Uncertainty in Phase-Resolving Numerical Modeling of Coastal Flooding and Erosion

Storm impact on beaches, including flooding and erosion, is commonly assessed with numerical models. In the last decade, these models have gained in accuracy, allowing application in more complex settings and/or over a longer period. However, uncertainty in model predictions has often been neglected. Prediction uncertainty can be inherent to the description of a process itself, known as stochastic uncertainty. For example, wave time series, required as input in phase-resolving models, are often reconstructed with randomly distributed phase. Also, different combinations of parameter values may all be plausible but give different predictions and hence cause uncertainty. We aim to quantify the contribution of stochastic and parameter uncertainty in coastal flooding and erosion predictions. A phase-resolving numerical model (XBeach) was applied to a terraced beach profile (Duck, NC) to compute wave transformation, runup and bed level change under storm and calm conditions. 100 realizations were run for each wave condition and combination of parameter values corresponding to wave breaking and bed friction. Numerical results indicate that the combined stochastic and parameter uncertainty is larger during a storm and affect different indicators than during calm conditions. In storm conditions uncertainty expresses predominantly in the wave setup η whereas in calm conditions rather in the sea-swell component of the swash contribution S_ss (coefficient of variation = std/mn is 0.25 m and 0.12 m in storm whereas 0.05 m and 0.10 m in calm conditions for η and S_ss, respectively). After an 8-hour storm, the averaged shoreline retreat and bed level change are 5 m and 0.5 m (erosion of upper beach) with a standard deviation, related to the stochastic and parameter uncertainty, of 1.0 m and 0.2 m, respectively. The relative contributions of the stochastic and parametric uncertainty and their implications will be further discussed at the conference. Financial support was provided by CONACYT CB-2016-284430.