Uncertainty in Nearshore Sand Bar Migration

A coupled nearshore hydrodynamics and sediment transport model with adjoint-based uncertainty quantification has been applied to sand bar morphodynamics in the Duck94 field experiment (Duck, NC). The model, which is based on previously established physical process formulas, is capable of producing accurate predictions of sand bar migration during the experiment. However, reasonable perturbations to the model inputs-namely initial conditions, boundary conditions, and calibration coefficients-can produce a range of model predictions. This range is taken to represent the potential for uncertainty of the model, for scenarios where one or more model inputs are poorly constrained. Experiments are presented that quantify the relative impact of different input types on uncertainty, and the results are contrasted for accretive versus erosive wave conditions. During accretive conditions (shoreward bar migration), uncertainty is dominated by the growth of errors in initial conditions, owing to positive feedbacks in the model. In contrast, during erosive conditions (seaward bar migration), there is a smaller impact of errors in initial conditions, and uncertainty is instead dominated by errors in sediment transport coefficients. The results thus point to qualitative differences in the nature of nearshore morphodynamic uncertainty under different wave conditions, with a greater role of internal error growth (positive feedbacks) during accretive conditions and errors in model physics being more important during erosive conditions.