Abstract
The shallow-water equations are often used as a classical simplified ocean model for barotropic ocean dynamics. The same equations can also be used to model simplified baroclinic dynamics through the reduced-gravity model. Herein, we propose to utilise a GPU-accelerated shallow-water simulation framework for representing two decoupled simplified ocean models for each of the barotropic and baroclinic dynamics, and use these models for ensemble prediction of short-term drift trajectories in coastal domains. This system can be used complementary to current operational systems as a lightweight tool for uncertainty quantification and in the future for ensemble-based data assimilation of in-situ observations. We show the relevance of our approach by demonstrating how the barotropic and baroclinic dynamical components are of varying importance at different geographical locations, and that one of the models can be used alone or in combination with the other. We benchmark our approach by showing the resulting ensemble trajectories in reference to deterministic trajectories produced by operational models for three Norwegian coastal regions.