LANDSLIDE SUSCEPTIBILITY ZONATION OF PUNGUT HILIR AREAS AND SUROUNDING USING WEIGHT OF EVIDENCE (WOE) AND ARTIFICIAL NEURAL NETWORK (ANN)

• Main Author: Zaky Khutby, Muhammad
• Format: Theses
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/55982
• Institution: Institut Teknologi Bandung

Landslides are one of the most frequent geological disasters in Indonesia every year. Landslides that occur are usually local or local but spread throughout Indonesia. The Pungut Hilir area in Kerinci Regency is one area that is prone to landslides. At this location, landslides often occur around the highway to cause material losses and cause fatalities. Therefore, it is crucial to conduct a landslide susceptibility assessment at this location. The landslide susceptibility assessment was carried out using the WoE and ANN methods. For landslide susceptibility analysis (ANN and WoE methods) to be carried out, several landslide point data are needed. In this study, 83 landslide point data were used, divided into two data sets (training set and testing set). The two data sets were then made into several simulations based on the percentage comparison of the two, namely DS 60:40, DS 70:30, and DS 85:15. As for the ANN method, it is necessary to add 323 non-landslide points and then divide by the same percent in the training set and testing set. Four factors were used during the landslide susceptibility analysis: topography, geology, hydrology, and human activities. The human activities factor consisted of 18 parameters (elevation, slope, slope direction, curvature, TWI, SPI, lithology, distance from fault, distance from straightness, earthquake, rainfall, distance from river, flow direction, flow accumulation, land use, distance from road, NDVI, and NDWI). Through the WoE and ANN methods, the effect of these parameters in each simulation on landslide events at the research site will be known. Based on the analysis using the WoE method, it is known that in DS 60:40 the most influential parameter is the aspect, in DS 70:30 the most influential factor is the distance from the road, and in DS 85:15 the most influential factor is the distance from the fault. Based on the analysis using the ANN method, it is known that the parameter that has the most influence is the slope in each simulation (DS 60:40, DS 70:30, dan DS 85:15). Then, validation was carried out on each simulation in both methods (WoE and ANN). Validation is based on the AUC value (success rate and prediction rate). In the research area, the validation results of the WoE method in DS 60:40 had AUC values of 0.67 and 0.64, DS 70:30 of 0.7 and 0.69, and DS 85:15 of 0.63 and 0.62. The validation of the ANN method in DS 60:40 has AUC values of 0.76 and 0.75, DS 70:30 of 0.76 and 0.76, and DS 85:15 of 0.71 and 074. In addition, validation is also carried out using the spatial domain. Based on this validation, it is known that the correct and acceptable pixels in DS 60:40 are 65.92%, in DS 70:30 it is 73.99%, and in DS 85:15 it is 76.91%. the result of the kappa coefficient value of the ANN method in DS 70:30 had a better level accuracy than other. Based on the results of the study, it was concluded that the ANN method in DS 70:30 had a better level of accuracy compared to other methods for landslide susceptibility maps in the Pungut Hilir area and its surroundings.