The reason for this failure (omitted due to time constraints in the talk) seems to be the 
poor way that models handle the surface mixed layer (and hence, by extension, deep convection). 
The usual belief is that tracers both passive and active are injected into deeper water by ver- 
tical Ekman pumping at the base of the mixed layer. A simple order-of-magnitude estimate, 
however, suggests that horizontal advection through the sloping mixed layer/stable fluid inter- 
face injects 2 to 5 times as much tracer as does vertical motion. 
Unfortunately, large-scale models do not include an accurate representation of the deep 
surface mixed layer (indeed, most cannot resolve the sorts of 20 km chimney we suspect to be 
responsible for deep convection in the Greenland Sea). Nor do they usually include seasonal 
effects, which Woods (1984) has demonstrated can dominate the production of both passive and 
active tracers in the subsurface ocean. 
Particle trajectories are only now beginning to be included as diagnostics in large-scale 
ocean models, and these will be vital if we are to understand where the deep water in the 
Atlantic finally ends up. Even fairly simple models (like a ‘thermocline equation' model I 
showed) can have remarkably convoluted particle paths. Fig. 2 shows a typical example. On the 
left are the horizontal velocity vectors (note change of scale) for a two-level ocean, which 
allows vertical velocities between the levels, which are of constant depth. The larger diagram 
shows the trajectory of a particle injected into the 4000 km square basin at the base of the 
subtropical Ekman layer at the point marked with a circle. The particle is then advected three- 
dimensionally with the steady flow on the left. Firm lines denote downward flow, dashed lines 
upward flow. The route which allows ventilation of the subtropical gyre involves recurring 
visits to the northern subpolar gyre and sinking on the northern boundary. Notice how little of 
the deep subtropics is actually ventilated. (When the crosses, placed every year, are far 
apart, the particle is in the upper level; when close, in the lower level). 
As with every other branch of the discipline, a great deal more work is still needed! 
