parameterization will be elusive. Because turbulence depends strongly on stratification, vertical resolution of turbulence parameters has the highest priority. Fine sampling in the horizontal will be more valuable than fine temporal sampling for resolving mixing issues.

Repeated, high, vertical-resolution sampling with a coarse, horizontal array of moored conductivity, temperature, and depth (CTD) and velocity profilers is now possible, and such an array can be used to examine mixing issues independent of a seafloor observatory program. To add value to these arrays, the seafloor observatory concept must provide the needed horizontal resolution over much finer scales than presently possible. This increased resolution would provide fully four-dimensional sampling and could be obtained with AUVs. Near real-time telemetry and an active modeling and analysis component could be employed to optimize the sampling strategy. Other methods of obtaining high-resolution spatial information should also be explored, such as horizontal mooring lines and acoustic Doppler technology.

The strategy outlined above could go far in terms of closing budgets or constraining numerical models. (Closure of heat, salt, momentum, and tracer budgets is important because it demonstrates sufficient observational accuracy and adequate resolution of advection and mixing.) A sequence of field programs in different dynamical regimes would provide us with a decadal leap in the understanding of mixing processes by covering the oceanic turbulence parameter space. A seafloor observatory program would be of great benefit to advancing our understanding and parameterization of mixing if it were to enable such a dedicated set of oceanic observations. Suggested dynamical regimes, which should not be considered exhaustive, are discussed in Box 2-9 below.

A similar problem exists for observing and parameterizing the effects of horizontal turbulent motions, in particular for mesoscale eddies, although the technical challenge of observing those scales of turbulence is not nearly so high as for the very small scales of vertical mixing.