lated from the volume of water 

 needed to create a Green Island to 

 Battery gradient (Hammond 1975). 

 However, since variations in Qgj 

 during this time are small, the 

 lag-related errors in Q f estimates 

 are also small and 1 day lag time 

 was used under all flow condi- 

 tions . 



SALINITY PROFILES 



Tidally averaged salinity in 

 the upper and lower layers are need- 

 ed to compute volume transports 

 between layers and boxes. Since 

 the required data are not avail- 

 able except occasionally at MP -7 

 and MP 18, an empirical model of 

 tidal variations in salinity was 

 used to estimate tidal averages 

 given a salinity profile and the 

 vertical boundary between surface 

 and bottom layers at each sta- 

 tion. The vertical boundary was 

 determined by examining the prin- 

 cipal components (Morrison 1976) 

 of velocity and salinity profiles 

 averaged over 2-m sections at each 

 station (Figure 3). Two components 

 accounted for most of the variance 

 (94%-98%) in both velocity and 

 salinity. Varimax rotation 

 (Morrison 1976) of the two-compo- 

 nent structure at each station showed 

 a "surface" and "bottom" component 

 with a mid-water column boundary 

 (Figure 3). The velocity and salin- 

 ity structure at each station was 

 similar, in agreement with the Hud- 

 son's classification as a partially 

 mixed estuary with dominantly two 

 layer flow. Using this analysis as 

 a guide, upper and lower layer mean 

 salinities were determined from 

 depths having a larger correlation 

 with component 1 or component 2, 

 respectively. 



These upper and lower layer 

 salinities were then used to esti- 

 mate tidally averaged salinities 

 by using them to fit an empirical 

 model for salinity distribution 

 in each layer. 



S (x,t) = So exp (-kx) + St 



(-COS (2 TVt)) (9) 



x = distance upstream 



from MP -7 

 t = proportion of time 



elapsed in tidal cycle 

 (LWS-LWS) at x (deter- 

 mined from USCG tide 

 tables) 

 S = half tidal range, S q = 

 mean salinity at MP -7 

 k = fitted advection- 

 diffusion parameter 



An estimate of tidally averaged 

 salinity was then obtained by in- 

 tegrating (9) over the tidal cycle 

 at each station. Since the equa- 

 tion is non-linear, Gauss-Newton 

 non-linear regression was used to 

 determine the least squares set 

 of parameters (Snedecor and Coch- 

 ran 1978) for each set of two 

 successive sampling times (1 day 

 apart 1977, 1 week apart 1978). 

 Data were not pooled if Q f changed 

 significantly during the sample 

 period, and data sets with less 

 than 2 degrees of freedom were 

 omitted. 



176 



