4.2. The relation between preparatory factors and landside distribution
The Preparatory factors (mechanisms) are cumulative events, which are responsible for the landslide initiation. Preparatory factors are like slope angle, slope aspect, geological rock formation, lithological structure (Faults/ thrust), drainage alignment, road distance, and pattern of land use. Assessment of the event controlling parameters should be basis of a systematic inventory assist to create a database. This database established to the co-relation between landslide initiation and characterization of event controlling parameters in hilly terrain. The following event controlling parameters and landslides distribution are examined below:
4.2.1. Slope Angle
The slope angle is important parameter in the slope stability analysis (Lee and Min 2001). The direction of the slope angle affects the landslide initiation (Gupta et al. 1999). Slope gradient frequently used in preparing LHZ maps (Ercanoglu et al. 2004; Saha et al. 2005; Lee 2004; Yao et al. 2008; Nandi and Shakoor 2009; Yalcin et al. 2011). For the hazard zonation study, the slope map is prepared with the help of ASTER DEM image using GIS domain and the map is classified into eight slope categories (Fig. 3a) of the study area.
Based on the landslide data analysis (Fig. 6), most landslides (124) had occurred in 30˚ − 40˚ of the slope angle. The data analysis revealed that the in the 60˚ − 70˚ of the slope angle area the occurrence rate is 32.4%. The slope angle between 50˚ – 60˚zone is covered 12.5% occurrence rate which is second highest and 10˚ – 20˚ of the slope angle zone is found to have third highest frequency rate (11.6%) respectively. Landslide frequency Rate distribution of the study area is shown in Fig. 6.
4.2.2. Slope Aspect
Slope Aspect is another major event conditioning parameter, and several researchers has considered this parameter (Nagarajan et al. 1998; Yalcin et al. 2011; Pourghasemi et al. 2013; Ghosh et al. 2014). Various hydro-meteorological events like amount of sunshine, precipitation and the topographic condition of the area effect landslides initiation (Pourghasemi et al. 2013). High amount of rainfall received on the hillsides of the slope is correlated to capacity of filtration of the slope and may be controlled by different parameters such as topography of the slope, permeability of the rock structure, porosity, moisture retention, organic ingredients, land use, and climatic season (Pourghasemi et al. 2013). Maximum amount of rainfall is received on south facing hill slope of the Himalayan mountain terrain and much higher than the north facing slope (Ghosh et al. 2014). For this study, the slope aspect map is prepared to represent the relationship between aspect facing and occurrence of landslides. The slope aspect map classified with the help of ASTER DEM using aspect tools in ArcGIS platform. The aspect map classified nine classes (Fig. 3b) such as flat, northwest, west, southwest, south, southeast, east, northeast, and north.
Based on the spatial analysis (Fig. 7) between the slope aspect and occurrence of landslide found that the maximum number of landslide (113) occurred in southwest facing slope and also have the maximum landslide frequency rate of 24.8% (highest). According to Fig. 7, the distribution of landslides occurring in different slopes are as follows; southwest (24.8%), southeast (17.6%), south (17.3%), northwest (8.6%), north (8.1%), northeast (8%), west (8%) and east (7.6%). The lowest frequency rate is found on the east aspect which is only 7.6%. Finally, it is noticed that frequency rate of the landslides on the southerly facing is very high, and it decreases towards the northerly facing accordingly (Fig. 7) in the study area.
The geological rock formation plays an important role in landslide initiation. Slope failure is controlled by the rock unit properties of any hill regions. Lithological structure serves significant function in landslide hazard zonation studies because various lithological rocks structure have different susceptibilities to active geomorphic processes (Pradhan et al. 2007, 2009). A lithological map (Fig. 3c) is prepared from the geological map, sourced from GSI (Kolkata). The present research area is represented by four major Formations like Chungthang, Kanchenjunga Gneiss/ Darjeeling Gneiss, Gorubathan Formation, and Granite Gneiss zone. The Kanchenjunga / Darjeeling Gneiss contribute the largest area, of about 251.13 sq. km comprising of augen bearing biotite gneiss with/without kyanite, banded/streaky migmatite, sillimanite / sillimanite granite gneiss/ biotite gneiss etc. The Gorubathan Formation contribute second largest area of about 132.50 sq. km, covered by biotite phyllite, biotite quartzite, interbedded chlolrite sericite schist, mica schist with garnet, metagreywacke, chlorite quartzite etc. The Chungthang Formation contributes an area of 88.72 sq. km, and is covered by quartz-chlorite-sericite schists, chlorite-phullite, and garnetiferous quartz mica scist.
According to the Landslide Frequency analysis shown in Fig. 8, the maximum landslide events (208) occurred in the Kanchenjunga / Darjeeling Gneiss group of rocks with a 29.5% frequency rate. The maximum landslide frequency rate of 46.1% is observed in the Chungthang Formation Group of rocks. Relatively, 24.4% landslide frequency rate is noticed in the Gorubathan Formation of Daling group of rocks. Only Granite Gneiss types of rock that covers a smaller area (9.66 sq. km) of the study region have no landslide event, with landslide frequency rate of 0% (Fig. 8).
4.2.4. Distance from the drainage lines
Rivers or streams alignment of any hilly mountains area plays a significant part in landslide imitation specially ‘bank erosion and toe cutting’ (Miller and Sias 1998). Drainage buffers are derived with the help of the shapefile of river lines using by Euclidean distance tool in ArcGIS. For this study, five buffer classes are prepared corresponding to distance from river at 200 m intervals (Fig. 3d).
The present study observed that the triggering of the landslides is in direct correlation with the distance from the drainage valley (Zhao et al. 2017; Zhao et al. 2018). Figure 9 shows that maximum numbers (303) of landslides have occurred within 200 m from the drainage valley beds; this zone covers the maximum landslide frequency rate (53.2%). As the distance from the drainage line increases, the landslide frequency rate is found to decrease. The study shows 22.9% of landslides occurred at 200–400 m distance, 13.3% of landslides occurred at 400–600 m distance, and 10.6% of landslides occurred at 600–800 m distance from the drainage valley bed (Fig. 9). Thus the frequency distribution shows that most of the landslides initiated near to the drainage valley of the study area.
4.2.5. Distance from the faults or thrusts
Structures of lithology like faults, thrusts, and shear zones are tectonic breaks which plays an important role in occurrence of landslide. These breaks generally decrease rock strength and are thus considered vulnerable points of landslide initiation (Pradhan and Youssef 2009). They have triggered several landslides in the study area. These structures are the most significant factors for assessing landslide hazard, and geospatial data can be useful to delineate them (Kanungo et al. 1995; Saha et al. 2005).
The distance from faults and thrusts are extracted from the geological map of study area and nine buffer classes are created around the faults and thrusts at 500 m interval each using Euclidean distance tools in ArcGIS (Fig. 3e). The spatial analysis shows that 123 landslides have occurred within 500 m from the faults or thrusts line and it contributed 12.28% of landslide frequency rate. The highest landslide frequency rate (43.54%) is observed at a distance between 500–1000 m from the faults or thrusts as shown in Fig. 10. Distribution of 17.45% landslide frequency rate is at a distance between 1000–1500 m from the faults or thrusts and it is decrease accordingly. It is thus noticed that most of the landslides have initiated near the faults or thrusts of the present study. Relation between distance from the faults or thrust and landslide distribution is shown in Fig. 10.
4.2.6. Distance from the road
The roads are another significant anthropogenic principal factor for generation of the landslide hazard zonation mapping (Yalcin 2008). Construction of road can trigger slope disturbance as it raises stress, strain on the slope. Several landslides are initiated along the road corridor due to uncontrolled rock cutting (Ayalew and Yamagishi 2005). In the present study, several landslides are found along the road corridors (Fig. 3f). The distance from the road is derived from Google earth satellite image and topographical (78 A/10 and 78 A/11) and nine buffer classes are prepared at 500 m interval using Euclidean distance tools in ArcGIS domain and is shown in Fig. 3f .
According to the geospatial data analysis, maximum number of landslides (318) has occurred within 500 m distance from the road and contributes 91.84% landslide frequency rate that is highest along the road corridor (Fig. 11). The study found the when the distance increases gradually from the road 500–1000 m, 1000–1500 m and 1500–2000 m, then the distribution of landslide frequency rate gradually decreases to 2.97%, 2.27% and 1.26% respectively. The relationship between the distance from the roads and landslide occurrences is shown in Fig. 11.
Land-use is also an important factor for landslide initiation (Nourani et al. 2013; Bchari 2019). This factor is considered as significant parameters in assessing landslide hazard mapping. For e.g. vegetation covers increase the strength of soil by root reinforcement. It has large potential to reduce the slope failure rate (Begueria 2006). Satellite imagery is helpful to directly record land features from the ground. For the present study such data is obtained from high-resolution satellite images. Based on the analysis of remote sensing data, the study area is divided into five land-use classes like snow cover, dense vegetation, sparse vegetation, scrub with rocky land, and built-up with barren land (Fig. 3g)
Analysis of spatial data interpretation, shows maximum number of landslides (244) are initiated in the sparsely vegetated areas and contributes for 14.96% landslide frequency rate (Fig. 12). Based on the landslide frequency rate, the maximum landslide frequency rate (69.18%) is found in built-up area with barren land and the lowest rate of frequency (3%) is recoded in scrub with rocky land area. Landslides are not found in the of snow-covered areas (Fig. 12). Thus it is noticed that the barren and sparsely vegetated land areas are most vulnerable to landslides, with very low frequency in the dense vegetated land area. The relation between land-use and landslide frequency distribution is shown in Fig. 12.
4.3. Landslide Hazard Zonation Index Mapping with Application of Weightage Overlay Method
Landslide initiation is controlled from the landslide preparatory factors (controlling factors) and triggering factors. Such landslides can be expected in the similar conditions in future. Using the probability method of assumption, the relationship between areas with landslide initiations and factors which is related to landslides are differentiated from areas without landslide initiations. The weightage overlay method is applied to quantify landslide hazard zonation for the present study area. The modeling, calculation, and analysis processes are same for the each and every parameter. The value of weightage shows the relative influence of event controlling factors in the landslide initiation.
Every preparatory parameter layer is prepared on a GIS domain for the preparation of hazard zonation mapping. Then, all spatial data layers are overlapped using weighted overlay tools in ArcGIS platform. After that, these layers are prepared a composite image that provides information on the area normalized landslide initiation hazards index. Finally, The Landslide Hazards Index (LHI) image are classified into four hazard classes (Fig. 13) like Low (2–3), Moderate (3.01–4), High (4.01–5), and Very High Hazard Zone (5.01–8). For the ground truth verification, the landslide data (point location) are overlaid on the hazard classes and observed that the landslide frequency (%) gradually rises from Low to Very High Hazard zone (Fig. 14).