Speaker
Description
Rogue waves are extreme manifestations of ocean dynamics, whose prediction is strongly affected by uncertainty in the characterization of sea states. This work develops and applies advanced Uncertainty Quantification (UQ) methodologies for the probabilistic assessment of extreme wave occurrence in nonlinear irregular seas. Sea states are described through parametric spectral models, whose defining parameters (e.g. directional spreading, wave steepness, peak enhancement factor) are treated as uncertain inputs. A non-intrusive UQ framework based on adaptive sparse grids and stochastic collocation methods is employed to propagate these uncertainties through a high-fidelity phase-resolving solver. Deterministic simulations rely on a High-Order Spectral formulation for deep-water gravity waves, that solves the time evolution of the free surface height on a periodic domain, considering non-linear interactions and dissipation due to wave breaking. In order to enhance local surrogate accuracy while limiting computational cost, an informed adaptive refinement strategy is introduced. The approach exploits the hierarchical structure of sparse grids by estimating the local discrepancy between interpolants constructed at consecutive levels. Rather than upgrading the full grid level, only nodes exceeding a dynamically defined relative-error threshold are selected for additional high-fidelity simulations. This targeted refinement enables the mitigation of local interpolation anomalies and improves convergence in critical parametric regions, while avoiding the prohibitive cost of a full-level refinement. The proposed framework facilitates the operational deployment of phase-resolving models and integrates them with phase-averaged approaches. This integration enables rapid, uncertainty-aware prediction of extreme wave statistics in realistic marine environments.