Abstract: Five different parameterizations of Langmuir turbulence (LT) effect are investigated in a realistic model of the North Atlantic and Arctic using realistic wave forcing from a global wave hindcast. The parameterizations mainly apply an enhancement to the turbulence velocity scale, and/or to the entrainment buoyancy flux in the surface boundary layer. An additional run is also performed with other wave effects to assess the relative importance of Langmuir turbulence, namely the Coriolis-Stokes forcing, Stokes tracer advection and wave-modified momentum fluxes. The default model (without wave effects) underestimates the mixed layer depth in summer and overestimates it at high latitudes in the winter. The results show that adding LT mixing reduces shallow mixed layer depth (MLD) biases, particularly in the subtropics all year-around, and in the Nordic Seas in summer. There is overall a stronger relative impact on the MLD during winter than during summer. In particular, the parameterization with the most vigorous LT effect causes winter MLD increases by more than 50% relative to a control run without Langmuir mixing. On the contrary, the parameterization which assumes LT effects on the entrainment buoyancy flux and accounts for the Stokes penetration depth is able to enhance the mixing in summer more than in winter. This parametrization is also distinct from the others because it restrains the LT mixing in regions of deep MLD biases, so it is the preferred choice for our purpose. The different parameterizations do not change the amplitude or phase of the seasonal cycle of heat content but do influence its long-term trend, which means that the LT can influence the drift of ocean models. The combined impact on water mass properties from the Coriolis-Stokes force, the Stokes drift tracer advection, and the wave-dependent momentum fluxes is negligible compared to the effect from the parameterized Langmuir turbulence.

Abstract: Twenty turbulent flux parameterizations are compared globally and regionally with a focus on the differences associated with episodic events. The regional focus is primarily upon the Gulf Stream and Drake Passage, as these two regions contain vastly different physical characteristics related to storm and frontal passages, varieties of sea-states, and atmospheric stability conditions. These turbulent flux parameterizations are comprised of six stress-related parameterizations [i.e., Large and Pond (1981), Large et al. (1994), Smith (1988), HEXOS (Smith et al. 1992, 1996), Taylor and Yelland (2001), and Bourassa (2006)] which are paired with a choice of three atmospheric stability parameterizations ['Neutral' assumption, Businger-Dyer (Businger 1966, Dyer 1967, Businger et al. 1971, and Dyer 1974) relations, and Beljaars-Holtslag (1991) with Benoit (1977)]. Two remaining turbulent flux algorithms are COARE version 3 (Fairall et al. 2003) and Kara et al. (2005), where Kara et al. is a polynomial curve fit approximation to COARE; these have their own separate stability considerations. The following data sets were used as a common input for parameterization: Coordinated Ocean Reference Experiment version 1.0, Reynolds daily SST, and NOAA WaveWatch III. The overlapping time period for these data sets is an eight year period (1997 through 2004). Four turbulent flux diagnostics (latent heat flux, sensible heat flux, stress, curl of the stress) are computed using the above parameterizations and analyzed by way of probability distribution functions (PDFs) and RMS analyses. The differences in modeled flux consistency are shown to vary by region and season. Modeled flux consistency is determined both qualitatively (using PDF diagrams) and quantitatively (using RMS differences), where the best consistencies are found during near-neutral atmospheric stratification. Drake Passage shows the least sensitivity (in terms of the change in the tails of PDFs) to seasonal change. Specific flux diagnostics show varying degrees of consistency between stability parameterizations. For example, the Gulf Stream's latent heat flux estimates are the most inconsistent (compared to any other flux diagnostic) during episodic and non-neutral conditions. In all stability conditions, stress and the curl of stress are the most consistent modeled flux diagnostics. Sea-state is also a very important source of modeled flux inconsistencies during episodic events for both regions.