598 



Fishery Bulletin 104(4) 



Transect 



I II III IV 



Transect 



O 



230 250 270 



Day of year 



Day of year 



Figure 3 



Spatial and temporal distribution (A and B) of Secchi depth (m), (C) water 

 depth (m), and (D) surface water temperature (°C). Mean (±1 standard 

 deviation) Secchi depth and water depth are displayed as a function of sta- 

 tion transects (I-IV; Fig. lA), calculated from ten sampling dates between 

 11 August and 9 October 1998 (days 223-282). Means with the same upper 

 case letters are not significantly different (Ryan's Q multiple comparison 

 test). Secchi depth and surface water temperature are also displayed as 

 function of station transects and sampling date (days 223-224). 



regression; F=11.34, df=l,135, P<0.001), and remain- 

 ing parameters were not significant at P<0.05 and 

 therefore not included in the model. The estimated 

 coefficient for Secchi depth was negative (parameter 

 estimate = -0.436), indicating an inverse relationship 

 between bluefish density and water clarity. Before ana- 

 lyzing spatial and temporal variations in Secchi depths 

 with an ANCOVA model, we examined the interaction 

 effect between transect and sampling date. There was 

 no significant transect-date interaction effect (two-way 

 ANOVA; transect X date: F=2.47, df=3,124, P=0.065), 

 and the assumption of equal slopes was met in the 

 full data set (Underwood, 1981). Secchi depth differed 

 significantly as a function of transect position (AN- 

 COVA; F=63.34, df=3,127, P<0.0001), whereby distance 

 from the coastline was positively correlated with Secchi 

 depth (Fig. 3A). Furthermore, Secchi depth significantly 

 decreased throughout the sampling period (ANCOVA; 

 F=30.20, df=l,127, P<0.0001) (Fig. 3B). 



The mean body size of bluefish collected from Au- 

 gust to October was 13.7 mm SL (range = 3-48 mm 

 SL) (Table 1; Fig. 4). Similar to spatial and temporal 



density patterns, bluefish mean body size differed sig- 

 nificantly by transect and sampling date (multivariate 

 repeated-measures ANOVA; transect: F=6.16. df=3,12, 

 P<0.01; date: F=7.50, df=9,50, P<0.0001) (Fig. 4). The 

 transect-sampling date interaction effect was again 

 significant (multivariate repeated-measures ANOVA; 

 transect X date: F=3.50, df=24,50, P<0.0001), thereby 

 precluding contrasts across main effects. The interac- 

 tion effect was caused by significantly larger bluefish 

 at transects I and II on day 254 compared to transects 

 III, IV, and transect III, respectively (Ryan's Q mul- 

 tiple comparison test). On day 268, significantly larger 

 bluefish were collected at transect IV than at all other 

 transects (Ryan's Q multiple comparison test). 



Surface water temperature and water depth were the 

 most significant factors affecting the size distribution 

 of bluefish (stepwise multiple regression; temperature: 

 F=14.07, df=2,94, P<0.0005; depth: P=5.50, df=2,94, P< 

 0.05), such that temperature and depth accounted for 

 a partial r- of 0.129 and 0.048, respectively (cumula- 

 tive 7-2=0.177). Moreover, estimated coefficients for both 

 variables were negative (parameter estimates = -1.121 



