Modeling of Landslide Hazard Assessment Using Geo Informatics Techniques

Abstract

Landslide is very common phenomenon all over the world causing loss of life, property, and infrastructure. One of the important ways to reduce these damages is to carry out landslide hazard assessment to identify areas where and when landslide will occur. A number of methods have been developed and applied for landslide hazard assessment but no agreement has been reached for the best method applicable to all areas. For this objective, the present study has been carried out in part of Utarakhand state, India and the Mu Cang Chai district, Viet Nam where experience a lot of landslides every year. A novel model namely Rotation Forest Fuzzy Rules Based Classifier Ensemble (RFCE) has been developed at the Uttarakhand area, India for landslide spatial prediction, which is based on hybrid intelligent approach of two state of the art machine learning methods namely Rotation Forest (RF) ensemble and Fuzzy Unordered Rules Induction Algorithm (FURIA) classifier. For doing this task, landslide inventory map with 930 landslide locations has been prepared with the help of Remote Sensing (RS) and field data. Landslide affecting factors (slope, aspect, elevation, curvature, plan curvature, profile curvature, lithology, soil, distance to lineaments, lineament density, land cover, rainfall, distance to roads, road density, river networks, distance to river, and river density) have been selected for landslide spatial prediction analysis using Geoinformatics techniques. Predictive capability of proposed RFCE model has been compared with other popular models such as Support Vector Machines (SVM), Logistic Regression (LR), Fisher s Linear Discriminant Analysis (FLDA), Naïve Bayes (NB), Bayesian Network (BN), Multilayer Perceptron neural network (MLPN), Radial Basis Function neural network (RBFN), and Vote Feature Intervals (VFI). Performance of the models has been validated using Receiver Operating Characteristics (ROC) curve and statistical analyzing methods. Analysis result shows that the novel landslide model RFCE has the highest p