Advanced GIS-Based Modeling for Flood hazards Mapping in Urban Semi-Arid Regions: Insights from Béni Mellal, Morocco

MOHAMED EL HAOU^1*Malika Ourribane¹Maryem ISMAILI¹Kamal Abdelrahman²Mohammed S Fnais²Samira Krimissa¹Hasna El Oudi³Sonia Hajji¹Meryem El Bouzkraoui¹Fatimazahra Tarchi¹Mustapha NAMOUS^1,4

• ¹Data Science for Sustainable Earth Laboratory (Data4Earth), Faculty of Sciences and Technics, Sultan Moulay Slimane University, Beni Mellal, Morocco
• ²King Saud University, Riyadh, Riyadh, Saudi Arabia
• ³Applied Geology and Geoenvironment Laboratory, Faculty of Sciences, Ibn Zohr University, 80000 Agadir, Morocco., Agqdir, Morocco
• ⁴Faculty of Arts and Science (FAFS), University of Saint-Boniface, 200 Cathedral Avenue, Winnipeg, MB R2H 0H7, Canada, manitoba, Canada

Floods are among the most destructive natural disasters, threatening people, the economy and cultural heritage. In Beni-Mellal, mountainous topography accentuates this risk by promoting the rapid flow of water to low-lying areas, where it accumulates more easily. This study maps the flood risk using three statistical methods: Information Value (IV), Weighting Factor (WF) and Weight of Evidence (WoE). A detailed database was built, combining an inventory of floods and key environmental variables, such as slope, proximity to rivers, land use and the Topographic Humidity Index (TWI). The database was built on pre-processed and standardized Sentinel-2 and Landsat 8 satellite images, as well as geological and soil maps, ensuring full coverage and high-definition resolution of 12.5 meters to ensure optimal spatial accuracy. The results show that 4.4% to 13.6% of the region is classified as very high risk, 13.8% to 31.1% at high risk, and 24.5% to 31.2% at moderate risk, with increased vulnerability in the southern areas, where land slope and occupation play a major role. The evaluation of model performance reveals that WoE has the highest accuracy and Kappa coefficient, demonstrating its robustness for flood classification. However, WF scores the best AUC scores (88.23% in training, 86.77% in test), making it the most effective model for prediction. The IV approach, although effective, is in third place. These results provide key information for policymakers and urban planners to improve flood risk management and develop appropriate planning strategies to limit flood impacts and build urban resilience to extreme weather events.