Exploring a form of pixel-based information value model for flood probability assessment and geo-visualization over an East African basin: a case of Nyabarongo in Rwanda.

The Nyabarongo basin in Rwanda is subjected to hydrometeorological hazards, particularly floods, which are the most prevailing and devastating. Therefore, understanding flood-controlling factors is so pertinent for the development of scientifically driven flood prevention strategies. This study aimed at exploring a form of pixel-based information value model integrated with remote sensing techniques and geo-information system to assess the probability of flood incidence and geo-visualize prone areas at basin's scale. To do this, a flood inventory was initially generated using 226 past flooded locations, which were split into a 75:25 ratio for model training and validation, respectively. Fourteen flood-controlling factors were selected after a multicollinearity diagnosis. The results unveiled that more than half of the basin's surface area is covered by very high (8.6%) and high (21.5%) to medium (31.8%) probability of flood incidence. This dispersion was mainly influenced by rainfall, proximity to rivers, Land Use/Land Cover, elevation, and SPI which influence the basin's hydrological behavior. The evaluated accuracy of the applied model using the Area Under Curve of the Receiver Operating Characteristics (AUC–ROC) highlighted a commendable and accurate performance of 0.883 and 0.828 for success and prediction rate, respectively. The study's findings provide a scientifically driven reference for flood mitigation plans and act as a benchmark for decision-making and policy updates regarding flood risk management toward the Nyabarongo basin and other basins with similar characteristics nationally or regionally. [ABSTRACT FROM AUTHOR]