The constants M, K, and L may then be successively evahiated using 

 the expressions: 



(m' + B) r Srimr + B)-Ss{ms+B) 



S,(m^+B)-S2(m,-^B) 



] 



(S,~S,)(m,+B){?n, + B)_ 

 (m' + B) r Sr S, 1 



Si \_{mi-i-B) J 



Practical consideration of the fact that values of S are known only to 

 1 part in oo.OOO and values of m are known only to about 1 part in 

 50,000, these uncertainties combined with the presence of small 

 (lift'erences between large numbers in the solution of the quadratics 

 mean that any evaluation of B by tliis method may result in errors 

 of an order of magnitude or more. 



In the derivation of the formula for the calibration curve based on a 

 linear relationship between conductivity and salinity, it is assumed 

 that C=C'(K-^LS) and the calibration ciu've becomes: 



m-\-B L 



where S, B, m' and m liave the same significance as before. B may 

 be evaluated from tlie expression: 



{7n,-m,){Si-S^)-ims-7n,){Si-S,) 



J. and /v may then be evaluated successivel\ from the expressions: 



(m' -{- B) (mo— mi) 



L= 



{Sr-S.2){mi + B){m,i-B) 



m-^B 



As will he shown later, 2 of the 11 drums seem to have atypical 

 condiictivity-chloriiiity rclationshijis. Excluding these 2, the remain- 

 ing 9 pi-ovide 84 cond)inati()ns for the evaluation of B above. These 

 were evaluated and a weighted nu'an (sum of the numerators divided 

 hy the sum of the dcnonnnators) B (h^rived as 195.77;^. Tlie 'M] 

 cond)inations for the cvnhialion of /. were weighted as before to 



36 



