Advanced modeling for flash flood susceptibility mapping using remote sensing and GIS techniques: a case study in Northeast Algeria
A. Mansour,D. Mrad,Y. Djebbar
DOI: https://doi.org/10.1007/s12665-023-11324-0
IF: 2.8
2024-01-01
Environmental Earth Sciences
Abstract:The most effective approach for assessing flash flood susceptibility, which has consistently demonstrated its utility in prior research for spatial planning and flood risk management, involves the utilization of the newest techniques to enhance prediction accuracy. In this study, we present four robust models that use remote sensing, the Geographic Information System ArcGIS10.3 environment, and TerrSet 19.0.6 to build maps of flash flood susceptibility. These models are frequency ratio (FR), logistic regression (LR), fuzzy logic, and weight of evidence (WoE). Our study focuses on the Boumerzoug sub-watershed in northeastern Algeria as a case study. We identified flood-prone areas through an exhaustive inventory of flood maps and selected twelve influential factors for analysis: land use and land cover (LULC), distance to the river, slope, flow accumulation, plan curvature, soil type, rainfall, stream power index (SPI), terrain roughness index (TRI), elevation, density of drainage, and topographic wetness index (TWI). These factors were employed to construct models and assess their correlations with flood occurrences. Crucially, our analysis revealed the absence of multicollinearity among these factors, as confirmed by the Variance Inflation Factor (VIF) and Tolerance (TOL). To gauge the performance of FR, LR, fuzzy logic, and WoE, we conducted a thorough validation process using various statistical measures, including Accuracy and the Area Under the Curve (AUC). For the training data set, the LR model achieved the highest AUC value (0.883), followed by WoE (0.879), FR (0.867), and fuzzy (0.865). As results, all models demonstrated excellent performance (AUC > 0.8) on the validation data set. These methodologies employed in our study provide valuable tools for mitigating flood risk. They enable the mapping of flash flood susceptibility at the study area level, offering critical initial insights for policymakers engaged in spatial planning and flood risk management.
environmental sciences,water resources,geosciences, multidisciplinary