Premium
Turbulent and numerical mixing in a salt wedge estuary: Dependence on grid resolution, bottom roughness, and turbulence closure
Author(s) -
Ralston David K.,
Cowles Geoffrey W.,
Geyer W. Rockwell,
Holleman Rusty C.
Publication year - 2017
Publication title -
journal of geophysical research: oceans
Language(s) - English
Resource type - Journals
eISSN - 2169-9291
pISSN - 2169-9275
DOI - 10.1002/2016jc011738
Subject(s) - turbulence , mechanics , geology , mixing (physics) , stratification (seeds) , bathymetry , advection , wedge (geometry) , richardson number , salinity , meteorology , geometry , physics , oceanography , mathematics , thermodynamics , seed dormancy , germination , botany , dormancy , biology , quantum mechanics
The Connecticut River is a tidal salt wedge estuary, where advection of sharp salinity gradients through channel constrictions and over steeply sloping bathymetry leads to spatially heterogeneous stratification and mixing. A 3‐D unstructured grid finite‐volume hydrodynamic model (FVCOM) was evaluated against shipboard and moored observations, and mixing by both the turbulent closure and numerical diffusion were calculated. Excessive numerical mixing in regions with strong velocities, sharp salinity gradients, and steep bathymetry reduced model skill for salinity. Model calibration was improved by optimizing both the bottom roughness ( z 0 ), based on comparison with the barotropic tidal propagation, and the mixing threshold in the turbulence closure (steady state Richardson number, Ri st ), based on comparison with salinity. Whereas a large body of evidence supports a value of Ri st ∼ 0.25, model skill for salinity improved with Ri st ∼ 0.1. With Ri st = 0.25, numerical mixing contributed about 1/2 the total mixing, while with Ri st = 0.10 it accounted for ∼2/3, but salinity structure was more accurately reproduced. The combined contributions of numerical and turbulent mixing were quantitatively consistent with high‐resolution measurements of turbulent mixing. A coarser grid had increased numerical mixing, requiring further reductions in turbulent mixing and greater bed friction to optimize skill. The optimal Ri st for the fine grid case was closer to 0.25 than for the coarse grid, suggesting that additional grid refinement might correspond with Ri st approaching the theoretical limit. Numerical mixing is rarely assessed in realistic models, but comparisons with high‐resolution observations in this study suggest it is an important factor.