Knowledge-driven method: a tool for landslide susceptibility zonation (LSZ)

ABSTRACT Gangtok, situated at the south-eastern part of the Sikkim state in India is the centre of the state’s administrative operations. Precambrian rocks comprising of foliated schists and phyllites are the major share of this city; slopes are therefore prone to regular landslides. There is a paradigm shift in disaster management policy of the city to emphasize disaster preparedness in order to ensure that human life and property are not harmed to the best possible. Apart from natural slope failure, now-a-days Gangtok city is more prone to frequent landslides as a result of recent road and building construction. Present research work deals with the development of Landslide susceptibility mapping (LSM) within the Gangtok Municipal Corporation (GMC) region by means of geospatial analysis. The weighted overlay method (WOM) is adopted for this purpose where, weights are assigned to various triggering factors based on expert opinion. Twelve triggering factors i.e., geology/lithology, slope morphometry, lineament density, water regime, rainfall, elevation, soil type, soil liquefaction, soil thickness, building density, relative relief, and land use/land covers (LULC) are considered for this analysis. Outcomes of LSM reveal that about 19 and 31 percent of the study area fall within very high and high susceptible zone respectively whereas, 30 and 18 percent of the city are categorized as medium to low susceptible categories respectively. The model generated LSM is further confirmed by using previous landslide events, yielding an overall accuracy of 80%.


Introduction
In the hilly terrain of India including the Himalayas and the Western Ghats, landslides have been a major and widely spread natural disaster. In India, landslides cause approximately US$ 4,500,000 economic damage and 2185 loss of life (Kaur, Gupta, Parkash, Thapa, & Mandal, 2017a). Sikkim Himalayas is a very hot bed for natural hazard especially for landslide and earthquake (Sikkim State Disaster Management Authority, 2012). Landslide hazard is the most common phenomenon in Sikkim Himalayas, owing to its critical location in an active fold thrust belt (FBT) and falls under seismic zone -IV. The land is continuously distressed due to landslides (Anbalagan, Kumar, Lakshmanan, Parida, & Neethu, 2015). Delineation of landslide susceptibility zones plays a crucial role in carrying out landslide hazard mitigation. Highlighting the landslide susceptible zones beforehand by identifying the potentially high-risk area can help in reduction of fatalities of people and animals, loss of properties, etc. (Arora, Das, & Gupta, 2004;Ercanoglu & Temiz, 2011;Guzzetti, Reichenbach, Cardinali, Galli, & Ardizzone, 2005;van Westen, Castellanos, & Kuriakose, 2008). These also serve as important information for government organization and decision-makers to carry out sustainable and appropriate urban development activities and land use planning. Generally, two sets of factors (namely naturally induced precondition factors and triggering factors induced by either natural or anthropogenic activities) determine the susceptibility to landslide hazard (Glade & Crozier, 2005;Marko, 2006).
Landslide inventory and susceptibility have great importance as they provide important information to decision-makers/planners and constitute basic information for hazard and risk analysis. According to Guzzetti, Reichenbach, Ardizzone, Cardinali, and Galli (2006), landslide susceptibility predicts the stable area in landslide-prone region and classifies the area from low to high hazard zone. Although the hazard and risk analysis are very much important, it is also true that data required for hazard and risk analysis are not usually possible to obtain and are very expensive for developing countries. Thus, in this research work, landslide susceptibility analysis is preferred for a landslide-prone area. The landslide conditioning factors like soil thickness, soil liquefaction, and building density factors are rarely used in study of LSM. Thus, one of the objectives of the research study is to find out an influence of the spatial distribution of these factors on landslide occurrence. The present study is based on a qualitative method (knowledge-driven /heuristic method) for assigning weights and ranks to the causative factors (twelve different input parameters) that can influence the occurrence of the landslide. As the susceptibility method is a modelbased approach rather than a data-driven method, so the accuracy of the model is based on input data with the addition of knowledge opinion. By observing the past incidences of landslides in the study area (Chakraborty, Basir, & Nath, 2012;Ghoshal & Sengupta, 2001;Rawat, Joshi, Rawat, & Kumar, 2011), we have chosen the study area to carry out the landslide susceptibility mapping. All they have published is the detailed geology of the landslides and hence landslide susceptibility mapping is missing. Therefore, there is a need to carry out a systematic approach to combine the quantitative method with knowledge-driven for landslide susceptibility mapping.

Study area
Gangtok is the landing placed on the ridge, running northeast to south-west of the Eastern Himalayas and falls within the lesser Himalaya zone. This region has erosional topography, surrounded and dissected by seasonal and perennial springs. Gangtok municipality, the headquarter of the East district of Sikkim, is located at an elevation of 1650 m (5410 ft) with 27.3325°N latitude to 88.6140°E longitude (Figure 1). The population of the town is approximately 10,000, belonging to dissimilar races of Nepali, Lepchas, and Bhutia. The city is fringed by two streams, namely Roro Chu in east and Ranikhola in West. These two rivers divide the natural drainage of the Gangtok area into two parts, eastern and western region. Both the streams meet at Ranipool and flow towards the south of the Ranikhola before and it joins the Teesta at Singtam. The geomorphic form of the slopes is complex. Slope modification has undergone due to landslide and anthropogenic activities in the study area. The study area has been characterized by two rock country, one is schistose and other is the gneissic area. (a) Tathangchen and Syari valley are formed on the lower and higher schist zone. Gentler slopes are formed over schistose in Tathangchen and the ridge axis passes through the schistose region. The topography of the area is generally flat or ascending which gives rise to the formation of lower schist zone known as "Deorali flat top" and upper schist zone resulted in the "Palace ridge road flat top." (b) Most of the scraps face area seen in the Gneissic region (two together Lingtse and Darjeeling gneiss). Whenever the ridge passes through it, topography become steeply ascending; for example, Lingtse gneiss has turned over to Namnang ridge and Darjeeling gneiss has given rise to Chandmari ridge. According to the Bureau of Indian Standards (BIS), the town is frequently subjected to earthquake and is categorized under seismic zone-IV, being situated near the tectonic plates of Eurasian and Indian convergent boundary. The world's third-highest peak, Mount Kanchenjunga (8598 m or 28,208 ft) is noticeable to the west of the town. The presences of steep slopes are responsible for landslides vulnerability. Gangtok is the centre of tourism in Sikkim state, India.

Episodes of landslide within GMC area
Recently, most of the landslides occurred within the GMC area are of anthropogenic in origin, i.e., due to the construction of roads and buildings, but there are also some well-known landslides of natural origin ( Figure 2).
Chandmari ward, which falls under the unstable zone, has been reactivated time and again (1975, 1984, and 1997). During the month of the rainy season (June), 1997, Chandmari landslide was again reactivated as an outcome of debris-cum-rockslide which caused severe damage to Sikkim Nationalised Transport (SNT) workshop, located at the bottom of the slide zone (Ghoshal & Sengupta, 2001).
Manvir Colony landslide is in the sinking zone which is bounded by rock ledge on its two sides.
Geological Survey of India (2010-2012) and Chakraborty et al. (2012) did a detailed study of this landslide. This sinking zone is restricted between EL ± 2022 m and EL ± 1890 m along Indira By-Pass road, west of Gangtok town. The main cause of sinking zone at Manvir is the piping out of finer material. Rawat et al. (2011) have studied on the Bathang landslide, situated in Gangtok town having geographical coordinates of 27°21ʹ27.21″ N and 88°3 7ʹ24.66″ E. Topographically this area is highly undulating in nature. Bathang landslide blocks the National highway and this site poses many problems in different magnitudes at different times.

Methodology
Landslides are generally controlled by geology, slope morphology, soil, and moisture conditions (Ohlmacher Gregory, 1930). Rainfall is a very crucial triggering factor for landslides, but ground conditions are also significant (Hong, Adler, & Huffman, 2007). In our present work, twelve landslide-related factors are examined which includes slope morphometry, elevation, geology/lithology, lineament, land use/land covers (LULC), building density, rainfall, water regime, soil type, soil thickness, soil liquefaction, and relative relief. These factors are considered for the present study based on a review of the literature and data availability (Agarwal and Garg, 2016;Chowdhury, Jha, Chowdary, & Mal, 2009;Jha, Chowdary, & Chowdhury, 2010;Kumar, Gopinath, & Seralathan, 2007). The data (used to generate landslide conditioning factors) specification and sources are mentioned in Tables 1 and 2. All input data are converted into raster format with a cell size of 20 m to carry out further modelling. Figure 3 represents the detailed methodologies adopted in carrying out the present research work. The expert-knowledge-based approach for landslide susceptibility generally consists of three steps, i.e., (i) extraction of knowledge from domain expert, (ii) characterization of conditioning factors, and lastly (iii) prediction of landslide susceptibility ( Figure 4).

Assigning weightage to various influencing factors
In the present study, the weights and ranks were assigned to the factors based on knowledge-driven method (Qualitative method). The knowledge often exists in the form of human expertise and has to be brought out and stored in a knowledge-based system. The extraction of knowledge from local domain expert is a crucial step and conducted by landslide experts. A landslide expert should be a person who has been trained as a geologist/geomorphologist and who has an extensive theoretical knowledge with field experience in landslide studies. Landslide experts need two type of knowledge namely (i) what are the conditioning factors that affect landslide susceptibility in the area and (ii) how does these set of conditioning factors affect landslide susceptibility. Various researchers in various fields (Bhasin et al., 2002;Kumar et al., 2007) have carried out assigning of weights to the conditional factors based on knowledge-driven method but its implementation on landslide susceptibility determination is rare. In this method, higher weights and ranks are assigned to those factors that have a higher influence on landslide occurrence (Table 1). The rank assigned to the factors ranges from 1 to 10 where 10 represents the highest influence and 1 represents the least influence. Similarly, weights to the factor subclasses are further assigned.

Weighted overlay method (WOM)
Weighted Overlay Analysis, based on knowledgedriven method, is an efficient analysis for integration of different parameters based on their relative importance (Nag, 2005) and various researchers in various studies (Nag, 2005;Riad, Billib, Hassan, & Omar, 2011a;Riad, Billib, Hassan, Salam, & El Din, 2011b) have implemented it. Generation of landslide susceptibility map of the study area was carried out by implementing the following Equation (1).
where F is the landslide susceptibility map of the study area, ith (x = 1, 2, 3 . . .. . .. . .. m number of) factor maps and nth (y = 1, 2, 3 . . .. . .. . .. n number of) factor classes, R and W are the rank and weight of factors and factors classes, respectively. The output map, based on the weighted overlay, is classified into four hazard zones such as low, medium, high, and very high. Figure 3 represents the detailed methodologies adapted for implementing WOM to derive the landslide susceptibility model in the GMC area.

Preparation of landslide incidence map
Landslide occurrence is based on the principle that "slope failure are more likely to occur under a condition that led to past instability" (Guzzetti et al. 2012). So, it is important to trace past landslides for the prediction of future landslide occurrences. In landslide studies, past data are used to prepare the inventory. Generally, two types of inventory are developed for landslide susceptibility zonation, i.e., landslide inventory map (Basharat et al. 2016;Kaur et al., 2018) and event landslide inventory map (Guzzetti et al. 2012;Mondini and Chang 2014). Landslide inventory map shows the location and spatial extent whereas event landslide inventory indicates the extent, location, and typology of landslide hazard caused by triggering factors. The research started with the analysis of existing Geological Survey of India (GSI) reports and archives focused on landslide events. They were: -Landslide compendium on Darjeeling Sikkim Himalayas (Geological Survey of India, 2016) -Archive of DST project (Yudhbir Gergan, 2004) -PhD thesis (Kalita, 2006) -Local newspapers (Sikkim express, Himali Bela) GSI carried out a nationwide inventory on landslides. Eventually, they proceeded with field-checks, surveying, landslide morphology study, and large-scale geological mapping which were documented. Some of the landslide inventories are prepared from high-resolution data like Resourcesat-2 (LISS IV Mx), Google Earth, and GSI reported data. Several field surveys are conducted to identify the landslide area as well as cross-check the prepared landslide location from satellite data. In some archive data only the name, location, and triggering factor are mentioned, whereas spatial extent is missing. So, to make the inventory homogeneous, we have used point location mapping for inventory.   respectively. Schistosity indicates easterly dips which vary between 15°to 60°and strikes N-S to NNW-SSE direction (GSI, 2014). According to GSI (2014), various kinds of rock outcrops occur in the GMC area, i.e., phyllitic quartzite, garnetiferous mica schist, quartzitic phyllite with mica schist, schist, quartz biotite schist with soil cover, high graded gneiss with soil cover, and Lingtse granite gneiss ( Figure 5 (a)). The phyllitic quartzite, quartzitic phyllite with mica schist, and quartz biotite schist with soil cover occupy 8.72 sq. km of the total study area. Lingtse granite occurs near lower Sichey, Arithang, and Deorali wards, encompassing 0.64 sq. km of the GMC area. The maximum study area is covered with schist and high graded gneiss of about 4.58 sq. km and 3.49 sq. km area, respectively. Garnetiferous mica schist (1.31 sq. km) is found mostly in Tatachen syari, and some parts of Daragoan, Deorali, and Ranipool ward. In Arithang, upper Sichey and Mahatma Gandhi Marg of GMC are occupied by banded gneiss (0.56 sq. km). Rocks show high weathering and tectonic activity near thrust and fault in the study area (Sarkar, Kanungo, Patra, & Kumar, 2008). High erosion and weathering process plays an important role in landslide occuring in the study area (Malladi, 2012).

Rainfall
The average annual rainfall of the Gangtok town is 350 cm (Bhasin et al., 2002). The spatial distribution of rainfall of the study is classified into two categories, i.e., below 300 cm and above 300 cm (SSDMA, 2012). GMC town has suffered from intense monsoonal rainfall from June to September with the highest recorded monthly average of 649.6 mm (25.6 in) in July. Approximately 58% of the study areas (11.20 sq. km) receive rainfall below 300 cm whereas about 42% of the area (8.06 sq. km) receives rainfall of above 300 cm (Figure5(c)). However, incessant rainfall is the main triggering factor for landslides.

Water regime
Presence of rivers, snow-covered mountain peaks, springs, large and small water bodies make this town wetland complex. Ratey Chu River is the only source of fresh water to Gangtok municipality. Tamze/Hans Pokhari is the largest lake in the valley lying athwart of south and east district of Sikkim, which is the major source of Ratey Chu River. Water regime is an important factor to measure landslide susceptibility as it determines the pore pressure and frictional force of the area. The regime in the study area is divided into four categories, i.e., subsurface/low surface regime, high surface water regime, medium surface water regime, and dry, covering an area of 6.96 sq. km (36.12%), 4.60 sq. km (23.85%), 7.17 sq. km (37.21%), 0.55 sq. km (2.83%), respectively ( Figure 5(d)).

Morphometric elements
Relief, elevation, and slope are the morphometric elements which are helpful to divide the landforms into different morphometric units and to define the relationship with the geology of the specific site. Relief map represents the difference between minimum elevation and maximum elevation point within an area and it has been highly implemented for landslide hazard zonation modelling (Gupta, Saha, Arora, & Kumar, 1999;Kanungo, Arrora, Sarkar, & Gupta, 2009). The elevation is defined as the vertical height of a point from sea level and relief is the difference between two elevation points. The slope may be defined as the vertical inclination between the hill top and valley bottom and is expressed generally in terms of degrees or percent. Relief itself does not directly influence the landslide occurrence but may trigger other landslide triggering factors. High relief becomes unstable due to the influence of triggering factors like rainfall precipitation, earthquake, etc. (Anbalagan et al., 2015). According to the relief map ( Figure 5(e)), the study area is categorized into five major classes, namely very low, low, medium, high, and very high relief accounting for 0.67 sq. km (3.49%), 0.42 sq. km (2.20%), 5.61 sq. km (29.06%), 6.33 sq. km (32.83 %), and 6.25 sq. km (32.42%), respectively. The slope is one of the critical factors for landslide occurrence investigation. Sidle and Ochiai (Sidle & Ochiai, 2006) observed that even though landslide occurs over a wide range of slope, it typically occurs between 20°and 90°slope. Slope also influences the development of vegetation and human settlement which in turn influences on the landslide occurrence (Guariguata & Larsen, 1990;Walker, Zarin, Fetcher, Myster, & Johnson, 1996). The slope of the study area is categorized into six major classes namely very gentle, gentle, moderate, steep, high, and critical, encompassing an area of 2.57 sq. km (13.32%), 8.73 sq. km (45.27%), 5.64 sq. km (29.22%), 0.76 sq. km (3.95%), 1.57 sq. km (8.15%), and 0.02 sq. km (0.1%), respectively ( Figure 5(f)).
3.1.6. Land use/land cover Different types of LULC have a direct influence on landslide occurrence by controlling the slope stability (Greenway, 1987). Anthropogenic activities significantly influence the reactivation or initiation of the landslide (Bruschi et al., 2013;Vanacker et al., 2003;van Den Eeckhaut, Poesen, van Gils, van Rompaey, & Vandekerckhove, 2009). The LULC map for 2016-2017 is developed through visual interpretation of multispectral LISS IV images (2006, 2011, and 2016) and QuickBird and later the interpretations are confirmed through field investigations. In the study area, the sparsely urbanized area encompasses an area of 4.39 sq. km accounting for approximately 22.79% of the total area; barren land with slope accounts for 4.36% of the total area (0.84 sq. km), heavy urbanized area with proper surface and subsurface drainage accounts for 13.33% (2.57 sq. km), heavily urbanized area with inadequate surface and subsurface drainage covers an area of 4.28 sq. km (22.19%), barren land with slide and slope cut accounts for an area and moderately vegetative area with thin grass cover occupy 1.69 sq. km (8.77%) and 3.85 sq. km (19.99%), respectively. Agricultural land with thinly populated area accounts for 1.65 sq. km (8.57%) of the study area ( Figure 5(h)).
3.1.7. Building density Rapid urbanization in the study area puts tremendous pressure on people for housing requirement which requires suitable locations for constructions. In the urban area, the majority of the settlements are unplanned and without proper supervision (Rai, Mondal, Singhal, Parool, & Pradhan, 2012). Being the capital of Sikkim state, Gangtok town is the most populated area in East district of Sikkim in addition to that due to the scenic beauty of the state, GMC area has well-occupied tourism infrastructure since the town is engaged in service sector like resorts, hotels, restaurants, travel agencies, etc. In Sikkim, East district has the highest population constituting about 46.29% of the total population of the state. The Gangtok subdivision has a total population of 179,376 (CENSUS, 2001). According to CENSUS report (2001), GMC with 15 wards has urbanpopulation of 43,711. Rapid migration of rural population to the urban areas of the city have led to its population growth in an unplanned and unorganized manner posing a severe threat in future perspective. Nevertheless, the Government has now stopped the allotment of land in the GMC area. With respect to the building density map (Figure 5(i)), 6.22 sq. km (32.06%), 8.03 sq. km (41.38%), 3.57 sq. km (18.40%), and 1.58 sq. km (8.16%) of the study area have very low, low, medium, and high building density, respectively. Sidle and Ochiai (2006) define static factors as those factors that are related to surface features. Landslide is also known as landslip in which sliding is found at the soil mantle. The soil matrix properties such as soil particle size, composition, pore distribution, etc. influence the slope stability (Seed, 1968) by influencing the water holding capacity and the rate of water movement. Finer texture soil can hold a higher volume of water than coarse-sized soil under unsaturated conditions. The soil type of the study area is broadly classified into five major classes, namely debris comprises mostly of rock pieces mixed with sands younger loose materials, debris comprises mostly of rock pieces along with older sand, sandy soil with the naturally formed surface, older well compacted mature soil formation, and clayey soil with a naturally formed surface ( Figure 5(j)) with the areal coverage of 7.23 sq. km (37.36%), 5.31 sq. km (27.43%), 19.74 sq. km (3.82%), 9.70 sq. km 1.88%), and 5.77 sq. km (1.12%), respectively.

Static factors of soil
On a hill slope, the soil thickness has a critical attribute for carrying out analysis of slope analysis. Within GMC area, soil thickness is divided into the following six classes (Figure 5(k)): thin soil cover, thin soil cover with slope, thin to medium thick soil cover, medium thick to thin soil cover, medium thick soil cover, and thick soil cover, encompassing an area of 1.27 sq. km (6.58%), 5.76 sq. km (29.91%), 4.30 sq. km (22.34%), 5.64 sq. km (29.27%), 1.7 sq. km (8.80%), and 0.60 sq. km (3.10%), respectively. The rank and weight to soil thickness map are assigned as per Joshi (2010).
Liquefaction by itself does not directly initiate any particular hazard but rather act as isolator during seismic shaking by impeding the transmission of vibrational energy (Ambraseys, 1973;Kaur et al., 2018). Soil liquefaction initiates some ground failures due to temporary or permanent movement of ground and becomes one of the most important attributes to landslide (Chung & Fabbri, 2008). Based on soil liquefaction property, the entire study area is divided into four classes ( Figure 5(l)), i.e., low, medium, high, and very high occupying an area of 2.99 sq. km (15.51%), 4.15 sq. km (21.54%), 5.18 sq. km (26.88%), and 6.95 sq. km (36.07%), respectively.

Susceptibility model of the GMC area
The final susceptibility map derived from the knowledge-driven method is classified into four classes, namely low (18.11%), medium (30.95%), high (31.78%), and very high (19.14%) hazard zone, respectively ( Figure 6). The final susceptibility map derived shows the imprint of hazard observed spatial pattern of major causative factors of landslide hazard, particularly slope morphometry, geology, and soil type in the study area.

Superimposition of reported landslides
Landslide events that have occurred in the study area, i.e., from the year 2000 to 2015, have been considered for validation (Kaur et al., 2017a). The final landslide susceptibility zonation map generated is overlain by recorded landslide events location ( Figure 6). It is observed that 51 events (82.26%) of the total landslide events falls under high and very high landslide susceptibility zone, 7 events (11.29 %) falls under medium landslide susceptibility zone, and about 6.45% of the landslide events falls under low landslide susceptibility zone.

Success rate curve (SRC) and prediction rate curve (PRC)
The final landslide susceptibility zonation map is also validated using SRC and PRC, which is a widely used technique for validations of the model (Kaur, Gupta, & Parkash, 2017b;Thapa et al., 2017aThapa et al., , 2017bThapa et al., , 2017cThapa et al., , 2017dThapa, Gupta, Kaur, & Mandal, 2018). The success rate of the goodness of fit of the model applied can be distinctly visualized using this validation method (Klose, Gruber, Damm, & Gerold, 2014). For generating success and prediction rate curve, reported landslide occurrence in the study area are grouped into two classes, namely train and test dataset. The success rate curve (SRC) generated by plotting cumulative percentage of landslide occurrence in the train dataset on the y-axis against cumulative percentage of susceptible areas on the x-axis represents the prediction rate but gives no information about the accuracy of prediction (Sterlacchini, Ballabio, Blahut, Masetti, & Sorichetta, 2011;van Westen et al., 2003). Hence, the prediction rate curve (PRC) is used to assess the accuracy of the future prediction of the model, the robustness of the prediction, the effectiveness of models, etc. (Fabbri & Chung, 2001). PRC is generated by a plot between cumulative percentages of landslide occurrence in the test dataset on the y-axis and cumulative percentage of susceptible areas on the x-axis. The area under the curve (AUC) obtained from SRCs and PRCs are calculated for easier interpretations of the results (Vijith, Rejith, & Madhu, 2009). Figure 7 represents the success and prediction rate of the model applied where the AUC of SRC is 81.18% and PRC is 80.10% indicating prediction accuracy of the model above 80%.

Discussion and conclusions
In this work, the KDM technique is found to be an effective tool for landslide susceptibility modelling, with an accuracy of around 80%. The bulks of landslides in GMC were triggered during the monsoon season (June to September) when the southwest monsoon is at its peak. Several variables have contributed to devastating slope failure in GMC, including frequent orographic rainfall, steep slope, loss of forest cover over rapidly growing urbanized areas, high surface water regime, and so on. According to the landslide susceptibility zonation in the GMC area geology, rainfall, water regime, slope, and soil type are found to have a cumulative influence.
The high zone of landslide susceptibility is characterized by quartzitic biotite, NE slope pattern, high graded gneiss rock type, elevation > 1800, high rainfall (342 mm), high relief, greater soil liquefaction, higher lineament density, thin soil cover, and so on. The landslide susceptibility class illustrates that places with a high or very high sensitive zone are more likely to have new landslides, demonstrating the validity of the current landslide susceptibility mapping technique.
Chandmari, Burtuk, Lower Sichey ward, as well as some areas of Arithang, Deorali, Tatanchen Syari, and Zero point are found to be more susceptible to landslide due to weak rocks, steeper slopes, deforestation, and continuous construction activities disrupting slope stability. The presence of metamorphic rocks such as the schist zone having 45 to 60-degree angle of foliation, Darjeeling gneiss, and phyllite are the reason for landslide occurrences in these zones. Foliation junction varies from. Sinking zones that are the manifestation of landslide are found in the Chandmari region (east of Gangtok) and the Indira Bypass route (west of Gangtok). Landslides along the Indira Bypass road are caused by the building of a weak wedge along the slope, which has resulted in the construction of massively stepped chute drains. For the stability of these chute drains, a solid base is required. In general, these chute drains have been constructed on lightly compacted slide debris. As a result, the majority of these chute drains is damaged or fails to form a number of gapping joints, transverse and longitudinal fractures in a variety of places. Seepage of water occurs through these fissures. During rainy season, unguided surface and subsurface drain channels become more energetic, and continuing piping activity becomes stronger, resulting in a high rate of creep and subsidence in this unstable slope. High and extremely high landslide susceptibility zones are also found around Ranipool (south of Gangtok city). This may be due to excessive saturation of unconsolidated and heterogeneous overburden material and road cutting along Ranipool-Pakyong road, ultimately creating a challenge owing to the influence of continual creep and subsidence along the slope. The outcome of this research particularly the LSM will be beneficial to the disaster management authority of the GMC area in the formulation of landslide hazard preparedness strategies. However, there are certain limitations of this research: (i) unavailability of landslide data prior to 2000-2015; just the GSI reported and Google Earth data are provided here. ii) unavailability of spatial extent of previous landslide; as a result, point location mapping methods are employed in the current study to create a homogeneous inventory. (iii) Subjectivity restriction of the knowledge-driven method.

Disclosure statement
No potential conflict of interest was reported by the authors.