ABSTRACT This study evaluates the use of synthetic aperture radar imagery for detecting and monitoring periodic flooding in Nechí, Colombia, where persistent cloud cover limits the use of optical images. A classification method combining autoencoder-based despeckling and Otsu thresholding was applied, achieving 91.16% overall accuracy, 95.33% specificity, 86.80% recall, and an F1-score of 90.58%. Flood mapping from 2017 to 2024 revealed strong interannual variability, with the largest event in 2017 (average extent 3.59 million m2; peak 12.95 million) and the second largest in 2020 (average 124,442 m2; peak 746,100 m2). Seasonal patterns were bimodal, with peaks in June (190,557 m2) and November (186,171 m2), coinciding with maximum river discharges above 3,500 m3 /s. Wet conditions associated with La Niña increased flood extent, while dry El Niño years reduced it. Spatial analysis showed that embankments influence flooding patterns and that urban expansion between 2017 and 2025, with an approximate increase of 0.33 km2 (26%), invaded flood-prone areas. The findings demonstrate the value of radar imagery for continuous flood monitoring and its potential to strengthen early warning and risk management strategies in tropical river basins.
Cortés-Arango et al. (Mon,) studied this question.