5.2 useless Rainier 



On the Rainier^ the shipboard observations were made forward of the mid- 

 section of the ship, either on the windward side of the pilot house or on the 

 flying bridge above it about 72 ft from the bow of the ship. Because of the 

 slope of the ship's deck, the pilot house was approximately the same height 

 above the sea surface as the boom; the flying bridge was about 8 ft higher. 



The nighttime differences between the shipboard and boom temperatures 

 shown in figure 11 were unexpected. When the wind is not across the ship to- 

 ward the boom, the shipboard temperature averages almost 0.4°C cooler during 

 the hours of darkness, while during daylight hours the two temperature aver- 

 ages are almost identical. Therefore, the data obtained when the wind was 

 not blowing across the ship toward the boom were subdivided into 1100-1959 

 and 2000-1059 GMT subgroups. When the wind blew across the ship toward the 

 boom, the shipboard temperature generally was between 0.2°C and 0.5°C cooler 

 than the boom temperature for the entire 24 hours. Because these data did not 

 readily lend themselves to subdivision, they were retained as a single group. 



Table 3 shows the sample means and variances for the distributions of 

 the observed temperatures used in each comparison. At night, and during the 

 day when the wind blew across the ship toward the boom, the shipboard temper- 

 atures also were colder than the adjusted rawinsonde temperatures. The reason 

 for this is not immediately apparent, but probably lies in the effect of the 

 ship on the environment. Another interesting feature is the higher variance 

 of the rawinsonde temperature during the day than at night, which could have 

 been caused by the scattered cloudiness and the daytime conditions that are 

 typical over this area of the Atlantic Ocean. 



Figure 12 was used to obtain the lag coefficients for determining the 

 independent sample size from which the critical values for the parametric 

 statistical tests were extracted. 



Results of the parametric and nonparametric tests applied to these data 

 are shown in table 4. The first point of interest is the different types of 

 decisions provided by these tests. The boom and rawinsonde data were plotted 

 for the comparisons and, as figure 13 illustrates, the data are apparently 

 not normally distributed. Thus, for the Rainier, the emphasis should be 

 placed on the K-S test results. The shipboard-boom comparisons for the 2000- 

 1059 GMT subgroup and the shipboard-rawinsonde comparisons for the 2000-1059 

 and 1100-1959 GMT subgroups show that this test rejects the null hypothesis 

 (see table 4). The usefulness of the K-S test is pointed up by the fact that 

 for the 1100-1959 GMT subgroup this test rejects the hypothesis for the ship- 

 board-rawinsonde comparison but does not do so for the boom-rawinsonde com- 

 parison, although table 3 shows the differences between the sample means to 

 be almoet identical (shipboard-rawinsonde being 0.15 and boom-rawinsonde 0.14) 

 Figure 14 indicates slight but important differences in the comparisons for 

 the 1100-1959 GMT subgroup. Thus, based on the K-S test, it appears that, 

 when the wind is not blowing across the ship toward the boom, the boom data 

 are the better measurement of the environment being sampled by the rawinsondes, 

 When the wind does blow toward the boom, this test does not reject the hypo- 

 thesis that the sensors are measuring the same environment. Therefore, it 

 seems to matter little whether the boom or the shipboard data are used with 

 the rawinsondes, although the boom does appear, on the average, to provide 

 nighttime temperatures 0.1°C to 0.2°C higher than those measured on board the 



ship . 



8 



