604 



ROSSITER 



[CHAP. 16 



the evidence given above, and, in view of the amount of scatter in the residuals 

 and the relatively shallow depths of water in the areas considered combined 

 with the remarks made on page 600, this is perhaps not surprising. It is barely 

 possible that some orderly picture of the distribution of the nodal tide may 

 emerge from a study of carefully selected data; these should preferably be 

 from deep-water gauges, and should be first treated so as to remove as many as 

 possible of the effects of water-density and air-pressure variations. The use of a 

 smoothing process to reduce the standard deviation of the observations before 



Newlyn 

 (England) 



1940 



Fig. 2. Deviations of annual mean heights of sea-level. 



as observed, 



— — • — — after eliminating meteorological effects. 



analyzing for the nodal tide must be viewed with suspicion; Jeffreys (1939, 

 p. 248) has shown that whilst smoothing can reduce the scatter of a series of 

 observations, it also introduces a spurious and undesirable correlation between 

 individual members. 



So far as secular variations are concerned, the preceding analysis strongly 

 suggests that future investigations will profit greatly if prior consideration is 

 given to the elimination of contributions from the apparent nodal tide and of 

 random contributions from air-pressure variations. 



