Statistical Approaches for Landslide Susceptibility Assessment and Prediction

Landslide is one of the destructive environmental hazards which causes a lot of damages to human lives and properties. Various approaches and techniques have been applied to assess the spatial distribution of landslides all over the world. Amongst them physical models, slope stability models, statistical and probabilistic models are very much important in the study of landslide assessment and prediction. In the present study, to assess the spatial distribution of landslide susceptibility in Darjeeling Himalaya several statistical models, i.e. frequency ratio (FR) model, modified information value (MIV) model, logistic regression (LR) model, artificial neural network (ANN) model, weighted overlay analysis (WOA) model, certainty factor (CF) model, analytical hierarchy process (AHP) model and fuzzy logic (FL) approach have been incorporated and finally a comparison has been made between the models on the basis of model validation results. Physical models with regard to landslides dealt with the assessment of various physical parameters of rocks and soil, i.e. shear stress, shear strength, cohesion, friction angle, pore-water pressure, grain size of soil, depth of the soil, saturated soil depth, density of water and soil, etc. All these parameters help to perform slope stability model as well as to identify the vulnerable slope in the mountain environment. In Darjeeling Himalaya, debris slide, rock fall, and earth slides are three common types of landslides. Statistical models based on RS and GIS help to identify susceptible landslide locations with accuracy. To perform all the statistical models, a landslide inventory was made based on historical landslides data, toposheet, Google earth image, and field investigation with GPS survey. The data layers, i.e. elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were integrated on GIS platform to assess pixel wise landslide susceptibility index, and finally a classification was made to prepare landslide susceptibility zonation map of Darjeeling Himalaya in connection to each model.

Landslides are caused due to the prevalence of geomorphic, tectonic, and hydrologic parameters which changes slope materials state and force them to move down slope. Darjeeling Himalaya exhibits a wide range of all these parameters and as a result of which slope failure has become a quite common phenomena in every monsoon which inflicts a great damage to human lives and properties. To mitigate the landslides and its destructive nature, the relationship between landslides location and various landslides occurrence factors are to be analysed scientifically. In the present work, various data layers such as elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were prepared in ARC GIS environment. To assess the probability of each class of the landslide causative factors, frequency ratio (FR) value was estimated considering both landslide affected pixels and landslide non-affected pixels. The derived frequency ratio established the relationship between the probability of landslide and each class of the landslide causative factor. The study revealed that steep slope zones; south, south-east, and south-west facing slope; high positive and high negative curvature areas; areas of weak lithology; places close to drainage; locations close to lineaments; high stream power index and high topographic wetness index areas; and land use character of wasteland dry/barren land, wastelands with scrub, forest, agricultural single crop and agricultural plantation are registered with high frequency ratio and high landslide probability.

The assessment of landslide susceptibility is closely associated with the spatial distribution of landslides. In the present study, both frequency ratio (FR) model and modified information value (MIV) model were applied to analyse landslide susceptibility in Darjeeling Himalaya. Both the models dealt with the relationship between landslide phenomena and landslide conditioning factors. To perform the models data layers, i.e. elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were taken into account. Each and every class/category of landslide conditioning factor contributes a relative importance in landslide occurrences. To prepare all the data layers, Landsat TM image, SRTM DEM, Google earth image, and some authorized maps were processed in accordance with ArcMap 10.1 and Erdas imagine 9.2. To obtain the relative significance of each class/category of landslide conditioning factors, frequency ratio (FR) value and modified information value (MIV) were estimated and accordingly the ranking values were assigned to each class/category to integrate all the data layers on GIS platform as well as to prepare landslide susceptibility map of Darjeeling Himalaya. The derived landslides susceptibility maps by using frequency ratio model and modified information value model were verified being considering the area under curve (AUC) of ROC curve and frequency ratio plot. The AUC value of ROC curve of FR model and MIV model was 0.746 and 0.769, respectively. The AUC value represents the prediction accuracy of landslide susceptibility map was 74.6% for frequency ratio model and 76.9% for modified information value model.

The application of geo-informatics has brought a new dimension in the study of landslide susceptibility assessment and prediction all over the world for regional development and planning of mountain terrain. The present study is dealt with the preparation of a landslide susceptibility map of Darjeeling Himalaya, a tectonically active section of Himalayan mountain range using logistic regression model on GIS environment. A landslide inventory map was developed in consultation with topographical maps, Google earth image, satellite images, and historical landslide events and was verified with the field data. In the study 15 landslide conditioning factors, i.e. elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were taken into account and finally their integration has been made on GIS environment with the help of estimated logistic regression co-efficient values to produce landslide susceptibility map of Darjeeling Himalaya. The produced susceptibility map satisfied the decision rules and the overall accuracy was acceptable. −2 Log likelihood, cox & Snell R Square, and Nagelkerke R Square values proved that the independent variables were statistically significant. The success rate curve showed the prediction accuracy of the landslide probability map which is also desirable (71.5).

The present study is dealt with the preparation of landslide susceptibility map of Darjeeling Himalaya with the help of GIS tools and artificial neural network (ANN) model. Fifteen landslide causative factors, i.e. elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were considered to produce the landslide susceptibility zonation map. To generate all these aforesaid causative factors map, topographical maps, geological map, soil map, satellite imageries, Google earth images and some other authorized maps were processed and constructed into a spatial data base using GIS and image processing techniques. The back-propagation method was applied to estimate factor’s weight and the landslide hazard indices were derived with the help of trained back-propagation weights. Then, the landslide susceptibility zonation map of Darjeeling Himalaya was made using GIS tool and classified into five, i.e. very low, low, moderate, high, and very low landslide susceptibility. To validate the prepared landslide susceptibility map, landslide inventory was used and accuracy result was obtained after processing ROC curve. The accuracy of the landslide susceptibility map was 81.5% which is desirable.

The present study is dealt with the application of weighted overlay analysis (WOA) model, certainty factor (CF) model, analytical hierarchy process (AHP) model for the preparation of landslide susceptibility zonation map of Darjeeling Himalaya. To perform three models, various data layers with regard to elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were taken into account. For the preparation of various data layers, topographical maps, Google earth images, SRTM DEM, http://​www.​worldclim.​org, satellite image (Landsat TM) and some authorized data were being processed on GIS environment (ArcMap 10.1). The prepared landslide susceptibility maps of Darjeeling Himalaya were classified into five, i.e. very low, low, moderate, high, and very high landslide susceptibility. To validate three landslide susceptibility zonation maps derived from WOA, CF, and AHP models, ROC Curve and frequency ratio plot methods were incorporated. ROC curve showed the level of accuracy of each landslide susceptibility map. The study revealed that WOA, CF, and AHP were with the accuracy level of 65.4%, 81.2%, and 67.5%. Frequency ratio plots sugessted that moderate, high, and very high landslide susceptibility zones in Darjeeling Himalaya are experienced with greater probability landslide phenomena.

The development of various models and their application in studies have brought a significant change in the subject discipline of geography. In the present study, various geomorphic and geohydrologic parameters, i.e. elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were considered, and their integration was made on GIS environment to prepare landslide susceptibility zonation map of Darjeeling Himalaya, India. To generate all data layers, Google earth imagery, toposheet and GPS field survey data (2015–2016); geological and soil map; SRTM DEM (30 m spatial resolution); Landsat TM Image, Feb. 2009 (30 m spatial resolution), rainfall data (1950–2010) and some other information were processed with the help of GIS. To integrate all the data layers and to prepare landslide susceptibility map, several models such as frequency ratio (FR) model, modified information value (MIV) model, logistic regression (LR) model, artificial neural network (ANN) model, weighted overlay analysis (WOA) model, certainty factor (CF) model, analytical hierarchy process (AHP) model and fuzzy logic (FL) approach were applied. The prepared landslide susceptibility maps using all the models were classified into five, i.e. very low, low, moderate, high, and very high. All the developed landslide susceptibility maps of Darjeeling Himalaya were being validated using receiver operating characteristics (ROC) curve). The study concluded that artificial neural network model (ANN), certainty factor (CF) model, and frequency ratio-based fuzzy logic approach are most reliable statistical techniques in the assessment and prediction of landslide susceptibility in Darjeeling Himalaya because of high level of accuracy in comparison to models applied in the study.