An examination of these statistics supports the conclusion 

 that a fifth-degree polynomial is an acceptable estimator of the 

 seasonal variation. Relatively, the fit is best for open-ocean data 

 (PAPA and ECHO), with 92 to 98 percent of the observed varia- 

 bility explained; next best at island locations (St. James Island and 

 Langara Island), with 83 to 96 percent of the variability explained; 

 and poorest at Scripps Pier, located about 1000 feet from shore, 

 with 83 to 92 percent of the variability explained. Equation (1) 

 has been fitted to many other years of data with similar results. 



On figure 1 the solid line is a plot of equation (1) using the 

 regression coefficients for the indicated location and year. The 

 histograms show the distribution of differences between the 

 observed and estimated sea-surface temperatures and the vertical 

 dotted lines indicate one standard deviation of these differences. 



It is of interest to note that from September 1 to January 1 

 the rate of cooling varies from 2. F per 30-day period at St. 

 James Island to 3. 4°F per 30-day period at PAPA. The rate of 

 warming from May 1 to July 1 varies from 3. 2 F per 30-day 

 period at Langara Island to 4. F per 30-day period at ECHO and 

 St. James Island. 



20 



