James 



Hiese errors axe considered realistic based on past 

 studies of water injection temperature reports. Saur (1963) investi- 

 gated the difference between injection temperatiires recorded in the 

 log and specially observed temperatures for 6826 pairs of observations 

 from 12 ships on 92 different trips. He found a generally normal 

 probability distribution of errors centered around a small positive 

 bias. Errors of 6°F were shown with a standard deviation of differ- 

 ences of 1.6°F. Gibson (i960) compared sea surface temperatures as 

 reported synoptically to those that had been recorded as a log entry 

 and found I3 percent of the observations were in error due to coding 

 discrepancies. These errors were small, however, generally not ex- 

 ceeding 2*^. Other investigations of injection temperatures by 

 Francesehini (1955), Kuhn and Farland (1963) and Beetham (I966) show 

 the assumed error function to be realistic. 



In addition to the error in the temperature observation 

 there is also a navigation error in present ship reports, which con- 

 tributes to analysis errors. This was applied by assuming that one 

 third of the reporting ships were within the quadrangle represented 

 by a position reported to the nearest tenth of a degree. ITiat is, 

 no position error within the capabilities of the code. A third of 

 the reports were assumed to be +6 miles from the true position and 

 another third with a +12 MM error. The magnitude of the movement 

 errors were applied randomly while the direction of the deviation 

 was uniformly distributed to the four cardinal directions. 



(2) Description of Error Tests 



Three cases of errors were treated using various per- 

 cents of the original data. In Case I the full range of accuracy 

 errors and movement errors were applied as described above. The 

 resulting data plots of "bad data" are considered equivalent to the 

 present sea surface temperature data used in preparing analyses. 

 The Case II tests were based on data in which the error was limited 

 to +1*^ although the navigation error was the same as in Case I. 

 This would be equivalent to using only temperatures reported from 

 ships with reliable instrumentation, such as the Near Surface Refer- 

 ence Temperature (NSRT) System described by Beetham (1966), or per- 

 haps airborne radiation thermometer observations when carefully 

 taken. 



A third case was rvin where it was assumed reliable 

 instruments were available so that the error would be +1°P and the 

 navigation error sxifficiently accurate so that the shap is actually 

 in the quadrangle of area reported. 



An example of the three cases are shown in Figure 6 

 along with typical analyses. The original data are altered in value 

 and position according to the three cases previously discussed. The 

 product of misplaced data and inaccurate readings obviously misleads 

 the analyst into the wrong orientation of the isotherms in some cases. 



134 



