You are on page 1of 19

Landslide susceptibility mapping along

national highway-1 in Jammu and


Kashmir State (India)

Gulzar Hussain, Yudhbir Singh,


Kanwarpreet Singh & G. M. Bhat

Innovative Infrastructure Solutions

ISSN 2364-4176
Volume 4
Number 1

Innov. Infrastruct. Solut. (2019) 4:1-17


DOI 10.1007/s41062-019-0245-9

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Nature Switzerland AG. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

1 23
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59
https://doi.org/10.1007/s41062-019-0245-9

TECHNICAL PAPERS

Landslide susceptibility mapping along national highway‑1 in Jammu


and Kashmir State (India)
Gulzar Hussain1 · Yudhbir Singh1 · Kanwarpreet Singh2 · G. M. Bhat1

Received: 5 June 2019 / Accepted: 9 October 2019


© Springer Nature Switzerland AG 2019

Abstract
A remote sensing and GIS-based landslide susceptibility mapping (LSM) has been carried out using frequency ratio (FR)
and weight of evidence (WoE) methods to identify and delineate the potential failure zones along National Highway – 1.
The thematic layers of various landslide causative factors have been generated for modeling in GIS. In addition, a landslide
inventory along the road network was prepared using satellite imagery, Google earth, and extensive field visits. LSM clas-
sified the area into five susceptibility classes: very low, low, moderate, high, and very high classes. The validation result
further substantiates the study and inferred that area under success rate curve shows a satisfactory relation between land-
slide affecting factors and landslide occurrences in case of FR and WoE models having an accuracy of 86.57% and 76.86%,
respectively. A landslide density method has also been adopted for validation of LSM which showed acceptable results with
decreasing trend of landslide density from very high to very low susceptible zone for both FR and WoE models. The LSM
generated will be helpful for various stakeholders like planners, engineers, designers, and local public for future construc-
tion and maintenance in the study area.

Keywords  Remote sensing · Arc GIS · Ladakh · Frequency ratio · Weight of evidence

Introduction of a number of factors such as the interaction of various


geological, geomorphological, physical, and anthropogenic
Landslide is one of the most common phenomena which activities [4, 5]. In highly mountainous areas, a landslide is
claims life, settlement, and environment globally. Landslides common which can be triggered because of various causa-
cause life and property losses around the world annually [1, tive factors like rainfall, seismicity, and rapid erosion by the
2]. Landslides have been recognized as a significant natural stream and water fluctuation [2, 6]. The gradual increase
disaster all over the world and particularly in hilly regions in the frequency of landslides events is due to a population
especially in the Himalaya [1]. The hilly areas in Himalaya, explosion which forces for rapid developmental activities
Western Ghats, and the Meghalaya Plateau in India cover in the area including road widening, construction of dams,
around 15% of the country (~ 0.49 million km2) which are and settlements, etc. The naturally existing slopes may get
highly prone to landslide [3]. The failure mechanism of converted into excavated slopes which are more prone to
slopes are not restricted to a single factor but is the interplay failure [7–9]. Both the intrinsic and extrinsic causative fac-
tors are required for the correct assessment of landslide sus-
* Kanwarpreet Singh ceptibility in an area [10–12]. For the hazard assessment
singhkanwarpreetnit@gmail.com of an area, most of the studies around the world involve
Gulzar Hussain only intrinsic factors because it is not easy to retrieve data
suhail490@gmail.com based on extrinsic factors. Hence, assessment of an area
G. M. Bhat in the context of landslide hazard where extrinsic factors
abhishek1@nith.ac.in are neglected is termed as landslide susceptibility mapping
[12–14]. The landslide study has been widely adopted during
1
Postgraduate Department of Geology, University of Jammu, recent years due to technology advancement related to the
Jammu, India
spatial information. The high-resolution satellite imageries
2
Civil Engineering Department, National Institute and GIS environment resulted in adaptation of the landslide
of Technology, Hamirpur, Himachal Pradesh 177005, India

13
Vol.:(0123456789)
Author's personal copy
59   Page 2 of 17 Innovative Infrastructure Solutions (2019) 4:59

susceptibility assessment on medium to large scale by giv- stretches for 120 km is confined to Kargil and Lamayuru sector
ing many reliable and appropriate models [15–21]. Various falls in the Indus tectonic zone and Trans-Himalaya zone of
qualitative and quantitative methods are available for land- Northwest Himalaya. The study area stands distinguishable
slides susceptibility assessment. from the rest of the state Jammu and Kashmir in terms of
The qualitative method is generally based on the knowl- topography and physiography. The region is bounded by the
edge of an expert and has been proved as highly produc- Karakoram to the north, and Himalayan mountains to the south
tive for the assessment of the regional area [1, 22, 23]. The act as an orographic barrier and created an arid environment
quantitative methods emphasize objective measurements and [49]. The average annual rainfall is very scanty during summer
the statistical, mathematical, or numerical analysis of data and experiences heavy snowfall in winter. However, there is
collected through polls, questionnaires, and surveys, or by a great variation in summer (34.8 °C) and winter temperature
manipulating the pre-existing statistical data using various (− 27.9 °C) below the freezing point [50]. The area is charac-
computational techniques. The quantitative-based frequency terized by two prominent geomorphic units, i.e., gorges and
ratio method is one of the simple probabilistic models based barren mountain ranges which goes up to 3000–4000 m from
on the observed relationships between each landslide causa- the mean sea level. The area is partly drained by the main river
tive factor [24] and landslides and is successfully used by “Wakha Chu” fed up by small streams of various order flowing
number of researchers around the world for landslide sus- in the north-westerly direction.
ceptibility mapping [25–40]. A quantitative-based statisti-
cal approach of weight of evidence (WoE) is a Bayesian Database preparation
approach which can assess the relationship of the dependent
variable to independent predictive variable. It is represented The collection and preparation of data is the basic funda-
by linear logarithm division of events (%) and non-events mental of any research. The landslide susceptibility map-
(%) to estimate the relative importance of evidence through ping of an area requires prior knowledge of its influencing
statistical means. This approach has been adopted worldwide factors which need to be identified and mapped [51]. The
by numerous researchers [41–48]. selection of landslide causative factors for susceptibility
The present study deals with the landslide susceptibility assessment depends upon the type, scale, availability, and
mapping along with the strategically important road net- method of data acquisition [52]. Nine intrinsic factors viz.
work of the country which is also the lifeline of the remote slope, aspect, curvature, relief, lithology, landuse landcover
districts of Jammu and Kashmir State including Ladakh (LULC), proximity to the river, proximity to the road, and
region. The selected road section is a major transport road proximity to lineament were converted into thematic maps.
corridor connecting the Ladakh region with the rest of India. A digital elevation model (DEM) of the study area was gen-
This route is historically and culturally important. The route erated from Survey of India (SOI) toposheets of 1: 50,000
passes near India–Pakistan border through high mountain scale at a contour interval of 20 m, by generating a tin of
passes and most of the road clings to mountainsides. Various 30 m resolution which finally converted into DEM in GIS.
types of landslides have been observed along this road sec- The slope gradient, slope aspect, curvature, and relief maps
tion which causes obstruction to the traffic flow every year are derived from the DEM in Arc GIS. The road corridor,
during monsoon season. A few of the man-made road cuts river, and its tributaries have also been digitized from SOI
were also observed along the stretch. toposheets. The geological map of the area has been digi-
The landslide susceptibility map along the NH -1 has tized as an input map for lithology based on the input base
been prepared using FR and WoE model, classifying the map as proposed by researcher [53].
selected region into different susceptibility zones ranging The lineaments of the area were delineated from the satel-
from very low to very high. These landslide susceptibility lite imagery and geological map. The LULC map is derived
maps will be helpful for various stakeholders like planners, from IRS LISS IV satellite imagery of 5.8 m resolution. All
engineers, geologist, and geotechnicians in future to pro- the above mentioned thematic data layers have been resam-
vide smooth traffic flow along the transport road corridor by pled into 30 m*30 m resolution raster format for analysis in
mitigating slopes present in high and very high susceptible GIS platform.
zones.

Landslide inventory
Study area
The landslide inventory map has been considered as a
The transport corridor of NH – 1 passes through an altitude basis for LSM [54, 55] showing the locations of landslides
variation between 2980 m in the valley and 4500 m on peaks is making their identification easy. In addition, it may be
the only all-weather surface link (Fig. 1). The study area which used as reducing landslide hazard and risk on a regional

13
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59 Page 3 of 17  59

Fig. 1  Digital elevation model of the study area with NH - 1 and landslide occurrences

scale [56]. A deep knowledge about the failure mechanism landslide inventory map. A huge damage to the property
of landslide can be obtained from a systematic landslide and three people have been killed because of landslide
inventory. There is no record available related to the past activities along the NH-1. A total of 60 landslides were
landslide events along the NH in any department; however, identified in the area and plotted as point feature on the
the extensive field visits have been conducted to formulate map (Fig. 2). The landslide affected area can be described
landslide inventory map by considering landslides which as point feature when the areal extent is small and cannot
occurred in the past and currently active landslides hav- be drawn on a map [57]. The different types of failures
ing failure potential. A LISS IV satellite imagery, Google observed in the area include rockfall, debris flow, and
earth, and field visits results into the preparation of the debris/earth slide (Fig. 2).

13
Author's personal copy
59   Page 4 of 17 Innovative Infrastructure Solutions (2019) 4:59

Fig. 2  Field photograph of
landslide occurrences a, b, c,
d showing translational slide
along the road, e, f, g rock
failure, h debris slide

Landslide causative factors the effect of soil moisture and vegetation pattern show a
strong influence on the slope aspect [58, 59]. The aspect
Aspect map of study area has been categorized into eight different
classes on the basis of direction, i.e., North (337.5°–22.5°),
The slope aspect plays an important role in LSM because it Northeast (22.5°–67.5°), East (67.5°–112.5°), South-
specifies the slope direction of the terrain and often shows east (112.5°–157.5°), South (157.5°–202.5°), Southwest
a good relation to the occurrence of landslides [52]. The (202.5°–247.5°), West (247.5°–292.5°), and Northwest
bedrock structure and exposition to the sun by combining (292.5°–337.5°) as shown in Fig. 3a.

13
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59 Page 5 of 17  59

Curvature the basis of natural break classification [64] into five


classes: 2596–2939 m; 2939–3241 m; 3241 m–3511 m;
The slope curvature is an important parameter in LSM as it 3511–3749 m; and 3749–4167 m (Fig. 3e).
controls the movement of surface hydrological regime, ero-
sion, and deposition rate as well as soil properties [60, 61].
The composite curvature has been used which is the combi- Proximity to the river
nation of both profile and plane curvature. The combination
curvature involves negative values which are classified as The authors [65] considered that the proximity of an area to
concave, positive values are classified as convex, and zero drainage plays important role in landslides because the water
values as flat as shown in Fig. 3b. along these networks have highly influenced the stability of
the slopes due to toe cutting of these slopes. The study area
has been classified into six drainage buffer classes, and these
Lithology are 0–25 m, 25–50 m, 50–150 m, 150–200 m, and rest of
the area (Fig. 3f).
Lithology of an area is a major intrinsic parameter for caus-
ing slope instability. The researchers [62] have considered
it as an important causative factor because unconsolidated Proximity to the road
materials are directly related to instability. Some of the
researchers inferred that a landslide problem varies differ- The proximity to the road has been considered as one of the
ently due to the diversity of lithological conditions along important parameters in landslide susceptibility mapping.
NH-1 [63]. Different lithological units in an area have a dif- The inattention of the geological and geotechnical factors
ferent level of instability [12, 54]. The geological map of during highway construction results into slope instability
the area proposed by authors [53] has been digitized for the [66]. Road construction without proper scientific study and
preparation of thematic layer map. However, the study is planning leads to landslide triggering due to human inter-
confined to the road network; hence, only those formations ference [67]. A road buffer has been created classifying the
which are exposed to the road network are considered. The study area into five classes 0–50 m, 50–100 m, 100–200, and
road network in the study area traverse from Ladakh plutonic rest of the area (Fig. 3g).
complex comprises of slightly to moderately weathered and
jointed granodiorite at its western extremities followed by
quarternary deposits comprise of moraines and glacioflu- Slope
vial deposits, Drass volcanic comprise basalt and andesite,
Sangelungma group comprise sandstone and shale, Undif- Slope gradient has a greater impact on the geomorphology
ferentiated Kuling–Lilang group comprise of phyllite, slate, and control over the surface and subsurface water flow [36,
and limestone as shown in Fig. 3c. 52, 68]. The slope map of the area has been categorized into
five classes: 0°–15°, 15°–25°, 25°–35°, 35°–45°, and > 45°
as (Fig. 3h). Slope gradient measures the rate of change of
Landuse landcover (LULC) elevation in the direction of the steepest fall. The gravity-
induced shear stress of unconsolidated material will be the
Landuse landcover represents the vegetation pattern and higher, and failure events will be more in case of steeper
constructional practices of the area. The landuse landcover slopes, but in the case of a moderate and gentle slope, the
map is derived from IRS LISS IV satellite imagery of 5.8 m shear stress of unconsolidated material will be less.
resolution by adopting band composition, supervised and
unsupervised classification method and categorized into five
types viz. sparse vegetation cover, dense vegetation cover, Proximity to lineament
water bodies, and settlements, and barren land. Most of the
area under investigation falls under barren land (Fig. 3d). Lineaments are structural features representing planes of
weakness. The occurrence of landslides is expected to be
close to lineaments because it helps in promoting selec-
Relief
tive erosion and confined water movement. The lineament
map was prepared from the geological map and high-res-
The study area comes under Trans-Himalaya and Indus
olution satellite imagery (IRS LISS-IV). The present area
tectonic zone and is devoid of vegetation cover and com-
in terms of proximity to lineament has been categorized
prised of incised valleys. The relief in the study area
into four groups viz. 0–100 m, 100–500 m, 500–1000 m,
is ranging between 2596–4167  m and is classified on
and > 1000 m (Fig. 3i).

13
Author's personal copy
59   Page 6 of 17 Innovative Infrastructure Solutions (2019) 4:59

Fig. 3  a Aspect map, b curvature map, c lithology map, d landuse landcover map, e proximity to river map, f proximity to road map, g relief
map, h slope gradient map, (i) proximity to lineament map

13
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59 Page 7 of 17  59

Fig. 3  (continued)

Frequency ratio model between the influencing factor and landslide gives frequency
ratio. A frequency ratio value greater than unity shows a
The present study is based on a frequency ratio model (FR) strong correlation for landslide occurrence, whereas
which is a probabilistic method based on the assumption that FR value less than unity represents lesser chance of
future landslide will occur under similar circumstances as landslide from given landslide initiating factor [24].
that of the past landslides. The relationship between the The frequency ratio (FR) can be expressed as:
observed landslide and influencing factor provides an estab- Frequency Ratio =  SlideRatio∕ClassRatioWhere Slide
lished correlation between landslide location and influencing ratio = Numberoflandslidepixel∕Totalnumberoflandslidepixel
factor in the study area [69, 70]. A quantitative relationship Class ratio = Numberofpixelineachclass∕

13
Author's personal copy
59   Page 8 of 17 Innovative Infrastructure Solutions (2019) 4:59

TotalnumberofpixelinwholeclassLandslide Susceptibility assessment for classifying an area into different susceptibil-


Index (LSI) =  FR
∑ ity cum hazard zones. The frequency ratio and weight of
evidence models have been utilized for the preparation of
LSM in the present study.
Weight of evidence model
Frequency ratio model
The WoE model depends upon the Bayesian probability
approach; initially, it has been originally developed for min- To analyze the role of each causative factor for landslide,
eral potential assessment [71]. The WoE approach is less the thematic maps have been generated in GIS environment
time-consuming as well as simple, and it can be applied into raster format of each 30 × 30 m size grids and further
easily using GIS packages. The weight for each landslide classified into different classes. The total number of pixels in
predictive factor (B) has been calculated based on the pres- each map is 128508 and the number of landslide inventory
ence or the absence of landslides (L) within an area. contains 60 pixels in the study area. For calculating the class
ratio, the number of pixels in the individual class is divided
by a total number of pixel in whole factor map. The slide
{ }
P BL
Wi+ = ln { } (1) ratio is obtained by division of number of pixels of the slide
P BL̄ in each class to total number pixels of the slide in whole fac-
tor map. The frequency ratio is then calculated by division
of slide ratio to the class ratio of each class of individual fac-
{̄} tor. The LSI is calculated using raster calculator in Arc GIS
P BL
by the summing up of frequency ratio of each factor and is
Wi− = ln { } (2)
̄ expressed as under:LSI = FR(slope) + FR(aspect) + FR(curvature) + 
P BL̄
FR(relief) + FR(proximity to river) + FR(proximity to road) + FR(lithology) 
+ FR(landuse landcover) + FR(proximity to lineament)
where P = Probability N  = Natural logB = Presence of The quantitative results of the FR approach are given in
potential landslide predictive factor B̄  = Absence of poten- Table 1. The relationship between slope aspect and landslide
tial landslide predictive factorL = Presence of landslide reveals that FR values greater than 1 for NE, E, SE, and S
L̄  = Absence of landslideWi+ = Predictable variable is pre- facing direction and FR values less than 1 for N, NW, W,
sent (Positive correlation) Wi− = Predictable variable is and SW facing directions clearly indicate NE, E, SE, and S
absent (Negative correlation) slope facing directions played an important role in initiating
The difference between the two weights is called as the the slope failures. The FR in case of curvature is greater than
weight contrast ( Wf  ), i.e., 1 for concave and convex and suggests that convex surface
Wf = Wi+ − Wi− is always exposed to the rain leading the continuous dila-
(3)
tion and contraction making the soil loose, while in case
The magnitude of the contrast depicts the overall spa- of concave surface, the rainwater is usually trapped for a
tial association between the predictable variable and the longer period causing infiltration and shear failure [72]. The
landslides. The weights have been attributed to the classes FR value in case of geology cum lithology map is the high-
of each selected thematic layer, respectively, to produce est for Nindum Formation (3.72) that comprises of weath-
weighted thematic maps in GIS, which were overlaid and ered shale, sandstone, siltstone, and limestone followed by
numerically added according to Eq. (4) to produce a land- Ladakh Granitic Complex (2.37) comprises of granodior-
slide susceptibility index (LSI) map. ite traversed by three to four major joint sets; Quarternary
∑ deposits (1.52) comprise of glaciofluvial deposits and Drass
LSI = Wf (4) volcanics (1.20). Nindum Formation comprised of shale,
sandstone with siltstone, etc., has shown the highest value of
FR, and these rocks are more prone to weathering and pro-
Results mote landslides. Ladakh Granitic Complex class of geology
which comprises of granodiorite rocks having average 3 to
An increase in the landslide intensity along NH-1 has been 4 major sets of joints has shown FR value less than Nindum
observed because of increase in the demand of development Formation but greater as compared to the remaining classes.
activities affecting life and property in the area. The trend LULC map has been prepared from IRS-P6 LISS IV data-
from traditional approaches to the most advanced approaches set having 5.8 m resolution, classified into five categories
can be seen all over the world for landslide susceptibility viz. water bodies, sparse vegetation, dense vegetation, set-
tlement, and barren land. The settlement has been shown

13
Table 1  Correlation between landslides and conditioning factors using the FR and WoE model
S. no. Factors Class Causative factor Landslide Slide ratio Class ratio FR W+ W− Wf
pixel count pixel count

1 Aspect North 14,627 2 0.03 0.11 0.29 − 1.27637 0.117613 − 1.39399


Northeast 20,673 16 0.27 0.16 1.66 0.45711 − 0.10409 0.561198
East 15,303 8 0.13 0.12 1.12 0.064742 0.014368 0.050375
Southeast 11,214 7 0.12 0.09 1.34 0.242097 − 0.00207 0.244164
South 19,758 13 0.22 0.15 1.41 0.29474 − 0.04658 0.34132
Innovative Infrastructure Solutions

Southwest 21,928 9 0.15 0.17 0.88 − 0.17719 0.055254 − 0.23244


West 15,341 4 0.07 0.12 0.56 − 0.63089 0.088811 − 0.7197
Northwest 9664 1 0.02 0.08 0.22 − 1.55506 0.09205 − 1.64711
2 Curvature Concave 70,396 33 0.55 0.55 1.00 − 0.04428 − 0.00489 − 0.03939
Flat 55,166 25 0.42 0.43 0.97 − 0.07812 0.021861 − 0.09998
(2019) 4:59

Convex 2946 2 0.03 0.02 1.45 0.326048 − 0.01071 0.336758


3 Geology/lithology Ladakh granitic complex 8114 9 0.15 0.06 2.50 0.816983 0.172944 0.644039
Kuksho Formation 636 0 0.00 0.00 0.00 0 0.275203 − 0.2752
Maklishun Formation 444 0 0.00 0.00 0.00 0 0.273702 − 0.2737
Karit Formation 502 0 0.00 0.00 0.00 0 0 0
Quaternary deposits 15,445 11 0.18 0.12 1.50 0.373959 0.195763 0.178196
Drass volcanics 17,800 10 0.16 0.14 1.14 0.136736 0.237015 − 0.10028
Nindum Formation 5177 9 0.15 0.04 3.75 1.266348 0.148842 1.117507
Ophiolite 1414 0 0.00 0.01 0.00 0 0.281306 − 0.28131
Shergole Formation 4926 0 0.00 0.04 0.00 0 0.309328 − 0.30933
Undifferentiated kuling group 74,050 21 0.35 0.58 0.60 − 0.54687 0.69802 − 1.24489
4 LULC Water bodies 1293 0 0.00 0.01 0.00 0 0.010113 − 0.01011
Author's personal copy

Dense vegetation 11,511 2 0.03 0.09 0.37 − 0.98851 0.059941 − 1.04845


Sparse vegetation 13,289 5 0.08 0.10 0.79 − 0.21585 0.022145 − 0.238
Barren area 101,282 52 0.87 0.79 1.08 0.094982 − 0.46308 0.558066
Settlements 1133 1 0.01 0.00 1.85 0.636778 − 0.00795 0.644729
5 Relief 2596–2925 m 21,464 20 0.33 0.17 2.00 0.642705 − 0.22271 0.865419
2925–3214 m 16,603 11 0.18 0.13 1.42 0.301661 − 0.06418 0.365845
3214–3484 m 30,942 16 0.27 0.24 1.11 0.053824 − 0.03469 0.088517
3484–3728 m 36,947 2 0.03 0.29 0.12 − 2.20299 0.305084 − 2.50807
3728–4167 m 22,552 11 0.18 0.18 1.04 0 − 0.00956 0.009557
Page 9 of 17 
59

13
Table 1  (continued)
59  

S. no. Factors Class Causative factor Landslide Slide ratio Class ratio FR W+ W− Wf
pixel count pixel count

13
6 Proximity to river 25 m 5609 0 0.00 0.04 0.00 0 0.096143 − 0.09614
50 m 5587 1 0.02 0.04 0.38 − 1.00709 0.079157 − 1.08625
Page 10 of 17

100 m 10,670 6 0.10 0.08 1.20 0.137673 0.032835 0.104838


150 m 9731 16 0.27 0.08 3.52 1.210622 − 0.1799 1.390518
200 m 8851 10 0.17 0.07 2.42 0.835404 − 0.05944 0.894849
Rest of the area 88,060 27 0.45 0.69 0.66 − 0.46883 0.609652 − 1.07848
7 Proximity to road 50 m 12,148 28 0.47 0.09 4.94 1.548389 − 0.52931 2.077696
100 m 11,130 21 0.35 0.09 4.04 1.348228 − 0.34019 1.688419
200 m 20,442 10 0.17 0.16 1.05 − 0.00166 − 0.00907 0.007416
Rest of the area 84,788 1 0.02 0.66 0.03 -3.7268 1.061378 − 4.78818
8 Slope 0°–15° 46,942 6 0.10 0.37 0.27 − 1.3438 0.349218 − 1.69302
15°–25° 47,080 25 0.42 0.37 1.14 0.080377 − 0.08272 0.163102
25°–35° 29,234 19 0.32 0.23 1.39 0.282456 − 0.12267 0.405121
35°–45° 5188 10 0.17 0.04 4.13 1.369586 − 0.14111 1.510699
45°–90° 64 0 0.00 0.00 0.00 0 0.000498 − 0.0005
9 Proximity to lineament 0–100 m 11,573 34 0.56 0.09 6.22 1.791035 − 0.74188 2.53291
100–500 m 37,636 9 0.15 0.29 0.51 − 0.71739 0.18402 − 0.90141
500–1000 m 28,521 10 0.16 0.22 0.74 − 0.33471 0.068629 − 0.40334
>1000 m 50,778 7 0.11 0.39 0.27 − 1.2682 0.378697 − 1.6469
Total no. of landslide pixels 60
Total no. of pixels in study area 128,508
Author's personal copy
Innovative Infrastructure Solutions
(2019) 4:59
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59 Page 11 of 17  59

Fig. 4  Landslide susceptibility map of NH-1A based on frequency ratio method

in the frequency ratio value of 1.0, and the barren area had classification (Fig. 4), because this method is a data clas-
frequency ratio value of 1.84. The FR value for the relation- sification method that minimizes and maximizes variance
ship between relief and landslides revealed that the higher between groups of data [64].
frequency ratio (2.0) is in 2596–2925  m class followed
by 1.42 and 1.1 in 2925–3214 m and 3214–3484 m class, Weight of evidence model
respectively. The higher value at low and mid-elevations
inferred that most of the debris material accumulated at The LSM map has also been delineated into different sus-
low relief areas. The snow melting, incessant rainfall, and ceptibiltiy classes by employing a method of WoE. The
anthropogenic activities cause slope failure. The proxim- weights for each classes of landslide causative factor have
ity to the drainage reveals that due to scanty rainfall in the been evaluated as per the methodology [Sect. 7].
area, river and nalas does not have a greater impact on the In case of slope gradient, the class 35°–45° is more
distribution of landslide. On the other hand, the relation- prone to landslides as compared to the other slope classes,
ship between proximity to road and landslide seems to be and it has also been observed in the field that steep slopes
very significant because the FR value is very higher (4.96) are the most affected because of slope failure activities in
for close proximity to the road (50 m) and decreases with the area (Table 1). In the case of slope aspect, the north-
increasing distance from the road. This clearly indicates that east, east, southeast, and south-facing slopes have higher
anthropogenic activities (road widening, etc.) along this road contrast for triggering landslides with positive weights
network have made the area more susceptible to landslides. which is higher as compared to other aspect classes. In
The FR value for the slope gradient factor is the highest case of slope curvature, the slopes having convex shape
(4.13) for 35°–45° class and substantial decrease in FR value are somewhat more prone to landslides as compared to
has been observed with decreasing slope gradient. flat and concave shaped slopes. The Nindum Formation
The landslide susceptibility index (LSI) calculated in case of geology cum lithology map has been observed
from the final landslide susceptibility map ranged between as major class for landslide prediction in the study area
2.61–21.21. The landslide susceptibility map thus prepared with weight 1.11 followed by Ladakh Granitic Complex
is categorized into five classes according to Jenks natural (0.65) and Quarternery deposits (0.17), respectively.

13
Author's personal copy
59   Page 12 of 17 Innovative Infrastructure Solutions (2019) 4:59

Fig. 5  Landslide susceptibility map of NH-1A based on the Weight of Evidence method

In case of LULC map, the barren land and settlement 100–500 m, 500–1000 m, and > 1000 m. In case of prox-
areas were observed as highly susceptible having posi- imity to river map, class 150 m showed greater value of
tive weight values distinctly greater than water bodies, weight (1.39) as compared to other classes. The proximity
dense, and sparse vegetation area with− 0.01,− 1.04, to road class 50 m has been observed as highly suscep-
and− 0.23, respectively. The proximity to lineament class tible to landslides with greater weight (2.07) followed
of 0–100 m has been observed as more landslide-prone by class 100 m as compared to the remaining classes. In
having weight 2.53 as compared to the other classes viz. case of relief map, the 2596–2925 m class has greater

Fig. 6  Percentage of land- (B)


(A)
slide susceptibility class and 50
landslide in LSM based on FR
method a pie chart showing 5%
40
Percentage (%)

landslide susceptibility area 14%


in LSM b bar chart showing 30% 30
landslide percentage in each
20
susceptible zone
19%
10

0
32% Very Low Medium High Very
Low High

Very low hazard Low hazard Susceptibility class area


Medium hazard High hazard Landslide area
Very high hazard Linear (Landslide area)

13
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59 Page 13 of 17  59

Fig. 7  Percentage of land- (B)


(A)
slide susceptibility class and 35
landslide in LSM based on WoE 5% 30
method a pie chart showing 14%

Percentage (%)
25
landslide susceptibility area 30%
in LSM b bar chart showing 20
landslide percentage in each 15
susceptible zone 10
20%
5
0
Very Low Medium High Very
31%
Low High

Very low hazard Low hazard Area


Medium hazard High hazard Landslide area
Very high hazard Linear (Landslide area)

landslide prediction weight as compared to higher class selected area, while high and very high susceptibility zone
viz. 2925–3214  m, 3214–3484  m, 3484–3728  m, and lies in 13.50 and 5.06% of the area. An increasing trend of
3728–4167 m. The calculated weights have been further landslide percentage has been observed from very low to
summed up in Arc GIS based on Eq. 4 using raster calcu- very high hazard zone as shown in Fig. 6b. The percentage
lator tool to delineate landslide susceptibility index map of landslides in case of very low susceptibility class is only
(Fig. 5). The landslide susceptibility index (LSI) calcu- 3%, while it has been evaluated as 43% in very high suscepti-
lated from final LSM ranged between − 14.93 and 8.43. bility class. The area covered by medium susceptibility zone
is 18.96%. The percentages of the landslide susceptibility
classes and landslides have also been shown in a similar way
Discussion and validation in previous studies by numerous researchers [9, 39].
The area occupied by landslide susceptibility classes
The landslide susceptibility maps were classified into very in LSM based on WoE method has been determined and
low, low, moderate, high, and very high susceptible class for shown in the form of Pie chart in Fig. 7a, showing high
both FR and WOE techniques. The percentages of the land- and very high susceptibility zone covered 13.53 and 5.34%
slide in each susceptibility class have been determined using of the study area, whereas very low and low susceptibility
Combine tool of spatial analyst in ArcGIS. The percentages zones occupied 30.18 and 31.22% of the area. The medium
of the landslide susceptibility zones for LSM based on the susceptibility class covered 19.7% of the LSM. The percent-
FR method are shown in Fig. 6a, showing that very low and age of the landslides in each susceptibility class has shown
low susceptibility zone occupied 30.24% and 32.22% of the increasing order from very low to very high susceptibility
class as shown in Fig. 7b.
The basic purpose of validation is to ascertain the effi-
100 ciency of the landslide susceptible models. The numer-
Cummulative percentage of landslide occurrence (%)

90 ous researchers all around the world used one or the other
80 approach to ascertain the effectiveness of the LSM. But in
70
general, the validation of landslide susceptible map must sat-
WoE = 76.86% FR = 86.57% isfy the two basic assumptions (a) Majority of the landslide
60
occurrences must be in the high susceptible zone; (b) High
50
susceptible zone should cover a small area [73, 74]. Further,
40 the analyzed landslide susceptibility map must be in good
30 agreement with the validation methods [22, 69, 75]. The
20 landslide susceptible map generated in GIS has been veri-
10
fied in the field. The validation is done at a different loca-
tion in case of high and very high susceptibility class. The
0
0 20 40 60 80 100 most prominent signatures of instability such as slides/slope
Cummulative area percentage (%) failures represent the high and very high susceptibility class.
The success rate curve explains how well the model and
Fig. 8  Success rate curve based on FR and WoE model the selected factors perform with respect to landslide [60]

13
Author's personal copy
59   Page 14 of 17 Innovative Infrastructure Solutions (2019) 4:59

which can be obtained by plotting the observed landslide density values in each susceptibility zone of these LSM
versus landslide susceptibility index (LSI). In the success maps have been compared map wise. An expected mismatch
rate curve, the index value of all grids in the LSI map is error has been observed revealing that the very low sus-
sorted in descending order and divided into 100 equal ceptibility zone in case of FR and WoE model-based LSM
classes. Then, the number of landslide grid in each class is maps covered 34.98 and 34.92 km2 area with 5.14549E − 05
calculated and converted into the cumulative percentage of and 0.00015 landslides density values, respectively, whereas
landslide occurrence which has been plotted against cumu- very high susceptibility zones in case of the LSM maps cov-
lative area percentage. The success rate curve in case of ered 5.8572 and 6.1839 km2 area with 0.00399 and 0.00276
FR model has been evaluated as 86.57% which is distinctly value of landslide densities, respectively.
greater than the WoE model with 76.86% success rate as An aerial mismatch between the landslide occupied area
shown in Fig. 8. and landslide susceptibility class area for FR and WoE-based
Another method of landslide density has been adopted LSM maps is analyzed from Table 2 because of differently
to check the accuracy of the LSM. The ratio of occurred adopted methodologies. The existing landslide conditions
landslides in a region to the spatially occupied areas by land- in the study area are reflected in the landslide susceptibility
slide susceptibility zone is known as landslide density which zonation maps indicating high and very high hazard zones.
depends upon the occurrence of landslide in each class of
a landslide causative factor [76]. For assessing the quality
of the landslide susceptibility map, the landslide density Conclusion
method is the suitable method [76]. Field-based informa-
tion of the past landslides is useful for verifying the LSM. The landslide susceptibility assessment is based on the role
A more accurate landslide hazard zonation map should have of the various causative factors as an important parameter
higher value of landslide density in high hazard zone and for initiating slope instability. The final LSM depends upon
lower in case of low hazard zone [77]. the different caustaive factors which can be selected accord-
The landslide dataset has been tabulated with LSM maps ing to an expert’s opinion. During preliminary survey, the
to evaluate the density of landslides in each hazard zone by identification and documentation of landslides has been car-
dividing landslide area with the susceptible class area. The ried out through extensive field visits for developing land-
outcome is shown in Table 2. It can be clearly seen from the slide inventory to further utilize it GIS modeling to delineate
table that the landslide density values in case of very high LSM. Thereafter, nine landslide causative intrinsic factors,
susceptible zone are 0.00399 and 0.00276 for FR and WoE i.e., aspect, curvature, relief, slope, lithology, proximity to
models, respectively, which are distinctly greater than the the river, proximity to the road, proximity to lineament, and
corresponding values for other zones. A regular decrease in LULC have been considered for LSM based on FR and WoE
landslide has also been observed from very high to very low modeling. The final LSZ map has been classified into five
susceptible zone with significant difference in landslide den- susceptibility classes: Very low, low, moderate, high, and
sity value. Hence, the prediction of the landslide susceptibil- very high based on natural break/ Jenks classification. The
ity zonation maps has been found in good agreement with success rates of both FR and WoE model-based LHZ maps
the currently existing field conditions and can be considered have been determined as 86.57and 76.86%, showing that
to be as per the concept given by [76]. The regions which the LSZ map based on FR model is more accurate. The out-
are depicted to have high susceptibility zone also indicate comes of landslide density method revealed that the land-
the potential of upcoming landslide events. The landslide slide density value decreased from very high to very low

Table 2  Landslide density in predicted landslide susceptible zones in FR and WoE-based LSM maps
Susceptible zones Frequency ratio model Weight of evidence model
2
Area ­(km ) Landslide area Landslide density Area ­(km2) Landslide area Landslide density
­(km2) ­(km2)

Very low 34.9821 0.0018 5.14549E−05 34.92 0.0054 0.00015


Low 37.2699 0.0045 0.00012 36.1188 0.0072 0.00019
Medium 21.9366 0.0063 0.00033 22.7898 0.0108 0.00047
High 15.6267 0.018 0.00112 15.66 0.0135 0.00086
Very high 5.8572 0.0234 0.00399 6.1839 0.0171 0.00276

13
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59 Page 15 of 17  59

susceptibility zone in case of FR and WoE-based LSM. As 14. Guzzetti F, Reichenbach P, Cardinali M, Galli M, Ardizzone F
the NH-1 is a strategically very important network of the (2005) Probabilistic landslide hazard assessment at the basin
scale. Geomorphology 72(1):272–299
country, the landslide susceptibility map of the area will 15. Carrara A, Cardinali M, Detti R, Guzzetti F, Pasqui V, Reichen-
play an important role during planning and execution of any bach P (1991) GIS techniques and statistical models in evaluating
development activities in the area and it will also help to landslide hazard. Earth Surf Proc Land 16(5):427–445
mitigate the failed slopes in cost-effective manners. 16. Brenning A (2005) Spatial prediction models for landslide haz-
ards: review, comparison and evaluation. Nat Hazard Earth Sys
5(6):853–862
Acknowledgements  One of the authors, Gulzar Hussain would like 17. Van Westen CJ, Van Asch TW, Soeters R (2006) Landslide hazard
to thank for the financial support received from the CSIR UGC fel- and risk zonation—why is it still so difficult? B Eng Geol Environ
lowship grant. 65(2):167–184
18. Chang KT, Merghadi A, Yunus AP, Pham BT, Dou J (2019)
Evaluating scale effects of topographic variables in landslide sus-
ceptibility models using GIS-based machine learning techniques.
Scientific Reports 9, Article number: 12296
References 19. Bui DT, Moayedi H, Kalantar B, Osouli A, Gor M, Pradhan B,
Nguyen H (2019) Rashid ASA (2019) Harris hawks optimization:
1. Aleotti P, Chowdhury R (1999) Landslide hazard assessment: a novel swarm intelligence technique for spatial assessment of
summary review and new perspectives. B Eng Geol Environ landslide susceptibility. Sensors 19:3590. https:​ //doi.org/10.3390/
58(1):21–44 s1916​3590
2. Dai FC, Lee CF (2002) Landslide characteristics and slope 20. Robiati C, Eyre M, Vanneschi C, Francioni M, Venn A, Coggan J
instability modeling using GIS, Lantau Island, Hong Kong. (2019) Application of remote sensing data for evaluation of rock-
Geomorphology 42(3):213–228 fall potential within a quarry slope. Int J Geo Inf 8:367. https​://
3. GSI (2005) A review of Bureau of Indian Standard guidelines doi.org/10.3390/ijgi8​09036​7
for preparation of landslide hazard zonation maps in mountain- 21. Milevski I, Dragicevic S (2019) Landslide susceptibility zonation
ous terrains and suggested guidelines (Macrozonation) of the republic of Macedonia with analytical hierarchy process
4. Varnes DJ (1978) Slope movement types and processes. Spec (AHP) approach. Sect Nat Math Biot Sci MASA. https​://doi.
Rep 176:11–33 org/10.13140​/RG.2.2.19484​.33923​
5. Cruden DM, Varnes DJ (1996) Landslide types and processes. 22. Van Westen CJ, Rengers N, Soeters R (2003) Use of geomorpho-
In: Turner AK, Schuster RL (eds) Landslides investigation and logical information in indirect landslide susceptibility assessment.
mitigation, special report, vol 247. National Research Council, Nat Hazards 30(3):399–419
Transportation Research Board, Washington, pp 36–75 23. Kumar A, Sharma RK, Bansal VK (2018) Landslide hazard zona-
6. Naranjo JL, Westen CJV, Soeters R (1994) Evaluating the use of tion using analytical hierarchy process along national highway-3
training areas in bivariate statistical landslide hazard analysis-a in mid himalayas of Himachal Pradesh, India. Environ Earth Sci
case study in Colombia. ITC J 3:292–300 77:719. https​://doi.org/10.1007/s1266​5-018-7896-2
7. Das I, Sahoo S, Van Westen C, Stein A, Hack R (2010) Land- 24. Lee S, Pradhan B (2006) Probabilistic landslide hazards and
slide susceptibility assessment using logistic regression and risk mapping on Penang Island, Malaysia. J Earth Syst Sci
its comparison with a rock mass classification system, along a 115(6):661–672
road section in the northern Himalayas (India). Geomorphology 25. Sarkar S, Kanungo DP (2004) An integrated approach for land-
114(4):627–637 slide susceptibility mapping using remote sensing and GIS. Pho-
8. Singh PK, Kainthola A, Singh TN (2015) Rock mass assessment togramm Eng Rem S 70(5):617–625
along the right bank of river sutlej, Luhri, Himachal Pradesh, 26. Lee S, Sambath T (2006) Landslide susceptibility mapping in the
India. Geo Nat Haz R 6(3):212–223 Damrei Romel area, Cambodia using frequency ratio and logistic
9. Singh K, Kumar V (2018) Hazard assessment of landslide disaster regression models. Environ Geo 50(6):847–855
using information value method and analytical hierarchy process 27. Vijith H, Madhu G (2008) Estimating potential landslide sites
in highly tectonic chamba region in bosom of Himalaya. J Mt Sci of an upland sub-watershed in Western Ghat’s of Kerala (India)
15(4):808–824. https​://doi.org/10.1007/s1162​9-017-4634-2 through frequency ratio and GIS. Environ Geol 55(7):1397–1405
10. Siddle HJ, Jones DB, Payne HR (1991) Development of a meth- 28. Bai SB, Wang J, Lü GN, Zhou PG, Hou SS, Xu SN (2010) GIS-
odology for landslip potential mapping in the Rhondda Valley. based logistic regression for landslide susceptibility mapping of
In: Slope stability engineering developments and applications: the Zhongxian segment in the Three Gorges area, China. Geomor-
proceedings of the international conference on slope stability phology 115(1):23–31
organized by the Institution of Civil Engineers and held on the 29. Erener A, Düzgün HSB (2010) Improvement of statistical land-
Isle of Wight on 15–18 April 1991 (pp. 137-142). Thomas Telford slide susceptibility mapping by using spatial and global regression
Publishing methods in the case of more and romsdal (Norway). Landslides
11. Wu W, Sidle RC (1995) A distributed slope stability model for 7(1):55–68
steep forested basins. Water Resour Res 31(8):2097–2110 30. Poudyal CP, Chang C, Oh HJ, Lee S (2010) Landslide suscepti-
12. Dai FC, Lee CF, Li J, Xu ZW (2001) Assessment of landslide bility maps comparing frequency ratio and artificial neural net-
susceptibility on the natural terrain of Lantau Island, Hong Kong. works: a case study from the Nepal Himalaya. Environ Earth Sci
Environ Geol 40(3):381–391 61(5):1049–1064
13. Ayalew L, Yamagishi H (2005) The application of GIS-based 31. Yilmaz I (2010) The effect of the sampling strategies on the land-
logistic regression for landslide susceptibility mapping in the slide susceptibility mapping by conditional probability and artifi-
Kakuda-Yahiko mountains central Japan. Geomorphology cial neural networks. Environ Earth Sci 60(3):505–519
65(1):15–31 32. Intarawichian N, Dasananda S (2011) Frequency ratio model based
landslide susceptibility mapping in lower Mae Chaem watershed,
Northern Thailand. Environ Earth Sci 64(8):2271–2285

13
Author's personal copy
59   Page 16 of 17 Innovative Infrastructure Solutions (2019) 4:59

33. Mezughi TH, Akhir JM, Rafek AG, Abdullah I (2011) Landslide tehri reservoir, Uttarakhand, India. J Remote Sens GIS 8:1. https​
susceptibility mapping using the statistical index method and fac- ://doi.org/10.4172/2469-4134.10002​54
tor effect analysis along the E–W highway (Gerik-Jeli), Malaysia. 49. Bookhagen B, Thiede RC, Strecker MR (2005) Late quaternary
Australian J Basic Appl Sci 5(6):847–857 intensified monsoon phases control landscape evolution in the
34. Khai-Wern KL, Tien TL, Lateh H (2011, May) Landslide haz- northwest Himalaya. Geology 33(2):149–152
ard mapping of Penang is land using probabilistic methods and 50. Chevuturi A, Dimri AP, Thayyen RJ (2018) Climate change over
logistic regression. In: 2011 IEEE international conference on Leh (Ladakh), India. Theor Appl Climatol 131(1–2):531–545
imaging systems and techniques (IST), (pp. 273–278). IEEE 51. Guzzetti F, Carrara A, Cardinali M, Reichenbach P (1999)
35. Lepore C, Kamal SA, Shanahan P, Bras RL (2012) Rainfall- Landslide hazard evaluation: a review of current techniques
induced landslide susceptibility zonation of Puerto Rico. Envi- and their application in a multi-scale study, central Italy. Geo-
ron Earth Sci 66(6):1667–1681 morphology 31(1):181–216
36. Kannan M, Saranathan E, Anabalagan R (2013) Landslide vul- 52. Ercanoglu M, Gokceoglu C (2004) Use of fuzzy relations to
nerability mapping using frequency ratio model: a geospatial produce landslide susceptibility map of a landslide prone area
approach in Bodi-Bodimettu Ghat section, Theni district, Tamil (West Black Sea Region, Turkey). Eng Geol 75(3):229–250
Nadu, India. Arab J Geosci 6(8):2901–2913 53. Srikantia SV, Razdan ML (1980) Geology of part of central
37. Sujatha ER, Rajamanickam V, Kumaravel P, Saranathan E Ladakh Himalaya with particular reference to Indus tectonic
(2013) Landslide susceptibility analysis using probabilistic like- zone. J Geol Soc India 21(11):523–545
lihood ratio model—a geospatial-based study. Arab J Geosci 54. Yalcin A, Reis S, Aydinoglu AC, Yomralioglu T (2011) A GIS-
6(2):429–440 based comparative study of frequency ratio, analytical hierarchy
38. Ramesh V, Anbazhagan S (2015) Landslide susceptibility map- process, bivariate statistics and logistics regression methods for
ping along Kolli hills Ghat road section (India) using frequency landslide susceptibility mapping in Trabzon, NE Turkey. CAT-
ratio, relative effect and fuzzy logic models. Environ Earth Sci ENA 85(3):274–287
73(12):8009–8021 55. Guzzetti F, Mondini AC, Cardinali M, Fiorucci F, Santangelo
39. Kumar A, Sharma RK, Bansal VK (2019) GIS-based compara- M, Chang KT (2012) Landslide inventory maps: new tools for
tive study of information value and frequency ratio method an old problem. Earth Sci Rev 112(1):42–66
for landslide hazard zonation in a part of mid-Himalaya in 56. Wieczorek GF (1984) Preparing a detailed landslide-inventory
Himachal Pradesh. Innov Infrastruct Solut 4:28. https​: //doi. map for hazard evaluation and reduction. Bull Assoc Eng Geol
org/10.1007/s4106​2-019-0215-2 21(3):337–342
40. Khan H, Shafique M, Khan A, Bacha MA, Shah SU, Calligaris 57. Yilmaz I (2009) Landslide susceptibility mapping using fre-
C (2019) Landslide susceptibility assessment using frequency quency ratio, logistic regression, artificial neural networks and
ratio, a case study of northern Pakistan. Egypt J Remote Sens their comparison: a case study from kat landslides (Tokat—Tur-
Space Sci. https​://doi.org/10.1016/j.ejrs.2018.03.004 key). Comput Geosci 35(6):1125–1138
41. Cerviet F, Berti M, Borgatti F, Manenti F, Corsini A (2010) 58. Wieczorek GF, Mandrone G, DeCola L (1997) The influence
Comparing predictive capability of statistical and deterministic of hillslope shape on debris-flow initiation. In: Debris-flow
methods for landslide susceptibility mapping: a case study in hazards mitigation: mechanics, prediction, and assessment (pp.
the northern Apennines (Reggio Emilia Province, Italy). Land- 21–31). ASCE
slides 7(4):433–445 59. Merodo JF, Pastor M, Mira P, Tonni L, Herreros MI, Gonzalez
42. Dahal RK, Hasegawa S, Nonomura S, Yamanaka M, Masuda E, Tamagnini R (2004) Modelling of diffuse failure mecha-
T, Nishino K (2008) GIS-based weights-of-evidence modelling nisms of catastrophic landslides. Comput Method Appl Mech
of rainfall-induced landslides in small catchments for landslide 193(27):2911–2939
susceptibility mapping‖. Environ Geol 54(2):314–332 60. Yesilnacar E, Topal T (2005) Landslide susceptibility mapping:
43. Lee S, Choi J (2004) Landslide susceptibility mapping using a comparison of logistic regression and neural networks meth-
GIS and the weight of evidence model. Int J Geogr Inf Sci ods in a medium scale study, Hendek region (Turkey). Eng Geol
18(8):789–814. https​: //doi.org/10.1080/13658​8 1041​0 0017​ 79(3):251–266
02003​ 61. Gorsevski PV, Gessler PE, Foltz RB, Elliot WJ (2006) Spa-
44. Regmi NR, Giardino JR, Vitek JD (2010) Modeling susceptibil- tial prediction of landslide hazard using logistic regression and
ity to landslides using the weight of evidence approach: western ROC analysis. Trans GIS 10(3):395–415
Colorado, USA. Geomorphology 115(1–2):172–187. https​://doi. 62. Sidle RC and Ochiai H (2006) Landslides: processes, prediction,
org/10.1016/j.geomo​rph.2009.10.002 and land use (Vol. 18). American Geophysical Union
45. Thiery YJP, Malet S, Sterlacchini A, Puissant Maquaire O (2007) 63. Singh Y, Bhat GM (2010) Role of basin morphometric param-
Landslide susceptibility assessment by bivariate methods at large eters in landslides along the national highway-1A between
scales: application to complex mountainous environment. Geo- udhampur and batote, Jammu and Kashmir, India: a case study.
morphology 92(1–2):38–59. https​://doi.org/10.1016/j.geomo​ Himal Geol 31(1):43–50
rph.2007.02.020 64. Jenks GF (1967) The data model concept in statistical mapping.
46. Pamela Pamela, Sadisun Imam, Yukni Arifianti (2018) Weights Int Year Book Cartogr 7:186–190
of evidence method for landslide susceptibility mapping in Tak- 65. Gokceoglu C, Aksoy H (1996) Landslide susceptibility mapping
engon, central aceh, Indonesia. IOP Conf Ser Earth Environ Sci. of the slopes in the residual soils of the Mengen region (Turkey)
https​://doi.org/10.1088/1755-1315/118/1/01203​7 by deterministic stability analyses and image processing tech-
47. Sifa SF, Mahmud T, Tarin MA, Haque DME (2019) Event-based niques. Eng Geol 44(1–4):147–161
landslide susceptibility mapping using weights of evidence 66. Singh Y, Bhat GM (2011) Landslide investigations: morpho-
(WoE) and modified frequency ratio (MFR) model: a case study metric and geotechnical approach-a case study from Northwest
of Rangamati district in Bangladesh. Geol Ecol Landsc. https​:// Himalaya, India. LAP LAMBERT Academic Publishing GmbH
doi.org/10.1080/24749​508.2019.16192​22 and Co.KG, Saarbrucken, pp 37–51
48. Ghildiyal B, Champati ray PK, Bisht MPS, Rawat GS (2019) 67. Singh Y, Bhat GM, Sharma V, Pandita SK, Thakur KK (2012)
Landslide susceptibility zonation using bivariate models, around Reservoir induced landslide at Assar, Jammu and Kashmir: a
case study. J Geol Soc India 80(3):435–439

13
Author's personal copy
Innovative Infrastructure Solutions (2019) 4:59 Page 17 of 17  59

68. Anbalagan R (1992) Landslide hazard evaluation and zonation 74. Duman TY, Can T, Gokceoglu C, Nefeslioglu HA, Sonmez H
mapping in mountainous terrain. Eng Geol 32(4):269–277 (2006) Application of logistic regression for landslide suscepti-
69. Chung CJF, Fabbri AG (1999) Probabilistic prediction mod- bility zoning of Cekmece Area, Istanbul, Turkey. Environ Geol
els for landslide hazard mapping. Photogrammeng rem s 51(2):241–256
65(12):1389–1399 75. Kayastha P, Dhital MR, De Smedt F (2013) Application of the
70. Lee S, Pradhan B (2007) Landslide hazard mapping at Selangor, analytical hierarchy process (AHP) for landslide susceptibility
Malaysia using frequency ratio and logistic regression models. mapping: a case study from the Tinau watershed, west Nepal.
Landslides 4(1):33–41 Comput Geosci 52:398–408. https​://doi.org/10.1016/j.cageo​
71. Bonham-Carter GF (1994) Geographic Information Systems for .2012.11.003
Geoscientists; modelling with GIS, Comp, Pergamon, ISBN: 76. Sarkar S, Kanungo DP (2004) An integrated approach for land-
9781483144948 slide susceptibility mapping using remote sensing and GIS. Pho-
72. Oh HJ, Lee S, Chotikasathien W, Kim CH, Kwon JH (2009) Pre- togramm Eng Rem Sens 70(5):617–625
dictive landslide susceptibility mapping using spatial information 77. Gupta RP, Kanungo DP, Arora MK, Sarkar S (2008) Approaches
in the Pechabun area of Thailand. Environ Geol 57(3):641 for comparative evaluation of raster GIS-based landslide suscep-
73. Can T, Nefeslioglu HA, Gokceoglu C, Sonmez H, Duman TY tibility zonation map. Int J Appl Earth Obs Geoinf 10:330–341
(2005) Susceptibility assessments of shallow earthflows triggered
by heavy rainfall at three catchments by logistic regression analy-
ses. Geomorphology 72(1):250–271

13

You might also like