ArticlePDF Available

Landslide Susceptibility Mapping in Darjeeling Himalayas, India

Wiley
Advances in Civil Engineering
Authors:

Abstract and Figures

Landslide susceptibility map aids decision makers and planners for the prevention and mitigation of landslide hazard. This study presents a methodology for the generation of landslide susceptibility mapping using remote sensing data and Geographic Information System technique for the part of the Darjeeling district, Eastern Himalaya, in India. Topographic, earthquake, and remote sensing data and published geology, soil, and rainfall maps were collected and processed using Geographic Information System. Landslide influencing factors in the study area are drainage, lineament, slope, rainfall, earthquake, lithology, land use/land cover, fault, valley, soil, relief, and aspect. These factors were evaluated for the generation of thematic data layers. Numerical weight and rating for each factor was assigned using the overlay analysis method for the generation of landslide susceptibility map in the Geographic Information System environment. The resulting landslide susceptibility zonation map demarcated the study area into four different susceptibility classes: very high, high, moderate, and low. Particle Swarm Optimization-Support Vector Machine technique was used for the prediction and classification of landslide susceptibility classes, and Genetic Programming method was used to generate models and to predict landslide susceptibility classes in conjunction with Geographic Information System output, respectively. Genetic Programming and Particle Swarm Optimization-Support Vector Machine have performed well with respect to overall prediction accuracy and validated the landslide susceptibility model generated in the Geographic Information System environment. The efficiency of the landslide susceptibility zonation map was also confirmed by correlating the landslide frequency between different susceptible classes.
This content is subject to copyright. Terms and conditions apply.
Research Article
Landslide Susceptibility Mapping in Darjeeling Himalayas, India
Amit Chawla
,
1
Sowmiya Chawla,
1
Srinivas Pasupuleti
,
1
A. C. S. Rao,
2
Kripamoy Sarkar,
3
and Rajesh Dwivedi
4
1
Department of Civil Engineering, IIT(ISM) Dhanbad, Jharkhand, India
2
Department of Computer Science & Engineering, IIT(ISM) Dhanbad, Jharkhand, India
3
Department of Applied Geology, IIT(ISM) Dhanbad, Jharkhand, India
4
Department of Computer Science & Engineering, Vignan’s Foundation for Science, Technology & Research,
Andhra Pradesh, India
Correspondence should be addressed to Amit Chawla; amit84_cha@yahoo.co.in
Received 30 March 2018; Revised 12 June 2018; Accepted 11 July 2018; Published 16 September 2018
Academic Editor: Haiyun Shi
Copyright ©2018 Amit Chawla et al. is is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Landslide susceptibility map aids decision makers and planners for the prevention and mitigation of landslide hazard. is study
presents a methodology for the generation of landslide susceptibility mapping using remote sensing data and Geographic
Information System technique for the part of the Darjeeling district, Eastern Himalaya, in India. Topographic, earthquake, and
remote sensing data and published geology, soil, and rainfall maps were collected and processed using Geographic Information
System. Landslide influencing factors in the study area are drainage, lineament, slope, rainfall, earthquake, lithology, land use/land
cover, fault, valley, soil, relief, and aspect. ese factors were evaluated for the generation of thematic data layers. Numerical
weight and rating for each factor was assigned using the overlay analysis method for the generation of landslide susceptibility map
in the Geographic Information System environment. e resulting landslide susceptibility zonation map demarcated the study
area into four different susceptibility classes: very high, high, moderate, and low. Particle Swarm Optimization-Support Vector
Machine technique was used for the prediction and classification of landslide susceptibility classes, and Genetic Programming
method was used to generate models and to predict landslide susceptibility classes in conjunction with Geographic Information
System output, respectively. Genetic Programming and Particle Swarm Optimization-Support Vector Machine have performed
well with respect to overall prediction accuracy and validated the landslide susceptibility model generated in the Geographic
Information System environment. e efficiency of the landslide susceptibility zonation map was also confirmed by correlating
the landslide frequency between different susceptible classes.
1. Introduction
Landslides are momentary and instantaneously happening
vandalize natural hazard in mountains; however, it turns in
to a disaster due to immature geology coupled with external
temporal triggering factors causing landscape changes and
direct and indirect losses. Landslide happens when the slope
changes from a steady to an unsteady state. e influence of
gravity is the key operating force for the landslide to happen.
It is a motion of mass of rock, earth, or debris along the slope
[1]. More than 15% of total land area in India is considered
to be affected by landslides [2]. Globally awareness has been
drawn on the study of landslides mostly because of growing
pressure of urbanization and its socioeconomic effects on the
terrain habitats [3]. is trend of landslides is expected to
continue in the time to come due to sustained deforestation,
increment of haphazard urbanization, and changing climatic
patterns in the landslide-prone areas [4, 5]. Landslide risk
cannot be utterly averted; however, the impact of its
acuteness and intensity may be reduced by identifying and
predicting the problems in advance. Despite advances in
science and technology, landslides continue to result in
economic, human, and environmental losses worldwide [6].
e external and temporal triggering factors like rainfall and
earthquakes, if happening individually or in combination
with some considerable time and magnitude, are directly
accountable for inducing the landslides [7]. Rainfall is an
inevitable triggering factor which could contribute to the
Hindawi
Advances in Civil Engineering
Volume 2018, Article ID 6416492, 17 pages
https://doi.org/10.1155/2018/6416492
occurrence of landslides [8]. e landslide susceptibility is
the probability of spatial phenomenon of slope failures [9].
Remote sensing can play a role in the generation of landslide
inventory map and thematic maps related to landslide oc-
currences [10]. Remote sensing (RS) data in conjunction
with data from other sources in digital form and their
analysis in Geographic Information System (GIS) envi-
ronment have made possible to generate different thematic
data layers corresponding to the contributing factors ac-
countable for the occurrence of landslides [11–13]. Geo-
matics by taking advantage of modern tools, such as remote
sensing and GIS, provides a perfect opportunity for usage,
validation, and comparison of different methods to produce
a landslide susceptibility map [14]. For the creation of the
LSZ map in the GIS environment, the integration of various
thematic data layers with weights was assigned with respect
to their relative significance [1518]. Production of the
landslide susceptibility map describes the prone area where
landslides may occur in the future [19]. e landslide
susceptibility analysis has been applied for the purpose of
assessing the degree of risk in landslide-prone areas [20].
Landslide susceptibility mapping is an important step prior
to landslide assessment planning, management, and di-
saster mitigation [21]. Landslide susceptibility zonation
(LSZ) maps are helpful in identifying the landslide vul-
nerable zone in advance, planning the future development
projects and mitigation programs. Hence, for effective and
efficient disaster management, there is an urgent re-
quirement of recognizing unstable slopes and mitigating its
effects, which may be attained with the assistance of LSZ
mapping.
e objective of Particle Swarm Optimization (PSO)-
Support Vector Machine (SVM) approach used in landslide
susceptibility mapping is to generate an accurate landslide
susceptibility map through classification technique [22].
Particle Swarm Optimization is a computational technique
that optimizes an issue towards iterative attempt to move
forward a particle solution with respect to a provided
measure about quality [23]. e advantages of PSO over
other soft computing models are as follows: PSO provides
more range and exploration for the population and the
movements of particles allows the fast convergence of
greater diversity in the search space. e principle behind
the Support Vector Machine classifier is setting a boundary
to an area of points which all are the same types.
Genetic Programming (GP) is an evolutionary com-
puting technique that is used to solve a variety of complex
problems. Koza [24] developed the Genetic Programming
approach, and since then, it has been widely used in science
and engineering applications. GP is a very popular method
with diverse applications. In the field of geosciences, GP has
been used by Litschert [25] to map landslide hazard zones in
California locations. In another study by Nourani et al. [26],
landslide susceptibility mapping was performed using GP
for Zonouz plain in Iran.
is study presents a methodology for the generation of
landslide susceptibility mapping using remote sensing data
and Geographic Information System technique for the part
of the Darjeeling district, Eastern Himalaya, in India.
e performance of this GIS output was evaluated by using
PSO-SVM and GP techniques, which aims to predict the
accuracy of landslide susceptibility classes. e study also
aims to determine the usefulness of remote sensing and GIS,
and an attempt was made to generate the landslide sus-
ceptibility map for effective and efficient disaster manage-
ment for the future.
2. Study Area
In India, about 12.6% of Indian landmass are in the Hi-
malayan terrain region and are landslide susceptible regions
[2]. is study focused on a part of Darjeeling hills, West
Bengal, Eastern Himalaya, in India, which lies within the
latitude 26°4931.910to 26°5638.366N and longitude
88°133.706to 88°2231.818E and reports an area of about
201sq. km (Figure 1). is study belongs to the steep and
rugged mountainous terrain, which falls at the alluvial plains
of north extreme of West Bengal, India, and is highly prone
to landslide. During monsoon, these regions observed
regular land sliding incidents activated by rainfall. e study
area also belongs to the high damage risk and severe
earthquake intensity zone in India, that is, Zone IV [27], and
is thus prone to earthquake-induced landslides. As per 2011
census, the study area has an inhabitant’s density of 586
inhabitants per square kilometer, and its inhabitant expansion
rate over the decade 20012011 was 14.77%. e study area is
dominated with the factors mostly favorable for occurrence of
landslide areas like slope of degree >45°, highly dissected hill
slopes, barren land, high rainfall, high earthquake-prone
areas, and relief >2000 m. e study area has a history of land
sliding events which resulted in loss of life and infrastructure.
In the past, the study area experiences earthquakes of high
intensity. From the history of landslide events in the Dar-
jeeling Himalayas, it is concluded that the study area is
vulnerable to landslides, and hence, the landslide suscepti-
bility zonation map for the study area is required to be
analyzed.
3. Data
In this study, the statistics utilized were IRS-RESOURCE-
SAT-II LISS-IV, CARTOSAT-I PAN satellite data (Table 1)
of the National Remote Sensing Centre, India; topographic
maps of the Survey of India (1 : 50,000 scale); earthquake
data of the National Centre for Seismology, India; and in-
formation from the published geology, soil, and rainfall
maps. Integration of LISS-IV and PAN data was attempted
to have the benefit of both high spatial and high spectral
resolution in a sole image. Figure 2 shows the satellite image
of the study area.
4. Methodology
e methodology adopted in this study (Figure 3) for the
generation of LSZ map in the GIS environment, involves the
selection of various factors, creation of various thematic
layers, assigning numerical rating to factors, blending of data
2Advances in Civil Engineering
Table 1: Details of satellite statistics used in this study.
Name of satellite Sensor Product ID Path/row Product type Resolution Date of satellite image
IRS-RESOURCESAT-II L4FX 174047811 Path-107 Orthorectified 5.0 m 15th Jan 2015
Row-52
CARTOSAT-I PAN_FORE 174047741 Path-0584 Standard geocorrected 2.5 m 19th Dec 2010
Row-0273
CARTOSAT-I PAN_FORE 174047751 Path-0584 Standard geocorrected 2.5 m 19th Dec 2010
Row-0274
India
Wes t B en gal
Study area
Darjeeling
1 0.5 01 km
88°140E 88°160E 88°180E 88°200E 88°220E
88°140E 88°160E 88°180E 88°200E 88°220E
26°500N 26°520N 26°540N 26°560N
26°500N 26°520N 26°540N 26°560N
Phulbari TG
Dhajia TG
Dhajia
Balasa TG
Bungkulung
Singbali TG
Gayabari TG Pankhabari
Urseong
Makaibari TG
Kailapani
Moonda Kotee
Chhota Ringtong
Dilaram Bagora
Maharani TG
Tung
Pomabang Majhua TG
Dilaram Deorali
Singeli TG
Jangpana TG
Mahanadi TG
Shivitar
Kurseong
Barbung
Sirubari
Springside TG
Gauri shankar TG
Gayabari
Tindharia
GaurigaonSelim Hill TGMithadanta Narbong TG
Simring TG
Gitingia TG
Malatar TG
Mahaldiram TG
Turuk
Sudung
Chunbhati
Rangtong
Kothidhura
Rohini TG
Marchebong
New Selim Hill TG
Phulbari TG
Dhajia TG
Dhajia
Balasa TG
Bungkulung
Singbali TG
Gayabari TG Pankhabari
Urse ong
Makaibari TG
Kailapani
Moonda Kotee
Chhota Ringtong
Dilaram Bagora
Maharani TG
Tun g
Pomabang Majhua TG
Dilaram Deorali
Singeli TG
Jangpana TG
Mahanadi TG
Shivitar
Kurs eong
Barbung
Sirubari
Springside TG
Gauri shankar TG
Gayabari
Tindharia
GaurigaonSelim Hill TGMithadanta Narbong TG
Simring TG
Gitingia TG
Malatar TG
Mahaldiram TG
Tur u k
Sudung
Chunbhati
Rangtong
Kothidhura
Rohini TG
Marchebong
New Selim Hill TG
Settlement
National highway
State highway
Railway
N
Figure 1: Location map of the study area.
0 0.5 1 2 3 4 5 (km)
0 50 100 200 300 400 500
(m)
Figure 2: (A) RESOURCESAT-II LISS-IV satellite image of the study area. (B) Illustration of the CARTOSAT-I PAN image of the
rectangular area marked in (A).
Advances in Civil Engineering 3
Relief
Slope
Aspect
DEM
Topographic
map Drainage Drainage buer
Remote sensing data Lineament Lineament buer
Lithology
Vall ey bue r
Land use/land
cover
Precipitation
Earthquake
Published
Maps and
others
Soil
Fault
Landslide Map validation Landslide
Susceptibility
Map
Integration
of data
Geospatial
overlay
analysis
Data layers
Weights and
ratings
Fault buer
Figure 3: Flow diagram showing different steps involved in the preparation of landslide susceptibility mapping.
N
S
E
W
Slope angle
0–15°
15°–25°
25°–35°
>45°
35°–45°
0 0.5 1 2 3 4 5 (km)
(a)
N
S
E
W
0 0.5 1 2 3 4 5 (km)
Aspect
Flat
North
Northeast
East
Southeast
South
Southwest
We st
Northwe st
North
(b)
Figure 4: Continued.
4Advances in Civil Engineering
N
S
E
W
Vall ey buff er
0–100 m
>100 m
0 0.5 1 2 3 4 5 (km)
(c)
N
S
E
W
0 0.5 1 2 3 4 5 (km)
Land use
Agriculture land
Barren land
Built up area
Scrub land
Sparse forest
Tea p lant atio n
ick forest
Water body
(d)
N
S
E
W
Soil type
Coarse loamy
Fine loamy
Loamy skeletal
(km)
0 0.5 1 2 3 4 5
(e)
N
S
E
W
Quaternary and
recent sediments
Geabdat sandstone
Chunabati formation
Damuda formation Graphite schist/gneiss
Darjeeling gneiss Rangit pebble slate
Lingtse granite gniess
Gorubathan formation
Feldspathic greywacke
Lithology
16
10
6
6g
Paro-subgroup
(parogniess)
2
Quartzite key beds
2q
3
2g
12
11
(km)
0 0.5 1 2 3 4 5
13
9
(f)
N
S
E
W
1st order
2nd order
Drainage buffer
(km)
0 0.5 1 2 3 4 5
(g)
N
S
E
W
0–125 m
125–250 m
250–375 m
375–500 m
>500 m
Lineament buer
(km)
0 0.5 1 2 3 4 5
(h)
Figure 4: Continued.
Advances in Civil Engineering 5
in GIS environment, calculating landslide potential index,
and classifying and validating landslide susceptibility map.
In this study, twelve different factors were assessed for
the generation of landslide susceptibility mapping in the GIS
environment. In this study, overlay analysis technique was
adopted in the GIS environment for the generation of
thematic data layers. For the study area, on the basis of field
knowledge, experience, and available literature, weight
values were allocated to the data layers/factors on a 1 to 10
numerical scale in series of their significance towards slope
instability, whereas the rating values were allocated to the
classes of the layers on a 0 to 9 numerical scale, in which
higher rating reflects greater influence on landslide event
compared to lower one. e numerical values adopted for
weights and ratings were allotted to the different factors.
ematic data layers were created by mathematically mul-
tiplying the weight of the layer with the ratings of the
correlating class of the individual layer. e outcome of
the final LSZ map was classified into various discrete
susceptibility classes. Validation of the landslide suscepti-
bility map was attempted with the help of landslide distri-
bution map and data of landslides. Twelve prime parameters
and the landslide potential index obtained from the pa-
rameters were involved as an input to PSO-SVM and GP
model.
5. Thematic Data Layers
For the preparation of the LSZ map, generation of various
thematic data layers is required, and the factors selected were
both preparatory and triggering. e layers were produced
in the GIS environment. CARTOSAT-1 DEM at a resolution
of pixel size 25 ×25 m was utilized to extract information on
slope, relief, and aspect layers.
5.1. Slope. e slope instability is directly proportional to the
angle of the slope. e slope map (Figure 4(a)) was derived
N
S
E
W
(km)
0 0.5 1 2 3 4 5
0–125 m
125–250 m
250–375 m
375–500 m
>500 m
Fault buffer
(i)
N
S
E
W
(km)
0 0.5 1 2 3 4 5
0–1000 m
1000–2000 m
>2000 m
Relief
(j)
N
S
E
W
(km)
0 0.5 1 2 3 4 5
High
Medium
Rainfall
(k)
N
S
E
W
(km)
0 0.5 1 2 3 4 5
High
Moderate
Earthquake
(l)
Figure 4: (a) Slope, (b) aspect, (c) valley buffer, (d) land use/land cover, (e) soil, (f ) lithology, (g) drainage buffer, (h) lineament buffer , (i)
fault buffer, (j) relief, (k) rainfall, and (l) earthquake maps of the study area.
6Advances in Civil Engineering
from the surface tool (spatial analysis) of DEM using GIS
software. Slope statistics of the study area computed were
23.66%, 32.21%, 26.91%, 13.31%, and 3.91% for the degree of
slope 0–15°, 15°–25°, 25°–35°, 35°–45°, and >45°, respectively.
5.2. Aspect. Generally, the south-facing slopes have lesser
vegetation density as compared to the north-facing slopes
[28]. With respect to the landslide distribution, south- and
east-facing slopes are further susceptible to landslides [29].
e aspect map (Figure 4(b)) is derived from the surface
(spatial analysis) tool of DEM using GIS software. Aspect
statistics of the study area computed were 0%, 7.76%, 9.66%,
14.17%, 11%, 15.24%, 13.73%, 13.34%, and 15.09% for the
flat, north, northeast, east, southeast, south, southwest, west,
and northwest directions, respectively.
5.3. Valley Buffer. A valley buffer of 100 meter was con-
sidered for the study area along the major streams, that is,
3rd and higher-order drainages, and accordingly, the valley
buffer map was generated (Figure 4(c)). Valley buffer sta-
tistics of the study area computed were 92.18% and 7.82%,
for the valley buffer >100 m and <100 m, respectively.
5.4. Land Use/Land Cover. Land use is an indirect measure
for the strength of the slope, as it commands the rate of
weathering and erosion. In this study, the LU/LC map
(Figure 4(d)) was generated by utilizing the IRS-
RESOURCESAT-II LISS-IV image with four bands along
with the CARTOSAT-I PAN image. e interpreted land
use/land cover has been digitized, and after that it was
rasterized on 25 ×25 m pixel size. Land use/land cover
statistics of the study area computed were 0.36%, 0.94%,
6.48%, 1.84%, 63.47%, 10.39%, 13.77%, and 2.75%, for the
land use class agriculture land, barren land, built-up area,
scrub land, sparse forest, tea plantation, thick forest, and
waterbody, respectively.
5.5. Soil. Top soil cover on a slope has an influence on the
occurrence of landslides [17]. e soil map (Figure 4(e)) was
extracted from a regional soil map published by National
Bureau of Soil Survey and Land use planning using the GIS
environment. Soil statistics of the study area computed were
45.16%, 16.46%, and 38.38%, for the soil unit coarse loamy,
fine loamy, and loamy skeletal, respectively.
5.6. Lithology. In contrast to the weaker rocks, the stronger
rocks give more resistance to the driving forces and conse-
quently are less susceptible to landslides and vice versa [30].
e lithology is a representation of the physical characteristics
of rock or soil. e lithology map (Figure 4(f )) was generated
from the geological map [31] of the Darjeeling area in GIS
environment. Lithology statistics of the study area computed
was 6.34%, 0.94%, 0.93%, 0.93%, 3.27%, 8.24%, 6.57%, 7.46%,
4.44%, 1.22%, 1.12%, 46.20%, 6.12%, 1.05%, and 5.17% for the
rock Chunabati formation, composite of Damuda formation
and Chunabati formation, composite of Rangit pebble slate
and Damuda formation, composite of Rangit pebble slate,
Damuda formation and Gorubathan formation, Damuda
formation, Darjeeling gneiss, feldspathic greywacke,
geabdat sandstone, Gorubathan formation, graphite
schist/gneiss, lingtse granite gniess, paro subgroup (paro
gneiss), quartzite key beds (paro quartzite), quaternary and
recent sediments, and Rangit pebble slate, respectively.
5.7. Drainage. Most common cause for landslide in terrain
region is soil erosion due to drainage activity. Increase of pore
water pressure and decrease in the shear strength due to water
infiltration will result in instability to the slope. Methodology
for generating the drainage buffer by correlating with the
landslide distribution has been suggested by Kanungo et al.
[30], to map the landslide hazard zones in the Darjeeling
Himalayas. From the topographic sheets of Survey of India in
the scale of 1 : 50,000, most of the drainage layers were
produced by digitization of the drainages and subsequently
they were updated with the aid of the PAN image and LISS-IV
image in the GIS environment. For all the drainage orders,
a 25 m buffer zone on both sides of the drainages was created.
After the spatial association of landslide distribution, it was
observed that most of the landslide pixels fall in the 1st- and
2nd-order drainage buffers only. Accordingly, a drainage
buffer map (Figure 4(g)) was generated in the GIS envi-
ronment by taking into account the 1st- and 2nd-order
streams only with 25 m buffer zones around these drainages.
Drainage statistics of the study area computed were 8.62%,
3.57%, and 87.81% for the drainage buffers 1st order, 2nd
order, and for the rest of the area, respectively.
5.8. Lineaments. Landslides are more prone in the jointed,
fractured, and faulted areas. Methodology for generating the
lineament buffer layers with equal interval buffer zones has
been suggested by Kanungo et al. [30], to map the landslide
hazard zones in the Darjeeling Himalayas. PAN and LISSIV
satellite images are used for the interpretation of lineaments.
Interpreted lineaments were processed for the generation of
lineament layers in the GIS environment. Lineament buffer
regions were generated at 250 m distance, and then they were
geospatially cross-tallied with the pixels of landslides. 84% of
the pixels had befallen in the primary two buffer regions
only. Hence, it has been evaluated to have five equal distance
lineament buffer zones at 125 m distance, and with respect to
these zones, the lineament buffer map (Figure 4(h)) was
generated. e lineament statistics of the study area com-
puted were 20.63%, 21.13%, 18.39%, 14.72%, and 25.13% for
the lineament buffers <125 m, 125–250 m, 250–375 m,
375–500 m, and >500 m, respectively.
5.9. Fault Buffer. Huge jointed area displaying offsets of
chain of hills are faults. e faults were obtained from the
geological map [31] of the Darjeeling area by using GIS
software. Hence for the study area, it has been considered to
have five equal distance buffer zones at 125 m distance for
the generation of the fault buffer map (Figure 4(i)), to study
the effect of faults on the landslide occurrence. e fault
Advances in Civil Engineering 7
buffer statistics of the study area computed were 0.73%,
0.88%, 0.98%, 1.14%, and 96.27% for the fault buffers
<125 m, 125–250 m, 250–375 m, 375–500 m, and >500 m,
respectively.
5.10. Relief. Relief represents the difference in altitude be-
tween two points. e lesser relief value nominates a mature
topography, whereas the higher relief value nominates
immature topography. e relief map (Figure 4(j)) was
extracted from DEM in the GIS environment. e relief
statistics of the study area computed were 56.10%, 41.07%,
and 2.83% for the reliefs 0–1000m, 1000–2000 m, and
>2000 m, respectively.
5.11. Rainfall. Rainfall plays a critical part in the unforeseen
events like landslides. It is an external temporal triggering
factor, and excess of it may make the slope become heavy
due to increase in pore water pressure and shall result in
slope slips. e rainfall map (Figure 4(k)) of the study area
was generated by digitizing the polygons from the rainfall
map of National Atlas and ematic Mapping Organisation,
Kolkata, India, in a vector layer. e rainfall map was
rasterized at 25 m ×25 m spatial resolution. e rainfall
statistics of the study area computed were 4.84% and 95.16%
for the rainfall classes medium and high, respectively.
5.12. Earthquake. Both the destructive natural disasters, that
is, earthquakes and landslides, are common in one sense that
both are scary and destructive in nature. e tremors
produced by the earthquakes do not only activate currently
developed landslides but also revive the older ones. e
earthquake map for the study area has been prepared by
using the point data of more than last 200 years, and these
point data were explained with the magnitude and epicenter
location. e data collected were obtained from National
Centre for Seismology, New Delhi, India. e earthquake
map (Figure 4(l)) has been generated by interpolation
technique utilizing the Inverse Distance Weighted (IDW)
tool in GIS environment; in which, the IDW tool forms
a surface with respect to the value of point data. e
earthquake map was subcategorized into two categories of
moderate- and high-prone zones with respect to the ref-
erence of past history of earthquake magnitudes ranging
from 3 to 5 (moderate) and above 5 (high) measured in the
Richter scale, respectively. Earthquake statistics of the study
area computed were 69.59% and 30.41% for the earthquake
classes high and moderate, respectively.
5.13. Landslide Distribution. e principal features for lo-
cating the landslides by using satellite images are spectral
attributes, size, shape, contrast, and so on; the landslides are
mostly bare of vegetation, and they showed high reflectance
[17]. In the past, for the detection of landslides, SPOT PAN
and ATM high-resolution satellite images of 10 m and 7.5 m
resolution, respectively, were used by Mason et al. [32]. For
the study area, landslides were detected by using the IRS-
RESOURCESAT-II LISS-IV and CARTOSAT-I PAN high-
resolution satellite images of 5 m and 2.5 m resolution,
respectively. Some of the extracted landslides from the
satellite images were crosschecked, and the landslide distri-
bution map (Figure 5) was modified accordingly. e mapping
of prevailing landslides is necessary to study the association
between the landslide distribution and the causative factors
[30]. History of landslide inventory of the study area was
prepared from the Special Publication Number 94, Geological
Survey of India. According to this landslide inventory, 17
major landslides were falling in our study area. When these 17
landslide locations were correlated with the landslide distri-
bution map and field verification, it was observed that most of
these major historical landslide locations are still falling in the
landslide distribution map of the study area. e landslide
distribution map contains 91 landslides which were scattered
over the complete study area. Most of the landslides falling are
in the varying span from 500 sq. m to 3000 sq. m.
In the study area, ground truth verification for some of
the significant landslides was carried out by using the global
positioning system (GPS) device. Four significant landslide
sites were selected within the study area (Figure 5), for
explaining the information on actual occurrence and type of
landslides with the help of field survey and Google Earth
satellite image (Figure 6). In location Site 1, the type of
landslide observed was translational debris slide, and the
state of this slide was enlarging and multiple. For Site 2,
the type of landslide observed was translational rock-cum-
debris slide, and the state of this slide was successive. For Site
3, the type of landslide observed was translational debris
slide, and the state of this slide was enlarging and single.
However, for Site 4, the type of landslide observed was
complex slide, where the slide is partly translational grading
to earthflow at toe, and the state of this slide was single.
5.14. Landslide Susceptibility Mapping. Identification of
potential landslide area can be achieved by developing
a rating scheme in which the factors and their classes were
assigned numerical values based on the associated causative
factors [17]. In this study, an attempt was made to generate
a grading system in which the factors and their classes have
been allocated numerical values. e thematic data layers
have been generated in the GIS environment. e weights
and ratings allotted to each factor and their classes are given
in Table 2. In this study, twelve thematic data layers were
generated, and they were overlaid and mathematically added
by using (1) for generating the landslide potential index
(LPI) in the GIS environment for each cell:
LPI Dr +Li +Sl +Pr +Ea +Lith
+Fb +Vb +La +So +Re +As,
(1)
where Dr,Li,Sl,Pr,Ea,Lith,Fb,Vb,La,So,Re,and As are the
representative symbols for the thematic weighted layers for
drainage buffer, lineament buffer, slope, precipitation,
earthquake, lithology, fault buffer, valley buffer, land use/
land cover, soil, relief, and aspect, respectively. e range for
LPI adjudged was between 171 and 502, which was later on
categorized to generate landslide susceptibility classes. For
8Advances in Civil Engineering
such classifications, a judicious way is to look for values of
sudden abrupt alterations [33]. Boundaries for different
susceptible classes were drawn at significant changes in
gradient [17]. Hence, a graph was plotted with LPI-frequency
values, emerging many swings. In this graph (Figure 7),
motion mean with the mean window lengths of 3, 7,
and 11 was utilized for smoothening the curve and finer
classification. To acquire respective susceptibility zones for
different categories, the borders were drawn at abrupt
alteration in gradient of the slope at the LPI values of 212,
302, and 381 (Table 3). e LSZ map (Figure 8) was
generated by using the LPI-class boundaries in the GIS
environment.
5.15. Classification and Prediction Using PSO-SVM Approach
Based on LPI. In PSO, the framework is started with
a populace of random solutions and scans for optima by
refreshing generations. In this approach, the potential ar-
rangements, called particles, fly through the issue space by
following the present ideal particles.
5.16. e Algorithm for PSO-SVM. PSO is introduced with
a gathering of random particles (solution) and afterward
looks for optima by refreshing generations. In each gen-
eration, every molecule is modified by following two “best”
parameters. e first is the best solution (fitness) it has
accomplished so far (fitness parameter is additionally put
away). is value is called pbest. Another “best” parameter
that is followed by the particle swarm analyzer is the best
value, got so far by any particle in the populace. is best
value is a global best and called gbest. e movements of the
particles are directed by the values pbest and gbest. At the
point when enhanced positions are being found, these
values will then come to direct the movements of the
swarm. e particle performs movement by updating its
velocity and positions by using (2) and (3), respectively.
Figure 9 describes the optimization of particles using PSO
algorithm.
Formula (2) is for velocity update operation:
vt+1
ivt
i+c1rand1pt
best ixt
i
􏼐 􏼑 +c2rand2gt
best xt
i
􏼐 􏼑.(2)
Formula (3) is for position update according to velocity:
xt+1
ixt
i+vt+1
i,(3)
where vt
iand xt
idenote the velocity and position of ith
particle, pt
best idenotes individual best known position of ith
particle, gt
best idenotes the entire swarm best known posi-
tion, and rand
1
and rand
2
are the two random variables,
whereas c
1
and c
2
are the learning parameters.
e flowchart presented in Figure 9 describes that for
every particle, if the calculated fitness value is superior than the
best fitness value calculated in the past, make the present
fitness value as new pbest. After that the gbest value is cal-
culated by assigning best particle’s pbest value to gbest.
5.17. Procedure for PSO-SVM. e flowchart describing the
working of PSO is shown in Figure 10. For the classifi-
cation task, optimal LPI values are identified using PSO
according to the minimum and maximum ranges for all
(km)
0 0.5 1 2 3 4 5
Ground truthing location
Landslide
Site 1
Site 3
Site 2
Site 4
Landslide distribution map of study area
88°120E88°140E 88°160E 88°180E 88°200E 88°220E 88°240E
88°120E
26°480N 26°580N26°560N26°540N26°520N26°500N
26°480N 26°560N26°540N26°520N26°500N
88°140E 88°160E 88°180E 88°200E 88°220E 88°240E
N
S
E
W
Figure 5: Landslide distribution map.
Advances in Civil Engineering 9
the 12 parameters by considering the numerical value of
weights and ratings as given in Table 2. is work started
with an initial population of “50” particles and processed
up to “500” maximum iterations (generation). e pro-
cedure for classification are as follows. Step 1: describe the
LPI function as fitness function and also initialize
“population” and “max generation”; Step 2: set the
minimum and maximum limits for each parameter used
in LPI function according to Table 2; Step 3: record the
movements of every particle in each generation in the
vector format containing the LPI value along with the
corresponding parameter values. In each generation,
population moves from one place to suitable place and
generates the new fitness values.
5.18. Classification System for PSO-SVM. In the classification,
data are loaded from the database and divided into 70%
training set and 30% testing set. In the data set, the parameter
values corresponding to fitness values are treated as a feature
vector. SVM looks generally advantageous hyperplanes that
give the maximum profit. Input: training set, class labels, and
testing set. Step 1: perform training of the SVM classifier with
training set and class labels. Step 2: perform testing with testing
set to test the accuracy of prediction.
5.19. Discussion for PSO-SVM Study. e PSO-SVM study
was carried out by taking selected “800” LPI values (200 for
each class) along with the corresponding parameters that
cover the entire range of classification for each class.
Training is carried by taking 600 LPIs (150 from each class),
and testing is performed using 200 LPIs (50 from each class).
e test accuracy described by the confusion matrix is
presented in Figure 11(a), and a comparison between
classifications based on LPI values and predicted classifi-
cation based on the SVM classifier the same as the confusion
matrix is presented in Figure 11(b). e confusion matrix is
used to explain the classification performance model on a set
of testing data for which the original labels are known. Each
row of the matrix represents the instances in a predicted
class, while each column represents the instances in an actual
class (or vice versa). In Figure 11(a), there are 5 columns and
5 rows, but classification in this task is limited to only 4
classes, and the fifth column and fifth row are filled with all
zeros. e first, second, third, and fourth columns and rows
26°5454N, 88°159E
Satellite image
In situ photo
(a)
26°5215N, 88°1830E
Satellite image
In situ photo
In situ photo
(b)
26°5352N, 88°1719E
Satellite image
In situ photo
(c)
26°5049N, 88°2019E
Satellite image
In situ photo
(d)
Figure 6: Field photographs, Google Earth images, and types of some of the major landslide events in the study area. (a) Site 1: translational
debris slide. (b) Site 2: translational rock-cum-debris slide. (c) Site 3: translational debris slide. (d) Site 4: complex slide.
10 Advances in Civil Engineering
correspond to low, moderate, high, and very high classes,
respectively. Each column related to these classes will have
a sum of 50 LPIs; that is, the testing is done by 200 LPIs, by
taking 50 LPIs from each class, which is the target class
(actual classification based on LPI values), and predicted
classification by the SVM classifier is represented by rows. In
the first row, a sum of 50 LPIs are presented that means out
of 50 low-class LPIs, and the system predicted that all are the
low class LPIs. In the second, third, and fourth rows, a sum
of 52, 51, and 47 is present that means out of 50 moderate, 50
high, and 50 very high-class LPIs, the system predicted that
52 belongs to moderate, 51 belongs to high, and 47 belongs
to very high. e prediction accuracy is also written in the
sixth row corresponding to each class. A comparison of the
measured LPI values generated by GIS and the predicted
Table 2: Weights and ratings for various factors and their classes.
Factors Classes Weights Ratings
Drainage
buffer
1st order 10 9
2nd order 5
Lineament
buffer
0–125 m
9
9
125–250 m 7
250–375 m 5
375–500 m 3
>500 m 1
Slope
0°–15°
8
1
15°–25°3
25°–35°5
35°–45°7
>45°9
Rainfall Moderate 76
High 9
Earthquake Moderate 76
High 9
Lithology
Paro-subgroup (parogneiss)
6
4
Darjeeling gneiss 6
Gorubathan formation 7
Feldspathic greywacke 2
Graphite schist/gneiss 7
Quartzite key beds
(paroquartzite) 1
Lingtse granite gneiss 8
Rangit pebble slate 8
Damuda formation 4
Geabdat sandstone 3
Chunabati formation 6
Quaternary and recent sediments 9
Composite of Rangit pebble slate
and Damuda formation 6
Composite of Rangit pebble slate,
Damuda formation, and
Gorubathan formation
7
Composite of Damuda formation
and Chunabati formation 5
Land use/
land cover
Agriculture land
5
5
Barren land 9
Built-up area 2
Scrub land 7
Sparse forest 6
Tea plantation 3
ick forest 1
Waterbody 0
Fault
buffer
0–125 m
5
9
125–250 m 7
250–375 m 5
375–500 m 3
>500 m 1
Valley
buffer
0–100 m 46
>100 m 0
Soil
Fine loamy
3
3
Coarse loamy 6
Loamy skeletal 9
Relief
<1000 m
3
3
1000–2000 m 6
>2000 m 9
Table 2: Continued.
Factors Classes Weights Ratings
Aspect
Flat
1
0
North 1
Northeast 4
East 7
Southeast 8
South 9
Southwest 6
West 3
Northwest 2
Table 3: Distribution of landslide susceptibility zones and landslide
potential index of the study area.
Susceptibility zone Area (sq. km) LPI
Low 7.267 171 to 212
Moderate 117.736 213 to 302
High 65.787 303 to 381
Very high 10.97 382 to 502
0
50
100
150
200
250
300
0
500
1000
1500
2000
2500
3000
3500
171 211 251 291 331 371 411 451 491
Frequency values for GIS
Low Moderate High
212 381
Landslide potential index (LPI)
Actual value
Window length 3
Window length 7
Window length 11
502
Very high
302
Predicted values using
PSO-SVM
Predicted frequency value
using PSO-SVM
Figure 7: Frequency division of landslide potential index and
predicted frequency using PSO-SVM.
Advances in Civil Engineering 11
(km)
0 0.5 1 2 3 4 5
Landslide distribution map of study area
88°120E 88°140E 88°160E 88°180E 88°200E 88°220E 88°240E
88°120E
26°480N 26°580N26°560N26°540N26°520N26°500N
26°480N 26°560N26°540N26°520N26°500N
88°140E 88°160E 88°180E 88°200E 88°220E 88°240E
N
S
E
W
Site–1
Site–3
Site–2
Site–4
Ground truthing location
Landslide
Low
Moderate
High
Very Hi gh
Figure 8: Landslide susceptibility map of the study area.
Calculate fitness values for each
particle
Assign best particle’s pbest value to gbest
Maximum
iteration
reached?
Ye s No
Initialize particles
Is current
fitness value
better than
pbest?
Assign current fitness
as new pbest Keep previous pbest
Calculate velocity for each particle
Use each particles velocity value to update its data values
No Ye s End
Figure 9: Flowchart for optimization of particles using PSO algorithm.
12 Advances in Civil Engineering
values generated by PSO-SVM in terms of frequency di-
vision is shown in Figure 7.
5.20. Classification and Prediction Using Genetic Pro-
gramming Approach Based on LPI. Genetic Programming
(GP) is based on Darwin’s theory of evolution. It uses the
principle of natural selection and genetic recombination.
is method transforms units that define a given problem,
and with the progress of number of iterations, they famil-
iarize themselves to their surroundings. A GP model consists
of variables and functions represented in a tree structure.
e leaf nodes of the tree are the variables that are user
defined like soil type and slope and interior nodes are the
functions like +, , sine, and cosine.
e basic steps involved in the operation of GP are the
creation of random set of population for the given variables
and functions which are then evaluated for fitness. e fittest
Start
Specify fitness function
Specify the parameters for the PSO
Generate initial population
Time-domain simulation
Find the fitness of each particle in the
current population
Iteration>max
iteration
or
Error<least
error
Stop
Modify position and velocity by
using (2) and (3)
Ye s
No
Gen=
gen+1
Figure 10: Flowchart describing the working of PSO.
50
25.0% 0
0.0% 0
0.0% 0
0.0% 0
0.0% 100%
0.0%
0
0.0% 48
24.0% 4
2.0% 0
0.0% 0
0.0% 92.3%
7.7%
0
0.0% 0
0.0% 46
23.0% 5
2.5% 0
0.0% 90.2%
9.8%
0
0.0% 2
1.0% 0
0.0% 45
22.5% 0
0.0% 95.7%
4.3%
0
0.0% 0
0.0% 0
0.0% 0
0.0% 0
0.0% NaN%
NaN%
100%
0.0% 96.0%
4.0% 92.0%
8.0% 90.0%
10.0% NaN%
NaN% 94.5%
5.5%
Confusion matrix
1
2
3
4
5
12345
Target class
Output class
(a)
44
45
46
47
48
49
50
51
52
53
Low Moderate High Very high
Actual
Predicted
Count of LPIs
Category of classes
(b)
Figure 11: Confusion matrix (a) and graph (b) describing the accuracy.
Advances in Civil Engineering 13
populations are passed on the next generation after applying
genetic operators like crossover, mutation, and duplication
and further checked for fitness. An illustration for crossover
operation is shown in Figure 12. As the number of gener-
ation increases, the possibility of more fit models enhances.
e GP tends to attain a model of 100% accuracy by iterating
for few numbers of generations or otherwise gives a model of
fairly high accuracy having reached the specified number of
generation or the time limit.
5.21. Genetic Programming Procedure. In this study,
a computer program code for GP was run in MATLAB
programming language. Before the run, training and
testing data sets as inputs to the GP code were created. e
input data sets were based on the variables and functions
that formed the original LPI equation given by the GIS
method. In this equation, the value for each variable to
create the data set was randomly chosen within the limits as
mentioned in Table 2. Out of the total number of data sets,
70% was used as the training set and remaining 30% as the
testing set.
To begin with, the number of generations was kept very
low as to check the accuracy that it attains at low generation
values. Gradually the number of generations was increased
and at 30th generation, the accuracy was found to be 100%,
but the complexity of the model obtained was fairly high. On
increasing further, the number of generations at 50th
generation, the model equation obtained was the simplest,
while maintaining the same accuracy of 100%. is equation
obtained was the closest match to the original LPI equation
given by the GIS method with a value of 0.000224 added at
the end. e run results for the various numbers of gen-
erations are shown in Table 4 below.
e equations obtained for each iteration are shown in
Table 5. A comparison of the measured LPI values generated
by GIS and the predicted values generated by GP in terms of
frequency division is shown in Figure 13.
e total number of data points is different for LPI values
of GIS and PSO-SVM methods (Figure 7) and GIS and GP
methods (Figure 13) of analysis. A particular LPI value has
a different number of occurrences while comparing the
performance of GIS and PSO-SVM methods and GIS and GP
methods; however, by comparing the trend at a suitable scale
in the frequency axis of the comparison chart, an overall
matching trend shall be observed. Hence, to compare the LPI-
frequency trend of both the methods (PSO-SVM and GP)
with the LPI-frequency trend using GIS (Figures 7 and 13), the
scales of the frequency axes have been kept different in left and
right ordinates.
5.22. Map Validation. Computation of landslide frequency
for each class is the dominant factor for evaluating the
quality of LSZ map [34]. e landslide susceptibility zones
reflect the existing field instability conditions [17]. While
Dr
(Dr + Li) + Sl
Li
+
+
Sl
Pr
+
+
Lith
(Pr + (Fb + Vb)) + Lith
Fb Vb
+Dr Li
+
+
Sl
Pr
+
+
Lith
Fb Vb
+
Crossover
(Pr + (Fb + Vb)) + Sl (Dr + Li) + Lith
Figure 12: Illustration of crossover operation.
Table 4: Genetic Programming run results.
Run number Population size Number or generations Training accuracy
#
Test accuracy
#
Function
1 500 2 0.9043 0.9016 +, , sine, square
2 500 3 0.9100 0.9089 +, sine, square
3 500 4 0.9693 0.9700 +, square,
4 500 5 0.9557 0.9560 +, square, , cosine
5 500 10 0.9707 0.9709 +, , cosine
6 500 20 0.9939 0.9942 +, square
7 500 30 1.0000 1.0000 +
8 500 40 1.0000 1.0000 +, square
9 500 50 1.0000 1.0000 +
10 1000 5001.0000 1.0000 +, , sine, square
Note.
#
e accuracy is measured in terms of coefficient of correlation (R). e iteration achieved best fitness before reaching 200th generation.
14 Advances in Civil Engineering
verifying the LSZ map, very high and high susceptible zones
show signs of soil erosion, slope instability, and so on. For
evaluating the effectiveness of the LSZ map, it is significant
to calculate the landslide frequency for each zone. e study
area has been classified into four zones, namely, low sus-
ceptibility zone, moderate susceptibility zone, high sus-
ceptibility zone, and very high susceptibility zone, and the
number of landslides and frequency per susceptible zone
have been computed (Table 6).
6. Conclusions
Based on the studies described above, the following con-
clusions are drawn:
Table 5: Prediction equations by GP.
Number of generation Functions used Model equations (for LPI)
2 +, , sine, square
1.239Dr +1.22As +0.9052Li +
0.9627Sl 0.2843 sin(sin(Sl)) +
0.01915 square(Ea) + 0.04053SoPr +110.9
3 +, sine, square
1.189So +0.0234Re +0.8868Li +0.9452Sl +
1.595Pr +1.189 sin(La) + 0.0003481 square(x1) +
0.0234Dr Ea +104.3
4 +, square, 0.867Li +1.043Sl +2.03Ea +0.8899Fb +1.043La +
0.007235 square(Dr +Lith) + 0.02292Li Vb +68.67
5 +, square, , cosine
0.9824Sl +1.315Pr +1.629Ea +1.315Vb +
0.01981 square(Lith) + 0.02173Dr Fb +
0.03422La(Li cos(La)) + 82.18
10 +, , cosine 1.164Dr +1.001Li +1.021Sl +1.869Ea +1.07Lith +
1.164La 1.021 cos(cos(Re)) + 0.06572Re Fb +34.1
20 +, square
0.4065Dr +1.005So +1.001Li +1.024Ea +
1.024Lith +1.005La +0.007404 square(Sl +Fb +
Vb +5.73168) + 0.01453Dr Pr +0.01453Re Pr +
72.64
30 + Dr +So +Re +As +Li +Sl +Pr +Ea +Lith +Fb +
Vb +La +1.044 ×1010
40 +, square Dr +So +As +Li +Sl +Pr +Lith +Fb +Vb +La +
2.458 square(0.6379)(Re +Ea) + 2.598 ×1010
50 + Dr +So + Re + As +Li + Sl + Pr + Ea + Lith + Fb +
Vb + La + 0.000224
500 +, , sine, square
Dr +So +Re +As +
Li +Sl +0.9998Pr +Ea +Lith +Fb +Vb +
La 9.609 sin(Pr +Lith) ×
1014 square(1.285362)3.783
(2Li sin(Vb))(2Lith Pr +Vb) × 1015 +1.652
0
20
40
60
80
100
120
0
500
1000
1500
2000
2500
3000
3500
171 211 251 291 331 371 411 451 491
Predicted frequency values using GP
Frequency values for GIS
Landslide potential index (LPI)
Low Moderate High
221 502
Very high
302
Predicted values using
genetic programming
381
Actual value
Window length 3
Window length 7
Window length 11
Figure 13: Frequency division of landslide potential index and predicted frequency using GP.
Advances in Civil Engineering 15
(1) e entire study area was divided in to four re-
spective susceptibility zones, that is, low (3.6%),
moderate (58.35%), high (32.61%), and very high
(5.44%) susceptibility zones, and landslide areas
occupied per zone are 0.34%, 14.35%, 33.71%, and
51.60% for low, moderate, high, and very high
susceptible zones, respectively; hence, this outcome
was validated on the reasoning of landslide distri-
bution, Genetic Programming method and Particle
Swarm Optimization (PSO)-Support Vector Ma-
chine (SVM) technique.
(2) Landslide frequency for the very high susceptibility
zone (1.82/sq. km) was significantly higher as
compared with that for high (0.73/sq. km), moderate
(0.19/sq. km), and low (0.14/sq. km) susceptibility
zones, and it was concluded that there is a gradual
increment and substantial detachment of landslide
frequency numerical values in between the suscep-
tibility classes. Hence due to prevailing field un-
reliable conditions, the area falling in very high and
high susceptible zones shall be treated as the po-
tential landslide-prone zone, and it is highly rec-
ommended to avoid those zones, and if not possible,
then immediate remedial measures may be taken to
diminish the impact of landslide events.
(3) e frequency division obtained by Particle Swarm
Optimization-Support Vector Machine technique
was a close match to the measured GIS values, and
PSO-SVM technique has performed well with overall
prediction accuracy of 94.5%, by considering accu-
racy average for each class.
(4) e Genetic Programming method has also per-
formed well with a nearly accuracy of 100% and
suggested a model equation which is very close to
the original equation given in the GIS environment.
e frequency division obtained by GP was also
a close match to the measured GIS values, thus
suggesting the same susceptibility zones as suggested
by the GIS model.
(5) It is high time that a diligent landslide hazard study of
very high and high susceptibility zones of the study
area should be executed, especially in suggesting the
geotechnical remedies for mitigating the slope in-
stability. e LSZ map of the study area shall help
planners in making decisions for future development
of projects and disaster mitigation measures.
Data Availability
e data like earthquake, satellite image, and toposheets
used to support the findings of this study were supplied by
the National Centre for Seismology, New Delhi, India;
National Remote Sensing Centre, Hyderabad, India; and
Survey of India, Kolkata, and cannot be made freely
available due to defence restricted area. e request for
access to these data should be made to the National Centre
for Seismology, New Delhi, India; National Remote
Sensing Centre, Hyderabad, India; and Survey of India,
Kolkata.
Conflicts of Interest
e authors declare that they have no conflicts of interest.
Acknowledgments
e authors thank the National Remote Sensing Centre,
Hyderabad, India; National Centre for Seismology, New
Delhi, India; Survey of India, Kolkata, India; Geological
Survey of India, Kolkata, India; National Atlas and ematic
Mapping Organisation, Kolkata, India; and National Bureau
of Soil Survey and Land Use Planning, Kolkata, India, for
providing various data used in this study. e authors
are grateful to Dr. Shanatanu Sarkar, CBRI, Roorkee, India;
Dr. V.V. Govind Kumar, IIT(ISM) Dhanbad, India; and
Mr. Harish Sinha, CHiPS, India, for their valuable sugges-
tions. Support from the Indian Institute of Technology (Indian
School of Mines) Dhanbad, Jharkhand, India; Public Works
Department, Delhi, India; and Central Public Works De-
partment, Delhi, India, is also acknowledged.
References
[1] D. M. Cruden, “A simple definition of a landslide,” Bulletin of
the International Association of Engineering Geology, vol. 43,
no. 1, pp. 27–29, 1991.
[2] NDMA, Management of Landslides and Snow Avalanches,
National Disaster Management Authority (NDMA), Gov-
ernment of India, New Delhi, India, 2009.
[3] P. Aleotti and R. Chowdhury, “Landslide hazard assessment:
summary, review and new perspectives,” Bulletin of Engi-
neering Geology and the Environment, vol. 58, no. 1, pp. 21–44,
1999.
[4] R. Schuster, “Socioeconomic significance of landslides,” in
Landslides: Investigation and Mitigation: Special Report,
Table 6: Details of landslides and frequency in different susceptible zones.
Landslide
susceptible zone
Area
(sq. km)
Area (%) of
total area
Number of
landslides
Landslide area occupied per
zone (sq. km)
Landslide area occupied
per zone (%)
Frequency
(per sq. km)
Low 7.267 3.6 1 0.003 0.34 0.14
Moderate 117.736 58.35 22 0.126 14.35 0.19
High 65.787 32.61 48 0.296 33.71 0.73
Very high 10.97 5.44 20 0.453 51.60 1.82
16 Advances in Civil Engineering
A. K. Turner and R. L. Schuster, Eds., Vol. 247, pp. 12–36,
National Academic Press, Washington, DC, USA, 1996.
[5] M. Ercanoglu and C. Gokceoglu, “Use of fuzzy relations to
produce landslide susceptibility map of a landslide prone area
(West Black Sea Region, Turkey),” Engineering Geology,
vol. 75, no. 3-4, pp. 229–250, 2004.
[6] C. Chalkias, M. Ferentinou, and C. Polykretis, “GIS based
landslide susceptibility mapping on the Peloponnese,
Peninsula, Greece,” Geosciences, vol. 4, no. 3, pp. 176190,
2014.
[7] S. Chawla, A. Chawla, and S. Pasupuleti, “A feasible approach
for landslide susceptibility map using GIS,” in Proceedings of
Geo-Risk ASCE 2017, pp. 101–110, Denver, CO, USA, June
2017.
[8] S. N. L. Taib, O. S. Selaman, C. L. Chen, R. Lim, and
D. S. Awang Ismail, “Landslide susceptibility in relation to
correlation of groundwater development and ground con-
dition,” Advances in Civil Engineering, vol. 2017, Article ID
4320340, 7 pages, 2017.
[9] F. Guzzeti, Landslide hazard and risk assessment, Ph.D.
disseration, University of Bonn, Bonn, Germany, 2005.
[10] H. Shahabi and M. Hashim, “Landslide susceptibility map-
ping using GIS-based statistical models and remote sensing
data in tropical environment,” Scientific Reports, vol. 5, no. 1,
pp. 1–14, 2015.
[11] R. P. Gupta and B. C. Joshi, “Landslide hazard zonation using
the GIS Approach-a case study from the Ramganga catch-
ment, Himalayas,” Engineering Geology, vol. 28, no. 1-2,
pp. 119–131, 1990.
[12] R. Nagarajan, A. Mukherjee, A. Roy, and M.V. Khire,
“Temporal remote sensing data and GIS application in
landslide hazard zonation of part of Western Ghat, India,”
International Journal of Remote Sensing, vol. 19, no. 4,
pp. 573–585, 1998.
[13] R. P. Gupta, Remote Sensing Geology, Springer-Verlag, Berlin,
Germany, 2nd edition, 2003.
[14] V. Vakhshoori and M. Zare, “Landslide susceptibility map-
ping by comparing weight of evidence, fuzzy logic, and fre-
quency ratio methods,” Geomatics, Natural Hazards and Risk,
vol. 7, no. 5, pp. 1731–1752, 2016.
[15] R. P. Gupta, A. K. Saha, M. K. Arora, and A. Kumar,
“Landslide hazard zonation in a part of the Bhagirathi Valley,
Garhwal Himalayas, using integrated remote sensing-GIS,”
Himalayan Geology, vol. 20, no. 2, pp. 71–85, 1999.
[16] A. K. Saha, R. P. Gupta, and M. K. Arora, “GIS-based landslide
hazard zonation in the Bhagirathi (Ganga) valley, Himalayas,”
International Journal of Remote Sensing, vol. 23, no. 2,
pp. 357–369, 2002.
[17] S. Sarkar and D. P. Kanungo, “An integrated approach for
landslide susceptibility mapping using remote sensing and
GIS,” Photogrammetric Engineering and Remote Sensing,
vol. 70, no. 5, pp. 617–625, 2004.
[18] A. K. Saha, R. P. Gupta, I. Sarkar, M. K. Arora, and
E. Csaplovics, “An approach for GIS-based statistical landslide
susceptibility zonation-with a case study in the Himalayas,”
Landslides, vol. 2, no. 1, pp. 61–69, 2005.
[19] M. Basharat, H. R. Shah, and N. Hameed, “Landslide sus-
ceptibility mapping using GIS and weighted overlay method:
a case study from NW Himalayas, Pakistan,” Arabian Journal
of Geosciences, vol. 9, no. 4, 2016.
[20] H. J. Oh, S. Lee, and S. M. Hong, “Landslide susceptibility
assessment using frequency ratio technique with iterative
random sampling,” Journal of Sensors, vol. 2017, Article ID
3730913, 21 pages, 2017.
[21] S. Kanwal, S. Atif, and M. Shafiq, “GIS based landslide sus-
ceptibility mapping of northern areas of Pakistan, a case study
of Shigar and Shyok basins,” Geomatics, Natural Hazards and
Risk, vol. 8, no. 2, pp. 348–366, 2016.
[22] C. Zhou, K. Yin, Y. Cao, and B. Ahmed, “Application of time
series analysis and PSO–SVM model in predicting the
Bazimen landslide in the ree Gorges Reservoir, China,”
Engineering Geology, vol. 204, pp. 108–120, 2016.
[23] Y. Shi and R. C. Eberhart, “Empirical study of particle swarm
optimization,” in Proceedings of 1999 Congress on Evolutionary
Computation-CEC99, vol. 3, pp. 19451950, Washington, DC,
USA, July 1999.
[24] J. R. Koza, Genetic Programming: A Paradigm for Genetically
Breeding Populations of Computer Programs to Solve Problems,
Stanford University, Department of Computer Science, Vol.
34, Stanford University, Department of Computer Science,
Stanford, CA, USA, 1990.
[25] S. Litschert, “Landslide hazard zoning using genetic pro-
gramming,” Physical Geography, vol. 25, no. 2, pp. 130–151,
2004.
[26] V. Nourani, B. Pradhan, H. Ghaffari, and S. S. Sharifi,
“Landslide susceptibility mapping at Zonouz Plain, Iran using
genetic programming and comparison with frequency ratio,
logistic regression, and artificial neural network models,”
Natural Hazards, vol. 71, no. 1, pp. 523–547, 2014.
[27] BIS, Criteria for Earthquake Resistant Design of Structures:
General Provisions and Buildings, Bureau of Indian Standards
(BIS), IS 1893:2002, Part-1, New Delhi, India, 2002.
[28] B. N. Sinha, R. S. Varma, and D. K. Paul, “Landslides in
Darjeeling district (West Bengal) and adjacent areas,” Bul-
letion of the geological Survey of India Series, vol. 36, pp. 1–45,
1975.
[29] A. S. Dhakal, T. Amada, and M. Aniya, “Landslide hazard
mapping and its evaluation using GIS: an investigation of
sampling schemes for a grid-cell based quantitative method,”
Photogrammetric Engineering and Remote Sensing, vol. 66,
no. 8, pp. 981–989, 2000.
[30] D. P. Kanungo, M. K. Arora, S. Sarkar, and R. P. Gupta, “A
comparative study of conventional, ANN black box, fuzzy and
combined neural and fuzzy weighting procedures for land-
slide susceptibility zonation in Darjeeling Himalayas,” Engi-
neering Geology, vol. 85, pp. 347–366, 2006.
[31] S. K. Acharya and K. K. Ray, Daling Group and Related Rocks,
Special Publication No. 22, Geological Survey of India,
Kolkata, India, 1989.
[32] P. J. Mason, M. S. Rosenbaum, and J. M. Moore, “Digital
image texture analysis for landslide hazard mapping,” Geo-
logical Society, London, Engineering Geology Special Publi-
cations, vol. 15, no. 1, pp. 297–305, 1998.
[33] J. C. Davis, Statistics and Data Analysis in Geology, John Wiley
& Sons, New York, NY, USA, 1986.
[34] S. Sarkar, Landslide hazard zonation and slope stability as-
sessment techniques: applications to Srinagar-Rudraprayag,
Garhwal Himalaya, Ph.D. thesis, University of Roorkee,
Roorkee, India, 1996.
Advances in Civil Engineering 17
... Landslides are the result of a combination of intrinsic conditions and external env ronmental factors. However, no standardized guidelines or criteria exist for determinin the influencing factors in landslide susceptibility studies [22][23][24][25]. Based on the findings [21,26], the influencing factors were categorized into four groups: morphological, ge logical, hydrological, and triggering factors. ...
... Landslides are the result of a combination of intrinsic conditions and external environmental factors. However, no standardized guidelines or criteria exist for determining the influencing factors in landslide susceptibility studies [22][23][24][25]. Based on the findings of [21,26], the influencing factors were categorized into four groups: morphological, geological, hydrological, and triggering factors. ...
Article
Full-text available
Landslides are common geological hazards worldwide, posing significant threats to both the environment and human lives. The preparation of a landslides susceptibility map is a major method to address the challenge related to sustainability. The study area, Nyingchi, is located in the southeastern region of the Qinghai-Tibet plateau, characterized by diverse terrain and complex geological formations. In this study, CNN was used to extract high-order features from the influencing factors, while BiLSTM was utilized to mine the historical data. Additionally, the attention mechanism was added to adjust the model weights dynamically. We constructed a hybrid CNN–BiLSTM-AM model to assess landslide susceptibility. A spatial database of 949 landslides was established using remote sensing images and field surveys. The effects of various feature selection methods were analyzed, and model performance was compared to that of six advanced models. The results show that the proposed model achieved a high prediction accuracy of 90.12% and exhibits strong generalization capabilities over large areas. It should be noted, however, that the influence of feature selection methods on model performance remains uncertain under complex conditions and is affected by multiple mechanisms.
... But the area suffers from a number of landslides every year that seriously affect locals as well as tourists, upsetting social and economic life in the area (Saha et al. 2023a). The landslides are largely due to the region's favorable terrain conditions, exacerbated by heavy rainfall, human activities, and occasional earthquakes (Sarkar et al. 2013;Pal et al. 2016;Chawla et al. 2018). This susceptibility to slope failure seriously threatens the region's stability, particularly affecting railway services (Samanta and Majumdar 2020). ...
Article
Full-text available
Landslide susceptibility mapping is crucial for reducing risks in culturally and historically significant areas like the Darjeeling Toy Train route, a UNESCO World Heritage site. In this study, the risk of landslides along this road is evaluated using Geographic Information System (GIS) tools and advanced machine learning models, such as Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Logistic Regression, and Classification and Regression Trees (CART). It uses a set of 512 landslide and non-landslide sites, with a 70:30 split between training and testing. Within the research area, thirteen topographical, hydrological, and geological factors linked to landslides are shown as GIS layers to make maps of landslide susceptibility (LSM). The study area particularly vulnerable to various types of landslides, including debris slides, rock falls, and soil slips. ROC–AUC results show that the SVM model did the best (0.813), followed by GBM (0.807), Logistic Regression (0.797), and CART (0.781). SVM had the highest accuracy rate at 83.2%, followed by GBM at 81.5% and LR at 80.3%. CART had the lowest overall accuracy rate at 78.6%. Furthermore, confusion matrix analysis showed that SVM and Logistic Regression were better at finding actual landslide-prone areas, with 84.6% and 82.1% recall rates, respectively. This made them more accurate in predicting high-risk areas. Susceptibility levels were categorized, revealing high-risk areas like Darjeeling and Rishihat and safer areas like Kurseong and Mohanbari. For lowering the risk of landslides and protecting this historic route, these results are very useful for land management and disaster preparation.
... Despite the frequent occurrence of landslides in the Kalimpong district, a comprehensive assessment of landslide susceptibility using multi-sensor datasets and advanced modelling techniques is lacking. Most of the existing research in this region has employed conventional qualitative methods (Sarkar and Kanungo 2004;Chawla et al. 2018Chawla et al. , 2019Saha et al. 2023aSaha et al. , 2023b to delineate landslide susceptibility zones by assigning weights to each conditioning factor. However, the process of assigning these weights is inherently subjective. ...
Article
This study delineates landslide susceptibility zones in the Kalimpong district by integrating multi-sensor datasets and assessing the effectiveness of statistical and machine learning models for precision mapping. The analysis utilises a comprehensive geospatial dataset, including remote sensing imagery, topographical, geological, and climatic factors. Four models were employed to generate landslide susceptibility maps (LSMs) using 16 influencing factors: two bivariate statistical models, frequency ratio (FR) and evidence belief function (EBF) and two machine learning models, random forest (RF) and support vector machine (SVM). Out of 1244 recorded landslide events, 871 events (70%) were used for training the models, and 373 events (30%) for validation. The distribution of susceptibility classes predicted by The RF and SVM models produced similar susceptibility distributions, predicting 13.30% and 14.30% of the area as highly susceptible, and 2.42% and 2.82% as very highly susceptible, respectively. In contrast, the FR model estimated 20.98% of the area as highly susceptible and 4.30% as very highly susceptible, whereas the EBF model predicted 17.42% and 5.89% for these categories, respectively. Model validation using receiver operating characteristic (ROC) curves revealed that the machine learning models (RF and SVM) had superior prediction accuracy with AUC values of 95.90% and 86.60%, respectively, compared to the statistical models (FR and EBF), which achieved AUC values of 74.30% and 76.80%. The findings indicate that Kalimpong-I is most vulnerable, with 6.76% of its area categorised as very high susceptibility and 24.80% as high susceptibility. Conversely, the Gorubathan block exhibited the least susceptible, with 0.95% and 6.48% of its area classified as very high and high susceptibility, respectively. This research provides essential insights for decision-makers and policy planners in landslide-prone regions and can be instrumental in developing early warning systems, which are vital for enhancing community safety through timely evacuations and preparedness measures.
... Roy et al. 2023;Poddar and Roy 2024,), Logistic Regression (LR) (Basu and Pal 2017), Support Vector Machine (SVM) (J. , Generalized Linear Model (GLM) (Saha et al. 2022), Random Forest (RF) (Saha et al. 2022), Convolutional Neural Network (CNN) (Saha et al. 2022;Moghimi et al. 2024) and Genetic Programming (GP) (Chawla et al. 2018). Soft-computing based ensembles with bi-variate statistical methods were also proposed by S. Das et al. (2024). ...
Article
Full-text available
In the field of landslide susceptibility, the utilization of data driven methodologies has seen a significant breakthrough. However, the performance of the models depends on the geo-environmental factors, and the selection of factors vary from one location to another, and this leads to a persistent lacuna for the present exploration. This study was aimed to assess landslide susceptibility for Darjeeling hills in Eastern Himalayan region with sixteen causative geo-environmental factors. The selection of causal factors was performed through a two-stage procedure, namely Pearson’s correlation coefficient (PCC) and Boruta algorithm (PCC-BA). The dataset associated with the research was split randomly into 70:30 ratio for train and test data. In addition, 30% of the training data was taken as validation dataset. Four advanced data-driven models namely K-nearest neighbour (KNN), Boosted Tree (BT), Gradient Boosting Machines (GBM) and ensembled Neural Network with Principal Component Analysis (PCA-NN) were taken up and four advanced novel ensembles namely KNN-BT, PCA-NN-BT, GBM-KNN and GBM-PCA-NN were constructed. The susceptibility maps were grouped into five divisions, viz., very low (VL), low (L), medium (M), high (H), and very high (VH) susceptibility. Through area under receiver operation characteristics curve, the accomplishment of constructed susceptibility models was substantiated with training, testing and validation dataset, where KNN-BT attained 0.943, 0.889 and 0.944 respectively, PCA-NN-BT attained 0.934, 0.876 and 0.943 respectively; GBM-KNN attained 0.959, 0.897 and 0.957 respectively; and GBM-PCA-NN attained 0.956, 0.889 and 0.962 respectively. The researchers have utilized an extensive explainable artificial intelligence (ex-AI) method, partial dependence profile (PDP) to quantify the effect of causal factors on all the four ensembled models. The study was aimed to demonstrate a significant capacity to substantially optimize disaster mitigation policies with a constituent endeavour to bridge the chasm between contemporary machine learning approaches and geo-spatial applications, and thereby paving the way to enhance the resilience of inhabitants in landslide prone areas of hilly portion of Darjeeling district.
... In particular, Darjeeling, a high altitudinal temperate, subtropical climatic region nestled in the Eastern Himalayan region, has been grappling with worsening air quality over the last few decades [25]. Darjeeling has become a popular tourist hub, mainly during summer [26], as the hills are prone to landslides due to heavy rainfall in the monsoon [27], and experience cold temperatures during winter [28]. The surge in tourism has also brought a plethora of vehicles, contributing significantly to pollution levels [25]. ...
Article
The present study focuses on the elemental characterization and contribution of prominent sources of particulate matter (PM) in Darjeeling, the high-altitudinal eastern Himalayan station. The concentration of PM10 and PM2.5 was exceeded the National Ambient Air Quality Standards (NAAQS) for 72% and 83% of the sampling days, respectively. Since the World Health Organization or other government organizations has not set any standards for PM1, the standards of PM10 and PM2.5 were considered as benchmarks. The concentration of PM1 exceeded the NAAQS for PM10 and PM2.5 on 57% and 85% of the days, respectively. The elemental characterization using wavelength dispersive X-ray fluorescence (WD-XRF) technique identified 21 elements with the dominance of Si, Na, B, Ba, Al, and K in PM10; while, Al, N, and B in PM2.5 and PM1. Principal component analysis depicted that biomass burning, fossil fuel combustion, crustal/soil dust, and industrial emissions were identified as primary contributors to PM10; PM2.5 was substantially attributed to industrial emissions, agricultural activities, biomass burning, vehicular activities and natural sources. Additionally, natural sources and anthropogenic activities like vehicular, agricultural, and industrial emissions, and combustion were identified as the major sources of PM1 in Darjeeling. The findings of this study could potentially raise awareness among researchers and policymakers, prompting them to develop sustainable strategies in hill regions across the globe.
Article
Full-text available
This study analyses trends in landslide publication in India, between 2010 and 2020, focusing on 79 studies, sourced from platforms like Google Scholar, ResearchGate, Web of Science, and ScienceDirect. The analysis reveals that approximately 65% of the publications were featured in Q1 and Q2 journals, with significant contributions from premium institutions like IIT-B (Mumbai), IIT-ISM (Dhanbad), and WIGH (Dehradun). The studies primarily focus on regions such as Uttarakhand, Himachal Pradesh, Tamil Nadu, and Kerala. Remote sensing and GIS emerged as the most frequently used approaches, appearing in 31 studies, followed by empirical methods (25 studies) and numerical techniques (20 studies). Researchers duly examined susceptibility, vulnerability, and risk using statistical techniques, with kinematic analysis and Slope Mass Rating (SMR) being notable empirical methods. The finite-element method was preferred for numerical slope failure analysis. The study found that 57% of the publications relied on existing data without conducting laboratory tests, while the remaining 43% conducted tests focusing on rock strength, shear strength, and physico-mechanical parameters. Direct shear tests and unconfined compressive strength (UCS) were the most commonly performed laboratory tests. Lithological analysis identified 20 different rock/soil types, with gneiss, quartzite, sandstone, and others frequently documented. Mitigation strategies such as informed decision-making, mechanical remediation, drainage system installation, and slope re-scaling were commonly suggested. Among the 31 computer-based programs used, ArcGIS, DIPS, and RS2 were the top tools for assessing mass-movement issues. The most common failure types reported were rock falls, debris flows, and rock slides, with rainfall identified as the primary landslide-triggering factor in 52 instances. The study highlights geology, geomorphology, and hydrology as key aggravating factors for landslide instances.
Article
Full-text available
Landslide Susceptibility Zonation is an efficient technique decision-makers use for disaster mitigation in landslide-prone regions. This study proposes an alternate approach for LSZ mapping, aiming to mitigate the limitations of the subjective expert opinion-based methods presently employed by disaster management authorities in India. Consequently, a GIS-based ensemble of Frequency Ratio and Analytical Hierarchy Process is employed, which offers a more robust and objective evaluation of Landslide Susceptibility. A landslide inventory of 592 incidents is processed using the database maintained by the Geological Survey of India, the national nodal agency for landslide studies. Then, LSZ mapping is conducted for a selected region in the Indian Himalayas using the processed inventory and nine causative factors (Elevation, Slope, Aspect, Curvature, Terrain Ruggedness Index (TRI), Distance to drainage, Land Use/Land Cover (LULC), Geology, and Lithology) as input. The generated LSZ map is evaluated using separate subsets of the inventory, yielding accuracies of 74.13% and 75.08%, respectively, during the training and testing stages. The study's findings hold potential implications for more effective disaster mitigation strategies and early warning systems in landslide-prone regions.
Article
Full-text available
The socioeconomic prosperity of a region is intricately tied to its infrastructure development, particularly the connectivity provided by highways. Preserving the economic well-being of an area demands meticulous planning and the execution of construction projects with minimal risk. Landslide hazard zonation (LHZ) is a crucial tool for assessing the likelihood of landslide occurrences in specific regions. This study delves into the LHZ of a 171.2 km 2 area, focusing on critical causative factors such as slope inclination and aspect, slope mass material, Land Use/Land Cover (LULC), distance from fault lines, roads, and drainage systems. The classification resulted in four distinct zones along the National Highway (NH-109) stretch from Bhowali to Almora. Notably, the moderate-hazard zone encompasses 60.4% of the area, followed by the high-hazard zone at 24.4%, the low-hazard zone at 11.9%, and the very high-hazard zone at 3.3%. To enhance the reliability of the study, 63 previously identified landslide locations were marked based on diverse sources, including field studies and landslide inventory data from the Geological Survey of India (GSI). In the validation process, 12 past landslide occurrences were situated within the very high-hazard zone, 32 in the high-hazard zone, 18 in the moderate-hazard zone, and only one in the low-hazard zone. The receiver operating characteristic (ROC) curve yielded a commendable quality with a % area under the curve (AUC) of 71.2%. In summary, this research underscores the importance of LHZ in assessing and mitigating landslide risks along a stretch of the NH-109 corridor. The findings provide valuable insights for informed decision-making in infrastructure development and risk management, contributing to the sustainable growth and protection of the region's livelihood.
Article
Full-text available
In recent years, earthquakes have been frequent near Xingwen County in southwestern China, where frequent landslides have continued to increase in the aftermath of earthquakes, causing severe infrastructure damage and economic losses. Therefore, tools to assess and map post-earthquake landslide susceptibility have become indispensable in today’s society. This study aims to integrate machine learning (ML) and statistical models to assess post-earthquake landslide susceptibility. We first established comprehensive framework by introducing surface deformation information as a seismic factor. The multicollinearity among the landslide condition factors (LCFs) was then assessed using the pearson correlation coefficient (PCC). The PCC values showed that only the altitude difference is redundant among all the LCFs. Finally, we integrated the Frequency Ratio (FR) and Index of Entropy (IoE) with the Extreme Learning Machine (ELM) to evaluate the post-earthquake landslide susceptibility. The results calculated by IoE revealed that the slope, elevation, plan curvature, lithology, and distance to fault are the essential causative factors for landslides. The final landslide susceptibility maps demonstrate that between 32.9% and 38.7% of the study area has high and very high landslide susceptibility. The values of several parameters proved that the hybrid FR–ELM and IoE–ELM models demonstrated superior performance compared to FR, IoE, and ELM. Thus, compared to the independent model, the hybrid model proposed in this study is more capable of identifying actual landslide areas due to its improved accuracy.
Article
Full-text available
Rainfall is inevitably one of the main factors that trigger landslides. However, not much study has been conducted on the impact of groundwater rise on slope stability. Thus, this study is intended to focus on the rise of the groundwater level from the bottom of the slope which would lead to landslides due to pore pressure development by eliminating other landslide-triggering factors (i.e., infiltration and surface runoff). Saturated sand was used for slope modeling, and sand densities of 1523 kg/m ³ , 1562 kg/m ³ , and 1592 kg/m ³ were tested with a constant slope angle of 45°. Another set of experiments was also performed on slopes having angles of 25°, 45°, and 60° and with a maintained density of sand at 1562 kg/m ³ . Through observation, failure was initiated first at the toe of the slope before minor and major slips or total collapse occurs. Dimensions of slip surfaces were measured and included in SLOPE/W for the computation of the safety factor. In conclusion, safety factors are found to be higher in denser soil and in the lowest slope angle. However, faster occurrence of collapse in denser soil was identified and could be contributed by the faster pore water pressure development.
Article
Full-text available
This paper assesses the performance of the landslide susceptibility analysis using frequency ratio (FR) with an iterative random sampling. A pair of before-and-after digital aerial photographs with 50 cm spatial resolution was used to detect landslide occurrences in Yongin area, Korea. Iterative random sampling was run ten times in total and each time it was applied to the training and validation datasets. Thirteen landslide causative factors were derived from the topographic, soil, forest, and geological maps. The FR scores were calculated from the causative factors and training occurrences repeatedly ten times. The ten landslide susceptibility maps were obtained from the integration of causative factors that assigned FR scores. The landslide susceptibility maps were validated by using each validation dataset. The FR method achieved susceptibility accuracies from 89.48% to 93.21%. And the landslide susceptibility accuracy of the FR method is higher than 89%. Moreover, the ten times iterative FR modeling may contribute to a better understanding of a regularized relationship between the causative factors and landslide susceptibility. This makes it possible to incorporate knowledge-driven considerations of the causative factors into the landslide susceptibility analysis and also be extensively used to other areas.
Article
Full-text available
This study presents the use of geographical information system (GIS) datasets and methods to investigate landslide susceptibility in the rugged mountainous terrain of Shigar and Shyok Basins in northern areas of Pakistan. Study area is situated in Karakorum mountainous region where catastrophic landslides occur frequently and pose a serious threat to local living conditions. Landslide susceptibility index maps were prepared by combining four main indicators groups: (1) human induced parameters, which contain Landsat 8 imagery and distance from roads; (2) topographical parameters, which include slope, aspect and plan curvature; (3) hydrological parameters, which contain stream power index, topographic wetness index and distance from drainage and (4) geological parameters, which consist of lithology and distance from major thrusts and faults. These layers were prepared in GIS environment, and analytical hierarchy process (AHP) based heuristic approach was adopted to generate final landslide susceptibility map for this preliminary regional level landslide hazard study. Three different susceptibility zones have been identified in the region. AHP weights with 0.04 consistency ratio were used in the preparation of final map. The final landslide susceptibility map of the study area indicates that the low, moderate, and high landslide susceptibility classes are respectively covering 23.2% (4745 km²), 45% (6737 km²), 23.2% (3444 km²) when distance from roads, distance from drainage and distance from major thrusts and faults parameters were not incorporated in the analysis. Including distance from roads, distance from drainage and distance from major thrusts and faults parameters layers to the final landslide susceptibility analysis shows that moderate susceptibility class is the predominant landslide susceptibility category covering 88% (13,412 km²) of the study area characterized by steep slopes (30°–45°), low positive plan curvature (0–1) and weak lithology.
Article
Full-text available
Landslide susceptibility map is very important for engineers, geologist, and land use planner for prevention and mitigation of landslide hazard in the area. This paper presents landslide susceptibility analysis for the part of tehsil Balakot, NW Himalayas, Pakistan. In this study, topographical, geological and remote sensing data were collected and processed using geographic information system (GIS) and ERDAS Imagine software. Nine influential causative factors for landslide occurrence were used for this purpose. The causative factors that influence the landslide occurrence include slope, aspect, curvature, elevation, lithology, land cover, faults, road network, and hydrology. These factors were analyzed for construction of thematic data layers. Numerical weight for each factor was assigned by the Analytic Hierarchy Process (AHP) using Pairwise Comparison Method. The landslide susceptibility indices were derived using weighted overlay method (WOM). As a result, landslide susceptibility map was produced in GIS. The susceptibility map classified the study area into very high, high, moderate, and low susceptible zones. The results of the susceptibility mapping were verified using the landslide occurrence. The verification results revealed 76 % accuracy. The validated results showed good agreement between landslide occurrence and produced susceptibility map of the area. Susceptibility map prepared by weighted overlay method is validated for landslide hazard, mitigations, and land use planning for future construction in the area.
Article
Full-text available
A regional scale basin susceptible to landslide located in Qaemshahr area in northern Iran was chosen for comparing the reliability of weight of evidence (WofE), fuzzy logic, and frequency ratio (FR) methods for landslide susceptibility mapping. The locations of 157 landslides were identified using Google Earth® or extracted from archived data, from which, 22 rockslides were eliminated from the data-set due to their different conditions. The 135 remaining landslides were randomly divided into two groups of modelling (70%) and validation (30%) data-sets. Elevation, slope degree, slope aspect, lithology, land use/cover, normalized difference vegetation index, rainfall, distance to drainage network, roads, and faults were considered as landslide causative factors. The landslide susceptibility maps were prepared using the three mentioned methods. The validation process was measured by the success and prediction rates calculated by area under receiver operating characteristic curve. The ‘OR’, ‘AND’, ‘SUM’, and ‘PRODUCT’ operators of the fuzzy logic method were unacceptable because these operators classify the target area into either very high or very low susceptible zones that are inconsistent with the physical conditions of the study area. The results of fuzzy ‘GAMMA’ operators were relatively reliable while, FR and WofE methods showed results that are more reliable.
Article
For nearly three decades there has been a phenomenal growth in the field of Remote Sensing. The second edition of this widely acclaimed book has been fully revised and updated. The reader will find a wide range of information on various aspects of geological remote sensing, ranging from laboratory spectra of minerals and rocks, ground truth, to aerial and space-borne remote sensing. This volume describes the integration of photogeology into remote sensing as well as how remote sensing is used as a tool of geo-exploration. It also covers a wide spectrum of geoscientific applications of remote sensing ranging from meso- to global scale. The subject matter is presented at a basic level, serving students as an introductory text on remote sensing. The main part of the book will also be of great value to active researchers.
Conference Paper
The slope changes from a stable to an unstable condition are the key factor for the occurrence of landslides. A change in the stability of a slope can be caused by a number of factors, acting together or alone. Natural causes of landslides include: pore water pressure, loss or absence of vertical vegetative structure, earthquakes adding loads and volcanic eruptions etc. Landslide susceptibility mapping is one of the crucial stages of landslide hazard and risk assessment. Accordingly, a methodology for landslide prone region using an integrated approach is presented. Our study area falls within the Eastern part of Indian Himalayas, in the northern most district of the state of West Bengal i.e. the Darjeeling District. Topographic sheets, field data and other informative maps were used as inputs to the study. Terrain factors which are important in contributing to the landslides occurrence in the region were identified and represent topographical, hydrological and geological conditions of the terrain. Terrain information such as lineament, geology, slope, relief, aspect etc.; have been derived from the satellite imageries and the existing thematic information have been updated to enable the quantification of parameters causing landslides. Map overlaying technique of spatial information has been used for qualitative landslide hazard susceptibility analysis. Very high and high potential zones have been identified. The results show a close agreement with the existing field instability condition.
Article
An application of GIS for landslide hazard assessment using multivariate statistical analysis, mapping, and the evaluation of the hazard maps is presented. The study area is the Kulekhani watershed (124 km2) located in the central region of Nepal. A distribution map of landslides was produced from aerial photo interpretation and field checking. To determine the factors and classes influencing landsliding, layers of topographic factors derived from a digital elevation model, geology, and land use/land cover were analyzed by quantification scaling type II (discriminant) analysis, and the results were used for hazard mapping. The effects of different samples of landslide and non-landslide groups on the critical factors and classes and subsequently on hazard maps were evaluated. Simple random sampling was used to obtain samples of the landslide group, and either an unaligned stratified random sampling or an aligned systematic sampling method generated the non-landslide group. For the analysis, one set of the landslide group was combined with each of five different sets of the non-landslide groups. Combinations of different samples yielded some minor differences in the critical factors and classes. The geology was found to be the most important factor for landslide hazard. The scores of the classes of the factors quantified by the five analyses were used for the hazard mapping in the GIS, with four levels of relative hazard classes: high, moderate, less, and least. The evaluation of five hazard maps indicated higher accuracy for the combinations in which the non-landslide group was generated by the unaligned stratified random sampling method. The agreements in the hazard maps, produced from different sample combinations using unaligned stratified random sampling for selecting nonlandslide group, were within the acceptable range for the practical use of a hazard map.