GIS-Based Landslide Susceptibility Mapping Using Frequency Ratio and Artificial Neural Networks

On Nepal’s steep hillsides, preventing landslides is an extremely challenging task. Every year, the country is affected by landslides, causing a significant number of deaths, loss of life and property, displacing families, and blocking the roads. Landslide susceptibility mapping is vital to safeguard life and property. This study focuses on preparing landslide susceptibility maps of two locations in Nepal using Frequency Ratio (FR) and Artificial Neural Network (ANN). A total of 329 landslides were mapped using historical landslide data, satellite images, and field surveys to generate the Landslide inventory maps. Landslide susceptibility mapping was performed based on twelve causative factors, namely slope gradient, slope aspect, curvature, distance to road, distance to stream, distance to fault, land cover and land use, geology, parent soil, average annual precipitation, Stream Power Index (SPI) and Topographic Wetness Index (TWI). The Receiver Operating Characteristic (ROC) and Area Under the Curve (AUC) analysis was performed to evaluate the accuracy of the models. The calculated AUC values for FR and ANN were found to be 85.62% and 87.7%, respectively, indicating that the landslide susceptibility prediction is fairly accurate. The landslide susceptibility maps produced using both methods were classified into five classes: very low, low, moderate, high, and very high. The final landslide susceptibility maps can be used for disaster risk reduction, land use planning, and early warning systems.