Introduction

Landslide is one of the most common natural hazards in hilly area like Nilgiri district. It can be devastating with massive destruction to life and property and may also lead to large scale landscape transformations. It is frequently responsible for considerable losses of both money and lives. The severity of the landslide problem worsens with increased urban development and change in land use. Landslide has been a perennial problem of the hill areas and various models have been designed for landslide vulnerabilities (Jibson et al. 2000; Luzi et al. 2000; Zhou et al. 2002; Carro et al. 2003, Lee 2007, Burrough and McDonnel 1998, Miles et al. 1999, Siddle et al. 1991, Lee et al. 1991, Hutchinson and Chandler 1991, Morgan et al. 1992, Carrara et al. 1991, 1992, Moon et al. 1992, Wadge 1988, Gupta and Joshi 1990, Wang Shu-Quiang and Unwin 1992, Pachauri and Pant 1992, Rahaman et al. 2014). Prevention of natural hazard can be achieved rarely with today’s technology and knowledge. However, it is possible to avoid or to diminish the impacts of disaster with effective disaster management strategies. The Geographical Information system (GIS), Remote Sensing (RS), and Global Positioning System (GPS) are the technologies which would helpful in studying and managing the natural hazard like landslide. In order to provide landslide vulnerability maps various methods such as fuzzy logic, statistic methods and Analytic Hierarchy Process (AHP) can be used. One of the method is the AHP, which is a theory of measurement for dealing with quantifiable and intangible criteria has been applied to numerous areas, such as decision theory and conflict resolution (Vargas 1990). Using this method, each factor is used in landslide vulnerability zoning is broken into smaller classes and then these classes are weighted based on their importance and eventually the prepared layers are assembled and the final map is produced. In this method, weight of each layer is depending up on the judgment of expert, with the compatible of map with reality.

The aim of the present study is to carry out a geoinformatic based vulnerable map for Kallar River sub watershed, to achieve aim following objective has followed in this study.

  1. I.

    To create database from GIS, RS and secondary data on various aspects.

  2. II.

    To produce reliable and updated thematic layers on topography, geology, hydrology, precipitation, geomorphology, land use, soil etc.

  3. III.

    Determination of weights and ratings for above mentioned thematic layers in the analysis by applying AHP method to identify landslide vulnerability zones.

  4. IV.

    Using landslide vulnerability map, to predict vulnerable areas, urban (suitable for building construction road and bridges) and agricultural development.

Study area

The study area the Kallar River Sub Watersheds is situated in eastern slope of Western Ghats stretching from West to the East. Study area comes under the part of the Nilgiri District. The Kallar River Sub Watershed is part of the Kallar Watershed. Its significant basin is Bhavani river basin, which is the main in moyar and Bhavani River. It is located between 11°18′N and 11°26′N latitude and 76°41′E and 76°55′E Longitudes, with an area of 207 sq.Km. It comprises of 2 districts (Nilgiris and Coimbatore), 4 taluks (Coonoor, Kothagiri, Udhagamandalam, Mettupalayam) and 24 Revenue Villages (Fig. 1). The elevation ranges from 550 to 2600 m (Fig. 2). About 90% of areas are mountains covered with diverse plant communities that form various types of forest and other agricultural activities, notably Tea, coffee plantation, vegetables and orchards. These are normally being cultivated in the upper and the lower area (Fig. 2). The climate of this area is temperate and salubrious for more than half of the year. The average day temperature of the sub watershed is 20 °C and the average mean annual rainfall is about 1120 mm. The winter is relatively cool. The maximum rainfall occurs during the month of October and November. The Study area falls in the following soil families: Hallimoyar, Attavalai, Millithenu, Terremia, and Karumpalam. The population is around 2.5 lakhs. A large number of beautiful spots such as kattery water falls, Log falls, Catherine falls, Lamb’s rock, Dolphin nose view point, Kallar, and Sim’s park of Coonoor are well known tourist attractions. In recent years, the study area has experienced several devastated landslide incidents that brought vast damage to properties and natural environment, and some loss of human life.

Fig. 1
figure 1

Location map

Full size image
Fig. 2
figure 2

Topography (left) and 3D view of Landsat 8 image (right)

Full size image

Data and Methodology

Data

The present study has made use of ten thematic layers (factors) such as elevation, aspect, slope angle, land use, lineament density, soil depth, drainage density, precipitation, distance to road, and NDVI for the preparation of the landslide vulnerability map. In this study the survey of India (SOI) toposheets 58 A/11 and 58 A/15, landsat 8 images (FCC), census data and precipitation data from Indian Meteorological Department (IMD) have been used to create these thematic layers. The working scale of geographic maps was chosen at 1:50,000. ArcGIS 10.1 and Erdas 9.2 were used to prepare thematic maps and layout. All the collected data were converted into a raster grid with 25 m × 25 m cells for the use. The total cell number is 2,73,869 for this study.

The contours with 20 m intervals were digitized from extracted SOI toposheet, and generated the digital elevation model (DEM). Using DEM as input the elevation (relief), aspect and slope angle maps were drawn. In addition road and drainage maps were also derived from the above toposheets. The drainage density was calculated. The distance to road buffer was calculated at 100 m interval. The lineament map was prepared from bhuvan portal, and lineament density was calculated. A map showing soil depth was prepared for block level. Using rainfall data of Indian Meteorological Department (IMD), rainfall intensity map of the area was produced by applying spline interpolation method. Land use/Land cover map was generated from Landsat-8 image using a hybrid method and field study where ten classes were identified. Finally, the NDVI was generated and its values range from −1 to +1 (pixel values 0–255) (Fig. 3).

Fig. 3
figure 3

a 20 m contour, b relief, c geomorphology, d drainage, e lineament, f rain gauge station, g soil type, h land use, i road and settlement

Full size image

Methodology

In the present study, the analytical hierarchy process (AHP) technique was used to produce landslide vulnerability zonation map for the Kallar River sub watershed, which is being one of the well-known landslide hotspot in Tamil Nadu. To achieve this, the relevant thematic layers pertaining causative factors were generated using remotely-sensed data, field surveys and GIS tools. Landslide vulnerability zonation map of the study area was eventually prepared using AHP method. In this method, the landslide vulnerability zone index (LVZI) value for each considered pixel was computed by summation of each factor’s weight multiplied by class weight of each referred factor (for that pixel) written as follows:

$$\text{LVZI}=\sum\nolimits_{i=j}^{n}{\left( {{W}_{i}}~{{R}_{j}} \right)}$$
(1)

where LVZI is the required landslide susceptibility index of the given pixel, R j and W i are class weight (or rating value) and the factor weight for factor i derived using AHP technique.

Analytical Hierarchy Process (AHP)

In order to prepare the landslide vulnerability zone map, the various methods such as fuzzy logic, statistic methods and AHP can be used. However in the present study AHP technique were followed. AHP involves building a hierarchy of decision elements (factors) and then making comparisons between possible pairs in a matrix to give a weight for each element and also a consistency ratio. AHP is a multi-objective, multi-criteria decision-making approach which enables the user to arrive at a scale of preference drawn from a set of alternatives; it is based on three principles: decomposition, comparative judgment and synthesis of priorities.

Analytical hierarchy process (AHP) is a semi-qualitative method, which involves a matrix-based pair-wise comparison of the contribution of different factors for land sliding. It was developed by Saaty (1980) and gained widespread attention later on. Factor weight of each criterion is determined by a pair-wise comparison matrix as described by Saaty (1990, 1994), and Saaty and Vargas (2001). Using this method, each factor layer is broken into smaller classes, and then these classes are compared based on their importance. For comparison of importance of classes relative to each other, each class is rated against every other class by assigning a relative dominant value between 1 and 9. This value and its description are shown in Table 1.

Table 1 Fundamental scales for pair-wise comparisons
Full size table

In order to establish a pair-wise comparison matrix (A), factors of each level and their weights are shown as: A 1 , A 2 , …, A n and w 1 , w 2 , …, w n . The relative importance of a i and a j is shown as a ij . The pair-wise comparison matrix of factors A 1 , A 2 , …, A n as A=[a ij ] is expressed as

$$A = \left\{ {\mathop a\nolimits_{ij} } \right\}n*n = \left( {\begin{array}{*{20}{c}} 1&{{a_{12}} \ldots }&{{a_{1n}}}\\ \vdots & \ddots & \vdots \\ {{a_{n1}}}&{{a_{n2}} \cdots }&1 \end{array}} \right) = \left( {\begin{array}{*{20}{c}} 1&{\frac{{{w_1}}}{{{w_2}}}....}&{\frac{{{w_1}}}{{{w_n}}}}\\ {\frac{{{w_2}}}{{{w_1}}}}&{1...}&{\frac{{{w_2}}}{{{w_n}}}}\\ {\frac{{{w_n}}}{{{w_1}}}}&{\frac{{{w_n}}}{{{w_2}}}}&1 \end{array}} \right)$$
(2)

In this matrix, the element, a ij  = 1/a ji and thus, when i = j, a ij  = 1. A matrix is normalized using Eq. 3 as

$$a_{ij}^{\prime} = \frac{{{a_{ij}}}}{{\sum\limits_{i = 1}^n {{a_{ij}}} }}$$
(3)

ij = 1,2,3,…….n.

And finally, weights of factors are computed using Eq. 4 as:

$$\mathop w\nolimits_i = \left( {\frac{1}{n}} \right)\left( {\frac{1}{n}} \right)\sum\limits_{i=1}^{n}{a_{ij}^{'}}$$
(4)

In matrix-based pair-wise comparison, if the factor on the horizontal axis is more important than the factor on the vertical axis and this value varies between 1 and 9. Conversely, the value varies between the reciprocals 1/2 and 1/9 (Table 1). In AHP, for checking consistency of matrix, consistency ratio is used, which depends on the number of parameters. The consistency ratio (CR) is obtained by comparing the consistency index (CI) with average random consistency index (RI).The consistency ratio is defined Eq. 5 as

$$\text{CR}=\frac{\text{CI}}{\text{RI}}$$
(5)

where, CI is the consistency index which is expressed as:

$$CI{\text{ }} = \frac{{{\lambda _{max}}{-}{\text{ }}n}}{{~n{\text{ }}{-}{\text{ }}1}}$$
(6)

where, λ max is the major or principal Eigen value of the matrix and it is computed from the matrix and n is the order of the matrix. And the average random consistency index (RI Table 2) is derived from a sample of randomly generated reciprocal matrices using the scales 1/9, 1/8, …, 8 and 9.

Table 2 Random consistency index (RI)
Full size table

The final result consists of the derived factor weights and class rating, and calculated consistency ratio (CR), as seen in (Table 3). In AHP, the consistency used to build a matrix is checked by a consistency ratio, which depends on the number of parameters. For a 10×10 matrix, the CR must be less than 0.1 to accept the computed weights. The models with a CR greater than 0.1 were automatically rejected, while a CR less than 0.1 were often acceptable.

Table 3 Factor weight
Full size table

In this study, the CR is 0.068, the ratio indicates a reasonable level of consistency in the pair-wise comparison, that is good enough to recognize the factor weights. Consequently, the weight corresponding to precipitation is highest, whereas elevation is lowest (Table 3). For all cases of the gained class weights, the CR is less than 0.1; the ratio indicates a reasonable level of consistency in the pair-wise comparison that was good enough to recognize the class weights.

Landslide vulnerability analysis

Relief

The area possesses high relative relief which refers to the difference between the highest and the lowest altitude in an area. The higher values indicate the rapid rise in altitude and presence of faults, lower relief signifies mature topography. As a risk agent, relative relief plays a vital role in the vulnerability of settlements, transport network and land (Chandel et al. 2011). In Kallar River Sub watersheds, there is a wide variation in relative relief (Fig. 4) ranging from low altitude to very high altitude. About 5.71, 6.36, and 2.07% areas have low (below 550 m), moderate (1300–1600 m) and very high (above 2600 m) relative relief respectively. The weight for relative relief (elevation) factor is 0.028 (0.28/10) and relative relief classes were rating based on influence to landslide vulnerability. The very high altitude above 2600 m was assigned higher rating of 0.377, moderate altitude 1300–1600 m as assigned medium rating of 0.064, and low altitude below 550 m were assigned very low rating as 0.019 (Table 4).

Fig. 4
figure 4

Elevation and aspect

Full size image
Table 4 Class rating
Full size table

Slope aspect

Aspect defines the down slope direction of the maximum rate of change or the direction of steepest slope in x-y plane. The aspect has significance in understanding the slope stability. Generally southeast (SE) to south (S) and southwest (SW) slopes are relatively more susceptible to slope failure and sliding activities. Present watershed area aspect classified in to 9 classes, such as flat 2.6%, north 14%, northeast 14.7%, east 7.8%, southeast 14.2%, south 22%, southeast 12.5%, west 6.5%, and northwest 5.3% (Fig. 4). The weight for aspect factor is 0.031 (0.31/10). The aspects classes were rated based on slope stability, the higher rating was assigned to southwest 0.353, followed by northeast 0.189, least rating was assigned for flat 0.029 (Table 4).

Slope

The slope refers to the degree of change in elevation over distance with lower the slope values indicate flatter terrain and higher values indicate steeper terrain. The slope map was generated from the 20 M contour. On the basis of which the Digital Elevation Model (DEM) has been generated. The DEM as an input parameter for slope. In the study area, there are 8 categories of slope classes and they are: 0–5°, 5°–10°, 10°–15°, 15°–20°, 20°–25°, 25°–30°, 30°–35°, and >35° (Fig. 5). The Weight for the slope factor is 0.164 (1.64/10) and the slope classes were rating accordingly to landslide vulnerability. The slope angle >35° was given higher rating of 0.377 which covers an area of 17.59 sq.km 8.5%, followed by 30°–35° rating as 0.206, covers 18.24 sq.km 8.8% and 0°–5° was assigned least rating as 0.023, with an area of 56.70 sq.km 27.4% (Table 4).

Fig. 5
figure 5

Slope and land use

Full size image

Land use/land cover

Generally, land use/land cover has effect on strength of slope materials against sliding and control of water content of slope. It reflects relationships between land use, risk and vulnerability to disaster events. Kallar river sub watershed area has different verities of land use and land cover features. About 70% of area is hilly terrain. The land use was classified in to 11 classes (Fig. 5). A larger part of the sub watershed occupies tea plantation which account for 73.2 sq.km (35.3%), followed by vegetation 49.5 sq.km (23.8%), dense forest 46.6 sq.km (22.4%), and settlement occupies 13.2 sq.km (6.5%) etc. Weight for the land use is 0.083 (0.83/10) and land use classes were rating based on influence to landslide vulnerability, the higher rating were assigned to tea plantation 0.315, and vegetation 0.219, followed by open dense forest 0.113, urban mixed 0.098, forest plantation 0.080, and the least rating was given to the water body 0.018 (Table 4).

Lineament density

Lineament density and distance to lineaments are two factors often used for GIS based landslide vulnerability analysis. Several studies have found that lineaments directly influence landslide occurrences either by distance between the lineament and landslides or by concentration of lineaments in a particular area, which also known as lineament density. (Atkinson and Massari 1998; Pachauri et al. 1998; Suzen and Doyuran 2004; Lee 2007; Lee and Pradhan 2006). In this study lineament density were calculated and brought under five classes, ranging from very low to very high density. Accordingly the very low density 52.4% (<42 m), low density (43–110 m) 40.2%, medium density (111–170 m) 6.1%, high density (171–240 m) 1.25% and very high density (241–380 m) 0.028% lineament density respectively (Fig. 6). Weights for the lineament density is 0.120 (1.2/10) and the rating for lineament density classes assigned based on influence to landslide vulnerability. The higher rating was given for very high density which is 0.431 and lower rating for very low density is 0.053 (Table 4).

Fig. 6
figure 6

Lineament density and soil depth

Full size image

Soil depth

Depth of the soil forms one of the important factors for assessing the stability of the soil and landslide susceptibility of the land. With the increase of soil depth, the tendency of soil to absorb moisture is increased, resulting in reduced runoff rate. Hence shallow soil is considered to be more unstable and prone to landslide than the deep soil (Sharma at el. 2009). The soil depths of the study area were divided in to five classes. The depth level of top soil 0 is covered an area of 11.30% (23.36 sq.km), followed by the shallow depth 50 m, covers 68.45% (141.48 sq.km), moderate shallow depth 150 m, occupies 15.29% (31.6 sq.km), deep 175 m, covers 4.42% (9.14 sq.km) and very deep 200 m covers 0.52% (1.08 sq.km) (Fig. 6). Weight for the soil depth is 0.055 (0.5/10). Rating of soil depth was assigned based on influence in landslide. The higher rating 0.480 were given to shallow depth, followed by 0.206 moderate shallow depth, 0.138 deep depth, 0.112 for very deep depth, and the least rating given to top soil 0.064 (Table 4).

Drainage density

The overall drainage pattern is dendritic and parallel pattern which covers larger area, and radial pattern in the central part of the area. A mathematical expression of drainage morphometry of an area is drainage density which is a measure of the length of stream channel per unit area of drainage basin [Drainage density (Dd) = Stream length (L)/Basin area (A)].

The measurement of drainage density is useful in determining landscape dissection and runoff potential. Higher values denote higher degree of dissection of land, as well as indicate the higher probability of slope failure. The drainage density in the study area can be classified into very low density (0.8 km2) to very high density (4 km2). In the study area about 78% of area comes under the low and medium density categories. Followed by very low and high density area (20%) and very high density area is (2%) (Fig. 7). Weight for the drainage density is 0.033 (0.33/10) and the drainage density classes were rating accordingly to landslide vulnerability. The very high density is more vulnerability in this area, so the rating is assigned for 0.445, followed by high density 0.297, moderate density 0.147, low density 0.073 and very low density 0.037 (Table 4).

Fig. 7
figure 7

Drainage density and rainfall

Full size image

Precipitation

Rainfall produces sudden floods which cause shallow landslides. Most of the landslides occur after the torrential rain. Thus the rainfall has been considered as one of the main parameters in producing landslide maps. Water infiltrates rapidly upon heavy rainfall and increases the degree of saturation and potential of landslide occurrence (Pourghasemi et al. 2009). In general the study area is mountainous with cool weather. In the present study 5 precipitation stations Coonoor, Kethi, Kothagiri, Udagamandalam, Kundha bridge and Mettupalayam were taken to study the rainfall. The rainfall has been brought under six classes, i.e., <1000 to >2300 mm. The <1000 (2.8% of area), followed by 1200–1400 mm (8%), 1400–1600 mm (15.8%), 1600–1800 mm (21.04%), 1800–2000 mm (26.10%), and 2000–2300 mm (25.46%). Weight given for the precipitation is 0.259 (2.59/10) and the total amount of precipitation classes were brought under rating based on landslide influences. The higher rating was given to 2000–2300 mm (0.401), followed by 1800–2000 mm (0.275) and very least rating was given to <1000 mm (0.029). Precipitation consistent ratio is 0.052 (Table 4).

Distance to road

The distance to road is one of the important parameters in preparing landslide vulnerability maps. Roads can be one of the reasons of occurring landslides (Yalcin 2008). Roads change the nature of topography and decrease the shear strength of toe of slope and cause the tensile stress. Naturally, slope may be stable, but after road construction, road can have undesirable effect on slope. In the present study, major roads are investigated, and this factor is divided into 5 classes (Fig. 8). The range distance from 0 to 800 m. Weight for distance from road is 0.170 (1.70/10). The class distances to road are rating based on vulnerability. The less distance from the major road is more prone to the landslides which is given for higher rating 0-–100 m (0.468), followed by 100–300 m (0.268) and least rating was given to 600–800 m (0.044).

Fig. 8
figure 8

Distance to road and NDVI

Full size image

NDVI

The vegetation index is also considered an influencing factor in landslide vulnerability assessment studies (Althuwaynee et al. 2012). NDVI was used in this study to reflect the vegetation density. In general, the value of NDVI ranges between (−)1 and 1; the higher the value of NDVI the denser of vegetation cover. The NDVI values were calculated by using the multi-spectrum information from the Landsat 8 image based on the following formula. NIR-R/NIR+R. The NDVI value of study area is 1. Therefore the NDVI is classified into five levels between 0 and 1 with a 0.2 interval. In this study NDVI range 0–0.2 covers area of 12.01%, followed by 0.2–0.4 (5.5%), 0.4–0.6 (82.4%) and range 0.8–1 (0.012%) (Fig. 8). Weight for NDVI is 0.057 (0.57/10). For NDVI class the rating was assigned. Most of the landslides were occurred in areas of low NDVI values, especially NDVI value of <0.2 is given for higher rating of 0.502, followed by 0.2–0.4 for 0.254 and least rating is 0.049 was assigned to NDVI value 0.8–1.

Result and discussion

In the present study, the analytical hierarchy process (AHP) was applied to develop a landslide vulnerability map for the Kallar River sub watershed, located in Nilgiri District. In order to achieve this, ten landslide inducing factors were taken into consideration. They are elevation, slope aspect, slope angle, distance from road, drainage density, lineament density, soil depth, precipitation, land use/land cover (LULC) and NDVI. The first eight parameters were extracted and calculated from their associated database while LULC and NDVI maps were derived from Landsat-8 satellite image. These factors were evaluated, by assigning weight for each factor and rating for each class. Based on the results given in Table 4, there are three most influencing factors to landslide activities (judged from their associated weights). They are precipitation (0.259), distance to road (0.170), and slope angle (0.164). And the three least influencing factors are elevation (0.028), slope aspect (0.031), and drainage density (0.033).

Applying the AHP, the LVZI values were computed by using Eq. (1). From the calculation, it is found that the LVZI had a minimum value of 0.04, and a maximum value of 0.31, with a mean value of 0.17 and a standard deviation of 0.04. The LVZI represents the relative vulnerability of a landslide occurrence. Therefore, the higher the index, the more vulnerable the area is to landslide. These LVZI values were then divided into five classes based on the range of natural breaks, which represent five different zones in the landslide vulnerability map. These are very high vulnerability (VHV) zone, high vulnerability (HV) zone, moderate vulnerability (MV) zone, low vulnerability (LV) zone and very low vulnerability (VLV) zone (Fig. 9).

Fig. 9
figure 9

Landslide vulnerability zone

Full size image

From the Table 5 one could infer that the area under High Vulnerability (HV) Very High Vulnerability (VHV) altogether occupies 33% of the study area. Hence about 149 sq.km is often prone for landslide and considered as high risk zone. However the moderate zone covers an area of 47 sq.km which account for 41%. These high vulnerability (HV) very high vulnerability (VHV) and Moderate Vulnerability (MV) are mainly occurred in larger areas covering central part by leaving about 26% of the area under Very low vulnerability (VLV) and low vulnerability (LV) in the western portion, eastern portion and a few pockets of mid-northern portion (Fig. 9). However it should be noted that the southern part of the study area is high and very high vulnerability as the area is under steep slope .

Table 5 Distribution of landslide locations with predicted landslide vulnerability zone classes
Full size table