Landslide Susceptibility Mapping by Using Logistic Regression Model with Neighborhood Analysis: A Case Study in Mizunami City

Landslides which affect human lives and economic losses are always attracted a lot of concerning in modern society. In order to identify the potential hazardous areas related to landslides, three methods have been used, such as qualitative or knowledge-based method, deterministic method and quantitative-based method. Geographical information system (GIS) technology and high computing ability provide a convenient tool to deal with landslide triggering factors and make the quantitative-based method achieve effectively. In this study, landslide-related factors such as topographical elevation, slope angle, slope aspect, topographical wetness index (TWI) and stream power index (SPI), were employed in the landslide susceptibility analysis. The logistical regression was used to obtain the relationships for landslide susceptibility between landslides and causative factors. The distributions of observed landslides were used to evaluate the performance of the susceptibility map. The approaches described in this paper showed us that the logistical regression and neighborhood can be used as simple tools to predict the potential landslide locations. This map will be helpful for city planning, infrastructure construction and agriculture developments in the future. 1. INTRODUCTION Landslides which cause the loss of human life and the damage to the social economy were attracted a lot of attention over the last decades. According a previous study Schuster (1996) [1] , landslides will increase in the next decades due to continued deforesting and the changing climatic patterns in landslide-prone areas. In order to identify the landslide-related area, there are three main approaches to assess the landslide susceptibility: qualitative methods, deterministic methods and quantitative methods. In the late 1970s, qualitative approaches were widely applied by engineering geologists to evaluate landslide. Deterministic approaches focus on slope geometry, shear strength data, and pore-water related data (Netra et al., 2010) [2] but lack of taking climatic and human induced factors into accounted. Nowadays, the rapid developments of computer technology and geographic information system (GIS) provide a convenient tool to deal with landslide triggering factors and make the quantitative-based method achieve effectively. Among a lot of quantitative methods, logistic model was recognized as the suitable approach to assess landslide susceptibility because it is free of data distribution and can handle a variety of datasets (Nandi et al., 2009) [3]. In this study, neighborhood analysis " seed cell approach " proposed by (Suzen and Doyuran., 2004) [4] and logistic model were applied to create the relationship between landslides and controlling factors with GIS in Mizunami