The study aims to apply statistical approaches for landslide susceptibility mapping for natural and cut slopes at two study areas in Vietnam. A landslide database from two study areas with 837 natural and 82 cut slopes was used to produce susceptibility maps to predict the landslide hazard in the future. The distribution of landslides was identified from field surveys, research reports and remote sensing images. By means of Likelihood Ratio (LR), Weight of Evidence (WoE) and Certainty Factor (CF) approaches, the tendency to landslide occurrences was assessed by relating landslide inventory (dependent variable) to a series of causal factors (independent variables) which were managed in the GIS environment. The developed models produced reliable susceptibility maps of study areas and the probability level of landslide can be divided by four different classes (low, medium, high and very high). The overall performance achieved by the LR, WoE and CF analyses was assessed on validation datasets in both two study areas with Kappa statistic (KIA) > 0.7, area under curve (AUC) > 0.85 could be considered very satisfactory for landslide susceptibility zonation. All three models give over 80% of accuracy, in which WoE give best results. Landslide zonation map shows high and very high classes account for only 25% of total area of Quang Ngai province, but they can explain for nearly 90% of existing landslide locations. The weights of statistical approaches can also provide the important level of causal factors, relatively. In which, 3 most affluent factors for natural slopes are geological engineering conditions, landuse and the rock type (lithology) of the slopes, for man-made slopes are the angle of cut slope, weathering depth and the strength of slope materials.
Landslides are more widespread than any other geological event and many factors can cause a slope to fail, such as natural occurrences or man-made activities. Landslide hazard mapping was defined as the quantitative prediction of the spatial distribution of slopes which are likely to be failure (Guzzetti et al., 1999). The causal factors that have been used for landslide hazard analysis can usually be grouped into geomorphology (topographic conditions), geology (rock types, structures, strength of slope materials), land use/land cover and hydrogeology (drainage state and ground water). However, the contributing factors are behaved differently for natural and cut slopes. A natural slope is different from a cut slope (road cuts, excavations, open-pit mining, etc.) in that the effects of rock types, fracture networks inside the rocks, the strength and weathering property of slope materials, the contribution of water on the surface or in the ground may undergone the test of time that will reveal tendencies of slumping, cracking and finally collapsing. On cut slopes, the slope angle, weathering depth, slope cover type, the property of upslope terrain, slope reinforcement, etc. may play an important role in their stability. A cut slope may expose soils that respond poorly to weathering elements, especially when the soil profile of slopes is not uniform and homogenous.