Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Landslide susceptibility mapping using knowledge driven statistical models in Darjeeling District, West Bengal, India

Landslide susceptibility mapping using knowledge driven statistical models in Darjeeling... Landslide is an important geological hazard in the large extent of geo-environment, damaging the human lives and properties. The present work, intends to identify the landslide susceptibility zones for Darjeeling, India, using the ensembles of important knowledge driven statistical technique i.e. fuzzy logic with Landslide Numerical Risk Factor (LNRF) and Analytical Hierarchical Process (AHP). In the study area, 326 landslides have been identified and a landslide inventory map has been prepared based on these landslides. The landslide inventory map has considered as the dependent factor and the geo-environmental factors like rainfall, slope, aspect, altitude, geology, soil texture, distance from river, lineament and road, land use/ land cover, NDVI and TWI have been considered as independent factors. Landslide susceptibility maps were prepared based on the Fuzzy- Landslide Numerical Risk Factor (LNRF) and Fuzzy- analytic hierarchy process (AHP) methods in a GIS environment. According to the results of LNRF and AHP based fuzzy logic 34 and 22% areas are highly susceptible to landslide in this district. The landslide maps of both models have been validated through ROC curve and RMSE. The areas under curves are 91% (for Fuzzy-LNRF) and 90% (for Fuzzy-AHP) and RMSE values of these models are 0.18 and 0.14 which are indicating the good accuracy of both models in the identification of landslide susceptibility zones. Moreover, the Fuzzy-LNRF model is promising and sufficient to be advised as a method to prepare landslide susceptibility map at regional scale. Keywords: Landslide numerical risk factor (LNRF), Fuzzy-AHP, Fuzzy logic (FL), Landslide susceptibility, GIS Introduction Rodriguez et al. 2008). Probably this rate of property The mountainous areas of the world are frequently af- damage will become more faster in the upcoming time fected by the occurrences of the landslide because of in parity with the gradual development of urban centers, high energy with variability and instability of masses economic and rising regional rainfall due to climatic (Gerrard 1994). From the environmental point of view, change in the landslide prone areas (Turner and Schus- different kind of problems such as loss of soil fertility, ter 1996). Landslide occurrence is a significant barrier to acceleration of deforestation rate etc. may be caused by the development in Darjeeling district. In Darjeeling dis- the landslides (Van Eynde et al. 2017). Most of the trict, landslides mainly take place due to heavy Mon- mountainous regions of India are characterized by the soonal rainfalls and seismicity (Panikkar and landslide disaster. A number of avalanche zones in the Subramanyan 1996). The Darjeeling district had been Indian Himalayan region are prominent, e.g. Jammu experienced major landslides in July–August, 1993, May Kashmir, Himachal Pradesh, Kumayun, Darjeeling and 2009 and September 2011 (Sarkar 1999). Massive rain Sikkim and North-eastern hilly states (Bhandari 2004). caused landslides at Darjeeling town, Mirik, Kurseong Landslide causes loss of property far greater than the and Kalimpong during June–July, 2015 and induced the any natural disaster (Turner and Schuster 1996; Garcia- loss of properties and lives. Reduction of effect of land- slide can be possible only with a comprehensive know- * Correspondence: sunilgeo.88@gmail.com ledge about the probability of occurrence, character and Department of Geography, University of Gour Banga, Malda, West Bengal, magnitude of landslide in an area. Therefore, delineation India of landslide vulnerable regions is indispensable for © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 2 of 18 carrying out safer alleviation programs, and future plan- and compare with them which have not been used in ning of the area (Fan et al. 2019). In the present paper this district previously. The main advantage of ensemble the main thrust has been given to delineate the landslide of fuzzy logic and LRNF is that we can use both expert susceptible zones and chalk out suitable method. knowledge as well as statistical method for predicting Landslide is the major hazardous phenomenon which landslide susceptibility using various causative factors. sometimes causes loss of human lives and properties. In this study, remote sensing data along with meta Therefore, any environmental factor may cause landslide data have been used to outline the landslide susceptibil- when soil resistance power is lower than the shear force ity areas for the Darjeeling District. Geo-environmental (Refahi 2000; Bouma and Imeson 2000). For landslide factors such as rainfall, slope, aspect, altitude, geology, hazard evaluation several qualitative and quantitative soil texture, distance from river, distance from linea- methods have been used (Aleotti and Chowdhury 1999; ment, distance from road, land use/ land cover, normal- Reichenbach et al. 2018). According to the qualitative ized difference vegetation index (NDVI) and method, the expert can evaluate the landslide suscepti- topographical wetness index (TWI) have been taken out bility zones in his own opinion. The expert also can as- to facilitate the quantification of landslide. Fuzzy-LRNF sess the vulnerable areas on the basis of similar and Fuzzy-AHP have been applied considering the ex- geological and geomorphological character using the tracted database. Using the LRNF models, the fuzzy landslide inventory maps or existing landslide areas membership value has been calculated and thereafter, (Ayalew and Yamagishi 2005). The multi-criteria deci- using the fuzzy gamma operator the membership values sion approach (MCDA) is an important way for indenti- of parameters have been assembled for producing the fying the potential landslide areas using proper database. landslides susceptible map of Darjeeling district. Simi- GIS based MCDA has been considered as the powerful larly using the Fuzzy-AHP method, another map has techniques and procedures for evaluating, designing and also been produced. Finally, the maps have been verified accuracy judgments’ of the results (Feizizadeh and Blas- and compared using known landslide locations based on chke 2011, 2013). The present study has followed the ROC and RMSE quantitative validation methods. The GIS based MCDA techniques like Fuzzy-Landslide Nu- main novelty is that the first time knowledge driven merical Risk Factor (LNRF) and Fuzzy-AHP for the technique (Fuzzy logic) has been assembled with LRNF landslide susceptibility mapping. Several other re- in this work to delineate the landslide susceptible zone searchers applied Fuzzy-AHP and LNRF. Torkashvand of Darjeeling district and compared with the Fuzzy-AHP et al. (2014) applied the Landslide Numerical Risk Factor method. Moreover, according to the previous literatures (LNRF) model using GIS in East of the Sabalan volcanic so many researchers used LRNF and AHP method for mass region in Iran. Mokarram and Zarei (2018), Feizi- mapping the landslide susceptibility but not a single re- zadeh et al. (2014), Mosavi et al. (2017), Hejazi (2015), searcher has used ensemble of fuzzy logic and LRNF Mirnazari et al. (2015) and Hembram and Saha (2018) model for predicting the spatial landslide probability and used Fuzzy-AHP model for their work and they got compared this ensemble method with fuzzy-AHP. fruitful result for susceptibility mapping. Various statis- tical methods have been used by the researchers for ana- Study area lyzing the spatial pattern of landslides and preparing the The Darjeeling district is located in the northernmost landslide susceptible map such as logistic regression part of the West Bengal in India. It is an important (Zêzere et al. 2017; Budimir et al. 2015), hierarchical ap- mountainous part of the eastern Himalaya. Geographic- proach (Youssef et al. 2014), statistical index (Dou et al., ally the study area is extended between the latitudes 2015), conditional analysis (Pourghasemi et al. 2012), 26°27″ to 27°13″N and longitudes 87°59″E–88°53″E. weight of evidence Pradhan and Lee (2010). In the re- The study area is covered with an area of 3149 sq.km cent years, different machine learning techniques have (Fig. 1). According to the census of 2011, the total popu- also been used by some scholars for mapping the land- lation of the district is 18, 46,823 with 9, 37,259 males slide disaster like decision tree (Pradhan 2013), random and 9, 09,564 females. The population density of the dis- forest (Dou et al. 2019), artificial neural network Prad- trict is 586 person/ km (District Statistical Handbook han and Lee (2010), support vector machine (Tien Bui 2013). The number of rural households and the urban et al. 2012) etc. For identifying the landslide susceptibil- households of Darjeeling district was 212000 and 89584 ity zones, they used some important factors such as the in 2001, but these have increased to 236000 and 154540 elevation, lithology, slope, land use, river, topographical in 2011 respectively (District Statistical Handbook 2011). wetness index, aspect, road, fault, and precipitation The total length of national highway, state highway, maps. The rationale of this work is to identify the land- major district road and other ordinary district road were slide susceptible areas using the ensemble models that 100, 80, 37 and 516 km respectively (District Gazetteer are fuzzy- LRNF and Fuzzy-AHP of Darjeeling district of Darjeeling District 1980). But the length of national Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 3 of 18 Fig. 1 Location map of study area showing district Darjeeling, India (a), and the location of 326 landslides (b) Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 4 of 18 highway, state highway and main district highway have Statistical Handbook, Census of India 2011, drainage increased to 111, 191 and 79 km respectively in 2011. and road networks from Open series Topographical The total length of surfaced and unsurfaced roads are Sheets (2015), images and ASTER DEM from USGS 3696.54 km and 1652.51 km approximately in 2013–14. Earth Explorer, Soil map from National Bureau of Soil The characteristics of both plain and mountainous top- Survey & Land use Planning and Geological map from ographies exit in this study area. The altitude of the Geological Survey of India. The summery of the proce- study area ranges from 15 m to 3602 m from mean sea dues followed is depicted in the flowchart (Fig. 2). level and the slope from 0° to 80° approximately. The major portion of the study area is covered with Triassic Software used rocks. The soil character of the study area is varied from To predict the landslide potential areas, the thematic one region to another region. The study area receives layers of selected geo-environmental parameters namely huge amount of rainfall in the monsoon season. The rainfall, slope, elevation, aspect, geology, soil texture, dis- average annual rainfall of the study area is about 3051 tance from lineament, distance from river, distance from mm. The major rivers of the study area are Mahananda, road, natural differential vegetation index (NDVI) and Tista, Mechi, Balason, Jaldhaka, Rammam and Rangit, TWI have been prepared with the help of ArcGIS 10.3.1, which are flowing from northern part. The district has ENVI 4.7, GEOMATICA and the mathematical calcula- some reputed eco-tourism sites and pilgrimage sites tions have been done with the help of SPSS softwares. namely like Tiger Hill, Rock Garden, Mahakal Temple, Dhirdham Temple, Batasia Loop, Ghoom Monastery and Happy Vally Tea Garden, etc. Historically, Tea Plan- Preparation of landslide inventory map tation and Cinchona are the main sources of livelihood The landslide inventory map is the vital part for analyz- in the Darjeeling district. The community of the study ing the landslide susceptibility, hazard and risk (Guzzetti area depends on the horticulture, tourism, and forestry. et al. 2006). Pradhan and Lee (2009) and Pourghasemi et Siliguri is the major town of the study area which is fa- al. (2012) prepared the landslide inventory map for iden- miliar with ‘gate way’ of eastern India. tifying the landslide hazards zones. Van Westen et al. (2000) remarked that the different data such as field in- Materials and methodology vestigations, historical landslide events and satellite Data sources image analysis can be used to prepare the landslide in- For the fulfillment of the present work, various import- ventory map (Fig. 1b). In the present study, 326 land- ant data have been collected from different sources e.g. slides have been identified from Google earth imagery rainfall and temperature data from Indian Meteoro- and multiple field visits. The landslide inventory map logical Department, population data from District has been prepared in the GIS environment for Fig. 2 Flowchart of methodology used in this study Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 5 of 18 calculating the LNRF to predict the potentials landslide annual rainfall data of last 5 years (Fig. 3a). The slope susceptibility areas in this district. map of the study area has been prepared from ASTER DEM with the help of 3D analyst tool of ArcGIS 10.3.1 Selection and Muliti-collinearity analysis of landslide software (Fig. 3b). In case of elevation and aspect, these causative factors data layers have been prepared from ASTER DEM with There is a variety of geo-environmental parameters that the help of 3D analyst tool of surface in GIS environ- have been used by the various researchers for producing ment (Fig. 3c, d). the landslide susceptibility map. Still there is no standard The geology map has been prepared with the help of guideline for selecting the landslide predictors. In this digitations process from referenced geological map research 12 landslide causative factors i.e. rainfall, slope, which has been collected from Geological Survey of aspect, altitude, geology, soil texture, distance from river, India (Fig. 3e). distance from lineament, distance from road, land use/ The thematic layer of soil texture has been generated land cover, NDVI and TWI have been selected based on with the help of the digitations process form referenced the multi-collinearity analysis for mapping the landslide soil map which has been collected from National Bureau susceptibility. In the landslide susceptibility analysis of Soil Survey and Land Use Department (Fig. 3f). The aforementioned causative factors are widely used (Dou distance from river and distance from road maps have et al. 2019; Arnone et al. 2016; Tien Bui et al. 2012). been prepared with the help of Euclidian distance buffer- Multi-collinearity analysis is an important way to ver- ing tool in GIS environment (Fig. 3g, h). The lineament ify the effectiveness of the landslide conditioning factors of the study area has been extracted from Landsat 8 OLI (Saha 2017). For the present study, the collinearity test (Optical land Imager) image with the help of the ENVI of landslide determining factors has been done in the 4.7 and GEOMATICA softwares. The distance from lin- SPSS software. A tolerance of less than 0.10 and and eament map has been prepared with the help of Euclid- variance inflation factors (VIF) 10 or above indicates ian distance buffering tool of ArcGIS software 10.3.1 multicollinearity problems (Dormann et al. 2012; Wang (Fig. 3i). The land use/ land cover (LULC) map has been et al. 2008). In the present study, the values of tolerances prepared from Landsat 8 OLI imagery with the help of and VIF of all the selected parameters are less than 10% maximum likelihood classification method (Fig. 3j). The (Table 1). So, there is no collinearity problem among the normalized differential vegetation index (NDVI) has selected landslide determining factors. been calculated from Landsat 8OLI image with the help of image analysis tool in ArcGIS 10.3.1 software (Fig. 3k). Preparation of thematic layers of selected parameters The thematic layer of TWI has been prepared from The average annual rainfall data of last 5 years, since ASTER DEM imagery in GIS environment using Eq. 1 2012 to 2017 have been collected from Indian Meteoro- (Fig. 3l) which was suggested by Moore et al. (1991). logical Department. The thematic layer of rainfall has been prepared with the help of the interpolation method of IDW in GIS environment based on the average TWI ¼ In ð1Þ tanβ Table 1 Collinearity statistics of landslide determining factors Where, TWI = topographic wetness index, α is cumu- Sl. Parameters Collinearity statistics lative upslope area draining through a point (per unit No Tolerance VIF contour length), β is the slope gradient (in degree). The 1 Rainfall 0.494 2.024 minimum, maximum, categorical classification and 2 Elevation 0.638 1.566 methods of the selected geo-environmental conditional factors have been done in GIS environment and men- 3 Slope 0.894 1.118 tioned in the following Table 2. 4 Aspect 0.742 1.348 5 Geology 0.743 1.346 Fuzzy method 6 Soil 0.858 1.166 For the present study, the Fuzzy method has been as- 7 Distance from River 0.635 1.575 sembled with AHP and LNRF methods. The fuzzy maps 8 Distance from Lineament 0.818 1.223 of selected parameters have been prepared with the help of membership function (MF) tool in GIS environment. 9 Distance from Road 0.527 1.898 The membership function (MF) values range between 0 10 LULC 0.833 1.201 and 1 (Zadeh 1965). The value 0 means that x is not a 11 NDVI 0.804 1.243 member of the fuzzy set, while the value 1 means that x 12 TWI 0.884 1.132 is a full member of the fuzzy set. A sample of fuzzy set Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 6 of 18 Fig. 3 Landslide conditioning factor maps used in this study: a Rainfall, b Slope, c Slope Elevation, d Aspect, e Geology, f Soil Map, g Distance from River, h Distance from Lineament, i Distance from Road, j Land Use/ Land Cover, k NDVI, l TWI Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 7 of 18 Table 2 Overview of selected parameters used for landslide susceptibility mapping Parameters Ranges Classes Methods Min Max Rainfall (mm) 1877 2334 1. 1877.38–1991.9 Natural Break 2. 1991.97–2090.54 3. 2090.45–2167.44 4. 2167.44–2239.06 5. 2239.06–2333.96 Slope (Degree) 0 79.2 1. 0–9.32 Natural Break 2. 9.32–18.64 3. 18.44–27.34 4. 27.34–36.66 5. 5. 36.66–79.23 Altitude (m) 15 3602 1. 15–422.93 Natural Break 2. 422.93–985.6 3. 985.6–1576.4 4. 1576.4–2279.73 5. 5. 2279.73–3602 Aspect –– 1. Flat (−1) Equal interval 2. North (0–22.5) 3. Northeast (22.5–67.5) 4. East (67.5–112.5) 5. Southeast (112.5–157.5) 6. South (157.5–202.5) 7. Southwest (202.5–247.5) 8. West (247.5–292.5) 9. Northwest (292.5–337.5) 10. north(337.5–360) Geology –– 1. Triassic lithological units 2. Cenozoic 3. Pliocene-Pleistocene, 4. Holocene 5. Middle-upper Pleistocene Soil texture –– 1. Gravelly loamy, soil texture classes 2. Fine loamy - Coarse Loamy 3. Gravelly loamy Skeletol 4. Gravelly Loam - Coarse Loamy 5. Coarse Loamy Distance from River (km) 0 4.33 1. 0–0.42 Natural Break 2. 0.42–1.10 3. 1.10–1.66 4. 1.66–2.26 5. 2.26–4.33 Distance from Lineament (km) 0 10.1 1. 0–1.54 Natural Break 2. 1.54–2.85 3. 2.85–4.20, 4 4. 4.20–5.75 5. 5.75–10.12 Distance from Road (km) 0 16.5 1. 0–1.74 Natural Break 2. 1.74–3.94 Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 8 of 18 Table 2 Overview of selected parameters used for landslide susceptibility mapping (Continued) Parameters Ranges Classes Methods Min Max 3. 3.94–6.72 4. 6.72–10.22 5. 10.22–16.49 Land use/Land cover –– 1. Water bodies supervised classification 2. Vegetation 3. Fallow land 4. Agricultural land 5. Settlement NDVI −0.07 0.49 1. −0.07-0.12 Natural Break 2. 0.12–0.17 3. 0.17–0.23 4. 0.23–0.29 5. 0.29–0.49 TWI 1.95 18.9 1. 1.95–7.37 Natural Break 2. 7.37–8.53 3. 8.53–9.76 4. 9.76–11.70 5. 11.70–18.91 is shown in the following equation (Mcbratney and GAMMA. In the present study, gamma operator has Odeh 1997). been used for combining membership functions. Fuzzy gamma operation has been calculated using eqs. A ¼ x; μ ðÞ x for each xεX ð2Þ (5). Where, μ is the MF (membership of x in fuzzy set A) so that: 1−γ μ ¼ðÞ μ k : μ : ð5Þ γ sum product If x does not belong to A then μ =0. If x belongs completely to A then μ =1. If x x belongs in a certain degree to A then The exponent γ, which is a number from < 0, 1 > inter- val, allows optimization of the membership combination. 0 < μ ðÞ x < 1 Setting it to the extremes of the interval give either fuzzy According to Eq. 3 MF was used for rainfall, elevation, algebraic sum (γ = 1) or fuzzy algebraic product (γ = 0). aspect, slope, NDVI, TWI [8]. To perform fuzzy gamma operation, several gamma op- 8 9 erator (k) values, i.e. 0.50, 0.70, 0.80, 0.90, 0.95, and 0x≤a < = 0.975 are there. In the present study, fuzzy gamma oper- μ ðÞ x ¼ fxðÞ ¼ x ‐a=b‐aa≻x≺b ð3Þ ator (k) value of 0.975 has been applied for producing : ; 1x≥b the landslide susceptibility map. Where x is the input data and a, b are the limit values. For geology, soil texture, distance from river, distance Landslide numerical risk factor (LNRF) model from lineament, LULC and distance from road the fol- Landslide Numerical Risk Factor (LNRF) model is an lowing MF has been used [4]. important method for identifying the landslide hazard 8 9 zones especially in the mountain region (Gupta and <0x≤a = Joshi 1990). According to this model, the LNRF > 1 value μ ðÞ x ¼ fxðÞ ¼ b ‐x=b‐aa≻x≺b ð4Þ : ; indicates that the geo-environmental factors have high 1x≥b responsibility for the occurrences of landslide. The LNRF < 1 values represent that the geo-environmental Fuzzy gamma operators factors are more stable and have less effect in landslides Several fuzzy operators exist for combining membership occurrences (Gupta and Joshi 1990). The LNRF has been functions such as AND, OR, SUM, PRODUCT and calculated through Eq. (6): Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 9 of 18 Table 3 Computation of LNRF model on landslide determining Factors Parameters Area Landslides Area LNRF Fuzzy membership In Hectare In Percentage In Hectare In Percentage values using LNRF model Rainfall (mm) Very Low (1877.38–1991.97) 27659.36 8.78 0 0 0 0 Low (1991.97–2090.54) 24856.96 7.89 0 0 0 0 Moderate (2090.45–2167.44) 80967.44 25.71 31.56 7.32 0.37 0.11 High (2167.44–2239.06) 114335.78 36.31 294.58 68.29 3.41 1.00 Very High (2239.06–2333.96) 67080.46 21.3 105.21 24.39 1.22 0.36 Slope (Degree) Very Low (0–9.32) 101505.25 32.23 14.5 3.36 0.17 0.11 Low (9.32–18.64) 58545.55 18.59 52.56 12.18 0.61 0.39 Medium (18.44–27.34) 70795.02 22.48 103.31 23.95 1.2 0.76 High (27.34–36.66) 57571.95 18.28 135.93 31.51 1.58 1.00 Very High (36.66–79.23) 26482.23 8.41 125.06 28.99 1.45 0.92 Altitude (m) Very Low (15–422.93) 115880.61 36.8 35.17 8.15 0.41 0.16 Low (422.93–985.6) 71784.89 22.8 224.01 51.93 2.6 1.00 Medium (985.6–1576.4) 63320.09 20.11 81.46 18.88 0.94 0.36 High (1576.4–2279.73) 44471.47 14.12 72.2 16.74 0.84 0.32 Very High(2279.73–3602) 19442.94 6.17 18.51 4.29 0.21 0.08 Aspect Flat (−1) 163.37 0.05 0 0 0 0 North (0–22.5) 20317.92 6.45 7.41 1.72 0.09 0.07 North-East (22.5–67.5) 39614.59 12.58 25.92 6.01 0.3 0.22 East (67.5–112.5) 39009.88 12.39 59.24 13.73 0.69 0.51 South-East (112.5–157.5) 44774.25 14.22 87.01 20.17 1.01 0.75 South (157.5–202.5) 45083.47 14.32 116.63 27.04 1.35 1.00 South-West (202.5–247.5) 39204.16 12.45 70.35 16.31 0.82 0.61 West (247.5–292.5) 32441.32 10.3 49.98 11.59 0.58 0.43 North-West (292.5–337.5) 35974.84 11.42 12.96 3 0.15 0.11 North (337.5–360) 18316.21 5.82 1.85 0.43 0.02 0.01 Geology Triassic 166113.98 52.75 302.86 70.21 3.51 1 Cenozoic 23200.35 7.37 82.6 19.15 0.96 0.27 Pliocene-Pleistocene 11258.98 3.58 45.89 10.64 0.53 0.15 Holocene 58152.53 18.47 0 0 0 0 Middle-upper Pleistocene 56174.15 17.84 0 0 0 0 Soil texture Gravelly loamy 23549.04 7.48 68.11 15.79 0.79 0.46 Fine loamy Coarse Loamy 126713.02 40.24 113.51 26.32 1.32 0.77 Gravelly loamy, Loamy Skeletal 38586.66 12.25 147.57 34.21 1.71 1 Gravelly Loamy Coarse Loamy 120449.25 38.25 90.81 21.05 1.05 0.61 Coarse Loamy 5602.03 1.78 11.35 2.63 0.13 0.08 Distance from River (km) 0–0.42 km 99542.45 31.61 73.65 17.07 0.85 0.37 0.42–1.10 km 110752.04 35.17 199.89 46.34 2.32 1 1.10–1.66 km 64340.57 20.43 115.73 26.83 1.34 0.58 1.66–2.26 km 32920.37 10.45 42.08 9.76 0.49 0.21 Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 10 of 18 Table 3 Computation of LNRF model on landslide determining Factors (Continued) Parameters Area Landslides Area LNRF Fuzzy membership In Hectare In Percentage In Hectare In Percentage values using LNRF model 2.26–4.33 km 7344.58 2.33 0 0 0 0 Distance from Lineament (km) 0–1.54 km 81563.6 25.9 110.6 25.64 1.28 0.77 1.54–2.85 km 95262.67 30.25 44.24 10.26 0.51 0.31 2.85–4.20 km 77028.98 24.46 143.78 33.33 1.67 1.00 4.20–5.75 km 44870.1 14.25 99.54 23.08 1.15 0.69 5.75–10.12 km 16174.66 5.14 33.18 7.69 0.38 0.23 Distance from Road (km) 0–1.74 km 140275.58 44.55 88.48 20.51 1.03 0.67 1.74–3.94 km 84741.4 26.91 44.24 10.26 0.51 0.33 3.94–6.72 km 50166.43 15.93 66.36 15.38 0.77 0.50 6.72–10.22 km 27267.21 8.66 132.72 30.77 1.54 1.00 10.22–16.49 12449.37 3.95 99.54 23.08 1.15 0.75 Land use/Land cover Water bodies 3466.31 1.1 39.87 9.24 0.46 0.16 Vegetation 227240.32 72.16 255.55 59.24 2.96 1.00 Fallow land 14437.29 4.58 29 6.72 0.34 0.11 Agricultural Land 65442.75 20.78 106.93 24.79 1.24 0.42 Settlement 4313.33 1.37 0 0 0 0.00 NDVI Very Low (−0.07–0.12) 37936.34 12.05 114.18 26.47 1.32 0.91 Low (0.12–0.17) 83384.9 26.48 125.06 28.99 1.45 1.00 Medium (0.17–0.23) 85506.39 27.15 96.06 22.27 1.11 0.77 High (0.23–0.29) 70015.86 22.23 65.25 15.13 0.76 0.52 Very High (0.29–0.49) 38056.51 12.09 30.81 7.14 0.36 0.25 TWI Very Low (1.95–7.37) 49986.44 15.87 63.12 14.63 0.73 0.35 Low (7.37–8.53) 113766.56 36.13 178.85 41.46 2.07 1.00 Medium (8.53–9.76) 93346.92 29.64 115.73 26.83 1.34 0.65 High (9.76–11.70) 46923.5 14.9 73.65 17.07 0.85 0.41 Very High (11.70–18.91) 10876.58 3.45 0 0 0 0 Gamma operators (k) values of 0.975 in GIS LNRF ¼ ð6Þ environment. Analytic hierarchy process (AHP) Where A: landslide area in every unit, E: mean area of Analytic hierarchy process is an important multi-criteria landslide in the whole unit. decision analysis (MCDA) method which can be applied The sub-parameters wise LNRF values have been cal- for assigning the weights to the individual parameter culated (Table 3). In the present study, fuzzy member- (Saaty and Vargas 1998). AHP method is a pair wise ship values have been allotted based on LNRF. To comparison matrix. When the matrix is formed, the transform the LNRF values to Fuzzy membership values consistency ratio (CR) value ranges from 0 to 1 (Saaty each sub-class has been divided by the maximum value 1980, 1990, 1994). To identify the potentiality index, of LNRF of individual parameter. The membership value general linear combination method can be performed ranges from 0 to 1. The fuzzy membership values using with the help of AHP method (Malczewski 1999). The LNRF model have been converted into a single layer to pair wise matrix has been formed with the help of the chalk out the landslide susceptibility zone using Fuzzy Saaty’s(1980) fundamental scale (Table 4). In the present Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 11 of 18 Table 4 Fundamental scale of Saaty’s(1980) upper Pleistocene (Fig. 3e). Triassic geological segment with 52.75% area is encompassed with 70.21% landslide Scale Description area. According to the LNRF model using fuzzy logic, 1 Equal Importance the Triassic geological segment has attained the max- 3 Moderate Importance imum LNRF and fuzzy membership value. The other 5 Strong Importance sub-layers of geological segment are indicating the less 7 Very Strong Importance probability of landslide occurrence. Pedagogically, the 9 Absolute Strong Importance region is composed with several soil textural classes 2,4,6,8 Intermediate values between the two adjacent judgments namely Gravelly loamy, Fine Loamy to Coarse Loamy, Gravelly Loamy to Loamy Skeletal, Gravelly Loamy to case, the weight of parameters calculated based on AHP Coarse Loamy and Coarse Loamy (Fig. 3f). The 34.21% method has been combined with fuzzy logic to prepare landslide areas are concentrated in Gravelly Loamy to landslide susceptibility map of the study area. Loamy skeletal soil texture class (Table 3) and the LNRF and fuzzy membership values are 1.71 and 1, which are Results representing high risk of landslides. The distance from Application of fuzzy-LNRF model river map (Fig. 3g) has been classified into five classes The spatial distribution of average annual rainfall of the such as 0–0.42 km, 0.42 km − 1.10 km, 1.10 km − 1.66 study area ranges from 1877.38 mm to 2333.96 mm km, 1.66 km − 2.26 km and 2.26 km to 4.33 km. The (Fig. 3a) respectively. The rainfall map has been catego- classes 0.42 km − 1.10 km and 1.10 km − 1.66 km of river rized into five classes such as very low (1877 mm–1991 buffering are covered with 46.34% and 26.83% landslide mm), low (1991 mm–2090 mm), moderate (2090 mm– area (Table 3). The LNRF values of 0.42 km − 1.10 km 2167 mm), high (2167 mm–2239 mm) and very high and 1.10 km − 1.66 km sub-layers are 2.32 and 1.34 as (2239 mm–2333 mm) respectively. The high sub-class of well as fuzzy membership values of the same layers are 1 rainfall with 36.31% area is covered the 68.29% land- and 0.58 which are indicating high landslide susceptibil- slides of the study area (Table 3). The LNRF values have ity. The lineament distance map (Fig. 3h) has been clas- been calculated and converted into fuzzy membership sified into five buffer zones such as 0–1.54 km, 1.54 km (FM) value. Here, high rainfall sub-class has attained the − 2.85 km, 2.85 km − 4.20 km, 4.20 km − 5.75 km and highest FM value i.e. 1, indicating the high risk of land- 5.75 km − 10.12 km. The 2.85–4.20 km buffer zone has slide than other sub-class of rainfall (Table 3). The higher LNRF value (Table 3) than the other buffer zones. spatially the slope of the study area ranges from 0 to Road building activity in mountain areas is regarded as 79.23° (Fig. 3b). It has been classified into five categories an infrastructure improvement, which may ground detri- such as very low (0°-9.32°), low (9.32°-18.64°), medium mental effects on slope steadiness; therefore, it can be (18.44° -27.34°), high (27.34° – 36.66°) and very high helpful for delineating the prone areas to landslide oc- (36.66°-79.23°) classes based natural break classification currence. The buffer layer 6.72 km to 10.22 km. distance method in GIS environment. The high and very high area from road is covered 30.77% landslide area (Table 3). slope classes are covered with 31% and 28% landslides This road buffer class has attained the fuzzy membership area. The fuzzy membership values of these sub-classes value 1. Other sub-layers of distance from road are are nearer to 1, representing as high landslides risk showing the low to medium landslide susceptibility. The areas. The altitude of the study area ranges from 15 m to five type of land use classes have been identified in this 3602 m (Fig. 3c) respectively. The altitudinal map has study area namely water bodies, natural vegetation, sand, been categorized into five classes such as very low (15 m agricultural land and settlement with the help maximum to 422.93 m), low (422.93 m to 985.6 m), medium (985.6 likelihood classification method in GIS environment m to 1576.4 m), high (1576.4 m to 2279.73 m) and very (Fig. 3j). The 72.16% area is covered with natural vegetation high (2279 m–3602 m). The low elevation class is cov- area in this district. The agricultural land and settlement ered with 51.93% landslide area (Table 3). The fuzzy areas are indicating less probability for the occurrence of membership value of 422.93 m–985.6 m elevation range landslide. Natural differential vegetation index (NDVI) is is 1, representing higher landslide susceptibility than one of the important factors of environment. The value of other sub-layers. Aspect of the study area has been clas- the NDVI of the study area ranges from − 0.07 to 0.49 sified into several categories such as flat, north, north- (Fig. 3k) respectively. The NDVI map of the study area has east, east, southeast, south, southwest, west and been classified into five classes such as very low (− 0.07– northwest. South sub-layer with 14.32% area is covered 0.12), low (0.12–0.17), medium (0.17–0.23), high (0.23– 27.04% landslide area. Geologically, the study area is 0.29) and very high (0.29–0.49). According toLNRFmodel, composed of five geological segments namely Triassic, the low sub-class of NDVI is attained the highest LNRF Cenozoic, Pliocene-Pleistocene, Holocene and Middle- and fuzzy membership values i.e. 1.45 and 1 which are also Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 12 of 18 Fig. 4 Landside Susceptibility Maps (a) based on Fuzzy-LNRF model & (b) based on Fuzzy-AHP model indicating the high landslide susceptibility. The spatial dis- 7.37), low (7.37–8.53), medium (8.53–9.76), high (9.76– tribution of TWI of the study area ranges from 1.95 to 11.70) and very high (11.70–18.91) (Table 4). The low and 18.91(Fig. 3l) respectively. TWI map of the study area has medium TWI classes area representing the high risk for been classified into five classes namely very low (1.95 to landslide than other sub-layers. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 13 of 18 The landslide susceptibility map (LSM) based on fuzzy raster calculator of spatial analyst tool in Arc- membership values has been prepared with the help of GIS10.3.1 software. The landslide susceptibility map fuzzy gamma operators (k) value of 0.975 and shown in (Fig. 4b) has been classified into three categories like Fig. 4a. It has been classified into three classes namely low, medium and high landslide susceptibility zones low, medium and high landslide susceptibility zones. covering an area of 37.88%, 39.29% and 22.83% re- Out of the total district 34% area is highly susceptible to spectively (Table 6). landslide (Table 6). Some landslides identifed from google earth imagery and directly from the field are mentioned below- (Fig. 5) Application of fuzzy-AHP model In the present work the Fuzzy set theory with AHP Validation is considered as the multi-criteria decision approach The landslide susceptibility maps of Darjeeling district, for identifying the landslide susceptibility zones. The prepared by Fuzzy-LNRF and Fuzzy-AHP models have integration of fuzzy set theory and AHP can be pro- been validated through the ROC curve and RMSE. vided a good and reliable technique for zoning land- Model based work may be possible to evaluate and jus- slide susceptibility. AHP is a single process which tify easily with the help of ROC curve (Chung and Fab- helps to determine the weight of different factors bri 2003). The ROC curve is drawn by X and Y axis. The based on the expert’s opinion and knowledge. For X axis represents the false positive rate (1-specificity) the present work, weights have been assigned to and the Y axis represents the true positive rate (sensitiv- landslide determining factors (Table 5)with the help ity) (Negnevitsky 2002). Mallick et al. (2018) and Feiziza- of AHP method. The highest weight has been deh et al. (2014) have applied the ROC curve for the assigned to rainfall (0.180) and followed by slope validation of the landslide susceptibility zone. In the (0.162), distance from the river (0.108), soil texture present study, 98 landslide patches have been selected (0.095) and elevation (0.086) for mapping the land- among the 326 landslide patches for validating the slide probability. Thereafter, the linear membership landslide susceptibility maps. The area under curve function (MF) has been used to prepare the fuzzy (AUC) can be drawn for the landslide susceptibility map of the selected parameters for landslide suscep- zones with the help of the Table 7. According to the tibility mapping. The value of fuzzy membership resultsof ROC curve, theareaunder curve (AUC) ranges from 0 to 1. Therefore, values of prepared values of landslide susceptibility maps, prepared by fuzzy maps of all selected parameters must be the Fuzzy-LNRF and Fuzzy-AHP models are 91% and ranged from 0 to 1. The weights of parameters, cal- 90% which are indicating the excellent potentiality of culated by AHP method, have been integrated with these models for landslide susceptibility mapping fuzzy maps of selected parameters to generate a sin- (Fig. 6a, b). The LSM of both models have also been gle layer of landslide susceptibility with the help of Table 5 Parameters wise weights, matrix and consistency ratio using AHP Parameters Rainfall Slope Elevation Aspect Geology Soil Distance Distance from Distance LULC NDVI TWI Weights Texture from River lineament from Road Rainfall 1 0.180 Slope 0.5 1 0.162 Elevation 0.33 0.2 1 0.086 Aspect 0.2 0.5 1 1 0.084 Geology 0.5 0.5 0.5 2 1 0.080 Soil Texture 0.5 0.2 2 1 0.5 1 0.095 Distance from 0.2 0.2 0.5 0.5 0.2 0.5 1 0.108 River Distance from 0.2 0.2 0.5 0.5 0.5 0.2 0.5 1 0.046 lineament Distance from 0.5 0.2 0.5 1 0.5 0.5 2 2 1 0.062 Road LULC 0.17 0.2 0.5 0.2 0.5 0.5 0.5 0.5 0.5 1 0.051 NDVI 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 2 1 0.025 TWI 0.14 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.5 2 1 1 0.021 Consistency Ratio = 0.078 Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 14 of 18 Table 6 Areal distributions of LSM based on LNRF and Fuzzy- validated through RMSE method. The RMSE is calcu- AHP models lated using the Eq.7. Landslide LNRF model Fuzzy-AHP model vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi susceptibility Area in sq. km % of Area Area in sq.km % of Area uX zones ðÞ P‐A Low 950.15 30.17 1192.97 37.88 i¼1 RMSE ¼ ð7Þ Medium 1114.40 35.39 1237.17 39.29 High 1084.46 34.44 718.86 22.83 Where, N is the numbers of samples, A is the ob- served values and P is the predicted values. Can et al. (2005) has considered the RMSE value of 0.5, as the cut- off value. The values of RMSE > 0.5 and < 0.5, indicate the bad predictive and good predictive model. In the Fig. 5 Picture showing the landslide sites – from the google earth image. a Lish catchment (26°57′N, 88°30′17“E), b Nimbong Khasmahal (26°58’04”N, 88°34′16“E), c Sittong (26°52’N, 88°22’30”E) and from the field, d Near Kurseong town (26°55′46.32″N, 88°19′50.38″E), e Near Darjeeling town (26°53′13.37″N, 88°17′46.45″E), f Nathula road (26°54′43.14″N, 88°18′21.10″E) Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 15 of 18 Table 7 Characteristics of AUC of ROC curve (Yesilnacar 2005) In this work, Fuzzy-AHP and Fuzzy-LNRF, the two ef- ficient and easily operate ensemble models have been AUC Values Characters used for delineating landslide susceptibility and com- 0.6–0.5 Average pared with them. In these two ensemble models Fuzzy- 0.7–0.6 Good AHP and Fuzzy-LNRF, input process, calculations and 0.8–0.9 Very Good output process are very simple and can be readily under- 1–0.9 Excellent stood. The geo-environmental factors like slope, aspect, altitude, geology, soil texture, distance from river, dis- present study, the results of RMSE are justified Fuzzy- tance from lineament, distance from road, rainfall, land LNRF (RMSE = 0.14) and Fuzzy-AHP (RMSE = 0.18) use/ land cover, NDVI and TWI have been considered models as good predicative models for landslides suscep- for the determination of the landslide susceptible area of tibility mapping. the Darjeeling district. According to the Fuzzy-LNRE and Fuzzy-AHP maps 34.44% and 22.83% areas (Fig. 7a and b) of the district fall under the high susceptibility of Discussions landslide. Based on these findings, it can be acknowl- In the landslide prone areas appropriate methods of edged that the high susceptibility zone delineated by the landslide susceptibility mapping play significant task in Fuzzy-LNRF method is forecasting greater percentage of providing a proper approach to authorities and decision the landslide area. Additionally, the reliability of the re- makers. Very fundamental information regarding the sults of two susceptibility maps has been validated based landslide conditioning factors can be acquired from the on the known landslide dataset by employing the area landslide susceptibility mapping (LSM) and it can be an under the curve (AUC) of the receiver operating charac- essential way in hazard mitigation measures and man- teristics (ROC) and RMSE. A landslide inventory map agement. There is a large number of weight combining has been prepared considering the multiple field works methods for preparing the landslide susceptibility map. and Google Earth Images. Out of the 326 landslides 246 The results of some multi-criteria decision analysis such (70%) locations have been used for training data and as AHP, LNRF, Fuzzy logic, Artificial Neural Network remaining 80 (30%) have been used as testing data. Re- Support Vector Machine, Logistic Regression and Fre- spective AUC values of 91% and 90% for fuzzy-LRNF quency Ratio are a little bit varied from region to region. and fuzzy-AHP proved that the map produced by the Some researchers such as, Malik et al. (2016), fuzzy-LRNF model looks like having a better accuracy Mohammadi et al. (2014), Shadfar et al. (2011)and than the fuzzy-AHP model. This finding is helpful for Gupta and Joshi (1990)pointed outthatLNRF emergency situation because time is very significant in method is suitable method in landslide susceptibility hazard studies. It could be assessed that the models ap- mapping. On the other hand, Abedini and Tulabi plied have relatively similar accuracies. Methods such as (2018) was indicated that LNRF is not suitable LNRF (Gupta and Joshi 1990 & Abedini and Tulabi method like frequency ratio and AHP. 2018) and fuzzy-AHP (Roodposhti et al. 2014) were used Fig. 6 Validation of LSMs by ROC curve a for Fuzzy-LNRF model b for Fuzzy-AHP model Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 16 of 18 Fig. 7 Graph showing the distribution of area under different susceptibility classes a Percentage of area and b Area in sq.km in mapping the landslide susceptibility. Compared to LSM: Landslide susceptible map; LULC: Land use/Land cover; MCDA: Multi- criteria decision approach; MF: Membership function; NDVI: Normalized these previous studies the used methods i.e. fuzzy-LNRF differential vegetation index; OLI: Operational land imager; RMSE: Root mean and Fuzzy-AHP in this study are showing better per- square error; ROC: Receiver operating characteristic; TM: Thematic mapper; formance in preparing the LSM. Taking into account the TWI: Topographical wetness index; VIF: Variance inflation factors performances, we suggest that both Fuzzy-LNRF and Acknowledgments Fuzzy-AHP models can be used in landslide studies, as Authors would like to thanks the inhabitants of Darjeeling District because they are capable of producing flawless and stable land- they have helped a lot during our field visit. The authors would like to give special thanks to two anonymous reviewers for their constructive and useful slide susceptible maps for mitigating the risk and man- comments during the review process. Also we would like to cordially thank agement planning. There are little differences in result the Mr. Arnab Chatterjee, Assistant Professor, Department of English, between the Fuzzy-LNRF and Fuzzy-AHP derived sus- Harishchandrapur College, Pipla, Malda for correcting the grammers and language. At last, authors would like to acknowledge all of the agencies and ceptibility maps. Moreover, the Fuzzy-LNRF model is individuals specially, Survey of India (SOI), Geological Survey of India (GSI) promising and sufficient to be advised as a method to and USGS for obtaining the maps and data required for the study. prepare landslide susceptibility map at regional scale. Authors’ contributions Both authors wrote the manuscript and developed the research Conclusions methodology. Both authors also read and approved the final manuscript. For the prevention of human lives and property, a short and long-term solution is necessary for mitigating the Funding No funding was received for this work. landslide risk in this region. At present day, landslide is to be considered the most serious natural hazards in the Availability of data and materials Darjeeling district. The study has been adopted the suit- Rainfall and Temperature were received from Indian metrological able multi-criteria decision making approaches like Department. Population data was collected from District Statistical Handbook, Census of India 2011. Landsat images and DEM will freely be Fuzzy-AHP and Fuzzy-LNRF to outline the landslide availed from https:// earthexplorer.usgs.gov/ website. Soil map was collected susceptibility zones. The landslide susceptibility maps of from National Bureau of Soil Survey and Land Use Planning. Geology map both models have been categorized into three classes was taken from Geological Survey of India. such as low, moderate and high landslide susceptibility Competing interests zones. The high landslide susceptibility zone has been The authors declare that they have no competing interests. found in the middle and northern portions of the study area because of the presence of fragile soil, high concen- Received: 23 November 2018 Accepted: 15 July 2019 tration of drainage, frequent heavy rainfall and sloppy land. Among these two ensemble models, Fuzzy-LNRF References is showing better acceptability than the other model. So, Abedini, M., and S. Tulabi. 2018. Assessing LNRF, FR, and AHP models in landslide susceptibility mapping index: A comparative study of Nojian watershed in the study has been done by these two ensemble models Lorestan province, Iran. Environmental Earth Sciences 77 (11): 405. help to understand the landslide hazards problem of an Aleotti, P., and R. Chowdhury. 1999. Landslide hazard assessment: Summary area. The study also provides the essential information review and new perspectives. Bulletin of Engineering Geology and the Environment 58 (1): 21–44. https://doi.org/10.1007/s100640050066. to the planner, government, local people and researchers Arnone, E., A. Francipane, A. Scarbaci, C. Puglisi, and L.V. Noto. 2016. Effect of to take suitable steps for the reduction and the mitiga- raster resolution and polygon-conversion algorithm on landslide tion of the landslides problems in the Darjeeling district. susceptibility mapping. Environmental Modelling & Software 84: 467–481 https://doi.org/10.1016/j.envsoft.2016.07.016. Abbreviations Ayalew, L., and H. Yamagishi. 2005. The application of GIS-based logistic AHP: Analytical hierarchical process; ASTER: Advanced space borne thermal regression for landslide susceptibility mapping in the Kakuda-Yahiko emission and reflection radiometer; DEM: Digital elevation map; Mountains, Central Japan. Geomorphology 65 (1–2): 15–31. https://doi.org/1 GIS: Geographical information system; LNRF: Landslide Numerical Risk Factor; 0.1016/j.geomorph.2004.06.010. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 17 of 18 Bhandari, R.K. 2004. Landslide hazard zonation: Some thoughts. In Coping with Mcbratney, A., and I.O. Odeh. 1997. Application of fuzzy sets in soil science: Fuzzy natural hazards: Indian context, ed. K.S. Valdiya, 134–152. Orient Longman: logic, fuzzy measurements and fuzzy decisions. Geoderma 77 (2–4): 85–113. Hyderabad. https://doi.org/10.1016/s0016-7061(97)00017-7. Bouma, N., and A. Imeson. 2000. Investigation of relationships between measured Mirnazari, J., S. Khezri, and H. Shahabi. 2015. Assessment and zoning of landslide field indicators and erosion processes on badland surfaces at Petrer, Spain. hazard using AHP model and fuzzy logic operators in Posht tang watershed Catena 40 (2): 147–171. https://doi.org/10.1016/s0341-8162(99)00046-6. of Sar pole Zahab (Kermanshah province). Geography and Development 12 (37): 53–70. Budimir, M.E.A., P.M. Atkinson, and H.G. Lewis. 2015. A systematic review of Mohammadi, A.T., A. Irani, and J. Sorur. 2014. The preparation of landslide map landslide probability mapping using logistic regression. Landslides 12 (3): by landslide numerical risk factor (LNRF) model and geographic information 419–436. Can, T., H.A. Nefeslioglu, C. Gokceoglu, H. Sonmez, and T.Y. Duman. 2005. system (GIS). The Egyptian Journal of Remote Sensing and Space Science 17 (2): 159–170. Susceptibility assessments of shallow earthflows triggered by heavy rainfall at three catchments by logistic regression analyses. Geomorphology 72 (1–4): Mokarram, M., and A.R. Zarei. 2018. Landslide susceptibility mapping using fuzzy- 250–271. https://doi.org/10.1016/j.geomorph.2005.05.011. AHP. Geotechnical and Geological Engineering. https://doi.org/10.1007/s10706- Chung, C., and A.G. Fabbri. 2003. Validation of spatial prediction models for 018-0583-y. landslide Hazard mapping. Natural Hazards 30 (3): 451–472. https://doi.org/1 Moore, I.D., R.B. Grayson, and A.R. Ladson. 1991. Digital terrain modelling: A 0.1023/b:nhaz.0000007172.62651.2b. review of hydrological, geomorphological, and biological applications. Malley, L.O. 1999. Bengal district gazetteer: Darjeeling, Government of West Hydrological Processes 5 (1): 3–30. https://doi.org/10.1002/hyp.3360050103. Bengal. pp. 2–22 Mosavi, S.M., M. Abedini, A. Esmali, and F. Madani. 2017. Landslide hazard zonation by using fuzzy MCDM models in the GIS (case study: Izeh urban District Statistical Handbook 2013. Bureau of Applied Economics and Statistics, watersheds of Khuzestan). Journal of Watershed Management Research 7 (14): Department of Statistics & Programme Implementation. Government of West 87–78. https://doi.org/10.29252/jwmr.7.14.87. Bengal, Kolkata. Dormann, C.F., J. Elith, S. Bacher, C. Buchmann, G. Carl, G. Carré, J.R.G. Marquéz, B. Negnevitsky, M. 2002. Artificial intelligence: A guide to intelligent systems, Gruber, B. Lafourcade, P.J. Leitão, T. Münkemüller, C. McClean, P.E. Osborne, B. AddisonWesley/Pearson education. Harlow. Reineking, B. Schröder, A.K. Skidmore, D. Zurell, and S. Lautenbach. 2012. Panikkar, S.V., and V. Subramanyan. 1996. A geomorphic evaluation of the Collinearity: A review of methods to deal with it and a simulation study landslides around Dehradun and Mussoorie, Uttar Pradesh, India. evaluating their performance. Ecography 36: 27–46 https://doi.org/10.1111/ Geomorphology 15: 169–181. j.1600-0587.2012.07348.x. Pourghasemi, H.R., M. Mohammady, and B. Pradhan. 2012. Landslide susceptibility Dou, J., A.P. Yunus, D.T. Bui, A. Merghadi, M. Sahana, Z. Zhu, et al. 2019. mapping using index of entropy and conditional probability models in GIS: Assessment of advanced random forest and decision tree algorithms for Safarood Basin, Iran. Catena 97: 71–84. https://doi.org/10.1016/j.catena.2012.05.005. modeling rainfall-induced landslide susceptibility in the Izu-Oshima Volcanic Pradhan, B. 2013. A comparative study on the predictive ability of the decision Island, Japan. Science of the Total Environment 662: 332–346. tree, support vector machine and neuro-fuzzy models in landslide susceptibility mapping using GIS. Computational Geosciences 51: 350–365 Fan, X., G. Scaringi, G. Domènech, F. Yang, X. Guo, L. Dai, et al. 2019. Two multi- https://doi.org/10.1016/j.cageo.2012.08.023. temporal datasets that track the enhanced landsliding after the 2008 Wenchuan earthquake. Earth System Science Data 11 (1): 35–55. Pradhan, B., and S. Lee. 2009. Regional landslide susceptibility analysis using back- Feizizadeh, B., and T. Blaschke. 2011. Landslide risk assessment based on GIS propagation neural network model at Cameron Highland, Malaysia. multi-criteria evaluation: A case study Boston Abad county, Iran. Journal of Landslides 7 (1): 13–30. https://doi.org/10.1007/s10346-009-0183-2. Earth Science and Engineering 1: 66–71. Pradhan, B., and S. Lee. 2010. Delineation of landslide hazard areas on Penang Feizizadeh, B., and T. Blaschke. 2013. GIS-multicriteria decision analysis for Island, Malaysia, by using frequency ratio, logistic regression, and artificial landslide susceptibility mapping: Comparing three methods for the Urmia neural network models. Environmental Earth Sciences 60 (5): 1037–1054. lake basin, Iran. Natural Hazards 65 (3): 2105–2128. Refahi, H.Gh. 2000. Soil Erosion by Water & Conservation. 2nd ed., Tehran Feizizadeh, B., M.S. Roodposhti, P. Jankowski, and T. Blaschke. 2014. A GIS-based University Publications, Tehran, Iran, pp. 551. extended fuzzy multi-criteria evaluation for landslide susceptibility mapping. Reichenbach, P., M. Rossi, B.D. Malamud, M. Mihir, and F. Guzzetti. 2018. A review Computers & Geosciences 73: 208–221. https://doi.org/10.1016/j.cageo.2014.08.001. of statistically-based landslide susceptibility models. Earth-Science Reviews 180: 60–91. Garcia-Rodriguez, M.J., J.A. Malpica, and B. Benito. 2008. Susceptibility assessment of earthquake-triggered landslides in El Salvador using logistic regression. Roodposhti, M.S., S. Rahimi, and M.J. Beglou. 2014. PROMETHEE II and fuzzy AHP: Geomorphology 95: 172–191. https://doi.org/10.1016/j.geomorph.2007.06.001. An enhanced GIS-based landslide susceptibility mapping. Natural Hazards 73 Gerrard, J. 1994. The landslide hazard in the imalayas: Geological control and (1): 77–95. human action. Geomorphology 10: 221–230. Saaty, T.L. 1980. The analytical hierarchy process, 350. New York: McGraw Hill. Gupta, R., and B. Joshi. 1990. Landslide hazard zoning using the GIS approach—A Saaty, T.L. 1990. The analytic hierarchy process: Planning, priority setting, resource case study from the Ramganga catchment, Himalayas. Engineering Geology allocation. 1st ed, 502. Pittsburgh: RWS publications. 28 (1–2): 119–131. https://doi.org/10.1016/0013-7952(90)90037-2. Saaty, T.L. 1994. Fundamentals of decision making and priority theory with analytic hierarchy process. 1st ed, 527. Pittsburgh: RWS publications. Guzzetti, F., P. Reichenbach, F. Ardizzone, M. Cardinali, and M. Galli. 2006. Estimating the quality of landslide susceptibility models. Geomorphology 81 Saaty, T.L., and L.G. Vargas. 1998. Diagnosis with dependent symptoms: Bayes (1–2): 166–184. https://doi.org/10.1016/j.geomorph.2006.04.007. theorem and the analytic hierarchy process. Operations Research 46 (4): 491–502. Hejazi, S.A. 2015. Landslide hazard mapping in Goijabel of Ahar using GIS. J Geogr Saha, S. 2017. Groundwater potential mapping using analytical hierarchical process: Plan 18 (20): 135–152. A study on Md. Bazar block of Birbhum District, West Bengal. Spatial Information Hembram, T.K., and S. Saha. 2018. Prioritization of sub-watersheds for soil Research 25 (4): 615–626. https://doi.org/10.1007/s41324-017-0127-1. erosion based on morphometric attributes using fuzzy AHP and Sarkar, S. 1999. Landslides in Darjiling Himalayas. Transactions Japanese compound factor in Jainti River basin, Jharkhand, eastern India. Geomorphological Union 20: 299–315. Environment Development and Sustainability:1–28. https://doi.org/10.1007/ Shadfar, S., M. Yamani, and M. Namaki. 2011. Zoning land slide hazard by s10668-018-0247-3. information value method (IVM), density area (DA) and landslide numerical Malczewski, J. 1999. GIS and multicriteria decision analysis. New York: Wiley. risk factor (LNRF) model in Chalkrood. Journal of Watershed Management in Engineering 3 (1): 40–47. Malik, R., S. Ghosh, S. Dhar, P. Singh, and M. Singh. 2016. GIS based landslide hazard from zonation along national highway AY-58, from Rishikesh to Tien Bui, D., B. Pradhan, O. Lofman, and I. Revhaug. 2012. Landslide susceptibility Joshimath, Uttarkhand, India. International Journal of Advanced Remote assessment in Vietnam using support vector machines, decision tree, and Sensing Geography 4 (2): 41–55. Naïve Bayes models. Mathematical Problems in Engineering 2012: 1–26. Mallick, J., R.K. Singh, M.A. Alawadh, S. Islam, R.A. Khan, and M.N. Qureshi. 2018. https://doi.org/10.1155/2012/974638. GIS-based landslide susceptibility evaluation using fuzzy-AHP multi- Torkashvand, A.M., A. Irani, and J. Sorur. 2014. The preparation of landslide map criteria decision-making techniques in the Abha watershed, Saudi by landslide numerical risk factor (LNRF) model and geographic information Arabia. Environmental Earth Sciences 77 (7). https://doi.org/10.1007/ system (GIS). The Egyptian Journal of Remote Sensing and Space Science 17 (2): s12665-018-7451-1. 159–170. https://doi.org/10.1016/j.ejrs.2014.08.001. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 18 of 18 Turner, A.K., and L.R. Schuster. 1996. Landslides: Investigation and mitigation. Washington, DC: National Academy Press. Van Eynde, E., S. Dondeyne, M. Isabirye, J. Deckers, and J. Poesen. 2017. Impact of landslides on soil characteristics: Implications for estimating their age. Catena. https://doi.org/10.1016/j.catena.2017.05.003. Van Westen, C.J., R. Soeters, and K. Sijmons. 2000. Digital geomorphological landslide hazard mapping of the Alpago area, Italy. International Journal of Applied Earth Observation and Geoinformation 2: 51–60. Wang, H., G. Wang, F. Wang, K. Sassa, and Y. Chen. 2008. Probabilistic modeling of seismically triggered landslides using Monte Carlo simulations. Landslides 5: 387–395. https://doi.org/10.1007/s10346-008-0131-6. Yesilnacar, E.K. 2005. The application of computational intelligence to landslide susceptibility mapping in Turkey, PhD Thesis, 423. Australia: University of Melbourne. Youssef, A.M., M. Al-kathery, and B. Pradhan. 2014. Landslide susceptibility mapping at Al-Hasher Area, Jizan (Saudi Arabia) using GIS based frequency ratio and index of entropy models. Geosciences Journal. https://doi.org/10.1 007/s12303-014-0032-8. Zadeh, L.A. 1965. Fuzzy sets. Information and Control 8 (3): 338–353. Zêzere, J.L., S. Pereira, R. Melo, S.C. Oliveira, and R.A. Garcia. 2017. Mapping landslide susceptibility using data-driven methods. Science of the Total Environment 589: 250–267. Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Geoenvironmental Disasters Springer Journals

Landslide susceptibility mapping using knowledge driven statistical models in Darjeeling District, West Bengal, India

Geoenvironmental Disasters , Volume 6 (1) – Aug 1, 2019

Loading next page...
 
/lp/springer-journals/landslide-susceptibility-mapping-using-knowledge-driven-statistical-ijQBMetV9f

References (62)

Publisher
Springer Journals
Copyright
Copyright © 2019 by The Author(s).
Subject
Environment; Environment, general; Earth Sciences, general; Geography, general; Geoecology/Natural Processes; Natural Hazards; Environmental Science and Engineering
eISSN
2197-8670
DOI
10.1186/s40677-019-0126-8
Publisher site
See Article on Publisher Site

Abstract

Landslide is an important geological hazard in the large extent of geo-environment, damaging the human lives and properties. The present work, intends to identify the landslide susceptibility zones for Darjeeling, India, using the ensembles of important knowledge driven statistical technique i.e. fuzzy logic with Landslide Numerical Risk Factor (LNRF) and Analytical Hierarchical Process (AHP). In the study area, 326 landslides have been identified and a landslide inventory map has been prepared based on these landslides. The landslide inventory map has considered as the dependent factor and the geo-environmental factors like rainfall, slope, aspect, altitude, geology, soil texture, distance from river, lineament and road, land use/ land cover, NDVI and TWI have been considered as independent factors. Landslide susceptibility maps were prepared based on the Fuzzy- Landslide Numerical Risk Factor (LNRF) and Fuzzy- analytic hierarchy process (AHP) methods in a GIS environment. According to the results of LNRF and AHP based fuzzy logic 34 and 22% areas are highly susceptible to landslide in this district. The landslide maps of both models have been validated through ROC curve and RMSE. The areas under curves are 91% (for Fuzzy-LNRF) and 90% (for Fuzzy-AHP) and RMSE values of these models are 0.18 and 0.14 which are indicating the good accuracy of both models in the identification of landslide susceptibility zones. Moreover, the Fuzzy-LNRF model is promising and sufficient to be advised as a method to prepare landslide susceptibility map at regional scale. Keywords: Landslide numerical risk factor (LNRF), Fuzzy-AHP, Fuzzy logic (FL), Landslide susceptibility, GIS Introduction Rodriguez et al. 2008). Probably this rate of property The mountainous areas of the world are frequently af- damage will become more faster in the upcoming time fected by the occurrences of the landslide because of in parity with the gradual development of urban centers, high energy with variability and instability of masses economic and rising regional rainfall due to climatic (Gerrard 1994). From the environmental point of view, change in the landslide prone areas (Turner and Schus- different kind of problems such as loss of soil fertility, ter 1996). Landslide occurrence is a significant barrier to acceleration of deforestation rate etc. may be caused by the development in Darjeeling district. In Darjeeling dis- the landslides (Van Eynde et al. 2017). Most of the trict, landslides mainly take place due to heavy Mon- mountainous regions of India are characterized by the soonal rainfalls and seismicity (Panikkar and landslide disaster. A number of avalanche zones in the Subramanyan 1996). The Darjeeling district had been Indian Himalayan region are prominent, e.g. Jammu experienced major landslides in July–August, 1993, May Kashmir, Himachal Pradesh, Kumayun, Darjeeling and 2009 and September 2011 (Sarkar 1999). Massive rain Sikkim and North-eastern hilly states (Bhandari 2004). caused landslides at Darjeeling town, Mirik, Kurseong Landslide causes loss of property far greater than the and Kalimpong during June–July, 2015 and induced the any natural disaster (Turner and Schuster 1996; Garcia- loss of properties and lives. Reduction of effect of land- slide can be possible only with a comprehensive know- * Correspondence: sunilgeo.88@gmail.com ledge about the probability of occurrence, character and Department of Geography, University of Gour Banga, Malda, West Bengal, magnitude of landslide in an area. Therefore, delineation India of landslide vulnerable regions is indispensable for © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 2 of 18 carrying out safer alleviation programs, and future plan- and compare with them which have not been used in ning of the area (Fan et al. 2019). In the present paper this district previously. The main advantage of ensemble the main thrust has been given to delineate the landslide of fuzzy logic and LRNF is that we can use both expert susceptible zones and chalk out suitable method. knowledge as well as statistical method for predicting Landslide is the major hazardous phenomenon which landslide susceptibility using various causative factors. sometimes causes loss of human lives and properties. In this study, remote sensing data along with meta Therefore, any environmental factor may cause landslide data have been used to outline the landslide susceptibil- when soil resistance power is lower than the shear force ity areas for the Darjeeling District. Geo-environmental (Refahi 2000; Bouma and Imeson 2000). For landslide factors such as rainfall, slope, aspect, altitude, geology, hazard evaluation several qualitative and quantitative soil texture, distance from river, distance from linea- methods have been used (Aleotti and Chowdhury 1999; ment, distance from road, land use/ land cover, normal- Reichenbach et al. 2018). According to the qualitative ized difference vegetation index (NDVI) and method, the expert can evaluate the landslide suscepti- topographical wetness index (TWI) have been taken out bility zones in his own opinion. The expert also can as- to facilitate the quantification of landslide. Fuzzy-LRNF sess the vulnerable areas on the basis of similar and Fuzzy-AHP have been applied considering the ex- geological and geomorphological character using the tracted database. Using the LRNF models, the fuzzy landslide inventory maps or existing landslide areas membership value has been calculated and thereafter, (Ayalew and Yamagishi 2005). The multi-criteria deci- using the fuzzy gamma operator the membership values sion approach (MCDA) is an important way for indenti- of parameters have been assembled for producing the fying the potential landslide areas using proper database. landslides susceptible map of Darjeeling district. Simi- GIS based MCDA has been considered as the powerful larly using the Fuzzy-AHP method, another map has techniques and procedures for evaluating, designing and also been produced. Finally, the maps have been verified accuracy judgments’ of the results (Feizizadeh and Blas- and compared using known landslide locations based on chke 2011, 2013). The present study has followed the ROC and RMSE quantitative validation methods. The GIS based MCDA techniques like Fuzzy-Landslide Nu- main novelty is that the first time knowledge driven merical Risk Factor (LNRF) and Fuzzy-AHP for the technique (Fuzzy logic) has been assembled with LRNF landslide susceptibility mapping. Several other re- in this work to delineate the landslide susceptible zone searchers applied Fuzzy-AHP and LNRF. Torkashvand of Darjeeling district and compared with the Fuzzy-AHP et al. (2014) applied the Landslide Numerical Risk Factor method. Moreover, according to the previous literatures (LNRF) model using GIS in East of the Sabalan volcanic so many researchers used LRNF and AHP method for mass region in Iran. Mokarram and Zarei (2018), Feizi- mapping the landslide susceptibility but not a single re- zadeh et al. (2014), Mosavi et al. (2017), Hejazi (2015), searcher has used ensemble of fuzzy logic and LRNF Mirnazari et al. (2015) and Hembram and Saha (2018) model for predicting the spatial landslide probability and used Fuzzy-AHP model for their work and they got compared this ensemble method with fuzzy-AHP. fruitful result for susceptibility mapping. Various statis- tical methods have been used by the researchers for ana- Study area lyzing the spatial pattern of landslides and preparing the The Darjeeling district is located in the northernmost landslide susceptible map such as logistic regression part of the West Bengal in India. It is an important (Zêzere et al. 2017; Budimir et al. 2015), hierarchical ap- mountainous part of the eastern Himalaya. Geographic- proach (Youssef et al. 2014), statistical index (Dou et al., ally the study area is extended between the latitudes 2015), conditional analysis (Pourghasemi et al. 2012), 26°27″ to 27°13″N and longitudes 87°59″E–88°53″E. weight of evidence Pradhan and Lee (2010). In the re- The study area is covered with an area of 3149 sq.km cent years, different machine learning techniques have (Fig. 1). According to the census of 2011, the total popu- also been used by some scholars for mapping the land- lation of the district is 18, 46,823 with 9, 37,259 males slide disaster like decision tree (Pradhan 2013), random and 9, 09,564 females. The population density of the dis- forest (Dou et al. 2019), artificial neural network Prad- trict is 586 person/ km (District Statistical Handbook han and Lee (2010), support vector machine (Tien Bui 2013). The number of rural households and the urban et al. 2012) etc. For identifying the landslide susceptibil- households of Darjeeling district was 212000 and 89584 ity zones, they used some important factors such as the in 2001, but these have increased to 236000 and 154540 elevation, lithology, slope, land use, river, topographical in 2011 respectively (District Statistical Handbook 2011). wetness index, aspect, road, fault, and precipitation The total length of national highway, state highway, maps. The rationale of this work is to identify the land- major district road and other ordinary district road were slide susceptible areas using the ensemble models that 100, 80, 37 and 516 km respectively (District Gazetteer are fuzzy- LRNF and Fuzzy-AHP of Darjeeling district of Darjeeling District 1980). But the length of national Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 3 of 18 Fig. 1 Location map of study area showing district Darjeeling, India (a), and the location of 326 landslides (b) Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 4 of 18 highway, state highway and main district highway have Statistical Handbook, Census of India 2011, drainage increased to 111, 191 and 79 km respectively in 2011. and road networks from Open series Topographical The total length of surfaced and unsurfaced roads are Sheets (2015), images and ASTER DEM from USGS 3696.54 km and 1652.51 km approximately in 2013–14. Earth Explorer, Soil map from National Bureau of Soil The characteristics of both plain and mountainous top- Survey & Land use Planning and Geological map from ographies exit in this study area. The altitude of the Geological Survey of India. The summery of the proce- study area ranges from 15 m to 3602 m from mean sea dues followed is depicted in the flowchart (Fig. 2). level and the slope from 0° to 80° approximately. The major portion of the study area is covered with Triassic Software used rocks. The soil character of the study area is varied from To predict the landslide potential areas, the thematic one region to another region. The study area receives layers of selected geo-environmental parameters namely huge amount of rainfall in the monsoon season. The rainfall, slope, elevation, aspect, geology, soil texture, dis- average annual rainfall of the study area is about 3051 tance from lineament, distance from river, distance from mm. The major rivers of the study area are Mahananda, road, natural differential vegetation index (NDVI) and Tista, Mechi, Balason, Jaldhaka, Rammam and Rangit, TWI have been prepared with the help of ArcGIS 10.3.1, which are flowing from northern part. The district has ENVI 4.7, GEOMATICA and the mathematical calcula- some reputed eco-tourism sites and pilgrimage sites tions have been done with the help of SPSS softwares. namely like Tiger Hill, Rock Garden, Mahakal Temple, Dhirdham Temple, Batasia Loop, Ghoom Monastery and Happy Vally Tea Garden, etc. Historically, Tea Plan- Preparation of landslide inventory map tation and Cinchona are the main sources of livelihood The landslide inventory map is the vital part for analyz- in the Darjeeling district. The community of the study ing the landslide susceptibility, hazard and risk (Guzzetti area depends on the horticulture, tourism, and forestry. et al. 2006). Pradhan and Lee (2009) and Pourghasemi et Siliguri is the major town of the study area which is fa- al. (2012) prepared the landslide inventory map for iden- miliar with ‘gate way’ of eastern India. tifying the landslide hazards zones. Van Westen et al. (2000) remarked that the different data such as field in- Materials and methodology vestigations, historical landslide events and satellite Data sources image analysis can be used to prepare the landslide in- For the fulfillment of the present work, various import- ventory map (Fig. 1b). In the present study, 326 land- ant data have been collected from different sources e.g. slides have been identified from Google earth imagery rainfall and temperature data from Indian Meteoro- and multiple field visits. The landslide inventory map logical Department, population data from District has been prepared in the GIS environment for Fig. 2 Flowchart of methodology used in this study Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 5 of 18 calculating the LNRF to predict the potentials landslide annual rainfall data of last 5 years (Fig. 3a). The slope susceptibility areas in this district. map of the study area has been prepared from ASTER DEM with the help of 3D analyst tool of ArcGIS 10.3.1 Selection and Muliti-collinearity analysis of landslide software (Fig. 3b). In case of elevation and aspect, these causative factors data layers have been prepared from ASTER DEM with There is a variety of geo-environmental parameters that the help of 3D analyst tool of surface in GIS environ- have been used by the various researchers for producing ment (Fig. 3c, d). the landslide susceptibility map. Still there is no standard The geology map has been prepared with the help of guideline for selecting the landslide predictors. In this digitations process from referenced geological map research 12 landslide causative factors i.e. rainfall, slope, which has been collected from Geological Survey of aspect, altitude, geology, soil texture, distance from river, India (Fig. 3e). distance from lineament, distance from road, land use/ The thematic layer of soil texture has been generated land cover, NDVI and TWI have been selected based on with the help of the digitations process form referenced the multi-collinearity analysis for mapping the landslide soil map which has been collected from National Bureau susceptibility. In the landslide susceptibility analysis of Soil Survey and Land Use Department (Fig. 3f). The aforementioned causative factors are widely used (Dou distance from river and distance from road maps have et al. 2019; Arnone et al. 2016; Tien Bui et al. 2012). been prepared with the help of Euclidian distance buffer- Multi-collinearity analysis is an important way to ver- ing tool in GIS environment (Fig. 3g, h). The lineament ify the effectiveness of the landslide conditioning factors of the study area has been extracted from Landsat 8 OLI (Saha 2017). For the present study, the collinearity test (Optical land Imager) image with the help of the ENVI of landslide determining factors has been done in the 4.7 and GEOMATICA softwares. The distance from lin- SPSS software. A tolerance of less than 0.10 and and eament map has been prepared with the help of Euclid- variance inflation factors (VIF) 10 or above indicates ian distance buffering tool of ArcGIS software 10.3.1 multicollinearity problems (Dormann et al. 2012; Wang (Fig. 3i). The land use/ land cover (LULC) map has been et al. 2008). In the present study, the values of tolerances prepared from Landsat 8 OLI imagery with the help of and VIF of all the selected parameters are less than 10% maximum likelihood classification method (Fig. 3j). The (Table 1). So, there is no collinearity problem among the normalized differential vegetation index (NDVI) has selected landslide determining factors. been calculated from Landsat 8OLI image with the help of image analysis tool in ArcGIS 10.3.1 software (Fig. 3k). Preparation of thematic layers of selected parameters The thematic layer of TWI has been prepared from The average annual rainfall data of last 5 years, since ASTER DEM imagery in GIS environment using Eq. 1 2012 to 2017 have been collected from Indian Meteoro- (Fig. 3l) which was suggested by Moore et al. (1991). logical Department. The thematic layer of rainfall has been prepared with the help of the interpolation method of IDW in GIS environment based on the average TWI ¼ In ð1Þ tanβ Table 1 Collinearity statistics of landslide determining factors Where, TWI = topographic wetness index, α is cumu- Sl. Parameters Collinearity statistics lative upslope area draining through a point (per unit No Tolerance VIF contour length), β is the slope gradient (in degree). The 1 Rainfall 0.494 2.024 minimum, maximum, categorical classification and 2 Elevation 0.638 1.566 methods of the selected geo-environmental conditional factors have been done in GIS environment and men- 3 Slope 0.894 1.118 tioned in the following Table 2. 4 Aspect 0.742 1.348 5 Geology 0.743 1.346 Fuzzy method 6 Soil 0.858 1.166 For the present study, the Fuzzy method has been as- 7 Distance from River 0.635 1.575 sembled with AHP and LNRF methods. The fuzzy maps 8 Distance from Lineament 0.818 1.223 of selected parameters have been prepared with the help of membership function (MF) tool in GIS environment. 9 Distance from Road 0.527 1.898 The membership function (MF) values range between 0 10 LULC 0.833 1.201 and 1 (Zadeh 1965). The value 0 means that x is not a 11 NDVI 0.804 1.243 member of the fuzzy set, while the value 1 means that x 12 TWI 0.884 1.132 is a full member of the fuzzy set. A sample of fuzzy set Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 6 of 18 Fig. 3 Landslide conditioning factor maps used in this study: a Rainfall, b Slope, c Slope Elevation, d Aspect, e Geology, f Soil Map, g Distance from River, h Distance from Lineament, i Distance from Road, j Land Use/ Land Cover, k NDVI, l TWI Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 7 of 18 Table 2 Overview of selected parameters used for landslide susceptibility mapping Parameters Ranges Classes Methods Min Max Rainfall (mm) 1877 2334 1. 1877.38–1991.9 Natural Break 2. 1991.97–2090.54 3. 2090.45–2167.44 4. 2167.44–2239.06 5. 2239.06–2333.96 Slope (Degree) 0 79.2 1. 0–9.32 Natural Break 2. 9.32–18.64 3. 18.44–27.34 4. 27.34–36.66 5. 5. 36.66–79.23 Altitude (m) 15 3602 1. 15–422.93 Natural Break 2. 422.93–985.6 3. 985.6–1576.4 4. 1576.4–2279.73 5. 5. 2279.73–3602 Aspect –– 1. Flat (−1) Equal interval 2. North (0–22.5) 3. Northeast (22.5–67.5) 4. East (67.5–112.5) 5. Southeast (112.5–157.5) 6. South (157.5–202.5) 7. Southwest (202.5–247.5) 8. West (247.5–292.5) 9. Northwest (292.5–337.5) 10. north(337.5–360) Geology –– 1. Triassic lithological units 2. Cenozoic 3. Pliocene-Pleistocene, 4. Holocene 5. Middle-upper Pleistocene Soil texture –– 1. Gravelly loamy, soil texture classes 2. Fine loamy - Coarse Loamy 3. Gravelly loamy Skeletol 4. Gravelly Loam - Coarse Loamy 5. Coarse Loamy Distance from River (km) 0 4.33 1. 0–0.42 Natural Break 2. 0.42–1.10 3. 1.10–1.66 4. 1.66–2.26 5. 2.26–4.33 Distance from Lineament (km) 0 10.1 1. 0–1.54 Natural Break 2. 1.54–2.85 3. 2.85–4.20, 4 4. 4.20–5.75 5. 5.75–10.12 Distance from Road (km) 0 16.5 1. 0–1.74 Natural Break 2. 1.74–3.94 Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 8 of 18 Table 2 Overview of selected parameters used for landslide susceptibility mapping (Continued) Parameters Ranges Classes Methods Min Max 3. 3.94–6.72 4. 6.72–10.22 5. 10.22–16.49 Land use/Land cover –– 1. Water bodies supervised classification 2. Vegetation 3. Fallow land 4. Agricultural land 5. Settlement NDVI −0.07 0.49 1. −0.07-0.12 Natural Break 2. 0.12–0.17 3. 0.17–0.23 4. 0.23–0.29 5. 0.29–0.49 TWI 1.95 18.9 1. 1.95–7.37 Natural Break 2. 7.37–8.53 3. 8.53–9.76 4. 9.76–11.70 5. 11.70–18.91 is shown in the following equation (Mcbratney and GAMMA. In the present study, gamma operator has Odeh 1997). been used for combining membership functions. Fuzzy gamma operation has been calculated using eqs. A ¼ x; μ ðÞ x for each xεX ð2Þ (5). Where, μ is the MF (membership of x in fuzzy set A) so that: 1−γ μ ¼ðÞ μ k : μ : ð5Þ γ sum product If x does not belong to A then μ =0. If x belongs completely to A then μ =1. If x x belongs in a certain degree to A then The exponent γ, which is a number from < 0, 1 > inter- val, allows optimization of the membership combination. 0 < μ ðÞ x < 1 Setting it to the extremes of the interval give either fuzzy According to Eq. 3 MF was used for rainfall, elevation, algebraic sum (γ = 1) or fuzzy algebraic product (γ = 0). aspect, slope, NDVI, TWI [8]. To perform fuzzy gamma operation, several gamma op- 8 9 erator (k) values, i.e. 0.50, 0.70, 0.80, 0.90, 0.95, and 0x≤a < = 0.975 are there. In the present study, fuzzy gamma oper- μ ðÞ x ¼ fxðÞ ¼ x ‐a=b‐aa≻x≺b ð3Þ ator (k) value of 0.975 has been applied for producing : ; 1x≥b the landslide susceptibility map. Where x is the input data and a, b are the limit values. For geology, soil texture, distance from river, distance Landslide numerical risk factor (LNRF) model from lineament, LULC and distance from road the fol- Landslide Numerical Risk Factor (LNRF) model is an lowing MF has been used [4]. important method for identifying the landslide hazard 8 9 zones especially in the mountain region (Gupta and <0x≤a = Joshi 1990). According to this model, the LNRF > 1 value μ ðÞ x ¼ fxðÞ ¼ b ‐x=b‐aa≻x≺b ð4Þ : ; indicates that the geo-environmental factors have high 1x≥b responsibility for the occurrences of landslide. The LNRF < 1 values represent that the geo-environmental Fuzzy gamma operators factors are more stable and have less effect in landslides Several fuzzy operators exist for combining membership occurrences (Gupta and Joshi 1990). The LNRF has been functions such as AND, OR, SUM, PRODUCT and calculated through Eq. (6): Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 9 of 18 Table 3 Computation of LNRF model on landslide determining Factors Parameters Area Landslides Area LNRF Fuzzy membership In Hectare In Percentage In Hectare In Percentage values using LNRF model Rainfall (mm) Very Low (1877.38–1991.97) 27659.36 8.78 0 0 0 0 Low (1991.97–2090.54) 24856.96 7.89 0 0 0 0 Moderate (2090.45–2167.44) 80967.44 25.71 31.56 7.32 0.37 0.11 High (2167.44–2239.06) 114335.78 36.31 294.58 68.29 3.41 1.00 Very High (2239.06–2333.96) 67080.46 21.3 105.21 24.39 1.22 0.36 Slope (Degree) Very Low (0–9.32) 101505.25 32.23 14.5 3.36 0.17 0.11 Low (9.32–18.64) 58545.55 18.59 52.56 12.18 0.61 0.39 Medium (18.44–27.34) 70795.02 22.48 103.31 23.95 1.2 0.76 High (27.34–36.66) 57571.95 18.28 135.93 31.51 1.58 1.00 Very High (36.66–79.23) 26482.23 8.41 125.06 28.99 1.45 0.92 Altitude (m) Very Low (15–422.93) 115880.61 36.8 35.17 8.15 0.41 0.16 Low (422.93–985.6) 71784.89 22.8 224.01 51.93 2.6 1.00 Medium (985.6–1576.4) 63320.09 20.11 81.46 18.88 0.94 0.36 High (1576.4–2279.73) 44471.47 14.12 72.2 16.74 0.84 0.32 Very High(2279.73–3602) 19442.94 6.17 18.51 4.29 0.21 0.08 Aspect Flat (−1) 163.37 0.05 0 0 0 0 North (0–22.5) 20317.92 6.45 7.41 1.72 0.09 0.07 North-East (22.5–67.5) 39614.59 12.58 25.92 6.01 0.3 0.22 East (67.5–112.5) 39009.88 12.39 59.24 13.73 0.69 0.51 South-East (112.5–157.5) 44774.25 14.22 87.01 20.17 1.01 0.75 South (157.5–202.5) 45083.47 14.32 116.63 27.04 1.35 1.00 South-West (202.5–247.5) 39204.16 12.45 70.35 16.31 0.82 0.61 West (247.5–292.5) 32441.32 10.3 49.98 11.59 0.58 0.43 North-West (292.5–337.5) 35974.84 11.42 12.96 3 0.15 0.11 North (337.5–360) 18316.21 5.82 1.85 0.43 0.02 0.01 Geology Triassic 166113.98 52.75 302.86 70.21 3.51 1 Cenozoic 23200.35 7.37 82.6 19.15 0.96 0.27 Pliocene-Pleistocene 11258.98 3.58 45.89 10.64 0.53 0.15 Holocene 58152.53 18.47 0 0 0 0 Middle-upper Pleistocene 56174.15 17.84 0 0 0 0 Soil texture Gravelly loamy 23549.04 7.48 68.11 15.79 0.79 0.46 Fine loamy Coarse Loamy 126713.02 40.24 113.51 26.32 1.32 0.77 Gravelly loamy, Loamy Skeletal 38586.66 12.25 147.57 34.21 1.71 1 Gravelly Loamy Coarse Loamy 120449.25 38.25 90.81 21.05 1.05 0.61 Coarse Loamy 5602.03 1.78 11.35 2.63 0.13 0.08 Distance from River (km) 0–0.42 km 99542.45 31.61 73.65 17.07 0.85 0.37 0.42–1.10 km 110752.04 35.17 199.89 46.34 2.32 1 1.10–1.66 km 64340.57 20.43 115.73 26.83 1.34 0.58 1.66–2.26 km 32920.37 10.45 42.08 9.76 0.49 0.21 Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 10 of 18 Table 3 Computation of LNRF model on landslide determining Factors (Continued) Parameters Area Landslides Area LNRF Fuzzy membership In Hectare In Percentage In Hectare In Percentage values using LNRF model 2.26–4.33 km 7344.58 2.33 0 0 0 0 Distance from Lineament (km) 0–1.54 km 81563.6 25.9 110.6 25.64 1.28 0.77 1.54–2.85 km 95262.67 30.25 44.24 10.26 0.51 0.31 2.85–4.20 km 77028.98 24.46 143.78 33.33 1.67 1.00 4.20–5.75 km 44870.1 14.25 99.54 23.08 1.15 0.69 5.75–10.12 km 16174.66 5.14 33.18 7.69 0.38 0.23 Distance from Road (km) 0–1.74 km 140275.58 44.55 88.48 20.51 1.03 0.67 1.74–3.94 km 84741.4 26.91 44.24 10.26 0.51 0.33 3.94–6.72 km 50166.43 15.93 66.36 15.38 0.77 0.50 6.72–10.22 km 27267.21 8.66 132.72 30.77 1.54 1.00 10.22–16.49 12449.37 3.95 99.54 23.08 1.15 0.75 Land use/Land cover Water bodies 3466.31 1.1 39.87 9.24 0.46 0.16 Vegetation 227240.32 72.16 255.55 59.24 2.96 1.00 Fallow land 14437.29 4.58 29 6.72 0.34 0.11 Agricultural Land 65442.75 20.78 106.93 24.79 1.24 0.42 Settlement 4313.33 1.37 0 0 0 0.00 NDVI Very Low (−0.07–0.12) 37936.34 12.05 114.18 26.47 1.32 0.91 Low (0.12–0.17) 83384.9 26.48 125.06 28.99 1.45 1.00 Medium (0.17–0.23) 85506.39 27.15 96.06 22.27 1.11 0.77 High (0.23–0.29) 70015.86 22.23 65.25 15.13 0.76 0.52 Very High (0.29–0.49) 38056.51 12.09 30.81 7.14 0.36 0.25 TWI Very Low (1.95–7.37) 49986.44 15.87 63.12 14.63 0.73 0.35 Low (7.37–8.53) 113766.56 36.13 178.85 41.46 2.07 1.00 Medium (8.53–9.76) 93346.92 29.64 115.73 26.83 1.34 0.65 High (9.76–11.70) 46923.5 14.9 73.65 17.07 0.85 0.41 Very High (11.70–18.91) 10876.58 3.45 0 0 0 0 Gamma operators (k) values of 0.975 in GIS LNRF ¼ ð6Þ environment. Analytic hierarchy process (AHP) Where A: landslide area in every unit, E: mean area of Analytic hierarchy process is an important multi-criteria landslide in the whole unit. decision analysis (MCDA) method which can be applied The sub-parameters wise LNRF values have been cal- for assigning the weights to the individual parameter culated (Table 3). In the present study, fuzzy member- (Saaty and Vargas 1998). AHP method is a pair wise ship values have been allotted based on LNRF. To comparison matrix. When the matrix is formed, the transform the LNRF values to Fuzzy membership values consistency ratio (CR) value ranges from 0 to 1 (Saaty each sub-class has been divided by the maximum value 1980, 1990, 1994). To identify the potentiality index, of LNRF of individual parameter. The membership value general linear combination method can be performed ranges from 0 to 1. The fuzzy membership values using with the help of AHP method (Malczewski 1999). The LNRF model have been converted into a single layer to pair wise matrix has been formed with the help of the chalk out the landslide susceptibility zone using Fuzzy Saaty’s(1980) fundamental scale (Table 4). In the present Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 11 of 18 Table 4 Fundamental scale of Saaty’s(1980) upper Pleistocene (Fig. 3e). Triassic geological segment with 52.75% area is encompassed with 70.21% landslide Scale Description area. According to the LNRF model using fuzzy logic, 1 Equal Importance the Triassic geological segment has attained the max- 3 Moderate Importance imum LNRF and fuzzy membership value. The other 5 Strong Importance sub-layers of geological segment are indicating the less 7 Very Strong Importance probability of landslide occurrence. Pedagogically, the 9 Absolute Strong Importance region is composed with several soil textural classes 2,4,6,8 Intermediate values between the two adjacent judgments namely Gravelly loamy, Fine Loamy to Coarse Loamy, Gravelly Loamy to Loamy Skeletal, Gravelly Loamy to case, the weight of parameters calculated based on AHP Coarse Loamy and Coarse Loamy (Fig. 3f). The 34.21% method has been combined with fuzzy logic to prepare landslide areas are concentrated in Gravelly Loamy to landslide susceptibility map of the study area. Loamy skeletal soil texture class (Table 3) and the LNRF and fuzzy membership values are 1.71 and 1, which are Results representing high risk of landslides. The distance from Application of fuzzy-LNRF model river map (Fig. 3g) has been classified into five classes The spatial distribution of average annual rainfall of the such as 0–0.42 km, 0.42 km − 1.10 km, 1.10 km − 1.66 study area ranges from 1877.38 mm to 2333.96 mm km, 1.66 km − 2.26 km and 2.26 km to 4.33 km. The (Fig. 3a) respectively. The rainfall map has been catego- classes 0.42 km − 1.10 km and 1.10 km − 1.66 km of river rized into five classes such as very low (1877 mm–1991 buffering are covered with 46.34% and 26.83% landslide mm), low (1991 mm–2090 mm), moderate (2090 mm– area (Table 3). The LNRF values of 0.42 km − 1.10 km 2167 mm), high (2167 mm–2239 mm) and very high and 1.10 km − 1.66 km sub-layers are 2.32 and 1.34 as (2239 mm–2333 mm) respectively. The high sub-class of well as fuzzy membership values of the same layers are 1 rainfall with 36.31% area is covered the 68.29% land- and 0.58 which are indicating high landslide susceptibil- slides of the study area (Table 3). The LNRF values have ity. The lineament distance map (Fig. 3h) has been clas- been calculated and converted into fuzzy membership sified into five buffer zones such as 0–1.54 km, 1.54 km (FM) value. Here, high rainfall sub-class has attained the − 2.85 km, 2.85 km − 4.20 km, 4.20 km − 5.75 km and highest FM value i.e. 1, indicating the high risk of land- 5.75 km − 10.12 km. The 2.85–4.20 km buffer zone has slide than other sub-class of rainfall (Table 3). The higher LNRF value (Table 3) than the other buffer zones. spatially the slope of the study area ranges from 0 to Road building activity in mountain areas is regarded as 79.23° (Fig. 3b). It has been classified into five categories an infrastructure improvement, which may ground detri- such as very low (0°-9.32°), low (9.32°-18.64°), medium mental effects on slope steadiness; therefore, it can be (18.44° -27.34°), high (27.34° – 36.66°) and very high helpful for delineating the prone areas to landslide oc- (36.66°-79.23°) classes based natural break classification currence. The buffer layer 6.72 km to 10.22 km. distance method in GIS environment. The high and very high area from road is covered 30.77% landslide area (Table 3). slope classes are covered with 31% and 28% landslides This road buffer class has attained the fuzzy membership area. The fuzzy membership values of these sub-classes value 1. Other sub-layers of distance from road are are nearer to 1, representing as high landslides risk showing the low to medium landslide susceptibility. The areas. The altitude of the study area ranges from 15 m to five type of land use classes have been identified in this 3602 m (Fig. 3c) respectively. The altitudinal map has study area namely water bodies, natural vegetation, sand, been categorized into five classes such as very low (15 m agricultural land and settlement with the help maximum to 422.93 m), low (422.93 m to 985.6 m), medium (985.6 likelihood classification method in GIS environment m to 1576.4 m), high (1576.4 m to 2279.73 m) and very (Fig. 3j). The 72.16% area is covered with natural vegetation high (2279 m–3602 m). The low elevation class is cov- area in this district. The agricultural land and settlement ered with 51.93% landslide area (Table 3). The fuzzy areas are indicating less probability for the occurrence of membership value of 422.93 m–985.6 m elevation range landslide. Natural differential vegetation index (NDVI) is is 1, representing higher landslide susceptibility than one of the important factors of environment. The value of other sub-layers. Aspect of the study area has been clas- the NDVI of the study area ranges from − 0.07 to 0.49 sified into several categories such as flat, north, north- (Fig. 3k) respectively. The NDVI map of the study area has east, east, southeast, south, southwest, west and been classified into five classes such as very low (− 0.07– northwest. South sub-layer with 14.32% area is covered 0.12), low (0.12–0.17), medium (0.17–0.23), high (0.23– 27.04% landslide area. Geologically, the study area is 0.29) and very high (0.29–0.49). According toLNRFmodel, composed of five geological segments namely Triassic, the low sub-class of NDVI is attained the highest LNRF Cenozoic, Pliocene-Pleistocene, Holocene and Middle- and fuzzy membership values i.e. 1.45 and 1 which are also Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 12 of 18 Fig. 4 Landside Susceptibility Maps (a) based on Fuzzy-LNRF model & (b) based on Fuzzy-AHP model indicating the high landslide susceptibility. The spatial dis- 7.37), low (7.37–8.53), medium (8.53–9.76), high (9.76– tribution of TWI of the study area ranges from 1.95 to 11.70) and very high (11.70–18.91) (Table 4). The low and 18.91(Fig. 3l) respectively. TWI map of the study area has medium TWI classes area representing the high risk for been classified into five classes namely very low (1.95 to landslide than other sub-layers. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 13 of 18 The landslide susceptibility map (LSM) based on fuzzy raster calculator of spatial analyst tool in Arc- membership values has been prepared with the help of GIS10.3.1 software. The landslide susceptibility map fuzzy gamma operators (k) value of 0.975 and shown in (Fig. 4b) has been classified into three categories like Fig. 4a. It has been classified into three classes namely low, medium and high landslide susceptibility zones low, medium and high landslide susceptibility zones. covering an area of 37.88%, 39.29% and 22.83% re- Out of the total district 34% area is highly susceptible to spectively (Table 6). landslide (Table 6). Some landslides identifed from google earth imagery and directly from the field are mentioned below- (Fig. 5) Application of fuzzy-AHP model In the present work the Fuzzy set theory with AHP Validation is considered as the multi-criteria decision approach The landslide susceptibility maps of Darjeeling district, for identifying the landslide susceptibility zones. The prepared by Fuzzy-LNRF and Fuzzy-AHP models have integration of fuzzy set theory and AHP can be pro- been validated through the ROC curve and RMSE. vided a good and reliable technique for zoning land- Model based work may be possible to evaluate and jus- slide susceptibility. AHP is a single process which tify easily with the help of ROC curve (Chung and Fab- helps to determine the weight of different factors bri 2003). The ROC curve is drawn by X and Y axis. The based on the expert’s opinion and knowledge. For X axis represents the false positive rate (1-specificity) the present work, weights have been assigned to and the Y axis represents the true positive rate (sensitiv- landslide determining factors (Table 5)with the help ity) (Negnevitsky 2002). Mallick et al. (2018) and Feiziza- of AHP method. The highest weight has been deh et al. (2014) have applied the ROC curve for the assigned to rainfall (0.180) and followed by slope validation of the landslide susceptibility zone. In the (0.162), distance from the river (0.108), soil texture present study, 98 landslide patches have been selected (0.095) and elevation (0.086) for mapping the land- among the 326 landslide patches for validating the slide probability. Thereafter, the linear membership landslide susceptibility maps. The area under curve function (MF) has been used to prepare the fuzzy (AUC) can be drawn for the landslide susceptibility map of the selected parameters for landslide suscep- zones with the help of the Table 7. According to the tibility mapping. The value of fuzzy membership resultsof ROC curve, theareaunder curve (AUC) ranges from 0 to 1. Therefore, values of prepared values of landslide susceptibility maps, prepared by fuzzy maps of all selected parameters must be the Fuzzy-LNRF and Fuzzy-AHP models are 91% and ranged from 0 to 1. The weights of parameters, cal- 90% which are indicating the excellent potentiality of culated by AHP method, have been integrated with these models for landslide susceptibility mapping fuzzy maps of selected parameters to generate a sin- (Fig. 6a, b). The LSM of both models have also been gle layer of landslide susceptibility with the help of Table 5 Parameters wise weights, matrix and consistency ratio using AHP Parameters Rainfall Slope Elevation Aspect Geology Soil Distance Distance from Distance LULC NDVI TWI Weights Texture from River lineament from Road Rainfall 1 0.180 Slope 0.5 1 0.162 Elevation 0.33 0.2 1 0.086 Aspect 0.2 0.5 1 1 0.084 Geology 0.5 0.5 0.5 2 1 0.080 Soil Texture 0.5 0.2 2 1 0.5 1 0.095 Distance from 0.2 0.2 0.5 0.5 0.2 0.5 1 0.108 River Distance from 0.2 0.2 0.5 0.5 0.5 0.2 0.5 1 0.046 lineament Distance from 0.5 0.2 0.5 1 0.5 0.5 2 2 1 0.062 Road LULC 0.17 0.2 0.5 0.2 0.5 0.5 0.5 0.5 0.5 1 0.051 NDVI 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 2 1 0.025 TWI 0.14 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.5 2 1 1 0.021 Consistency Ratio = 0.078 Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 14 of 18 Table 6 Areal distributions of LSM based on LNRF and Fuzzy- validated through RMSE method. The RMSE is calcu- AHP models lated using the Eq.7. Landslide LNRF model Fuzzy-AHP model vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi susceptibility Area in sq. km % of Area Area in sq.km % of Area uX zones ðÞ P‐A Low 950.15 30.17 1192.97 37.88 i¼1 RMSE ¼ ð7Þ Medium 1114.40 35.39 1237.17 39.29 High 1084.46 34.44 718.86 22.83 Where, N is the numbers of samples, A is the ob- served values and P is the predicted values. Can et al. (2005) has considered the RMSE value of 0.5, as the cut- off value. The values of RMSE > 0.5 and < 0.5, indicate the bad predictive and good predictive model. In the Fig. 5 Picture showing the landslide sites – from the google earth image. a Lish catchment (26°57′N, 88°30′17“E), b Nimbong Khasmahal (26°58’04”N, 88°34′16“E), c Sittong (26°52’N, 88°22’30”E) and from the field, d Near Kurseong town (26°55′46.32″N, 88°19′50.38″E), e Near Darjeeling town (26°53′13.37″N, 88°17′46.45″E), f Nathula road (26°54′43.14″N, 88°18′21.10″E) Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 15 of 18 Table 7 Characteristics of AUC of ROC curve (Yesilnacar 2005) In this work, Fuzzy-AHP and Fuzzy-LNRF, the two ef- ficient and easily operate ensemble models have been AUC Values Characters used for delineating landslide susceptibility and com- 0.6–0.5 Average pared with them. In these two ensemble models Fuzzy- 0.7–0.6 Good AHP and Fuzzy-LNRF, input process, calculations and 0.8–0.9 Very Good output process are very simple and can be readily under- 1–0.9 Excellent stood. The geo-environmental factors like slope, aspect, altitude, geology, soil texture, distance from river, dis- present study, the results of RMSE are justified Fuzzy- tance from lineament, distance from road, rainfall, land LNRF (RMSE = 0.14) and Fuzzy-AHP (RMSE = 0.18) use/ land cover, NDVI and TWI have been considered models as good predicative models for landslides suscep- for the determination of the landslide susceptible area of tibility mapping. the Darjeeling district. According to the Fuzzy-LNRE and Fuzzy-AHP maps 34.44% and 22.83% areas (Fig. 7a and b) of the district fall under the high susceptibility of Discussions landslide. Based on these findings, it can be acknowl- In the landslide prone areas appropriate methods of edged that the high susceptibility zone delineated by the landslide susceptibility mapping play significant task in Fuzzy-LNRF method is forecasting greater percentage of providing a proper approach to authorities and decision the landslide area. Additionally, the reliability of the re- makers. Very fundamental information regarding the sults of two susceptibility maps has been validated based landslide conditioning factors can be acquired from the on the known landslide dataset by employing the area landslide susceptibility mapping (LSM) and it can be an under the curve (AUC) of the receiver operating charac- essential way in hazard mitigation measures and man- teristics (ROC) and RMSE. A landslide inventory map agement. There is a large number of weight combining has been prepared considering the multiple field works methods for preparing the landslide susceptibility map. and Google Earth Images. Out of the 326 landslides 246 The results of some multi-criteria decision analysis such (70%) locations have been used for training data and as AHP, LNRF, Fuzzy logic, Artificial Neural Network remaining 80 (30%) have been used as testing data. Re- Support Vector Machine, Logistic Regression and Fre- spective AUC values of 91% and 90% for fuzzy-LRNF quency Ratio are a little bit varied from region to region. and fuzzy-AHP proved that the map produced by the Some researchers such as, Malik et al. (2016), fuzzy-LRNF model looks like having a better accuracy Mohammadi et al. (2014), Shadfar et al. (2011)and than the fuzzy-AHP model. This finding is helpful for Gupta and Joshi (1990)pointed outthatLNRF emergency situation because time is very significant in method is suitable method in landslide susceptibility hazard studies. It could be assessed that the models ap- mapping. On the other hand, Abedini and Tulabi plied have relatively similar accuracies. Methods such as (2018) was indicated that LNRF is not suitable LNRF (Gupta and Joshi 1990 & Abedini and Tulabi method like frequency ratio and AHP. 2018) and fuzzy-AHP (Roodposhti et al. 2014) were used Fig. 6 Validation of LSMs by ROC curve a for Fuzzy-LNRF model b for Fuzzy-AHP model Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 16 of 18 Fig. 7 Graph showing the distribution of area under different susceptibility classes a Percentage of area and b Area in sq.km in mapping the landslide susceptibility. Compared to LSM: Landslide susceptible map; LULC: Land use/Land cover; MCDA: Multi- criteria decision approach; MF: Membership function; NDVI: Normalized these previous studies the used methods i.e. fuzzy-LNRF differential vegetation index; OLI: Operational land imager; RMSE: Root mean and Fuzzy-AHP in this study are showing better per- square error; ROC: Receiver operating characteristic; TM: Thematic mapper; formance in preparing the LSM. Taking into account the TWI: Topographical wetness index; VIF: Variance inflation factors performances, we suggest that both Fuzzy-LNRF and Acknowledgments Fuzzy-AHP models can be used in landslide studies, as Authors would like to thanks the inhabitants of Darjeeling District because they are capable of producing flawless and stable land- they have helped a lot during our field visit. The authors would like to give special thanks to two anonymous reviewers for their constructive and useful slide susceptible maps for mitigating the risk and man- comments during the review process. Also we would like to cordially thank agement planning. There are little differences in result the Mr. Arnab Chatterjee, Assistant Professor, Department of English, between the Fuzzy-LNRF and Fuzzy-AHP derived sus- Harishchandrapur College, Pipla, Malda for correcting the grammers and language. At last, authors would like to acknowledge all of the agencies and ceptibility maps. Moreover, the Fuzzy-LNRF model is individuals specially, Survey of India (SOI), Geological Survey of India (GSI) promising and sufficient to be advised as a method to and USGS for obtaining the maps and data required for the study. prepare landslide susceptibility map at regional scale. Authors’ contributions Both authors wrote the manuscript and developed the research Conclusions methodology. Both authors also read and approved the final manuscript. For the prevention of human lives and property, a short and long-term solution is necessary for mitigating the Funding No funding was received for this work. landslide risk in this region. At present day, landslide is to be considered the most serious natural hazards in the Availability of data and materials Darjeeling district. The study has been adopted the suit- Rainfall and Temperature were received from Indian metrological able multi-criteria decision making approaches like Department. Population data was collected from District Statistical Handbook, Census of India 2011. Landsat images and DEM will freely be Fuzzy-AHP and Fuzzy-LNRF to outline the landslide availed from https:// earthexplorer.usgs.gov/ website. Soil map was collected susceptibility zones. The landslide susceptibility maps of from National Bureau of Soil Survey and Land Use Planning. Geology map both models have been categorized into three classes was taken from Geological Survey of India. such as low, moderate and high landslide susceptibility Competing interests zones. The high landslide susceptibility zone has been The authors declare that they have no competing interests. found in the middle and northern portions of the study area because of the presence of fragile soil, high concen- Received: 23 November 2018 Accepted: 15 July 2019 tration of drainage, frequent heavy rainfall and sloppy land. Among these two ensemble models, Fuzzy-LNRF References is showing better acceptability than the other model. So, Abedini, M., and S. Tulabi. 2018. Assessing LNRF, FR, and AHP models in landslide susceptibility mapping index: A comparative study of Nojian watershed in the study has been done by these two ensemble models Lorestan province, Iran. Environmental Earth Sciences 77 (11): 405. help to understand the landslide hazards problem of an Aleotti, P., and R. Chowdhury. 1999. Landslide hazard assessment: Summary area. The study also provides the essential information review and new perspectives. Bulletin of Engineering Geology and the Environment 58 (1): 21–44. https://doi.org/10.1007/s100640050066. to the planner, government, local people and researchers Arnone, E., A. Francipane, A. Scarbaci, C. Puglisi, and L.V. Noto. 2016. Effect of to take suitable steps for the reduction and the mitiga- raster resolution and polygon-conversion algorithm on landslide tion of the landslides problems in the Darjeeling district. susceptibility mapping. Environmental Modelling & Software 84: 467–481 https://doi.org/10.1016/j.envsoft.2016.07.016. Abbreviations Ayalew, L., and H. Yamagishi. 2005. The application of GIS-based logistic AHP: Analytical hierarchical process; ASTER: Advanced space borne thermal regression for landslide susceptibility mapping in the Kakuda-Yahiko emission and reflection radiometer; DEM: Digital elevation map; Mountains, Central Japan. Geomorphology 65 (1–2): 15–31. https://doi.org/1 GIS: Geographical information system; LNRF: Landslide Numerical Risk Factor; 0.1016/j.geomorph.2004.06.010. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 17 of 18 Bhandari, R.K. 2004. Landslide hazard zonation: Some thoughts. In Coping with Mcbratney, A., and I.O. Odeh. 1997. Application of fuzzy sets in soil science: Fuzzy natural hazards: Indian context, ed. K.S. Valdiya, 134–152. Orient Longman: logic, fuzzy measurements and fuzzy decisions. Geoderma 77 (2–4): 85–113. Hyderabad. https://doi.org/10.1016/s0016-7061(97)00017-7. Bouma, N., and A. Imeson. 2000. Investigation of relationships between measured Mirnazari, J., S. Khezri, and H. Shahabi. 2015. Assessment and zoning of landslide field indicators and erosion processes on badland surfaces at Petrer, Spain. hazard using AHP model and fuzzy logic operators in Posht tang watershed Catena 40 (2): 147–171. https://doi.org/10.1016/s0341-8162(99)00046-6. of Sar pole Zahab (Kermanshah province). Geography and Development 12 (37): 53–70. Budimir, M.E.A., P.M. Atkinson, and H.G. Lewis. 2015. A systematic review of Mohammadi, A.T., A. Irani, and J. Sorur. 2014. The preparation of landslide map landslide probability mapping using logistic regression. Landslides 12 (3): by landslide numerical risk factor (LNRF) model and geographic information 419–436. Can, T., H.A. Nefeslioglu, C. Gokceoglu, H. Sonmez, and T.Y. Duman. 2005. system (GIS). The Egyptian Journal of Remote Sensing and Space Science 17 (2): 159–170. Susceptibility assessments of shallow earthflows triggered by heavy rainfall at three catchments by logistic regression analyses. Geomorphology 72 (1–4): Mokarram, M., and A.R. Zarei. 2018. Landslide susceptibility mapping using fuzzy- 250–271. https://doi.org/10.1016/j.geomorph.2005.05.011. AHP. Geotechnical and Geological Engineering. https://doi.org/10.1007/s10706- Chung, C., and A.G. Fabbri. 2003. Validation of spatial prediction models for 018-0583-y. landslide Hazard mapping. Natural Hazards 30 (3): 451–472. https://doi.org/1 Moore, I.D., R.B. Grayson, and A.R. Ladson. 1991. Digital terrain modelling: A 0.1023/b:nhaz.0000007172.62651.2b. review of hydrological, geomorphological, and biological applications. Malley, L.O. 1999. Bengal district gazetteer: Darjeeling, Government of West Hydrological Processes 5 (1): 3–30. https://doi.org/10.1002/hyp.3360050103. Bengal. pp. 2–22 Mosavi, S.M., M. Abedini, A. Esmali, and F. Madani. 2017. Landslide hazard zonation by using fuzzy MCDM models in the GIS (case study: Izeh urban District Statistical Handbook 2013. Bureau of Applied Economics and Statistics, watersheds of Khuzestan). Journal of Watershed Management Research 7 (14): Department of Statistics & Programme Implementation. Government of West 87–78. https://doi.org/10.29252/jwmr.7.14.87. Bengal, Kolkata. Dormann, C.F., J. Elith, S. Bacher, C. Buchmann, G. Carl, G. Carré, J.R.G. Marquéz, B. Negnevitsky, M. 2002. Artificial intelligence: A guide to intelligent systems, Gruber, B. Lafourcade, P.J. Leitão, T. Münkemüller, C. McClean, P.E. Osborne, B. AddisonWesley/Pearson education. Harlow. Reineking, B. Schröder, A.K. Skidmore, D. Zurell, and S. Lautenbach. 2012. Panikkar, S.V., and V. Subramanyan. 1996. A geomorphic evaluation of the Collinearity: A review of methods to deal with it and a simulation study landslides around Dehradun and Mussoorie, Uttar Pradesh, India. evaluating their performance. Ecography 36: 27–46 https://doi.org/10.1111/ Geomorphology 15: 169–181. j.1600-0587.2012.07348.x. Pourghasemi, H.R., M. Mohammady, and B. Pradhan. 2012. Landslide susceptibility Dou, J., A.P. Yunus, D.T. Bui, A. Merghadi, M. Sahana, Z. Zhu, et al. 2019. mapping using index of entropy and conditional probability models in GIS: Assessment of advanced random forest and decision tree algorithms for Safarood Basin, Iran. Catena 97: 71–84. https://doi.org/10.1016/j.catena.2012.05.005. modeling rainfall-induced landslide susceptibility in the Izu-Oshima Volcanic Pradhan, B. 2013. A comparative study on the predictive ability of the decision Island, Japan. Science of the Total Environment 662: 332–346. tree, support vector machine and neuro-fuzzy models in landslide susceptibility mapping using GIS. Computational Geosciences 51: 350–365 Fan, X., G. Scaringi, G. Domènech, F. Yang, X. Guo, L. Dai, et al. 2019. Two multi- https://doi.org/10.1016/j.cageo.2012.08.023. temporal datasets that track the enhanced landsliding after the 2008 Wenchuan earthquake. Earth System Science Data 11 (1): 35–55. Pradhan, B., and S. Lee. 2009. Regional landslide susceptibility analysis using back- Feizizadeh, B., and T. Blaschke. 2011. Landslide risk assessment based on GIS propagation neural network model at Cameron Highland, Malaysia. multi-criteria evaluation: A case study Boston Abad county, Iran. Journal of Landslides 7 (1): 13–30. https://doi.org/10.1007/s10346-009-0183-2. Earth Science and Engineering 1: 66–71. Pradhan, B., and S. Lee. 2010. Delineation of landslide hazard areas on Penang Feizizadeh, B., and T. Blaschke. 2013. GIS-multicriteria decision analysis for Island, Malaysia, by using frequency ratio, logistic regression, and artificial landslide susceptibility mapping: Comparing three methods for the Urmia neural network models. Environmental Earth Sciences 60 (5): 1037–1054. lake basin, Iran. Natural Hazards 65 (3): 2105–2128. Refahi, H.Gh. 2000. Soil Erosion by Water & Conservation. 2nd ed., Tehran Feizizadeh, B., M.S. Roodposhti, P. Jankowski, and T. Blaschke. 2014. A GIS-based University Publications, Tehran, Iran, pp. 551. extended fuzzy multi-criteria evaluation for landslide susceptibility mapping. Reichenbach, P., M. Rossi, B.D. Malamud, M. Mihir, and F. Guzzetti. 2018. A review Computers & Geosciences 73: 208–221. https://doi.org/10.1016/j.cageo.2014.08.001. of statistically-based landslide susceptibility models. Earth-Science Reviews 180: 60–91. Garcia-Rodriguez, M.J., J.A. Malpica, and B. Benito. 2008. Susceptibility assessment of earthquake-triggered landslides in El Salvador using logistic regression. Roodposhti, M.S., S. Rahimi, and M.J. Beglou. 2014. PROMETHEE II and fuzzy AHP: Geomorphology 95: 172–191. https://doi.org/10.1016/j.geomorph.2007.06.001. An enhanced GIS-based landslide susceptibility mapping. Natural Hazards 73 Gerrard, J. 1994. The landslide hazard in the imalayas: Geological control and (1): 77–95. human action. Geomorphology 10: 221–230. Saaty, T.L. 1980. The analytical hierarchy process, 350. New York: McGraw Hill. Gupta, R., and B. Joshi. 1990. Landslide hazard zoning using the GIS approach—A Saaty, T.L. 1990. The analytic hierarchy process: Planning, priority setting, resource case study from the Ramganga catchment, Himalayas. Engineering Geology allocation. 1st ed, 502. Pittsburgh: RWS publications. 28 (1–2): 119–131. https://doi.org/10.1016/0013-7952(90)90037-2. Saaty, T.L. 1994. Fundamentals of decision making and priority theory with analytic hierarchy process. 1st ed, 527. Pittsburgh: RWS publications. Guzzetti, F., P. Reichenbach, F. Ardizzone, M. Cardinali, and M. Galli. 2006. Estimating the quality of landslide susceptibility models. Geomorphology 81 Saaty, T.L., and L.G. Vargas. 1998. Diagnosis with dependent symptoms: Bayes (1–2): 166–184. https://doi.org/10.1016/j.geomorph.2006.04.007. theorem and the analytic hierarchy process. Operations Research 46 (4): 491–502. Hejazi, S.A. 2015. Landslide hazard mapping in Goijabel of Ahar using GIS. J Geogr Saha, S. 2017. Groundwater potential mapping using analytical hierarchical process: Plan 18 (20): 135–152. A study on Md. Bazar block of Birbhum District, West Bengal. Spatial Information Hembram, T.K., and S. Saha. 2018. Prioritization of sub-watersheds for soil Research 25 (4): 615–626. https://doi.org/10.1007/s41324-017-0127-1. erosion based on morphometric attributes using fuzzy AHP and Sarkar, S. 1999. Landslides in Darjiling Himalayas. Transactions Japanese compound factor in Jainti River basin, Jharkhand, eastern India. Geomorphological Union 20: 299–315. Environment Development and Sustainability:1–28. https://doi.org/10.1007/ Shadfar, S., M. Yamani, and M. Namaki. 2011. Zoning land slide hazard by s10668-018-0247-3. information value method (IVM), density area (DA) and landslide numerical Malczewski, J. 1999. GIS and multicriteria decision analysis. New York: Wiley. risk factor (LNRF) model in Chalkrood. Journal of Watershed Management in Engineering 3 (1): 40–47. Malik, R., S. Ghosh, S. Dhar, P. Singh, and M. Singh. 2016. GIS based landslide hazard from zonation along national highway AY-58, from Rishikesh to Tien Bui, D., B. Pradhan, O. Lofman, and I. Revhaug. 2012. Landslide susceptibility Joshimath, Uttarkhand, India. International Journal of Advanced Remote assessment in Vietnam using support vector machines, decision tree, and Sensing Geography 4 (2): 41–55. Naïve Bayes models. Mathematical Problems in Engineering 2012: 1–26. Mallick, J., R.K. Singh, M.A. Alawadh, S. Islam, R.A. Khan, and M.N. Qureshi. 2018. https://doi.org/10.1155/2012/974638. GIS-based landslide susceptibility evaluation using fuzzy-AHP multi- Torkashvand, A.M., A. Irani, and J. Sorur. 2014. The preparation of landslide map criteria decision-making techniques in the Abha watershed, Saudi by landslide numerical risk factor (LNRF) model and geographic information Arabia. Environmental Earth Sciences 77 (7). https://doi.org/10.1007/ system (GIS). The Egyptian Journal of Remote Sensing and Space Science 17 (2): s12665-018-7451-1. 159–170. https://doi.org/10.1016/j.ejrs.2014.08.001. Roy and Saha Geoenvironmental Disasters (2019) 6:11 Page 18 of 18 Turner, A.K., and L.R. Schuster. 1996. Landslides: Investigation and mitigation. Washington, DC: National Academy Press. Van Eynde, E., S. Dondeyne, M. Isabirye, J. Deckers, and J. Poesen. 2017. Impact of landslides on soil characteristics: Implications for estimating their age. Catena. https://doi.org/10.1016/j.catena.2017.05.003. Van Westen, C.J., R. Soeters, and K. Sijmons. 2000. Digital geomorphological landslide hazard mapping of the Alpago area, Italy. International Journal of Applied Earth Observation and Geoinformation 2: 51–60. Wang, H., G. Wang, F. Wang, K. Sassa, and Y. Chen. 2008. Probabilistic modeling of seismically triggered landslides using Monte Carlo simulations. Landslides 5: 387–395. https://doi.org/10.1007/s10346-008-0131-6. Yesilnacar, E.K. 2005. The application of computational intelligence to landslide susceptibility mapping in Turkey, PhD Thesis, 423. Australia: University of Melbourne. Youssef, A.M., M. Al-kathery, and B. Pradhan. 2014. Landslide susceptibility mapping at Al-Hasher Area, Jizan (Saudi Arabia) using GIS based frequency ratio and index of entropy models. Geosciences Journal. https://doi.org/10.1 007/s12303-014-0032-8. Zadeh, L.A. 1965. Fuzzy sets. Information and Control 8 (3): 338–353. Zêzere, J.L., S. Pereira, R. Melo, S.C. Oliveira, and R.A. Garcia. 2017. Mapping landslide susceptibility using data-driven methods. Science of the Total Environment 589: 250–267. Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Journal

Geoenvironmental DisastersSpringer Journals

Published: Aug 1, 2019

There are no references for this article.