The Gulf Stream has a very welldefined
structure that remains largely invariant as it meanders about.


The downstream velocity can be
characterized quite accurately as two backtoback exponentials with
scalewidths quite comparable to the radius of deformation set by the
depth of the corresponding pycnocline.

This doubleexponential
structure with an almost discontinuous shear in the center of the
current characterizes the velocity field at all sampled depths, 52 to
252 m.

It seems
likely that this stiffness is a general characteristic of other
separated western boundary currents such as the Kuroshio Extension
(Hall, 1989) and Agulhas Retroflection current (Lutjeharms and
Ansorge, 2001).

The transition from one side of the velocity maximum to the other
takes place over a scale that is an order of magnitude less than the
width of the Gulf Stream itself, about 6 km to either side.
This blurring of the velocity maximum, which according to the exponential
fit which might be described
statistically as a consequence of submesoscale mixing.

The traditional
mixing length argument says that
K_{h} ~ < v'^{2}> ^{1/2}
l where v' represents
perturbations normal to the mean flow in stream coordinates
and l is the mixing length.
We do not have a direct measure of l so
we use the mean (~6 km) of the two scalewidths in the second terms
of the equation for mean flow in stream coordinates.
This gives us K_{h} 1000 m^{2}s^{1}.

In another approach the momentum deficit can be
thought of as a balance between the forcing that seeks to establish a
purely exponential profile on one hand, and its loss through
submesoscale diffusion on the other resulting in the observed
velocity profile.
We compute the loss as the difference
between a purely exponential profile and the one determined
from the observations.
This loss takes place through lateral mixing to
both sides (hence the factor 2) ~2<v'^{2}>^{1/2} l =
2K_{h}.
Integration
yields
and estimate of ~3000 m^{2}s^{1}
for the peak as a whole.

In summary, the smoothing of the velocity peak
can be interpreted as due to a submesoscale diffusivity of about
13x10^{3} m^{2}s^{1}.

This can be compared to mesoscale isopycnal
diffusivity obtained from the SYNOP float trajectory data which is
about an order of magnitude larger at
O(1030) x10^{3} m^{2}s^{1} (Zhang et al., 2001).

At this location the uv
covariances to both sides of the Gulf Stream suggest a conversion of
kinetic energy from the eddy to mean flow. We interpret this as a
geometric result of the downstream decrease in meandering approaching
the Oleander line. It appears that patterns of in and outflow and
energetics can be quite site specific, reflecting, we think,
preferred states or patterns of the meandering of the Gulf Stream.

