the excess of one deviation over the other was attributed to causes other 

 than space and time variations, e.g., instrument error or data processing 

 error. Normally, both experimental and historical data might contain 

 instrument error. The standard deviations were not directly comparable, 

 because the time periods of the experimental data were only two-thirds 

 those of the historical data. 



The ratios (table 25) for various parameters related only to depth 

 varied from 0.95 to l.l4, well within the 33 percent variation that can be 

 attributed to the difference in length of the time periods. Thus, the 

 magnitude of error incorporated into the standard deviations is approximately 

 the same for the historical and experimental data. However, the standard 

 deviation ratios of various parameters related to temperature consistently 

 fell beyond the 33 percent tolerance and varied between 1.45 and 2.66. This 

 indicates that standard deviations of historical data related to temperature 

 contain a large error factor. 



CONCLUSIONS 



Conclusions drawn from analysis of the data are pertinent to vari- 

 ations in the thermal structure only at ocean weather station ECHO. Stud- 

 ies are contemplated for other stations. The following specific conclu- 

 sions are made: 



1. Because experimental cruise data are of much greater value than 

 historical data, samples should be collected at other ocean weather stations 

 for comparison purposes. 



2. Monthly means of parameters of the thermal structure show sinus- 

 oidal annual curves. 



3. The temperature element of the 450-foot BT appears to be subject 

 to considerable error; the pressure element appears to be relatively stable. 

 Temperature errors may indicate reference temperature errors. 



k» The probability of persistence for all thermal structure parameters 

 is highest when comparing means of all observations taken over a period of 

 a month within adjacent 1-degree quadrangles. This result supports the 

 persistence theory explained by Perlroth and Simpson (5). 



5. The probability of persistence is quite low when comparing year- 

 to-year mean monthly values of all observations within a 1-degree quadrangle. 



6. The probability of persistence for any parameter during the same 

 5-day period is generally greater between adjacent lO-minute quadrangles 

 than it is between consecutive 5-^y time periods within the same lO-minute 

 quadrangle . 



7. The gradients of the first and maximum gradients and the thickness 

 of the maximxim gradient exhibit greatest spatial persistence. 



8. The only parameter exhibiting high temporal persistence is the 

 thickness of the first gradient. 



11 



