Advanced flood risk mapping in Bouarfa watershed using integrated machine learning, GIS, and MCDM

In Morocco, floods occur frequently, often causing significant damage to infrastructure and the environment due to a lack of adequate protective measures. The unpredictability of these events is attributable to climate change and the irregular nature of weather conditions. However, determining flood susceptibility can facilitate the mitigation and prevention of risk. This study aims at mapping flood-susceptible areas in the Bouarfa watershed using a multi-criteria decision analysis (MCDA) approach integrated within a Geographic Information System (GIS). Seven key conditioning parameters were considered: altitude, slope, geology, drainage density, flow accumulation, land use/land cover (LULC), and soil. The Analytical Hierarchy Process (AHP) was used to assign weights to these factors. The results obtained demonstrate that 39.53% (546.24 km²) of the territory is exposed to a very low to moderate flood risk and 60.47% (835.59 km²) to a high to very high flood risk. The model's accuracy was validated using historical flood locations and the Area Under the Curve (AUC) method, which yielded a value of 84.5%, indicating very good performance. This map serves as a critical tool for decision-makers for risk mitigation and land-use planning in this arid region.