James 



A. Area A - Mantial Analysis of Perfect Data 



Area A was selected to be representative of the complex 

 isotherm patterns found north of the Gfulf Stream. Figure 1 shows the 

 original data plot emd analysis for Area A. Sea surface temperatures 

 are considered perfect values, since the data were fitted to the iso- 

 therm pattern. The grid spacing of 12.5 NM covers an area equivalent 

 to a five degree square hut a geographic fix is not indicated since 

 the isotherm pattern is considered typical of thermal conditions over 

 a wide area north of the Gulf Stream. The major features of Area A 

 are two warm tongues separated by a cold intrusion of water from the 

 northwest. Colder water is also evident in the northeast portion of 

 the area. 



The first test was to randomly select ten percent of the 

 grid data and to analyze this new temperature field. This procedure 

 was followed to see if the essence of the isotherm pattern could be 

 maintained with only ten percent of the original observations. Tables 

 of random values were used to select the i<-0 observations that would 

 be utilized in the analysis. Figure 2 shows the resulting data plot 

 of 10 percent data by a trained analyst with no knowledge of the 

 original isotherm pattern. Comparison of Figures 1 and 2 indicates 

 that although some distortion occurs in the pattern the major fea- 

 tures are well described and the more prominent tongue to the east 

 was still drawn as the main feature. A mean absolute error was com- 

 puted by overlaying the new analysis over the original data. An 

 error of 1.30°F was obtained, whidi is small compared to values nor- 

 mally found for analyses in this area. This is not surprising in 

 that there was an adequate input of observations and the data were of 

 perfect accuracy. 



Three more tests of this nature were made using 5^ 3*5 and 

 2 percent of the perfect data shown in Figure 1. Tba-t is, the data 

 plot consisted of 20, ik and 8 obsejrvations respectively. In each 

 case the obsei*vations utilized were randomly selected and the analyst 

 preparing the analysis did not have a pre- conception of the pattern. 

 Different analysts were used to avoid continuity of analysis. 



Figures 3, 4, and 5 show the resvilting analyses. With 5 

 percent of data the isotherm pattern is still well defined although 

 the source of the cold water intrusion is misplaced. Even with 3*5 

 percent of data the two warm tongues are properly located, although 

 they are considerably^smoothed. For optimum sonar routing, however, 

 the pattern shown would still be of value. Using the 8 observations 

 available with 2 percent of data the analysis does not describe any 

 of the original pattern. The pattern consists of one smoothed warm 

 tongue which is not too accurately centered. Unfortunately, as it 

 will be shown later, present sea surface temperature analyses are 

 based on data input comparable to the 2 percent chart but of much 

 lower quality than represented by the observations utilized in Fig- 

 ure 5» Only through composite groupings of observations for 3 to 5 



127 



