Prediction of Dam-Breach Characteristics Using Multivariate Regression Analysis

Earthen dams are utilized to mitigate flood hazards. Due to climate change, the frequency and intensity of floods and the risk of dam failure from overtopping have increased during the last few decades. Accurate prediction of dam-breach characteristics and the resulting impacts on the downstream areas are essential for developing flood hazard countermeasures and emergency plans. In this study, multivariate regression analyses are employed to predict dam-breach characteristics using datasets generated from a physics-based numerical model developed in-house to simulate dam breach. Key predicted dam-breach characteristics include maximum downstream water depth, time to peak breach flow, and breach widening and deepening rates. The dam parameters are the inflow discharge, tailwater depth, dam dimensions, reservoir volume, and initial breach dimensions. The significance of each parameter on the dam-breach characteristics is analyzed using quality measures, such as the adjusted coefficient of correlation, Bayesian information criterion, and root mean squared error. The best regression model for each dam-breach characteristic is identified through backward elimination of dam parameters, starting with the least to the most significant parameter, while observing the model’s quality at each elimination step. The best regression models are presented, highlighting the most influential parameters for each dam-breach characteristic. Finally, the integration of these regression models into the National Water Model is proposed for future work.