Although the absolute differences are small, table 9 indicates a diurnal 

 variation of 0.5°C in the shipboard temperature when the wind is not from the 

 ship toward the boom, while the boom temperatures vary by about half that 

 amount, which agrees closely with the diurnal variation in the rawinsonde 

 temperatures . 



As is evident from table 10, the K-S test does not reject the null hypo- 

 thesis in any of the comparisons, but the results of the parametric Student's 

 t test disagrees with the K-S test results in two of the comparisons. Although 

 statistically either choice would be acceptable, it probably would be prefer- 

 able to use the boom data with the rawinsonde data when the wind is not blowing 

 across the ship toward the boom and shipboard data when the wind is across the 

 ship toward the boom. 



Figure 23 shows the autocorrelation functions for the boom, shipboard, 

 and rawinsonde temperatures for the Roakauay . As this figure indicates, the 

 autocorrelation function goes to zero after the second lag for the daytime 

 data and after the fourth lag for the nighttime data. Here, as for the other 

 ships, division of the number of observations N by 2 or 3 will provide an esti- 

 mate of n, the number of independent samples to be used in making the various 

 tests applied here. 



6. CONCLUSIONS 



It has been established, primarily by use of the Kolmogorov-Smirnov 

 nonparametric test, that for the majority of the cases examined here the 

 manually recorded shipboard temperatures were not from the same environment 

 as the electronically measured boom temperatures. The reason for this does 

 not lie in the sensors, but primarily in the ship's influence on the environ- 

 ment as indicated by the fact that the rawinsonde and boom instruments appar- 

 ently were measuring the same environment. The difference between these 

 measurements and the manually recorded shipboard temperatures showed a sig- 

 nificant dependence on wind direction relative to the ship-boom axis. 



This study, then, shows that if only one set of BOMEX temperature data 

 is to be used with the rawinsonde measurements, that set should be from the 

 boom observations. If it proves possible to choose data for each rawinsonde, 

 a procedure similar to that presented in this paper could be used to make 

 the proper choice. 



Use of the boom data alone would probably lead to biased results, as 

 this study indicates that a ship does influence its temperature environment 

 to a considerable extent, which in turn implies that the boom instrumentation 

 is similarly affected. Future studies should consider redesign of the boom 

 or replacement of the boom instrumentation by other sensors unless the boom 

 can be oriented into the wind at all times. 



The temperature of the air over the ship is influenced at all hours, in- 

 cluding nighttime. The ship's observed air temperatures are from 0.1°C to 

 0.2°C colder than those from the boom instruments. A possible, and plausible, 

 explanation lies in the thermal capacity of the ship, with its superstructure 

 and deck. The differences in the nighttime cooling - ship's temperature being 

 lower than the boom temperature - are caused by the fact that the radiational 

 cooling over the ship's structure produces a lower temperature than at the boom, 

 which is located over the open sea, primarily because of the difference in 



11 



