i.e., by considering the tidal force Itself. This would be 

 a purely deductive method stemming directly from the differ- 

 ential equation. 



e. Another possible approach now being considered is an 

 attempt to develop a purely noncausal technique of predicting 

 the thermal structure by using the principles of statistical 

 mechanics. This would probably employ the largest possible 

 collection of BT time-series data and would be developed by 

 using standard techniques of time-series prediction. In such a 

 study it would be necessary to assume some particular statisti- 

 cal form of the distribution of the components of the periodic 

 oscillations. An interesting treatment of this type was applied 

 by Pierson (1952) who demonstrated prediction in time and space 

 of a wave train whose displacements from the mean displayed a 

 Gaussian distribution, 



f , All the BT data at the Hydrographic Office are now being 

 analyzed for the occurrence of synoptic situations, or situations 

 in which there was a reasonably large horizontal distribution 



of observations for a given period of time. If good examples of 

 such distributions are found, it is intended to attempt to apply 

 wel]^-known techniques of extrapolation and kinematic analysis, 

 with a view to developing a useful prediction tool. 



73 



