RAPLOC/DKEPTOW experiments. These measurements were all made from the ORB at 

 anchor. This data set consists of 1 1 STD/SV profiles made over the same time interval that 

 107 XBT profiles were obtained with two independent XBT systems. 



The differences for the five sets of comparisons are summarized in appendix B, 

 which contains tabulated statistical summaries for each of the data sets and plots which 

 present 200-, 300-, and 400-m-depth positive and negative ogives for three subsets of differ- 

 ences — all visually acceptable profiles that include the profiles made when the Gulf of 

 Alaska and CAPER XBT systems malfunctioned; profile set I; and profile set lA.* Listed in 

 the tabulations are the number of differences, average differences, standard deviations, and 

 the percentage of positive and negative differences. Table 28 is an attempt to summarize 

 the Gulf of Alaska and CAPER comparisons for the sets of profiles that consist of all 

 visually acceptable profiles. Table 28 supports the following observations: 



a. The average differences are positive at all three depths, and are an 

 increasing function of depth. 



b. The standard deviation is about the same at all three depths. 



c. At all three depths, more of the differences are positive than negative. 



d. At all three depths, the largest differences are positive. Thus, for the 

 data set of visually acceptable XBT profiles, it is concluded that, on the 

 average, the recorded temperatures are higher and the vertical temper- 

 ature gradients larger** than those measured by the hydrocast and 

 STD/SV systems. These are the expected results if malfunctioning 



XBT systems are undetected. 



The differences for profile set I, which consists of all XBT profiles except those 

 differences associated with the malfunctioning system, are summarized in table 29. An 

 inspection of the table 29 data for the Gulf of Alaska water mass 2 and transition water 

 mass, and for the CAPER water mass 2, leads to the same observations. Thus, even with the 

 omission of the profiles made with the two malfunctioning XBT systems, the average 

 temperature and the average vertical temperature gradients remain systematically biased. 

 For the SUDS I and the RAPLOC/DEEPTOW comparisons, the average difference was 

 negative at one of the three depths; for the RAPLOC/DEEPTOW comparisons, the average 

 difference was a decreasing, rather than increasing, function of depth. However, for both 

 data sets, at all three depths, the largest differences were positive. This later observation 

 tends to support the previous conclusions. 



In an effort to eliminate the biasing profile, profile set I was separated into two 

 subsets by using the accuracy criterion defined in the XBT system errors section. The 

 criterion was modified by using the final system errors in place of the preliminary system 

 errors. One subset included all differences that satisfied the criterion (profile set lA), and 

 the other all differences that did not satisfy the criterion (profile set IB). The differences 

 for profile set lA are summarized in table 30. An inspection of these data shows that for 

 the Gulf of Alaska data set, the average differences still show a slight increase with depth 

 and, for the SUDS I data set, a slight decrease with depth. For all five sets of data, the 



"Profile set lA is a subset of profile set 1. It is defined above. 



"The mathematical convention that a positive number is larger than a negative number is used in this 

 discussion. For example, a +0.12°C/100 m gradient is larger than a -O.22°C/10O m gradient. 



64 



