Small arrays of sensors were set on buoys off Castle Harbor 

 and maintained for 2 days each. The sensors were primitive 

 strip chart recorders and metal thermometers. Instruments 

 were placed on the bottom and several higher up in the water column. 

 The details were reported elsewhere (Refs. 3, 9), but in summary, 

 a high degree of coherence was found at horizontal distances up to 

 2 miles, for waves of period about 5 hours. The higher frequency 

 motion (all results are qualitative) are coherent at 300 meters in the 

 horizontal. The fluctuations of temperature were greatest at the 

 bottom, and there is a suggestion that the slope (20 degrees) not only 

 amplifies the motion, but also degrades the incoming energy into 

 higher frequencies, thus accounting for the low coherence found by 

 Haurwitz, Stommel and Munk. Some theoretical work has been done 

 on the effects of bottom slope (Refs. 4, 7). 



The results of the buoy measurements were considered encouraging 

 enough to warrant proceeding to the actual construction of an array. 

 Our design uses Stommel's cable technique, but suitably modified to 

 avoid the limitation of a bottom mounting. Argus I., Bermuda was 

 originally chosen as the site for installation because the bottom drops 

 away so steeply that comparatively deep water is reached quickly. 

 However, logistic considerations dictated the use of the slope off 

 St. Davids, Bermuda. The array is confined to one dimension (hori- 

 zontal and perpendicular to the contours). We later added more di- 

 mensions using short-term buoys. A long-term buoy is planned to 

 extend the aperture. For greatest directionality of the linear array 

 (i.e., the main lobe of the antenna function) the sensor cable points 

 toward the deep South Atlantic. The center of the main thermocline 

 is at approximately 800 m. near Bermuda, and this is the optimum 

 experiment depth. A working depth of 700 m. was used howeven in 

 order to increase the safety margin of the floats. 



The optimum configuration of the sensors is the subject of con- 

 siderable guesswork. The sampling interval is dictated by the time- 

 constant of the instrument, which in turn is dictated by the phenomena 

 of interest. In principle, there is no propagation of gravity wave 

 disturbances with frequencies greater than the maximum of the 

 Brunt-Vaisala frequency. Using hydrographic data (and there is 

 15 years of twice-a-month Panulirus data) estimates can be made 

 of the high-frequency cut-off. These estimates are somewhat crude 

 due to the nature of water-bottle sampling. However, it appears 

 that the high-frequency cut-off is in the neighborhood of 15 minutes, 

 dictating samples of not less than one every 5 minutes. 



196 



