DETERMINATION OF SALINITY 



73 



solutions, the slide-wire readings were slightly different 

 for the final standard than for the initial standard. If it 

 is assumed that these changes were permanent, the 

 slide-wire readings for the final standard should be used 

 whereas if these changes are assumed to have been tern 

 porary and to have disappeared (such as might be the 

 case if part of the auxiliary cell solution vaporized dur- 

 ing the run and condensed again afterward), then the 

 slide-wire readings of the final standard should be used. 

 Following; the remeasurement of the final standards at 

 40°C, they were withdrawn and replaced by other sam- 

 ples originally having the same salinity, and another 

 series of slide-wire readings taken. Assuming that no 

 change in salinity of the final standards had occured, 

 the final standard as reroeasured should be used. If it 

 be assumed that the final standard had changed in salin- 

 ity, then the fresh standard should be used. Thus there 

 are four combinations per cell which wil! give a temper- 

 ature coefficient of salinity. Their means have been 

 taken as shown in table 4. 



These temperature coefficients, even if accurately 

 determined, would only apply with the same settings of 

 the auxiliary resistance Q. As the settings given above 

 approximately represent ohms and as the resistance of 

 the cells were about 250 to 300 ohms each, it is seen 

 that the uncompensated sea-water resistance was about 

 1 per cent of the resistance of the unknown. Taking the 

 temperature coefficient of electrical conductivity of sea 

 water as 3 per cent per degree centigrade, the tempera- 

 ture coefficient of salinity of the bridge would have been 

 expected to be of the order of 0.0003 x 35.00 or about 

 0.01 per mille per degree centigrade. The general 

 agreement between the experimental and calculated val- 

 ues indicates that the temperature coefficient of the 

 bridge arm containing the Y cell differed from that of 

 the arm containing the X cell by not more than 3 parts 

 in 10,000. This would not be true generally, but would 

 depend on the difference in cell constants of the X and 

 Y cells and on the ratio of the resistance of Q to the re- 

 sistance of the sea water in the Y cell. It may be noted, 

 however, that had a wire resistance been used in place 

 of sea water in the Y cell, the temperature coefficient 

 would have been in the neighborhood of 0.03 x 35 or about 



1 per mille per degree centigrade. It should be under- 

 stood that the wire resistances in the bridge were of 

 manganin, having a negligible temperature coefficient, 

 and that when measurements were made the X and Y 

 cells were accurately at the same temperature. 



Copenhagen standard water was used at every third 

 station, substandards being used at the intermediate sta- 

 tions. At a station where Copenhagen water was used, 

 three large samples were taken, measured, and the sur- 

 plus kept until the ne.xt station (usually two days later), 

 when they were used as standards in the same manner 

 as the Copenhagen water was used. At this second sta- 

 tion large samples were again measured for use as 

 standards at the third station. At the next succeeding 

 station, Copenhagen water was again used and the cycle 

 repeated. It will be seen then that the possible errors . 

 of a single determination of salinity at successive sta- 

 tions are 1, 2, and 3 times the error of a single deter- 

 mination made at a station where Copenhagen standard 

 water was used. The bridge could be balanced to about 

 0.002 per mille salinity but the accuracy of the measure- 

 ment was not as great as this precision because of the 

 uncertainty of the resistance Q, errors in the assump- 

 tion that a shearing correction compensated for the dif- 

 ference between initial and final standards, and other 

 minor factors such as unequal heating caused by the 

 test current. Individual measurements were therefore 

 accurate to within about 0.02 per mille salinity in terms 

 of that of the standard used. Thus, if the salinity of the 

 Copenhagen standard water was accurately known, meas- 

 urements against such a standard were good to about 

 0.02 per mille salinity. Consequently it is possible that 

 at stations where a first substandard was used the meas- 

 urements might have been in error by 0.04 per mille and 

 at stations where a second substandard was used an er- 

 ror of 0.06 per mille was possible. This is highly im- 

 probable, however, inasmuch as such a situation would 

 require all the errors to be made in the same direction 

 and, as three different cells were used for the measure- 

 ment of samples from each station, such discrepancies 

 would probably have been detected in plotting vertical 

 distribution curves, unless the standards~for all three 

 cells were in error by similar amounts and in the same 



Table 4. Data for temperature coefficients of salinity for cells A, B, and C 



Cell 



Obser- 

 vation 



Temper- 

 ature 



Standard 



Q 



Slide-wire 

 reading 



Temperature coefficient 



From 



Value 



Mean for cell A 



B 



Mean for cellB 



Mean for cellC 



I 

 II 



ni 



IV 



I 



II 



III 



IV 



I 

 II 



ni 



IV 



30 

 30 

 40 

 40 



30 

 30 

 40 

 40 



30 

 30 

 40 

 40 



I and III 



I and IV 



II and III 

 II and IV 



I and III 



I and IV 



II and III 

 II and IV 



I and III 



I and IV 



II and III 

 II and IV 



°/ooper ° 

 0.0078 

 0.0047 

 0.0082 

 0.0051 



0.0064 



0.0041 

 0.0067 

 0.0044 

 0.0070 



0.0056 



0.0138 

 0.0116 

 0.0136 

 0.0114 



0.0126 



