GIS-Based Landslide Susceptibility Modeling Using Frequency Ratio and Logistic Regression at Abuna Gindeberet area, West Shewa Zone, Central Ethiopia

Abstract

Landslide is one of the greatest disasters that cause the damages to properties and loss of life. Landslide susceptibility mapping was carried out for the Abuna Gindeberet area of the West Shewa Zone in central Ethiopia. The main objective of this study is to prepare causative factors and landslide susceptibility maps using GIS-based frequency ratios and logistic regression models. Based on Google Earth imagery, about 1222 landslides were identified and classified randomly into training landslide datasets (70%) for model development and the remaining (30%) validation landslide datasets in the ArcGIS environment. The eight landslide causative factors like slope, aspect, curvature, lithology, land use/land cover, distance to lineament, distance to stream, and rainfall, were identified and integrated with training landslides to determine the weights of each factor class and landslide causative factors using frequency ratio and logistic regression models, respectively. The landslide susceptibility maps were produced by overlaying the weights of all the landslide causative factors using the raster calculator of the spatial analyst tool in ArcGIS 10.8. The final landslide susceptibility maps were reclassified as very low, low, moderate, high, and very high susceptibility classes in both frequency ratio and logistic regression models. The results of landslide susceptibility maps produced by the frequency ratio model indicate that the very low, low, moderate, high, and very high susceptibility classes cover 9%, 21%, 24%, 34%, and 12% of the area, respectively. Similarly, the landslide susceptibility maps produced from the logistic regression model revealed that 7%, 33%, 9%, 28%, and 23% of the area fall in the very low, low, moderate, high, and very high classes respectively. These susceptibility maps were validated by the receiver operating characteristic (ROC) of the area under the curve (AUC). The AUC results of the frequency ratio and logistic regression models showed success rates of 0.857 and 0.849, respectively, while the prediction rates for these models are 0.828 and 0.831, respectively. Thus, the landslide susceptibility map (LSM) using the validation dataset provided acceptable results, and the different classes that are delineated using these models can be used in the future for safe development planning in the study area.