performance evaluation of the random forest model in flood hazard assessment of the kashkan watershed

Objective: flooding is one of the most dangerous natural events worldwide, caused by a combination of climatic, hydrological, geomorphological, and geological factors. floods can occur due to heavy rainfall, prolonged rainfall, rapid snowmelt, or dam failure. regardless of the cause, floods lead to widespread destruction and damage to human societies and infrastructure. given the severe risks, assessing flood hazards has become essential. flood sensitivity maps are useful tools to analyze and manage flood-prone areas. material and methods: this study aims to identify flood-sensitive zones using the random forest (rf) model in the kashkan basin, lorestan province. thirteen flood-related factors and a map of past flood events were used. of the 58 recorded flood locations, 73% were used for model training and 27% for validation. results and discusion: the analysis revealed that proximity to rivers, elevation, slope, and roughness index are the most influential factors in the region's flooding. the rf model's performance was evaluated using the roc index, which scored 0.97, indicating excellent model accuracy in generating the flood sensitivity map. conclusions: flooding is driven by various environmental and human factors. based on the flood risk prediction map, proper management strategies can be implemented to reduce damage and casualties caused by floods.