604 



FIGURE 9. The flow produced by releasing salt 

 solution into a gradient of sugar solution, 

 using conditions comparable with, but the in- 

 verse of those shown in Figure 8. 



fingers and above the diffusive interfaces, indica- 

 ting again that there is an increase in density due 

 to the continuing flux in the fingers. 



Two other features of the laboratory observations 

 which have important implications for the ocean 

 should also be mentioned. The most rapid formation 

 of layers in the series of experiments reported by 

 Turner (1978) occurred when the tank was stratified 

 in the "finger" sense, and the fingers were allowed 

 to run down towards a marginally stable state. 

 When source fluid was introduced, layers formed 

 more rapidly and regularly than before, because of 

 the potential energy already available in the 

 ambient fluid. This implies that "reactivation" 

 of layers in a region where they have previously 

 formed will proceed more quickly than the original 

 layering process. It suggests that the patches of 

 strong layers, under the Mediterranean outflow 

 into the Atlantic for example, are associated with 

 the arrival of a fresh pulse of intruding fluid. 

 The second related observation is that the further 

 stage of overturning to produce nearly uniformly- 

 mixed layers, bounded above and below by finger 

 interfaces is also more likely to be reached near 

 the source of the intruding water. The relationship 

 between the two types of layering has been demon- 

 strated directly in the measurements of Gregg (1975) , 

 which show that inversions of intrusive origin can 

 in the course of time break down to form well-mixed 

 layers . 



Layer Formation at Fronts 



An important geometry which merits separate study 

 is a discontinuity of T-S properties over a vertical 

 or inclined surface, i.e., a front. The motions 

 produced when an inclined boundary is inserted into 

 a fluid stratified with opposing gradients (Section 

 3) have some of the required features, but the 

 presence of the solid wall is clearly undesirable. 



Fronts can be set up in the laboratory in several 

 ways. Large horizontal T and S gradients can, for 

 example, be produced just by pouring fluid with 

 contrasting properties into one end of a stratified 

 tank at several levels, or by stirring it in 

 throughout the depth. A somewhat more controllable 

 method is to insert a vertical barrier in a previ- 

 ously stratified tank, to introduce the extra fluid 

 on one side of it, and allow the disturbances to 



die away before removing the barrier. This tech- 

 nique has been used in the experiment shown in 

 Figure 10. It is difficult to get the two vertical 

 gradients exactly matched, and so when the barrier 

 is removed internal waves are set up, which soon 

 die away, leaving the isopycnals horizontal but 

 the front distorted. The initial state illustrated 

 in Figure 10a is completely determined by the 

 readjustment of the density field, but note that 

 diffusive interfaces have already developed in the 

 sense to be expected with an excess of S on the 

 left. At a later stage (Figure 10b) the frontal 

 surface is spread out horizontally by the inter- 

 leaving of inclined layers, the scale of which is 

 unrelated to that of the initial adjustment, and 

 which are driven entirely by the local density 

 anomalies produced by double-diffusive transports. 



A more sophisticated version of this experiment 

 is currently being studied by Ruddick (personal 

 communication) . He has set up identical vertical 

 density distributions on two sides of a barrier, 

 using sugar (S) in one half and salt (T) in the 

 other. When the barrier is withdrawn, there is 

 some small scale mixing, but virtually no larger 

 scale distortion. A series of regular, interleaving 

 layers then develops, with a spacing and speed of 

 advance which are systematically related to the 

 horizontal property differences. 



There are now many measurements which support 

 the view that the prominence and strength of 

 layering in the ocean are related to the magnitude 

 of the horizontal gradients of properties. To 

 cite just two examples: profiles across the Antarc- 

 tic polar front [Gordon et al. (1977)] reveal 

 inversions which decrease in strength with increasing 

 distance away from the front. Coastal fronts 

 between colder fresh water on a continental shelf 

 and warmer salty water offshore also exhibit strong 

 interleaving [Voorhis, Webb, and Millard (1976)]. 



A general conclusion which can already be drawn 

 from the laboratory experiments described in this 

 section is that the formation and propagation of 

 interleaving double-diffusive layers is a self- 

 driven process, sustained by local density anomalies 

 due to double-diffusive transports. Once a series 

 of noses and layers has formed, the changes of T 

 and S within them can be described in terms of the 

 one-dimensional (vertical) transport processes 

 previously studied. It should eventually be possible 



