaolin landslide
calibrated to the morphology of the landslide deposit. Engineering Geology 123:22–
39
Potyondy DO, Cundall PA (2004) A bonded-particle model for rock. Int J Rock Mech Min
Sci 41:1239–1364
Steven NW, Simon D (2006) Particulate kinematic simulations of debris avalanches:
interpretation of deposits and landslide seismic signals of Mount Saint Helens, 1980
May 18. Int J Geophys 167:991–1004
Taipei Association of Applied Geology Engineers (2011) Comprehensive geological safety
evaluation report for Tungaoling Area. Ministry of National Defense, R.O.C
C.-M. Lo ())
Department of Civil Engineering,
Chienkuo Technology University,
Kaohsiung, Taiwan, Republic of China
e-mail: ppb428@yahoo.com.tw
C.-F. Lee
Disaster Prevention Technology Research Center,
Sinotech Engineering Consultants, Inc.,
Taipei, Taiwan, Republic of China
H.-T. Chou
Department of Civil Engineering,
National Central University,
Chung-Li, Taiwan, Republic of China
M.-L. Lin
Department of Civil Engineering,
National Taiwan University,
Taipei, Taiwan, Republic of China
Technical Note
Landslides
Author's personal copy
