Landslide Hazard Assessment Using Probabilistic and Statistical Approaches: A Case Study of Chamba Region, Himachal Pradesh, India

.


Introduction
Landslide is a significant environmental risk, instigating destruction to life and infrastructure.Several authors categorize landslides such as: fall, topple and flow, spread, slide etc. in terms of rock, debris, and earth under the influence of the gravitational forces in the downward and outward direction of the slope (Varnes and IAEG, 1984; Crozier 1986; Couture 2011).The strength of the slope forming materials and slope angle controls the stability of the slope.Past landslide records, intensive field work and professional experience play important role in the selection of the essential factors.The Indian states of Jammu and Kashmir, Himachal Pradesh, Uttarakhand, and Sikkim often get disturbed due to landslides during every monsoon.Districts like Chamba, Kangra, Mandi, Kullu, Bilaspur, and Shimla are highly affected by landslides in the state of Himachal Pradesh.Hence, a vulnerable stretch from Lahru to Chamba along NH-154A via Banikhet was selected for LHZ assessment in the present study.
Landslides associated with the river beds are affected due to the undercutting of the toe of slopes.The economy of Himachal Pradesh and Government of India is adversely affected by these slope failures and in some cases many lives were also lost.Mitigation and remedies for the slope failures in study area has been useful after identification of landslide hazard zones.
The mountainous terrain of district Chamba in Himachal Pradesh is suffering from landslide disaster mainly in the region where slope gradient is greater than 40° which is harming lives and million dollars property across the region.
An earthquake triggered landslides caused death of around 500 people in Chamba in year 1905.The inherent and extrinsic landslide causative factors like geomorphological, geological, hydrology, soil, landuse land cover (LULC), rainfall, seismic etc. play important role for causing landslide movement (Cruden and Varnes 1996; Courture 2011)from which the geology, soil, and LULC are the major factors contributing in landslides (Varnes 1984; Anbalagan 1992; Hutchinson1995; Achour et al.2017).The constructional activities of infrastructural mega projects are also disturbing hilly terrains in this region.Hence, the landslide hazard assessment is essential in the region for future development and construction.

Darla volcanic rocks
Basaltic and andesitic flows

Data Source
Selection of the rightful elements with their assistance in the landslide occurrences is necessary ( The rainfall record from January 2000 to February 2019 was collected from Indian meteorological center Shimla. Other maps were generated from data provided by the concerned governmental and non-governmental agencies (Table 2).Figure 3 shows all the landslide causative factor maps which were correlated with the mass movements of the slopes.Weight of each class of landslide causative factor is analyzed using AHP approach through the comparison of each pair in the form of pairwise comparison matrix.Weights were assigned in the scale of 1-9 (equal importance 1, moderate importance 3, strong importance 5, very strong importance 7 and extremely importance as 9) on the basis of importance of each factor in triggering the failure of the slopes.Results of inverse comparison are reciprocal (1/2 to 1/9) for the given weight to mandatory triggering element.Consistency of each matrix was validated through the consistency ratio (Satty 1990(Satty , 1994) ) which is analyzed for each factor from the weights of each element through equation 1 and 2. 1 Where, order of matrix as 'n' and 'λmax' denoted as matrix principle eigen value and Random index values were used from table 3. Weight values assigned for the different factors on their importance in the form of matrix are shown in Table 3.

Thematic Layers (a) Landslide Inventory
Interpretation of satellite images, past landslide records from different agencies and field surveys play significant role in the development of landslide inventory (Kayastha et al. 2013).In the present study, inventory of 184 landslides was developed by collecting data from the past records of border road organization (BRO), public works department (PWD), National Highway Authority of India (NHAI) and LISS IV satellite images and Google earth images which were validated through field visits and updated to present day status (Figure 4a).The cause of maximum numbers of landslides (debris and rock falls) are correlated with thrusts and lineaments.Rock fall near Chowari town and 20 km before Chamba town weretriggered by the seismic shocks and rainfall (Fig 1b and f), debris flow at 500m from Lahru town and Banikhet landslides took place due to heavy rainfall which is triggering factor for 75% of the total landslides in the area.

(b) Lithology
Lithology holds an important role in the stability of the mass movements, in minimizing the erosions and in weathering of rocks as well (Anabalgan 1992).Upper Shiwalik, Pindru, Mandi, Khokan, Khalel (Lilang group), Ghar, Dharmsala and Chamba formation, Dalhousie granite with alluvium rock bodies and their sequence of succession has been presented in Figure 4

(d) Soil
Five types of soils were found in the study area like loamy skeletal soil with loamy surface, rocky outcrops, coarse loamy, loamy and sandy skeletal soils (Figure 4d).Coarse loamy soil covered a minimum area of 3.49% while 35.47% maximum area was found to be covered with loamy skeletal soil.Landslides were found to be more prone in sandy skeletal soil.

(h) Geomorphology
The geomorphology map of the study area was delineated at the scale of 1:50,000 based on the Landsat, Cartosat 1D and Google earth images.Structural, Denudational, low and moderate dissected hills were identified and validated through filed visits.River and flood plains of river covers only 4.56% of the area while 95.44% of area is covered by the structural and denudational hills as shown in Figure 4h.

(i) Slope Aspect
The aspect of a slope is known as direction of the slope facet with respect to north.Aspect of slopes depends upon the dryness of air, soil moisture and solar heat effect (Yalcin 2008).The slope aspect map was classified into eight classes viz.north, north east, east, south east, south, south west, west and north west as mentioned in Figure 4i.

(j) Rainfall
Rainfall is an important causative factor in the occurrences of slope failures.Rainfall data of different metrological stations in the region was collected and analyzed from January 2000 to February 2019.Lowest rainfall record is 1437mm in year 2002 and maximum rainfall 1691mm in year 2017 was recorded in region.Kriging tool of GIS is used for development of rainfall map.Rainfall map of the study area was reclassified into four class'sviz.1437-1547mm, 1548-1627mm, 1628-1691mm, and 1692-1752mm as shown in Figure 4(j).

(k) Relative Relief
Slopes angles and direction of slopes depend upon altitude rangeof the area.Altitude ofLahru town is 656m and Kalatopis 2757m from mean sea level.Relative relief map was developed from the contour map which digitized from SOI topographic sheets at the scale of 1:50,000.Reclassify tool of GIS has been used for reclassification of relative relief map into four classes (656-1190m, 1191-1581m, 1582-2041m, 2042-2757m) on the basis of slope failure occurrences, type and size in relative reliefs (Figure 4k).

(a) Landslide hazard zonation using analytical hierarchy process (AHP) method
Landslide hazard zonation maps are prepared with impact from different causative factor maps on the scale of 1:10,000 in GIS environment.Digital elevation model of study area is obtained after the conversion of vector data Conditions of landslide hazards is also affected by density of drainage networks and lineaments.In the study area very low and low classes of lineament (0.080, 0.77) and drainage density (0.067, 0.076) have less impact on landslide incidences respectively.High density of lineament (0.618) and drainage networks (0.608) have high impact followed by moderate class of both factors (0.226 and 0.216).In the case of relative relief, high altitude (2042-2757m) of terrain is highly effected by landslide resulted high AHP value of 0.606 followed by less (1582-2041m) altitude of terrains with less number of landslides (0.229) while 656-1190m of altitude have average of least number of landslides with AHP value 0.054.Geomorphology of the terrain classified into six categories in which low dissected structural hills (0.326) are most effected by slope failures due to exposure for weathering processes followed by moderately dissected denudation hills (0.        Successive rate curve method for validation of both AHP and Inf.V methods ; Suzen and Doyuran, 2004; Ercanoglu and Gokceoglu, 2004; Yesilnacar and Topal, 2005; Kanungo et al., 2006; García-Rodriguez et al., 2008; Nefeslioglu et al., 2008; Nandi and Shakoor, 2009; Dongyeob Kim et al., 2010; Pradhan B., 2012).Distribution of landslide events and landslide controlling parameters are correlated with historical events by using statistical methodology (Refice and Capolongo, 2002; Zhou et al., 2003).The influence of rainfall has also been analyzed in terms of thresholds for triggering landslides by various researchers (Iverson 2000; Crosta and Frattini 2003; Brooks et al. 2004; Montrasio et al. 2012; Salciarini et al.2012; Ran et al. 2012; Marchi et al. 2002; Floris et al.2004; Giannecchini2005; Kanungo and Sharma 2014; and Chung et al. 2016).

Figure 1 :
Figure 1: Photographs of critical slopes taken during March 2019 field visit along SH28 and NH -154A (a) Earth debris failure 500m from Lahru at SH28 (b) Transitional and Rock debris landslide near Chowari at SH28, (c) Debris flow landslide neat Jot at SH28 (d) Rock fall and Earth flow landslide 250m before Ghatasani bridge at NH145-A (e) Earth Debris flow landslide near Banikhet at NH-154A (f) Rock fall landslide near Chamba at NH154A.Study area Study area covers 27355.47m 2 area enveloped between 80.84 Km NH-154A from Lahru to Chamba via Banikhet (76º10'E, 32º 35'N) and 54.90 Km along the state highway 28 from Lahru to Chamba via Jot (75º 51' E, 32 º 26' N).The area has a dendritic drainage pattern and can be easily identified on using Survey of India 1986 topographic sheets (43P/14, 43P/15, 52D/2 and 52D/3) on the scale of 1:50,000.The study area has high relief with altitude ranging from 656 feet near Lahru to 2757 feet in Dalhousie.Maximum temperature in summer is 30ºC whereas minimum is 14ºC while in winter season temperature varies between 2ºC to 11ºC.Rainfall of 186.42mm was recorded during the year 2018 in this region (Indian meteorological department).Climate conditions and physiography are the root cause of slope failures in the study area.On 2 nd August 2018 one bus was washed out and road was blocked for the almost one day reported in several local newspapers and media.During the monsoon period of 2017 and 2018 landslide measuring 434m length and 136m wide has taken place below the Bindgi village, Bhattiyat tehsil near hair pin bend (75°57'11.70"E,32°26'19.21"Nto 75°57'16.49"E,32°26'20.11"Nand 75°57'12.99"E,32°26'17.81"Nto 75°57'16.83"E,32°26'19.33"N).Mountain slopes are affected by human interferences for development and unplanned construction although rainfall and seismic activities also play vital role for triggering landslides.Transitional landslide was found along NH-154A at a distance of about 250m Ghatasani Bridge damaging around 800 m road stretch as reported by Hindustan, Times of India and as per records of H.P Public Works Department.Debris flow caused damage to NH nearby 500m from Lahru at July 14, 2017 as reported by the Tribune and Times of India.Road construction initiated huge translational slide due to movement of thrust near Chowari has also been reported (Singh and Thakur 1989, Keshar Singh 2011).

Figure 2 :
Figure 2: Study area Lahru to Chamba, Chamba District, and Himachal Pradesh, India Geological Setting Regional geological setup in study area consists of a large syncline called Chamba syncline, having metasedimentary rocks with 8-10km thickness belonging to Cambrian to Triassic era.Subordinate phyllites with laminated cross-bedded quartzite are main subsistence of the Chamba formations.Phyllites, slates and limestones are foundation of the Shali formation which is overlain by the alluvium.PindruFormation rock bodies are

Figure 3 :
Figure 3: Flow chart for landslide hazard zonation mapping (a) Analytical Hierarchy Process (AHP) Analytical Hierarchy Process technique was utilized by various researchers for landslide hazard zonation.The AHP technique is useful in evaluation of different parameters role in slope failure (Varnes 1990 and 1980, Satty 2005, Feizizadeh and Blaschke 2012).It is based on three principles: decomposition, comparative judgment and synthesis of priorities (Pandey et al. 2016, Singh and Sarda2018).
Reciprocal weight values were given for inverse comparison of factors based on dominant causative factor.CR value less than 0.1 indicates reasonable consistency level of weight values (Satty and Vergas 2000).Finally to analyze landslide hazard index (LHI) values, all weights were combined by using equation 3. 3 Where, LHI = landslide hazard index values; Wik=weight of subclass i in factor k; Wi = weight of the factor k; n = total number of factors.(b) Information value model Based on assumption factors responsible for slope failures in past was the same that will cause landslides in future.Distributions of landslides based on their location, size and type were compared with each causative factor for assigning the weight values.Assignments of weight values were depend upon the density of landslides in each class of the factor (Lin and Tung 2003).Information value model (IVM), weight of evidence (WOE), frequency ratio (FR) and weighted overlay etc. statistical methods were used for the analysis of weights values.Van Westen 1993 modified the information value model proposed by Yin and Yan 1988.Researcher has used information value model for landslide hazard zonation mapping in such study.Rainfall, slope gradient, slope aspect, geology, drainage density, lineament density, geomorphology, relative relief, land use/cover soil and curvature factors plays very important role in failures of slopes.This method aims to find the probability of a landslide event based on the comprehensive information available of the significant factors by using equation 4. 4 Where, factor of class information value l(Lxi), number of landslide pixels in each class Npix (Si), total number of pixels in each class Npix(Ni), sum of landslide pixels in the study area ∑Npix(Si), sum of total pixels in the study area ∑Npix(Ni).Equation 5 analyzes the landslide hazard index values through summing of all weight values.5 Where LHI indicates the landslide hazard index values, n denotes the total number of factors.LSI<0 indicates the less than likelihood average, LSI=0 indicates the landslide likelihood equal to the average, LSI>0 indicates the average of landslides.
(b) and Table 2. Lithology map for the study was developed from the articles of Keshar Singh 2011 and 1993 Keshar Singh and V. C. Thakur 1989.(c) Land use and land cover Contribution of land use/ cover in slope failures has been reported well in literature by several researchers(Lee et al. 2013).Landslide prone areas are the evidences of land use changes like road expansions, new construction activities (homes, bridges and reservoirs), deforestation and rapid expansion in agriculture fields (Jaiswal et al. 2010).Comparison of past data (SOI topographic sheets(43P14, 43P15, 52D2 and 52D3)) with the present data (Google earth, LISS IV satellite imageries) has been employed in the preparation of LULC map through digitization and change detection tool as presented in Figure 4(c).Forest, sparse vegetation and agriculture classes of the region have been found to cover 79.63% while water bodies and snow occupied only 1.08% area.
(e) Drainage DensityDrainage density plays an important role in the weathering and soil erosion on the slopes(Demir et al. 2013).Orientation of lineaments was followed by the stream networks in region.Density of stream networks depends upon the ground water conditions(Kavzoglu et al. 2014).Stream networks were extracted from satellite imagery of Cartosat -1D sensor by using hydrology tool of spatial analyst in ESRI (Arc GIS) software and digitization of survey of India topographic sheets at the scale of 1:50,000.Drainage density map was developed by using equation 6 and reclassified into four classes (Very low, low, moderate and high drainage density) Figure4(e).

Figure 3 :
Figure 3:Thematic data layers (a) Landslide inventory map, (b)Lithology map, (c) Land use and cover map (d) Drainage density(e) Lineament Density map (f) Lineament density map.(f)Lineament Density Role of rock cleavages, joints and fault planes are very important for controllingpore pressure (Ramakrishnan et al. 2013).Intense shearing of rocks is directly affected by the active fault planes which leads rock masses of mountains towards failures (Leir et al. 2004).Lineament density map was prepared from the various literatures and maps at the scale of 1:50,000, National remote sensing center (NRSC), Indian space research organization (ISRO).Lineament density map has been classified into three classes viz.low, moderate and high density based on equation 7 (Figure 4f).
Curvature is the intersection of random planes with surface (Ramesh and Anbazhagan 2015).Frictional and cohesion forces act differently in different types of surfaces.Inflow and outflow of water or drainage systems is controlled by these surfaces.Flat, Convex and Concave surfaces are normally found in hilly terrain which plays important role in erosion and weathering processes of slopes.Curvature map was developed from the digital elevation model (DEM) of the region (1:50,000) and classified into three classes (Flat, Convex and Concave) Figure 4(l).
Chamba and Dalhousie.There are none evidence of slope failure into Chamba and Dalhousie formation while into upper Siwalik rock formation has high number of slope failures resulted high weight value of 1.362 followed by Khalel and Mandi rock formations.Geomorphology of terrain classified into six various classes like structural hillsmoderately dissected (SHM), structural hills-low dissected, and denudational hills -moderately dissected (DHM), river, flood plains and valley fills.Weight values of low dissected structural hills are analyzed high weight value of 0.326 which followed by moderately dissected denudational (0.32) and structural hills (0.199) while river, flood plain and valley fills are zero weight value due to none presence of landslides.In the case of land use and cover, forest (-1.338) and agriculture lands (-0.028) of region have negative value while slopes which correlated with settlement areas have high weight value 0.945.Road networks (0.482), sparse vegetation (0.266), rocky terrain (0.184) and shrub (0.149) have also follow the same trend of information weight values.Among all variables rainfall is most important and essential factor which recorded by Indian metrological department from 1437mm to 1752 mm.low regions of rainfall 1437-1547mm (-0.652) and 1548-1627mm (-0.225) have less weight values while high regions of rainfall 1628-1691mm has high weight value of 0.598 which followed by 1692-1752mm (0.012).

Figure 5 : 1
Figure 5: Landslide hazard zonation maps using AHP and Information value methods: (a) Analytical hierarchy process (AHP) hazard map (b) Information value (Inf.V) hazard map Finally to develop landslide hazard zonation of the region, landslide hazard index values are analyzed from summation of all information weight values by equation 9. Landslide hazard index values were classified into very low (1.2%), low (5.31%), moderate (20.03%), high (29.26%)and very high (44.2%)by using reclassify tool of GIS.

Table 1 .
Tectonic activity of the region is the cause of frequent the seismic activities in the region.Seismic activity PindruFormation rock bodies are subsistence with interstratified schist, quartzite, staurolite, schist and conglomerate.KhokanFormation rock bodies are surrounded by Vaikrita thrust.Kalhel rock bodies of Kalhel formation (Lilang Group) is found in pockets which were overlaid by GharFormation.Upper Siwaliks rock bodies are youngest formation of rocks in the region shown in

Table 2 :
Data Type/Source for LHZ Mapping Geological Survey of India and articles of Keshar Singh 2011 and 1993 Keshar Singh and V. C. Thakur 1989 Geology, Lineament Density 4 Geomorphology Scale 1:50,000 Bhuvan, National Remote Sensing Centre (NRSC), Hyderabad, India Geomorphology 5 Soil Map Scale 1:50,000 Survey of India Soil 6 Field Data GPS Locations -Landslide Inventory 7 Rainfall Data Scale 1:50,000 Indian Metrological Centre, Shimla, H.P., India Rainfall Map

Table 4 :
Random Index values for different matrix orders (N) 31) and moderate dissected structural hills (0.199).Land covered by shrub have high AHP value (0.235) followed by road (0.183) and settlement (0.134) while other classes of LULC have less AHP values.Moisture containment play very important role in formation of slope curvatures.Flat type of slope surfaces have lesser AHP value of 0.083 other than concave (0.723) and convex (0.194) curvature type of land surfaces.Rainfall in the study area given the 0.606, 0.216, 0.091and 0.042 AHP values for 1692- area of very high, high and moderate hazard zones of landslides are prevailing in south east (Lahru, Banikhet and Dalhousie) and south west (Lahru) direction of selected examined area of Chamba district while very low and low hazard zones lies near Chamba city in north direction of the study area.Mostly very high and high hazard zones lies near lineament density high and exposed land cover for erosion and weathering processes while very low and low hazard zones are located near water bodies, densely vegetated covered and flood plains.It can be clearly observed that 71.33% of high and very high hazard zones are covered while moderate 24.40%, low 4.27% and very low 2.95% hazard zones are covered respectively (Figure4c).Density of landslides also increases while frequency of landslides has increased.Consistency index (CI) values are calculated for the weight values accuracy of all factors subclass through the MS excel sheet and random index values obtained from table 3. Consistency ratio (CR) value 0.074 is obtained for the weight value of prevailing factor for pair through pairwise comparison (equation 2).

Table 4 :
Analysis of thematic layers and subclasses causative factor

Table 5 :
Spatial relationship between each factor causing landslide and landslide using the information value model 372 , sandy skeletal soil (0.614) has evidence of maximum number of landslide which was followed by loamy skeletal soil (-0.50).Zero weight value was obtained for rocky outcrops and coarse loamy soil classes due to absence of landslides.Sandy skeletal soil has maximum exposed for weathering process due to river and gullies.Among all factors lithology terrain play very important role.Mainly igneous (Micaceous Granite, Mandi Volcanes-Basaltic and andesitic flows) sedimentary (sandstones, clay, sandstone with shale, limestone with calcareous shale and ferruginous chert, conglomerate boulder) and metamorphic (Slate, Phyllite, Carbonaceous and Quartzite, Paragneisses, Chlorite, Mica Schist, Slate and Quartzite, Interstratified Schists and Quartzite) rocks were found into various formations like Alliuvium, Upper Siwalik, Pindru, Mandi, Khokan, Khalel (Lilang group), Ghar, Dharmsala, Analysis of correlation between slope failure locations and thematic layers (slope gradient, slope aspect, lithology, 366 lineament density, geomorphology, drainage density, rainfall, soil, land use/ cover, relative relief and curvature) landside frequency very low (-0.545),density of drainage networks would also low.Tributaries and gullies of main 393 rivers are contributed in toe cutting of slopes and bank erosion.Lineaments play an important role for initiating 394 landslides.Planar and toppling such types of slope failures are managed by different types of lineaments such as 395 fault, joint etc. Initiation of landslides are commonly affected by strength and orientation of discontinuities, greater 396 information values were evaluated for high (-0.188)and very high lineament density class (0.404).In case of soil 397 map