ability of the instrument to measure water prop- 

 erties with sufficient accuracy. 



The continuous sampHng devices that are avail- 

 able on the market today provide a measurement 

 accuracy from 0.01° C. to 0.05° C in temperature 

 and about 0.03°/oo in salinity. Kollmeyer (1964) 

 pointed out that errors of ±0.02° C. and 0.02°/oo 

 can, under some circumstances, result in a maxi- 

 mum possible error of 0.0169 dynamic meters in 

 1,000 meters. To evaluate the final accuracies in 

 dynamic heights obtained from a continuous 

 sampler, a more realistic analysis is required here. 

 A selected group of stations, some of which were 

 used in the comparative anomaly plots above, were 

 processed several additional times on the computer 

 using different error values of temperature and 

 salinity. The errors used were designed to be 

 additive therefore increasing the anomaly of 

 specific volume and giving a maximum error. As 

 mentioned above, there were a great number of 

 points used such that they do approach the con- 

 cept of continuous samples. The results are shown 

 in table IVD, with errors given in dynamic meters 

 difference from the true values. 



The instrument error is also present when using 

 Nansen bottles and reversing thermometers. 

 Whitney (1957) indicates that accuracies of 

 ±0.024° C. for 0.05° C. scale graduations and 

 ±0.036° C. for0.1° C. graduations can be expected, 

 and Kollmeyer (1964) indicates ±0.01° C. and 

 ±0.02° C. respectively. During Ice Patrol 1965 

 some 813 thermometer comparisons were made 

 during the surveys. Each Nansen bottle con- 

 tained two thermometers with the accepted 

 temperature value being the average figure ex- 

 cept in cases of obvious malfunction. Since the 

 true value of the temperature is not known, the 

 error in these thermometer readings can be 

 represented by the standard deviation of each 

 thermometer from its paired average. Table 

 VD shows this deviation for each survey during 

 the 1965 Ice Patrol season giving the largest 

 95-percent confidence limit of ±0.016° C. ob- 

 tained during the third survey. An analysis 

 conducted this year on Ice Patrol between two 

 different inductive salinometers. Industrial In- 

 struments Model RS-7A and Hytech Model 6210 

 using 3 different operators and duplicate bottled 

 salinity samples yielded a standard deviation 

 between duplicate samples of 0.007°/oo over 

 41 comparisons. This gives a 95-percent confi- 

 dence limit of 0.014°/oo for a laboratory salinometer 

 at sea. 



From table IVD it can be seen that the average 

 of the maximum instrument error in dynamic 

 heights over 1,000 meters for point sampling is 

 ±0.0124 DM using measurement accuracies of 

 ±0.016° C. and ±0.014°/oo. The average of the 

 maximum instrument error for a continuous 

 sampler ranges from ±0.0234 DM to ±0.0286 

 DM over measurement accuracies of ±0.01° C. 

 and ±0.05° C. respectively and with a salinity 

 error of ±0.03°/oo. It is obvious that the errors 

 from both instruments suites will be random, 

 tending to result in a bell-shaped distribution 

 curve with the maximum errors of table IVD as 

 its practical end points. With this assumption 

 it is then possible to compute the 2<r or 95-percent 

 confidence limit of measurement accuracy for these 

 dynamic heights. Table IVD also lists this 2(r 

 value. 



It can be observed from table IVD that the 

 greatest dynamic height error contribution is 

 from the salinity measurement error which is 

 about five times as great as the error in temper- 

 ature measurement. A temperature measure- 

 ment error of ±0.05° C only accounts for about 

 0.005 dynamic meters while a salinity measure- 

 ment error of ±0.03°/oo accounts for about 0.025 

 dynamic meters over 1,000 meters. 



In order to evaluate the accuracy of a continuous 

 sampler versus the Nansen bottle point sampling 

 method for dynamic height calculations, the 

 errors must be considered as follows: Total error 

 (5)= instrument error (5t)+missed gradient 

 error (Sg). 



The total error for the continuous sampler falls 

 within a range of ±0.0231 DM to ±0.0282 DM, 

 depending on the temperature error, and using 

 the average 95-percent confidence figures for 

 instrument error from table IVD and a missed 

 gradient error of 0. For point sampling, the 

 missed gradient error value is so variable and 

 dependent on the water mass and density gradient 

 levels that it is difficult to assign a fixed error, 

 however, from the comparisons of the fixed 

 sampling method and the gradient sampling 

 method presented herein, it has been shown that 

 over 1,000 meters of integration, the 95-percent 

 confidence limit of point sampling error is ±0.0133 

 DM due to missed gradients. Using this figure 

 and the average 95-percent confidence limit 

 mstrument error of ±0.0122 DM from table IVD 

 for Nansen bottles, the error resulting from the 

 use of standard depth point sampling is equal to 



69 



