coastal location, Scripps Pier. It is also recognized that there is 

 a diurnal variability present in these data since the daily observa- 

 tions were taken at random times during the day. The latter 

 short -period variations will not be included in the model, with the 

 result that their effect will contribute to the unexplained variance. 



Since the seasonal variation in sea-surface temperature is 

 not symmetrical about an origin, the use of the following fifth- 

 degree polynomial is suggested by the scatter diagrams: 



r ■- f/^ . o,D ^ aJJ- - o,,D'' ^ a^D'^ ^ ajf (1) 



where l> is measured in days from some arbitrary origin, T is 

 the least-squares fitted value, or estimate, of surface tempera- 

 ture, and the sub scripted n's are regression coefficients to be 

 estimated. 



Equation (1) was fitted to 5 years of data taken at each of the 

 five locations listed above to demonstrate the adequacy of the 

 fifth-degree polynomial as an estimator of the seasonal sea- 

 surface temperature variation. The origin of time was taken as 

 July 1 and the years referred to are fiscal years. The notation 

 "1954" refers to a fiscal year 1954 starting 1 July 1953, and end- 

 ing 30 Jmie 1954. 



The following related quantities are used as measures of the 

 "goodness of fit" of equation (1) to the observed data: /' , multiple 

 correlation coefficient: 100K-. percent variance explained by 

 regression; and «, standard deviation in degrees Fahrenheit of the 

 observations about the regression curve. 



17 



