Machine learning driven geospatial techniques for integrated urban flood susceptibility mapping in Jimma City Ethiopia

Urban flooding is a growing global concern, driven by rapid urbanization, inadequate infrastructure, land use changes, and climate variability. Identification and mapping of flood-prone areas are critical for effective flood risk management, urban planning, and disaster mitigation. Flood susceptibility assessment forms the foundation for proactive flood prevention strategies. This study employed five machine learning (ML) algorithms: Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Logistic Regression (LR), and Extreme Gradient Boosting (XGBoost) to map urban flood susceptibility across heterogeneous geographical settings, integrating multiple geo-environmental and socio-economic flood-contributing factors. Model performance was evaluated using metrics including accuracy, precision, recall, F1-score, mean squared error (MSE), and the area under the receiver operating characteristic curve (AUC-ROC). Twelve flood-conditioning factors were carefully selected based on the scale and characteristics of the study area to evaluate flood susceptibility. The flood inventory was randomly partitioned into training (70%) and validation (30%) datasets to ensure robust model development and validation based on historical records, field survey, and high-resolution satellite imagery. Analysis of the resulting flood susceptibility map revealed that the majority of the study area falls within the moderate susceptibility category, while only a small portion exhibits very low susceptibility. Among the five machine learning models evaluated, XGBoost demonstrated superior predictive performance, achieving the highest accuracy (0.88) and AUC (0.93), whereas the K-Nearest Neighbors (KNN) model showed comparatively lower performance with an accuracy of 0.76 and an AUC of 0.80. Despite differences in model complexity and computational requirements, machine learning-based flood susceptibility mapping offers a rapid, scalable, and cost-effective approach, particularly advantageous in data-scarce regions. The findings provide actionable insights for urban planners and flood management authorities, supporting evidence-based decision-making and enabling the design and implementation of effective strategies to mitigate socio-economic impacts and strengthen urban resilience against flooding.