An Empirical Evaluation of Turbulence Closure Models in the Coastal Ocean

• Published in: Journal of geophysical research. Oceans Vol. 127; no. 4
• Main Authors: Savelyev, I. B., Martin, P. J., Fan, Y., Savidge, D. K., Shearman, R. K., Haack, T., Paolo, T., Terrill, E. J., Wang, Q.
• Format: Journal Article
• Language: English
• Published: 01.04.2022
• Subjects: dissipation; ocean mixing; ocean turbulence; turbulence closure; turbulent kinetic energy; upper ocean
• ISSN: 2169-9275; 2169-9291
• DOI: 10.1029/2021JC017588

A comprehensive in situ data set was obtained for the purpose of testing upper ocean turbulence models. The data set was collected in 38 m deep water over the North Carolina shelf. Available time series of surface meteorological forcing, including waves, winds, and heat fluxes, as well as underwater profiles of temperature, salinity, and horizontal velocity, were found suitable for constraining and testing numerical models in realistic environmental scenarios of the coastal ocean. The Navy Coastal Ocean Model was tested in a vertical 1‐D mode with a suite of previously incorporated subgrid turbulence closure models. Modeled output of turbulent quantities, such as the turbulent kinetic energy and its dissipation rate, were evaluated against comparable turbulence observations from a bottom mounted acoustic velocity profiler and a glider mounted turbulent shear probe. The results demonstrate a steady incremental skill increase among available turbulence closure models over several decades of their development, however the correlation with observed turbulence remains weak. The analysis of remaining model errors identified a need to add wave‐dependence to the air‐sea momentum flux formulation to account for waves that are out of balance or misaligned with the wind. Plain Language Summary Ocean circulation contains an interconnected cascade of fluid motions with lengthscales ranging from the basin scales (∼1,000 km), down to the smallest dissipative scales (∼10 mm). Resolving all these motions is prohibitively computationally expensive for a numerical ocean circulation model, even on the world's fastest supercomputers. Therefore, the model grid usually covers the basin with spacing only down to ∼1 km. Smaller scale turbulent processes are averaged within each grid cell and are parameterized and solved for by means of a simplified subgrid turbulence model. This study takes on the challenge of testing such subgrid model against real ocean observations. The turbulent quantities computed by the model, such as the amount of small‐scale turbulent kinetic energy and the rate of its dissipation, were evaluated against similar quantities observed at sea. The results demonstrate a steady improvement in model performance from earlier versions a few decades ago, to the modern version. However, a recommendation is made to modify the formulation of air‐sea boundary condition to achieve further improvements in the model performance. In summary, this study confirms the usefulness of subgrid models to simplify ocean modeling, and offers a direction and an empirical testbed for further development. Key Points A comprehensive coastal ocean small‐scale turbulence data set was collected Turbulence closure models were tested against the in‐situ data set Model performance was evaluated and remaining modeling errors quantified