Identifying flash flood-prone watersheds using geomorphological and radar-derived rainfall features: a case study in Antioquia, Colombia

Abstract Flash floods represent one of the most destructive hydrometeorological hazards due to their sudden occurrence, localized nature, and severe socio-economic impacts. With a combination of steep slopes and intense rainfall events, flash floods in the mountainous region of Antioquia, Colombia, have resulted in a cost of around 200 M USD and 1000 fatalities over the last ten decades. Identifying watersheds prone to flash flood events (defined as “flashy” here) is crucial for planning and implementing early warning systems. Nevertheless, due to the complex nature of the processes involved, there is a large uncertainty when it comes to identifying flashy watersheds. Considering the importance of the topic, we present a methodology for identifying flashy watersheds by leveraging accessible geomorphological features and high-resolution quantitative precipitation estimates (QPEs). Our research evaluates the combined influence of rainfall intensity, antecedent soil moisture, terrain features, and land use on flash flood susceptibility. We analyzed 63 watersheds with documented flash flood events between 2012 and 2022, along with an additional validation set of approximately 100 randomly selected watersheds. Using statistical methods, we identified which features are more characteristic of flashy watersheds. We identified that features such as the slope, the basin relief, the Melton Index, the HAND index, and precipitation intensity are key determinants of flash flood susceptibility. Our findings emphasize the significance of both geomorphological indices and convective precipitation metrics in identifying watersheds with high flash flood risk. The proposed classification framework, validated through a bootstrap methodology, effectively discriminates between flashy and non-flashy watersheds, offering valuable insights for disaster risk management and flood preparedness strategies. Our results underscore the necessity of integrating detailed, local-scale rainfall and geomorphological data for accurate flash flood prediction, enabling more targeted mitigation and management practices in flash flood-prone regions, such as Antioquia.