to produce a regression equation that would represent the temporal 

 and spatial variation of sea-surface temperature independently of 

 year-to-year effects. This would be analogous to the climatic 

 charts of the meteorologist. 



In the absence of any other information this regression 

 equation will give the best estimate of sea-surface temperature 

 and its variance for any latitude, longitude, and day-of-year. 

 Year-to-year variations, which of course do exist, are neglected. 

 To improve on this estimate it is necessary to consider the year- 

 to-year variation. This might be done as follows: Assume that 

 some observations have been made during the past several months 

 over the area. The composite surface could be adjusted to the 

 new data by a least-squares adjustment of the origin of the regres- 

 sion equation to pass the surface through the currently observed 

 data. The adjusted surface will then be the best estimate of sea- 

 surface temperature for any future day. 



For any particular day of the year contour charts , such as 

 illustrated by figure 20, could if desired be prepared. This par- 

 ticular chart was prepared, using equation (4) fitted to the data 

 taken in fiscal year 1950 in the large area, for 8 November 1950. 

 A comparison of this chart with a chart prepared using more 

 classical techniques shows excellent agreement. 



76 



