1

JAXA Sponsored Mini Project on Utilization of Space Technology for Disaster Mitigation
2005/2006


Application of Remote Sensing & GIS Technology for Landslide
Susceptibility Analysis

Introduction
Most of the reported severe natural disasters in Sri Lanka are rain induced landslides and flooding.
Recently, in May 2003, torrential rains accompanied by heavy winds had left an estimated 247 dead and
over 200,000 families displaced due to landslides and floods. The losses from landslides are termed
vulnerability. Other component is landslide hazard. Hazard assessments are estimations of an areas
susceptibility to landslides based on seven inherent physical factors distribution of old landslides,
combined soil cohesion, slope steepness, hydrology, landuse, human interventions (road buffer) and type
of bedrock and its structure. Development of Landslide Susceptibility Map for the Kalawana District
Secretariat Division in the Ratnapura District of Sri Lanka of a scale 1:50,000 developed as a part of the
main research on Application of Remote Sensing and GIS Technology for Landslide Risk Assessment in
Sri Lanka. The study was made to develop multi-temporal GIS database assessment including
deterministic slope stability model SINMAP (Stability Index MAP) and statistical approach of
Weighted Analysis Method for the identification of landslide susceptibility rating. In SINMAP model
accuracy significantly depended with the DEM and input soil parameters. The factor overlay criteria or
the weighted average probabilistic model indicated various conveniences in incorporation of uncertainty
of combined weighted factors like human intervention in road construction in hill slopes.

Study Area
The area covers 387 km
2
and extents from latitude 6
0
35 30 N to 6
0
22 20 N and longitude 80
0
38 25
E to 80
0
17 23 E in the Ratnapura district. The area selected for the study is belongs to the Kalawana
district secretariat division and is having relatively low population of about 48,201 due to undulating
terrains, steep slopes and dense forests.
Objectives
The main objective of the study is to identify the area susceptible to landslides with the use of a GIS
database. The specific objectives are fall in to two major categories. One, is to use deterministic slope
stability model SINMAP (Stability Index MAP) with due consideration of intrinsic triggering factors of
slopes in the region. Secondly, number of extrinsic triggering factors also creates instability of slopes
which is difficult to address in analytical modeling. Therefore, Weight Average Analysis Method was
used to incorporate a number of intrinsic and extrinsic factors for the evaluation of areas susceptible to
landslides. Final outcome is to develop more reliable model for the landslide susceptibility assessment
using GIS.

Methodology
Two types of approaches were considered in this study for landslide susceptibility assessment. First,
deterministic modeling approach consists of number of intrinsic variables such as soil strength
parameters, root cohesion and wetness index. Second, weighted average analysis considers a number of
extrinsic factors which covers human intervention, landuse change, rainfall and intrinsic factors of
geology, slope and stream order. Figure 1 shows a flowchart of methodologies for both methods.
2

The first method was analytical approach, the evaluation of stability index determined using the ArcView
extension, SINMAP. The model uses the formula for the factor of safety (FS) for the infinite slope
stability model (ratio of stabilizing to destabilizing forces). In the FS formula, terrain stability model
SINMAP can compute all variables from the topography, except for combined cohesion; C, tan of
effective angle of internal friction of soil; tan() and wetness index, T/R. The DEM and appropriate soil
parameters were the fundamental input parameters for the model. For the purpose of detailed calibration
of the model, initially the Kalawana division divided in to 10 number of sub watersheds using hydrology
extension of the ArcView 3.2. The analysis was further extended by sub dividing to more sub watersheds
in old landslide proven areas. The evaluation of the T/R ratio and the combined cohesion ( i.e. soil
cohesion and the root cohesion) was determined considering supportive landuse pattern of a watershed.
The uncertainty of these parameters is incorporated through the use of uniform probability distributions
with lower and upper bounds. The stability index (SI) is defined as the probability that a location is stable,
calculated by considering the most and least favorable situations (i.e. either lower or upper bounds) and
the probability that a certain factor of safety may be attained.

In the second method, statistical approach, a number of factor maps were used for the study. These were
slope map derived from the DEM, bedrock geology map, landuse map derived from the aerial photos, soil
map derived from the average combined cohesion concept (i.e. root cohesion and soil cohesion), derived
map of road buffer from slopes, rainfall map and stream density map. Rainfall is the most triggering
factor in most slope instability problems in hill slopes. Therefore, hydrological parameters (rainfall and
stream density) were considered to obtain reasonably accurate weight for the analysis. Three day
cumulative rainfall distribution pattern in a particular watershed and total number of 1
st
order streams in
the watershed were considered. The evaluation of weighted average was made through the Analytical
Hierarchy Process (AHP). Firstly, factor preference from 1 to 9 was allocated for each factors map
convert them in to quantitative value through the model calibration using known landslide data base.

After the valuing of area with regard to seven factors, at present, the values of seven factors classes X are
multiplied by derived individual weights for each factor (w1..w7) and then are summed together. Then
the total value M1 for each pixel and the regional model will be derived:

M1 = w1X1 + w2X2 + w3X3 + w4X4 + w5X5 + w6X6 + w7X7
And with replacing the combined weights (w1..w7) that had been earned previously, the final model
was derived:
M = w1X1 + w2X2 + w3X3 + w4X4 + w5X5 + w6X6 + w7X7

Where: M = Susceptibility coefficient
X1X7 = orderly are related to slope, soil, geology, landuse, hydrology 1(rainfall), hydrology 2(1
st
order
streams) and road buffer and, w1. w7 = are the weights related to each X1. X7 factors. M variations
from 0 to 1, thereafter 5 number of susceptibility classes were assigned as high values attain high
susceptibility landslides.

Field Survey & Laboratory Findings
Laboratory tests of soil reveled that liquid limits were in between 27% to 44%. Water content and the
plasticity index decreases with the depth. The fine content of soils ranges from 4% to 60%. The dry density
increases marginally with depth due to in-situ and completely weathered rock with an average value of
1.5~1.6 and 1.89 Mg/m
3
, respectively. The specific gravity of the soil averages from 2.6 to 2.9. Laboratory
test for shear strength showed an average effective cohesion, C, of 6.5 kPa to 15.5 kPa with effective angle
of internal friction, from 26 to 35. At shear boundary effective cohesion, C = 0 kPa and effective angle
of internal friction, = 22.
3
Results & Discussion
The deterministic method of approach was based on the use of the terrain stability model SINMAP. The
analysis fairly well defines areas that intuitively appear to be susceptible to landslides. It was noted that
few landslides occurred on slope that would not normally be recognized as susceptible to landsliding area.
This means that methodology missed classifying several of these sites as being landslide-prone due to the
site-specific geologic conditions and inaccurate input parameters of soil. The wetness index is indicated
more inaccuracy during preliminary modeling of the area and therefore, re-calibration was made by
further subdividing of landslide proven watershed areas of the main Kalawana watershed. The out put
results were shown in the Table 1 and 2 before and after recalibration.
Table 1: Comparisons of SINMAP results of the model after re- calibration of the watershed number 5 of
the main watershed (Kalawana Division)
Results of SI of the sub watershed number
05 of Kalawana main watershed before re-
calibration of results
Results of the SI after re-calibration of
watershed number 05 in to further 07 sub
watersheds
Stability Index
(SI) and
Class of
stability against
landslides
occurrence
% Area fall in
the stability
class
No. of
Landslides
fall in the
category

Density of
Landslides

LS/km
2

% Area fall in
the stability
class
No. of
Landslides
fall in the
category

Density of
Landslides

LS/km
2

Stable 50.1% 4 0.2% 45.6% 4 0.1%
Moderately
Stable 7.9% 1 0.5% 6.4% 0 0.0%
Quasi-stable 12.6% 7 0.2% 10.4% 2 0.3%
Lower
Threshold 28.1% 3 0.2% 35.0% 8 0.4%
Upper
Threshold 1.3% 1 1.4% 2.8% 2 1.3%
Total 16 0.2% 16 0.2%
Table 2: Comparisons of SINMAP results of the model after re- calibration of the watershed number 08
of the main watershed (Kalawana Division)
Results of the SI of sub watershed number
08 of Kalawana main watershed before re-
calibration of results
Results of the SI after re-calibration of
watershed number 08 in to further 07 sub
watersheds
Stability Index
(SI) and
Class of
stability against
landslides
occurrence
% Area fall in
the stability
class
No. of
Landslides
fall in the
category

Density of
Landslides

LS/km
2

% Area fall in
the stability
class
No. of
Landslides
fall in the
category

Density of
Landslides

LS/km
2

Stable 45.6% 3 0.1% 40.1% 2 0.04%
Moderately
Stable 9.1% 2 0.2% 7.4% 0 0.0%
Quasi-stable 12.9% 1 0.1% 11.2% 5 0.4%
Lower
Threshold 29.1% 16 0.5% 35.4% 12 0.3%
Upper
Threshold 3.1% 0 0.0% 5.0% 3 0.5%
Total 22 0.2% 22 0.2%

4
Table 3: Results of the Landslide Susceptibility Analysis of the Kalawana Division
% Area fall in the equivalent stability class according to the
susceptibility criteria
Sub-watershed
number 05 of main
watershed
Sub-Watershed
number 08 of main
watershed

Overall results of the
Kalawana Main
Watershed

RESULTS

WAA Use of Weighted Average
Analysis method of assessment
model
SINMAP- Use of terrain stability
model of Stability Index Mapping
(SINMAP) WAA SINMAP WAA SINMAP WAA SINMAP
No Susceptibility
46.45
45.6
22.07
40.1
38.67 52.49
Low Susceptibility
8.09
6.4
25.53
7.4
18.99 8.90
Moderately Susceptibility
26.29
10.4
25.44
11.2
22.80 12.11
High Susceptibility 15.78 35.0 21.04 35.4 16.30 24.11
Very High Susceptibility 3.39 2.8 5.91 5.0 3.24 2.19

The above finding indicates the variability of conceptual models and their relative importance with the
input databases. The detailed and comprehensive geographical information databases are required to
obtained the statistical interpretation and calibration of the model. Similarly, soil saturation conditions
and the validity of wetness indices within a watershed also create another avenue to calibrate the
analytical model with the inputs of other soil parameters.

Conclusions & Recommendations
Space technology using satellite and aerial remote sensing plays a very important role in terrain mapping
and scientific assessment of the ground conditions. This technology is ideally suitable for inaccessible
mountainous regions where majority of old landslides were identified. By using multi-temporal satellite
data the areas of landslide every year can be compared with other prediction variables such as altitude,
slope, nearness to settlements, road access and grave sites, and the areas most susceptible to landslide in a
particular year can be flagged for extra precautionary measures to be taken. One of the critical
observations of the study is a lack of complete data base of recent occurrences of old landslides on May,
2003. Only 39 landslides were used out of 78 case records at Kalawana to calibrate models due to
inaccessibility at field. Therefore, multi-temporal satellite data may be only solution to overcome such
problems and acquire more geographical information. In landslide study high spatial resolution satellite
photo images are recommended. The satellite imagery data allowed generation of high-resolution Digital
Elevation Models (DEM). The derived relief parameters can be analyzed in a GIS together with other
information obtained from remote sensing data, thematic maps and field observation for a spatially
differentiated terrain properties as a basis for further assessment of landslide hazard.

The further study should not be restricted to ALOS, LANDSAT TM and IKONOS data, but there would
appear to be great potential in using this and other remotely sensed data, such as airborne radiometrics
(Cranfield, 2003, pers. comm.) to map specific ground conditions including the identification of areas of
alteration and deep weathering that may be additional predisposing factors for landsliding. In the longer-
term, such work will serve to inform relevant sectors of local government of the potential risks associated
with major land development projects.