Predicting Landslide Susceptibility based on Naïve Bayes and Information Value Models in Shibam and AL-Taweelah District, Yemen

Landslides are one of the common natural hazards in Yemen, particularly in recent years, causing annual human and material damages. However, studies on landslide susceptibility mapping (LSMs) in Yemen are still limited. This study focuses on assessing the susceptibility to landslides in the Shibam and Al-Taweelah districts using the bivariate statistical model Information value (InV) and the machine learning model (Naïve byes) NB. The study identified 98 landslide event locations collected through field surveys and Google Earth Pro, which were then split into a 70/30 ratio for training and testing; 11 factors contributing to landslides were recognized. The results showed that the northern and northeastern parts are the most susceptible to landslides. Additionally, the NB model outperformed the InV model in predicting landslide susceptibility, achieving an AUC value of 0.785, while the InV model had a value of 0.674. The results of this research will assist engineers, decision-makers, and the community in understanding the risks of landslides in Shibam and Al-Tawilah, selecting suitable locations for construction and development purposes, and supporting proactive measures to prevent landslide-related disasters.