LANDSLIDE SUSCEPTIBILITY ZONATION OF SUNGAI PENUH WATERSHED AND SUROUNDING AREAS USING FREQUENCY RASIO (FR) AND ARTIFICIAL NEURAL NETWORK (ANN)

• Main Author: Miyafto Prabowo, Roishe
• Format: Theses
• Language: Indonesia
• Subjects: Teknik tertentu, alat, perlengkapan, materi
• Online Access: https://digilib.itb.ac.id/gdl/view/55975
• Institution: Institut Teknologi Bandung

The Sungai Penuh watershed and surroundings are part of the Sungai Penuh City, Jambi Province. This area is included in the medium to high of landslide susceptibility. The morphological condition of the area is in the form of steep hills. This is one of the controlling factor for landslides. The landslide susceptibility maps currently available are still at a provincial scale. The purpose of this research to determine landslide susceptibility maps on a more detailed scale, namely the sub-watershed scale. The method used to analyze the susceptibility of landslides consists of two types, namely bivariate statistical method and machine learning method. The bivariate statistical method is the frequency ratio (FR) and the machine learning method is the artificial neural network (ANN). The machine learning method is expected to take advantage of the speed and accuracy of calculating the landslide susceptibility map. Based on field observations and identification by remote sensing on the Google Earth Pro image, there were 98 landslides in the study area. The landslide event data is then divided into two groups randomly with three data set distribution, namely scenario A training data by 60% with testing data by 40%, scenario B training data by 70% with testing data 30%, with scenario C training data by 85% with testing data by 15%. Futhermore, twenty parameters were tested to determine their effect on landslides, namely elevation, slope, aspect, curvature, topography wetness index (TWI), stream power index (SPI), lithology, distance to fault, distance to lineament, lineament density, earthquake, rainfall, distance to drainage, drainage density, flow direction, springs density, land use, distance to road, NDVI, and NDWI. Based on the results of the evaluation of the area under (AUC), kappa coefficient, seed cell area indeks (SCAI), and spatial domain the best method is the artificial neural network (ANN) method with scenario B.