LANDSLIDE SUSCEPTIBILITY ANALYSIS USING WEIGHT OF EVIDENCE (WOE) AND LOGISTIC REGRESSION (LR) METHODS IN SALEM DISTRICT, BREBES REGENCY, CENTRAL JAVA

• Main Author: Balhum Rachmany, Balqy
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69739
• Institution: Institut Teknologi Bandung

The research area located at Salem District, Brebes Regency, Central Java. This area has many landslides phenomena, one of them was included in 2018 national disaster in Pasir Panjang Village. Hills and valleys morphological conditions are one of controlling factors for landslides. Because of the available landslide susceptibility map is still on a provincial scale, this study aims to determine the zoning for more detailed scale. Used method consists of two types, there are Weight of Evidence (WoE) method which is included in bivariate method and Logistic Regression (LR) method which is included in multivariate method. Based on field observations and remote sensing identification using Google Earth imagery, there were 257 landslide events in the research area. The incident data were randomly divided into two groups with percentage of 70% (183 landslide) as training data and 30% (74 landslide) as test data. Parameters tested include elevation, slope, slope direction, flow direction, lithology, curvature (plan, profile, and total), elevation, Normalized Difference Index (NDVI), land cover, distance from fault, distance from river, and distance from road. Based on the area under curve (AUC) value, the parameters that have the dominant influence on the occurrence of landslides using the WoE method are distance from fault, rainfall, land cover, slope, elevation, distance from road, distance from river, and NDVI. The dominant parameters that have an active influence on the occurrence of landslides in the research area using the LR method are distance from fault, elevation, rainfall, slope, and distance from river. The results of AUC evaluation of success rate showed that LR method (0.92) was better than WoE method (0.77). The results of AUC evaluation of prediction rate value show that LR method (0.90) is better than WoE method (0.67). The Landslides Susceptibility Zoning Map of Salem District using WoE method has high susceptibility zone percentage of 19.5%, a moderate susceptibility zone of 11.5%, a low susceptibility zone of 11%, and very low susceptibility zone of 58%. The results of zoning using LR method have high susceptibility zone percentage of 13%, moderate susceptibility zone of 15.5%, low susceptibility zone of 17.5%, and very low susceptibility zone of 54%.