A review of empirical dam breach models

Abstract

Dam breach modelling is central to flood risk assessment, yet the empirical models commonly used for estimating breach parameters display considerable variability and uncertainty. This review evaluates the performance of widely applied empirical dam breach models and identifies those that consistently show lower uncertainty, such as the Froehlich model, while highlighting recognised limitations in older models, including the United States Bureau of Reclamation and MacDonald and Langridge‐Monopolis models. Although physically based and artificial neural network‐driven models are emerging, empirical approaches remain dominant in regulatory practice, including national reservoir flood mapping in the United Kingdom. Persistent challenges include inconsistent parameter definitions, variations in units, and limited historical datasets for calibration. The review identifies the need to update empirical relationships using expanded failure databases and emphasises the importance of transparent model selection procedures. Instantaneous failure scenarios are recommended as a conservative benchmark for worst‐case analysis. The findings underscore the necessity of improved breach prediction methodologies to improve emergency planning, hazard mitigation, and infrastructure resilience.