2258
Proceedings of the 18
th
International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013
Water systems strongly develop with Cai river and Ho Lo
river in the west and the east respectively. Both rivers form a
hydrological network distributed evenly in the study area.
1.2
Geological characteristics
A general stratigraphy of the study area composes of following
formations:
Table 1. Brief description of geological formations
Formation
Composition
Phong Hanh formation
(
)
Sph
Quartzite of blocky structure, quartz-sericite
schist
Mang Yang formation
(
)
2
T my
Boulders, cobbles, sandstone, siltstone and
claystone, blocky structure, thick bedding
Kon Tum formation
(
)
2
N kt
Green, dark grey, greenish grey basalt,
bubble or porosity structure, fracturing;
diatomite, clay with diatomite, basalt with
interbedded clay
Dai Nga formation
(
)
2
1
N Q đn
Greenish grey, dark grey basalt, blocky
structure, fracturing
Quaternary formation
Boulders, cobbles, gravel, sand, silt, clay
1.3
Landslide characteristics
Landslide occurs mainly at low areas with elevations less than
200m and gentle slopes (slope angles vary from 5° to 20°) in
heavily weathered rocks which compose of clay, swelling clay
with diatomite… Landscape after sliding forms terraces with
many fractures on the ground surface. Landslide exposures
usually locate at positions associated with hydrological factors.
There are water flows at the toe of slopes in massive landslides.
The shear strength of soils within landslide areas is rather low
(cohesions vary from 12 to 17kPa, friction angles from 10° to
13°).
1.4
Landslide causes
The fieldwork study shows some of causes of landslides in the
study area:
- Heavy landslides ususally occur at geological structures
which compose of clay, clay with diatomite underlying thin
heavily weathered basaltic layers. Soils and rocks at landslide
areas are different in permeability, swelling, strength (weak
soils overlay hard rocks).
- The erosion process occurs at the toe of slopes and piping
develops due to the surface run off and underground water flow.
- Heavy and long rains cause rocks and soils swelling or
liquefaction that in turn cause the volume change suddenly.
- Human activities also take part in the landslide trigger: the
forest clearance, reservoir and rice field preparation… Those
activities result in loosing strength at the toe of slopes.
2 LANDSLIDE SUSCEPTIBILITY MAPPING
2.1
Theoretical Background
Nowadays there are two common groups of methods for
landslide prediction:
a) Knowledge-driven methods: those methods are based on
experiences of experts and rely on empirical relationships.
b) Data-driven methods: those methods analyse collected
data and establish models based on the inherent relationship of
data. This approach, especially, Bayesian conditional
probability method will be applied for the landslide
susceptibility mapping in this research. In Bayesian conditional
probability method, the likelihood of event D given the presence
of factor B can be expressed as the prior probability:
P D B
P B | D
P D | B
P D
P B
P B
(1)
On the contrary, the unlikelihood of event D given the
presence of factor B is the probability:
P D B
P B | D
P D | B
P D
P B
P B
(2)
where
P D | B
the likelihood of event D given the presence of
factor B
P D | B
the unlikelihood of event D given the presence of
factor B
P B | D
the likelihood of factor B given the presence of
event D
P B | D
the likelihood of factor B given the absence of
event D
P D B
the probability of the presence of both event D
and factor B
P B
the probability of the presence of factor B
If there are more datasets (more B factors) involved in the
mapping prediction and those datasets are conditionally
independent to each other with respect to a set of landslide
occurences, the Bayesian conditional probability of event D is
expressed by the following formula:
k
k
logit D W
1
2
n
logit D W
e
P D | B B
B
1 e
(3)
where
P D
logit D ln
1 P D
(4)
P B | D
W log
P B | D
(5)
P B | D
W log
P B | D
(6)
k
W
: positive weight (
W
) or negative weight (
W
) of
factor B depending on the presence or absence of factor
B
The total weight of all B factors at each pixel on the
prediction map is the sum of
, in which
will be the
positive weight
W
k
W
k
W
if the factor B is present or negative weight
W
if B is absent.
The constrast of weight is:
W
C W W
(7)
2.2
Landslide susceptibility mapping
After the landslide survey at the end of year 2010, 142 landslide
occurences are recorded in which 103 positions are used for the
landslide susceptibility prediction by Bayesian conditional
probability and 39 occurences are used for the validation. The
location of 103 landslide occurrences is presented in Fig 2.
Black plus symbols on this figure are landslide occurences.
There are seven factors chosen for the calculation of the
landslide probability: elevation (DEM), slope direction
(ASPECT), slope angle (SLOPE), geology (GEO), distance to
road (DTR), distance to drainage (DTD), topographical wetness
