Rigorous and defensible in silico simulation for New York Harbor

Despite the tremendous progress made over the past two decades, in silico oceanography still suffers many shortcomings and lacks scientific rigor due to the proliferation of many ‘gray’ areas. Those gray areas have historically hampered the greater adoption of unstructured-grid (UG) models but have also plagued structured-grid models. In fact, error compensation is a more common problem than many researchers realize. This is an urgent issue facing the entire community especially in light of the rapid advancement of precision in coastal observation. We expound some of those critical issues here, with particular focus on what we think are the most fundamental issues, e.g., on the representation of bathy-topo DEMs in those models. Consequently, we seek to establish the following principles for in silico oceanography:

1. Bathymetry is a first order forcing in coastal domains. Hence, observation-based DEM data should not be smoothed or otherwise manipulated beyond the resolution of the numerical grid.
2. Oceanographic processes are driven across multiple scales. Hence, grids must be as high resolution and extend over as large a domain as required by the processes and known forcing (and allowed by computational limits). This includes some ultra localized processes (e.g., small-scale shears).
3. Assessment should focus on processes. Hence, traditional quantitative error metrics, while useful, are not a substitute for feature-based metrics and should not distract or mislead high-fidelity representation of processes.

Using the New York Harbor UFS testbed as an example, we describe in unprecedented detail how we conduct UG modeling from beginning to end: starting from DEM processing, mesh generation, model setup, all the way to post-processing and data analysis. The details are summarized in a companion paper (Zhang et al. 2024, Ocean Modelling Special Issue), and serve to debunk many myths that exist in the coastal modeling community. We demonstrate that objective, rigorous and defensible UG modeling is achievable at the moment if one acknowledges the central role that the DEM plays in coastal modeling.