Systematic microstructure variability in double‐diffusively stable coastal waters of nonuniform density gradient

Abstract

[1] Conductivity microstructure, water velocity, and stratification were measured during a tow-yo transect near the New England shelf/slope front in early August 1997. Velocity data were collected with an acoustic Doppler profiler on the ship. The other data were collected with a towed platform. Estimates of c, the rate of dissipation of temperature variance, were computed from the conductivity data with vertical resolution of 0.3 m. Relationships between c and shear, temperature gradient, buoyancy frequency (N), and gradient Richardson number (Ri) were explored, with special focus on measurements taken in waters stable to double-diffusive processes (to avoid ambiguity of interpretation) and exhibiting variable density gradient (N ranging from 5 to 40 cph). For this subset of data, c computed from data grouped into classes of local mean temperature gradient (d " T/dz) was proportional to d " T/dz to the 0.7 power, which is consistent with diapycnal thermal eddy diffusivity K being proportional to (d " T /dz) À1.3 within the framework of the Osborn-Cox model that relates c and the mean temperature gradient to the heat flux. No correlation between K and Ri was observed, with Ri computed at 4 m vertical scale, so that systematic inhomogeneous largescale forcing is not responsible for a false correlation of K and d " T/dz. Water mass salinity characteristics in the area caused N 2 to be proportional to (d " T/dz)4/5 , rather than to d " T /dz as in the isohaline case, giving rise to the steep inverse relation K = 10 À10 N À3.3 , with N in radians/ s and diffusivity in m 2 /s. The fit K = 2 Â 10 À9 N À2.5 results if one questionable data ensemble is disregarded. These relations are comparable to results obtained previously from the nearbottom tow data. They are not intended to be universal formulae but are meant to describe the conditions we encountered. They are not expected to hold at high and low values of N outside of our measurement range. An interpretation is that under these conditions the less strongly stratified (lower N) layers in this shelf area are more prone to instability of the larger-scale shear than the intervening interfaces, with the subsequent greater energy dissipation in the layers leading to higher buoyancy flux KN 2 in the layers than in the interfaces.

Microscale, finescale, and mesoscale measurements made during the 2004 Structured Mixing Project (Micro-Tow 04) Cruise

“…For eighteen of the twenty-five transects K(N) is steeper than N -2 (or equally steep), so F is a decreasing function of density gradient; agreeing with the prior published results (Rehmann and Duda, 2000;Duda and Rehmann 2002). Figure 7 shows K versus N computed for all the selected doubly-stable 2-s intervals for Transects 1-25.…”

“…K is proportional to N -2.2 , so that flux F is proportional to N −1/5 . The minus 2.2 exponent is a bit less negative than the -2.5 to -3.3 estimates in the Duda and Rehmann (2002) paper, so that F as reported here decreases less rapidly with increasing N than the prior result.…”

“…Using processing code MT04 Version 8, flux as a function of density gradient is computed as described below. The methods used are very smilar to those used in Rehmann and Duda (2000) and Duda and Rehmann (2002).…”

“…This can also be stated as K being inversely related to N and falling more steeply than K(N) ∝ N -2 . Some evidence of gradient-dependent K behaving in this way was found in the ONR Coastal Mixing and Optics (CMO) program dye diffusion studies and towed-microscale conductivity measurements (Duda and Rehmann, 2002). We sought to expand on the CMO work by collecting and analyzing more data of better quality than previously available.…”

Cryptic peaks: invisible vertical structure in fluorescent particles revealed using a planar laser imaging fluorometer

“…N was calculated from the smoothed SCAMP density data by calculating the change in depth necessary to yield density intervals of Dr 5 0.005 kg m 23 , as outlined in Duda and Rehmann (2002). This method is numerically stable because it avoids N values of 0 s 21 .…”