Abstract In May 2024, Porto Alegre, in southern Brazil, experienced the most devastating flood event ever recorded, affecting over 160,000 people and causing extensive damage to buildings and essential services. The city was not adequately prepared to respond to an event of this magnitude, and many residents are still recovering from its impacts. This study proposes a hydrological hazard map for Porto Alegre as a reliable and accessible tool to support government planning and risk management. The proposed hazard map was developed from three precursor maps: (i) the Flood Susceptibility Map (FSM), (ii) the Flood Frequency Map (FFM), and (iii) the Historic Flood Depth Map (FDM). Both FSM and FFM were generated using artificial neural network models trained on historical records of urban flooding and seven thematic maps representing key terrain characteristics. The FDM was created by integrating pre- and post-disaster optical satellite imagery, a high-resolution Digital Elevation Model, and field-surveyed flood marks collected during or near the flood peak. The final Hydrological Hazard Map was obtained by combining these three components using different weighting schemes, selecting the configuration that best represented observed flood patterns. Model performance was satisfactory: the FSM validation yielded an AUC of 0.9484, the FFM showed strong predictive agreement with observed values, and the FDM achieved an AUC above 90% with an average spatial displacement of less than 40 m. The final hazard map aligns with previous studies and consistently identifies the highest hazard levels in densely populated and highly urbanized areas with intense human activity.
Bengua et al. (Wed,) studied this question.