To determine whether the effect of the ship environment was noticeable, 

 the first 240 m of a set of rawinsonde flights for each ship were analyzed. 

 The set consisted of 5-sec averages of temperature and humidity - derived from 

 the SCARD analog recording sampled at the rate of 10 times per second - 

 plotted against height. If the flight contained noise or bad references in 

 the first 200 m it was not used. 



Several flights showed a strong lapse rate of from 3°C to 6°C per 100 m 

 in the lowest 50 to 100 m, followed by a decrease to about 1°C per 100 m, 

 which usually remained relatively uniform to the 200-m level. Therefore, the 

 temperature value selected for comparison was the first value above this change 

 in lapse rate. When such a change was not apparent, a temperature value near 

 the 50-m level was selected, and the data were then extrapolated to deck 

 height (9 m above the ocean surface) based on a lapse rate of 0.9°C per 100 m 

 as determined for a homogeneous layer over the Caribbean Sea by Malkus (1958) . 



To reduce the bias that would be introduced by using observations record- 

 ed at different times by the temperature and humidity sensors, the value from 

 one sensor was discarded when the corresponding value from the other was 

 missing . 



4 . THEORY 



The superadiabatic lapse rate evident in the analyzed rawinsonde data 

 [fig. 4) apparently was missing in soundings analyzed by other investigators 

 of tropical phenomena. Brocks (Roll, 1965) showed that the superadiabatic 

 layer just above the sea surface rarely extends above 23 m, and Malkus (1958) 

 has given evidence that the layer above the superadiabatic is generally homo- 

 geneous where the temperature profile roughly is characterized by an adiabatic 

 lapse rate. The superadiabatic lapse rates in the analyzed BOMEX rawinsonde 

 data may be artificial. The reason for such spurious rates may lie, for ex- 

 ample, in that the ship heated its immediate environment or possibly in that 

 rawinsonde instruments had been placed on the deck in sunlight and inadver- 

 tently had been allowed to "bake" before being released. Future comparison 

 of these rawinsonde data with data obtained from dropsondes and from the 

 boundary layer instrument package (BLIP) flown from the Ooeanographer and 

 Mt. Mitchell may help clarify this problem. 



If a ship were a contaminator of the environment in which temperature 

 was measured on board, a wind across the ship toward the boom might spread 

 contamination into the vicinity of the boom. The data were therefore grouped 

 according to wind directions (fig. 5). For each scheduled rawinsonde release 

 time, the differences between the shipboard and boom temperatures were then 

 averaged and the release times subdivided into groups having similar differ- 

 ences. For simplicity, and in order to keep the sample size of the groups 

 large enough for significant results, the data from each ship were divided 

 into two groups. 



The hypothesis that each of these data samples are representative mea- 

 surements of the same environment could be tested statistically in various 

 ways. For this study, the variance-ratio or F test, the Student's t test, 

 and the Kolmogorov-Smirnov test were chosen. 



