the Scripps Pier data shown on figure 6 there is a suggestion of a cooling trend 

 for the years 1939 through 1956. Applying the theory of runs to this period 3 the 

 following sequence is obtained: 



AAAAA BB AA BBB A BB A BB 



This sequence has eight runs. The critical number of runs at the 5 percent prob- 

 ability level for 18 observations is six, and it is concluded that no trend exists 

 even in this selected period of time. The applicable median /3 is 16.60° C. 



The years 1957 and 1958 have been recognized from a consideration of 

 several natural science parameters as the changing years. 5 These years appeared 

 to conclude the 1947-to-1957 decade — a decade of below-median sea-surface 

 temperatures. On figure 6, the /3 's for the succeeding years, 1959 to 1962 in- 

 clusive, have been included for Scripps Pier. In 1959 the annual average tempera- 

 ture continued to increase for the third successive year followed by an abrupt de- 

 crease in 1960 and lesser decreases in 1961 and 1962. This same pattern, though 

 less marked, occurred for the three island locations. Thus, it appears that in 

 1956 a long-term oscillatory variation, of period at least 6 years, began and that 

 it was still in progress in 1962. 



In examining the Scripps Pier /3's in figure 6, we find that the period 1930 

 to 1935 also appears to contain an oscillatory term with a period of several years. 

 In figure 3 are plotted the eight correlograms, obtained after removing the annual 

 and semiannual oscillatory terms, for 5-year periods at Scripps Pier. Although 

 these correlograms are plotted out to lags of only 150 days, they were computed 

 out to lags of 900 days. A study of the eight 900-day lag correlograms shows one 

 to be somewhat different from the remaining seven — the correlogram for the 1931- 

 to-1935 period. The correlogram for this period for lags out to 1400 days (fig. 9) 

 shows a peak in the autocorrelation coefficient of 0.46 at a lag of about 1170 days, 

 or a period of about 3.2 years. Since the points are scattered in the neighborhood 

 of this peak as well as in the neighborhood of the minimum at lag about 750 days, 

 another estimate of period length is given by twice the difference in lags between 

 the up-crossing at lag 960 days and the down-crossing at lag 265 days, or about 

 3.8 years. Either period confirms the intuitive conclusion reached from an exami- 

 nation of figure 6. 



It thus appears reasonable to conclude that sea-surface temperatures will 

 vary significantly over periods of several years and that these occurrences are not 

 unexpected or improbable. 



REVERSE SIDE BLANK 19 



