XBT SYSTEM ERRORS 



It is common practice to read temperatures visually from the XBT analog recording. 

 With practice, it is possible to read the profiles to a temperature accuracy of ±0.2°C and 

 a depth accuracy of ±2 m. However, for this analysis, the profiles were all read by means 

 of a Hewlett-Packard digitizing system that consisted of an HP9830A calculator, an 

 HP9864A digitizer, and an HP9866A printer. The HP9864A digitizer measures to the 

 nearest 0.01 inch in x and y. Over a temperature range of -2.0°C to 35.0°C, this translates 

 to a temperature digitizing accuracy of 0.055°C; and over a depth range of to 460 m or 

 1830 m, to a depth digitizing accuracy of 0.95 m. All temperatures were recorded to the 

 nearest 0.01°C and the depth to the nearest 1 .0 m. The digitizing programs are listed in 

 appendix A. 



According to Sippican Corporation (personal communication): "No trace should be 

 used if the calibration hne deviates from 62.0°F (or 16.7°C) by more than 0.5°F (or 0.3°C). 

 If the deviation falls between 0.2°F and 0.5°F (or 0.1°C and 0.3°C), the calibration line 

 must be used as reference to avoid significant errors in temperature computation." The 

 digitizing programs listed in appendix A use the calibration line as a reference. In the data 

 set used in this study only seven Gulf of Alaska, four SUDS I, and four RAPLOC/DEEPTOW 

 profiles had calibration line deviations that exceeded ±0.34°C (±0.3°C to the nearest 

 0.1°C). These profiles will be noted at appropriate places in the analysis to follow. 



The manufacturer states that the absolute value of the accuracy of the XBT system, 

 to the nearest 0.1°C, is ±0.2°C. In any statistical analysis that requires an estimate of the 

 accuracy of a measurement, it is more useful to have the accuracy specified as a standard 

 deviation. The following rationale is used to obtain an estimate of the standard deviation 

 of the XBT system. Since, in this study, all XBT temperatures are read and recorded to the 

 nearest 0.0 1°C, the absolute accuracy may be interpreted to be ±0.24°C (±0.2°C to the 

 nearest 0.1°C). For a sample drawn from a normal population, plus or minus two standard 

 deviations includes 95% of the sample. If it is assumed that ±0.24°C is equal to two 

 standard deviations, then an estimate of the standard deviation for XBT measured temper- 

 atures is ±0.1 2°C. In the discussion to follow, the standard deviation of the XBT system is 

 assumed to be ±0.12°C. 



Procedure for Determining XBT System Error 



Perhaps the most accurate way to determine the XBT system error in the field is to 

 make several simultaneous XBT and STD/SV profiles. Since the temperature measurements 

 used in these studies were made in support of acoustic propagation loss experiments, the 

 making of several simultaneous profiles was not generally possible. As a result, it was 

 necessary to devise an alternative method. All the visually acceptable XBT profiles made in 

 water masses where hydrocast and/or STD/SV measurements were made are used to estab- 

 lish the XBT system errors. These profiles were processed as follows: 



1 . Compute the average 200-, 300-, and 400-m hydrocast and/or STD/SV 

 temperatures for all hydrocast and STD/SV temperature measurements made in the same 

 near-surface water mass. 



2. Read XBT profile temperature at 200, 300, and 400 m, applying any indicated 

 temperature and/or depth corrections. 



25 



