Representing effects of subgrid-scale topography on coarse-grid hydrodynamic models for shallow coastal marshes

Hydrodynamic modeling of flow and salinity transport through shallow coastal marshes often suffers from errors introduced through inadequate representation of the underlying bathymetry. Although high-resolution topographic data has become available through lidar, it generally requires upscaling to a coarser resolution to maintain practical computational costs for large coastal marshes. The effects of simulating at coarse resolution, along methods to improve such simulations, are illustrated in this study by comparing model results with field data for the Nueces River delta in Texas (USA). It is shown that with simple upscaling techniques surface connectivity of the model domain is altered, where existing flow paths are smoothed and new paths are created. Narrow channels are widened to the grid scale, leading to discrepancies in modeled flux and salinity. Subgrid topography models have been used to reproduce effects of high-resolution topographic features on computationally-efficient coarse grids, but four major issues are associated with existing subgrid models: (i) surface connectivity is not always maintained, (ii) salinity transport is rarely modeled, (iii) effects of large topographic features in the cell interior are often neglected, and (iv) sensitivity of subgrid model to mesh design is strong. This research presents a new high-performance subgrid model to address these issues associated with grid-coarsening. In this new model, The high-resolution topographic data is parametrized into the model equations to scale grid cell storage and flow rate across cell faces. A block-checking procedure is designed to maintain surface connectivity during coarsening. The existence of interior features generates additional reaction forces and enhances longitudinal dispersion, which are modeled by reducing grid cell volumes and face areas. A mesh-shifting method is used to alleviate sensitivity of model performance to mesh design. The new subgrid model is tested on both synthetic domains and real Nueces Delta bathymetry. Compare to simple upscaling, using the proposed subgrid method better approximates fine-grid simulation results for surface elevation, inundation area, in-channel flow rate and salinity with negligible additional computational cost. Model-data agreements for salinity is also improved. This new model can be applied to large domains where coarse-grid model performance significantly deteriorates due to interrupted surface connectivity. Compare to existing approaches that model all interior topographic features as drag effects, the new subgrid model builds stronger physical connections between the geometry of the features and their effects on the flow fields. The use of the mesh-shifting method restrains numerical diffusion caused by misalignment between channel and grids, making model results less sensitive to mesh design.