fore e ee tee 
4 
\ 
\ 
\ 
\ 
‘ 
‘ 
‘ 
‘ 
Us 
Fig. 2. Horizontal velocity vectors in a two-level ocean model (left figures) and trajectory of a 
particle injected at the base of the Ekman layer at the point marked by a circle (right figure). 
Small horizontal arrows show 10 cm s~! in upper level, 1 em s-1 in lower level. Coordinates 
show distance in km. 
The question was then asked: Why does deep water form in the Atlantic anyway? As an 
illustration, the flow of a dense tongue or plume of water over a parabolic depth profile was 
discussed. It was shown that the asymptotic depth of descent of such a plume (assumed to 
emerge from a very deep reservoir) is roughly proportional to the product Cg/f2 (C=curvature) 
times the fractional buoyancy difference between plume and surrounding waters. This fraction is 
minute for the Denmark Strait (0.007) but quite large for Gibraltar (0.11). This suggests that 
Denmark Strait overflow should not form bottom water, but should simply turn and flow horizon- 
tally along depth contours at a shallow depth, with forces balanced between Coriolis upward and 
gravity downward. However, recent laboratory and theoretical work by Griffiths and Linden 
(1982) and Killworth, Paldor and Stern (1984) suggests that such plumes are highly unstable, and 
form large-scale lenses of dense water. Positive curvature of the bottom is a slightly stabi- 
lizing feature here. Since rotation is not dominant in the Denmark Strait, there is little small- 
scale mixing, so that any other frictional mechanism will provide an extra component to the 
triangle of forces acting on the plume and allow the plume/lenses to slowly descend the slope to 
the bottom. Conversely, the Gibraltar outflow is dominated by rotation, and the resulting small- 
scale mixing diffuses the plume away before it can reach the bottom, as Smith's (1975) model 
shows. 
Once it has reached the bottom, dense water is believed to spread largely in western boun- 
dary layers. Models of this phenomenon reproduce the feature fairly well, but the vertical 
depth of penetration of surface injected tracers like tritium is badly reproduced. The same is 
true for active tracers like potential vorticity. 
