Peak discharge estimation for ungauged basins: a review

Abstract

ABSTRACT Peak discharge estimation is a critical component of water resource management, particularly in ungauged basins where direct hydrological records are lacking. Reliable prediction of flood peaks is essential for flood warnings, infrastructure design, and sustainable planning. Traditional approaches such as empirical formulas, rainfall-runoff models, and regionalization techniques have long been applied, yet their effectiveness is limited in data-scarce environments. To address these challenges, recent research has explored both process-based and data-driven alternatives. Widely used hydrological platforms such as SWAT and HEC-HMS provide valuable insights but require substantial calibration data, constraining application in ungauged catchments. In parallel, machine learning frameworks, including artificial neural networks, support vector machines, and gradient boosting, have emerged as promising tools due to their ability to capture nonlinear relationships and adapt to diverse conditions. Recent advances highlight a shift towards integrating artificial intelligence (AI) and remote sensing. AI-based forecasting has demonstrated lead times comparable to operational flood nowcasts, while hybrid models combining satellite observations with hydrological simulations have improved flood depth estimation and discharge reconstruction. Novel global deep learning frameworks and explainable AI tools for flood susceptibility mapping further emphasize the growing role of intelligent, climate-resilient modelling in hydrology.