Multi-decadal simulation of estuarine sedimentation under sea level rise with a response-surface surrogate model

Multi-decadal prediction of estuarine sedimentation with high-fidelity hydromorphodynamic models presents high computation costs, especially when accounting for stochasticity and uncertainty. A StochAstic model for Multi-decadaL Estuarine Sedimentation (SeAMLESS) is formulated here to support a specific decision-need related to resilience planning and coastal management: estimating future sedimentation and dredging within a sedimentation basin for different scenarios of sea level rise and rules for dredging. SeAMLESS combines a reduced-dimension process model and a response-surface surrogate model to yield an ordinary differential equation that can be integrated over stochastic time series of storm events. Applications show that SeAMLESS can predict probabilities and amounts of future basin sedimentation and dredging with minimal loss of accuracy, compared to a high-fidelity model, while delivering O(10⁴−10⁵) reduction in computational costs.