Uncertainty propagation and its impact on earth-fill dam failures

Abstract

Dam failures represent a significant hazard to downstream, particularly when combined with extreme hydrological inputs and uncertainties in breach development. In this study, a comprehensive hydrological and dam breach modeling framework was applied to the Kanlikoy Dam in Cyprus. The HEC-HMS model was calibrated with the 2010 Flood event. Subsequently, a comprehensive sensitivity analysis identified the Curve Number (CN) and Muskingum K as the most influential parameters. Uncertainty analysis produced maximum and minimum inflow hydrographs for both the calibrated hydrograph and 100-year return period hydrograph. These hydrographs were coupled with probabilistic overtopping and piping failure simulations in McBreach, where breach width, side slopes, failure mode, and formation time were sampled stochastically. The probabilistic dam breach framework was then applied to evaluate flood responses at probabilistic breach scenarios of 1%, 10%, 50%, and 90%, capturing variability in breach processes. The resulting flood extent, flood depth, flow velocity, and flood hazard maps were generated as direct outputs of these simulations, driven by inflow conditions derived from a 100-year rainfall event and accounting for uncertainty through minimum and maximum inflow hydrographs. Maximum inflow conditions generated significantly larger inundation areas and higher flood depths, especially for low-probability events. However, the results also showed that frequent events exhibited greater sensitivity in hazard classification, as small inflow variations can lead to notable changes in hazard levels. Overall, the hydrological uncertainty affects both extreme-event consequences and the reliability of flood hazard mapping for more frequent events.