Sensitivity analysis of three-dimensional salinity simulations in North San Francisco Bay using the unstructured-grid SUNTANS model

The unstructured-grid SUNTANS model is applied to San Francisco Bay using a grid with an average resolution of 50 m. This accurately resolves tidal hydrodynamics in a domain that extends from the Pacific Ocean to the western portion of the Delta region, the flow through which is approximated with two rectangular boxes as a “false delta”. A detailed calibration is performed, and we show that the model accurately predicts tidal heights, currents, and salinity at several locations throughout the Bay. We perform a sensitivity study to understand the effects of grid resolution, the turbulence model, and the scalar transport scheme. Three levels of grid refinement are performed, and the results of a second-order accurate, TVD scalar transport scheme are compared to those with first-order upwinding. We find that the best convergence rate with respect to grid refinement occurs when the TVD scheme is employed. This accuracy degrades when the turbulence model is not employed due to a lack of feedback between vertical turbulent mixing and stratification. Significant horizontal diffusion associated with first-order upwinding eliminates the necessary horizontal salinity gradients required to induce baroclinic circulation, and renders the results less sensitive to the turbulence model or grid refinement.

► Performed calibration of San Francisco Bay SUNTANS model.
► Sensitivity study of grid resolution, turbulence model, and scalar transport scheme.
► Second-order accurate convergence using TVD degrades without turbulence model.
► First-order upwind renders results insensitive to turbulence model or grid refinement.