ANALYZE SUSCEPTIBILITY TO URBAN FLOODING USING MULTIPLE MACHINELEARNING ALGORITHMS IN AMAZON

Over the past five decades, Brazil has faced increasing flood events, prompting attention from both the public and civil protection authorities. This research models flood susceptibility in the urban area of Belém, in the eastern Brazilian Amazon, using Machine Learning (ML) algorithms. The analysis is based on flood records from 2010 to 2020 and geographic factors like altitude, slope, flow power, Height Above the Nearest Drainage (HAND), and proximity to water channels. Five distinct flood susceptibility models were generated using ML algorithms. The Random Forest (RF) model achieved the highest accuracy, with an Area Under the Curve (AUC) rate above 90%. Key factors influencing flooding included altitude, HAND, soil profiles, and precipitation, providing valuable insights for flood management decisions.