Ensemble Multi-Criteria Flood Susceptibility Modelling with Spatial Uncertainty Quantification: A Provincial-Scale Application in KwaZulu-Natal, South Africa

Flooding remains a major hydro-meteorological hazard in KwaZulu-Natal, yet province-wide susceptibility assessments incorporating modelling uncertainty are limited. This study develops an ensemble multi-criteria flood susceptibility framework integrating Analytical Hierarchy Process (AHP), fuzzy logic transformation, and frequency ratio (FR) modelling within a cloud-based geospatial environment. Twelve hydro-geomorphological and environmental conditioning factors, including topography, rainfall, land cover, hydrology, and soil proxies, were normalized using percentile scaling. Three independent flood susceptibility models were generated and combined using ensemble mean aggregation, while pixel-wise standard deviation quantified spatial uncertainty. Model validation employed a 10-year historical flood inventory (2015–2025) comprising 65 documented flood locations. The ensemble flood susceptibility index (FSI) ranged from 0.05 to 1.00, with moderate susceptibility zones covering 51.08% of the province. High and very high susceptibility classes occupied 8.36%, indicating spatially concentrated but hydrologically significant risk hotspots. Uncertainty analysis showed low inter-model variability (0.00–0.11), demonstrating strong methodological stability. Validation results confirmed that 73.85% of historical flood points were located within high susceptibility zones, with over 90% captured within overall susceptible classes. The study introduces a hybrid deterministic–fuzzy–probabilistic ensemble modelling approach combined with pixel-level uncertainty mapping and scalable cloud computation. Findings support disaster risk reduction, urban and catchment planning, and early warning system optimization in flood-prone regions. The framework provides a transferable methodology for data-limited environments requiring reliable and uncertainty-aware flood hazard assessment.