LANDSLIDE SUSCEPTIBILITY EVALUATION IN SUKAMAKMUR AND ITS SURROUNDING AREA, BOGOR DISTRICT, USING WEIGHT OF EVIDENCE (WOE), LOGISTIC REGRESSION (LR) AND THEIR COMBINATIONS

• Main Author: Anindy Ismiralda, Dinta
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/53856
• Institution: Institut Teknologi Bandung

Bogor district is an area that has a high level of landslide susceptibility. The areas most frequently affected by this disaster are Sukamakmur and its surroundings. Landslide susceptibility is influenced by several factors such as slope, slope direction, lithology, soil type, land cover, elevation, rainfall, straightness, curvature, distance from river and roads. The purpose of this research is to find out a method that has the best accurancy level for evaluating the level of landslide susceptibility in Sukamakmur area and its surroundings, also to analyze the factors causing landslides. The method used in this research is a quantitative method including Weight of Evidence (WoE), Logisitic Regression (LR), and their combinations WoE is a quantitative technique that uses a number of data combinations to produce map of data marking based on the level of its probability. Logistic Regression (LR) is a multivariate statistical method that can explain the connection between the response variable and the independent variable. The LR method can only analyze numerical data. The combination method is a combination of the bivariate (WoE) and multivariate (LR) methods. The combination method can evaluate the classification of each parameter through WoE weighting and analyze the correlation from each parameter using LR method. The landslide susceptibility map generated from the three methods will be validated using the Area Under Curve, Seed Cell Area Index, and Spatial Domain methods Based on the results of field research and observations of Landsat 8, there are 249 landslide points. That data was divided into two groups, 70% of the data were used to create models and 30% of the data were used to validate the previously created models. The research stages include weighting of 12 parameters, there are slope, slope direction, lithology, soil type, land cover, elevation, rainfall,vegetation density (NDVI) distance from lineament, curvature, distance from river, and distance from roads using the WoE weighting method. Each paramet that has been weighted will be arranged by its parameter members from the largest to the smallest weight values to calculate the AUC value. Based on the AUC value, the parameters that have an effect and can be used in to make landslide susceptibility models are nine parameters, there are slope direction, elevation, rainfall, lithology, soil type, slope, distance from lineament, distance from rivers, and land cover. The most influential parameter is the rainfall with an AUC of 0.886 and the distance from straightness with an AUC value of 0.838. From AUC evaluation results, the combination method has a value of 0.806. it is greatest than the AUC value of the WoE method of 0.782 and LR 0.700. The evaluation results with the SCAI (Seed Cell Area Index) value of the landslide susceptibility model analyzed by the combination method have better results than another models, SCAI value for WoE and Combination method is systematic. The model analyzed by the LR method has systematic Value, but it has lower value than WoE and combination method. The combination method creates the best results for all level susceptibility. Based on the results of Spatial Domain analysis, the combination method and the WoE method create a higher level of pixel correctness compared to the analysis results of the WoE method with LR method and the LR method with WoE-LR method. The result of this research shows that combination method is the most accurate and effective method compared to the WoE method and LR method to analyze the lanslide susceptibility in research areas.