Lough and Potter Vertical distribution of Melanogrammus aeglefinus and Gadus morhua 



283 



then subsequently preserved in 70% ethyl alcohol and 

 water. Temperature and salinity profiles were obtained 

 from the MOCNESS, XBT drop, or CTD cast. 



Data analysis 



Larvae and pelagic juveniles were measured to the 

 nearest O.lmmSL. Demersal juveniles from the bot- 

 tom trawl were measured to the nearest mm. Length 

 of haddock and cod were corrected for shrinkage to 

 live-length using the method of Bolz & Lough (1983). 

 Catches from the lm 2 MOCNESS were initially stan- 

 dardized to n/100 m 3 of water, and 10 m 2 MOCNESS 

 catches to n/10,000 m 3 . The fish were grouped into size- 

 classes based on developmental characters (Auditore 

 et al. 1993): 2-5, 6-8, 9-13, 14-19, 20-29, 30-39 

 40-49, 50-59, and 60-69 mm. 



As an initial step, the percent of a size-class at each 

 depth was plotted with the temperature and salinity 

 profile for each tow (figures not shown). Tows were 

 classified as day, night, or (in a few cases) twilight, 

 i.e., lh before and after sunrise and sunset. For each 

 pelagic tow, the mean number of fish within a time- 

 block and size-class was calculated for each depth by 

 summing all hauls at that site. Day and night vertical 

 profiles are plotted from these arithmetic means. Also, 

 abundance of larvae was integrated over the water 

 column, in terms of n/10 m 2 for the 1 m 2 MOCNESS 

 catches and h/1000 m 2 for the 10 m 2 MOCNESS catches. 

 Abundance-weighted mean depth or center of density 

 distributions of size-classes were calculated using the 

 method of Miller et al. ( 1963). To examine trends, mean 

 depths of selected sites were plotted in time-series with 

 water column thermoclines. Since salinity changes 

 throughout the water column are generally small 

 (<lpsu), temperature usually has a dominant effect 

 on density and stratification of the surface water on 

 Georges Bank (Flagg 1987). 



To test for differences in population means between 

 day and night vertical distributions and for interac- 

 tions among depth, time of day, and size of fish, an 

 unbalanced ANOVA Type-Ill model was used on all 

 tow data (Dunn & Clark 1974). The linear model is: 



Y = b„ + b, (D) + b 2 (T) + b, (S) + b 4 (DxT) + 

 b 5 (DxS) + b 6 (TxS) + b 7 (DxTxS) + e 



(1) 



where Y is fish density (rc/m 3 ), D is depth effect (coded 

 depth level), T is time effect (coded night or day), S is 

 size effect (coded size-class), b is a constant, and e is 

 the normally-distributed residual error term. Trans- 

 formation of the fish density data by log in (x+0.1) re- 

 sulted in a normal distribution of the residuals. Al- 

 though the interpretation of vertical migration based 

 on ANOVA can be difficult, the depth x time interac- 

 tion is generally considered the most useful factor for 



detecting diel migrations (Sokal & Rohlf 1969). Sum- 

 mary tables are presented of the factor /^-values and 

 their level of significance. 



For each station, an average temperature and salin- 

 ity profile was made by summing all the time-series 

 data across the sampled depth levels and calculating 

 the mean, standard deviation, and 95% confidence lim- 

 its. At sites where stratification of the water column 

 occurred primarily from solar insolation, upper and 

 lower bounds of the thermocline region were deter- 

 mined by inspection for each tow profile. The thermo- 

 cline was defined as the region of greatest tempera- 

 ture change. In a well-stratified water column, the zone 

 of greatest temperature change was easily identified 

 between the relatively uniform water above and below 

 this zone. In a weakly-stratified water column, deter- 

 mination of the thermocline zone was more subjective. 

 The distribution of larvae within, above, and below 

 the thermocline was estimated for each size-class 

 within a tow at Sites I&II for May 1983. The mean 

 percentage distribution and standard error of each re- 

 gion were calculated on the angular trans formation of 

 the tow percentage data: angle = arcsin ^proportion. 

 Bottom-trawl catches of recently-settled juvenile had- 

 dock and cod were standardized to «/30min haul. 

 Length data were grouped by 1cm size-classes and then 

 grouped by day, night, and twilight catches. Standard 

 statistics (x, SD, CV) were calculated on the untrans- 

 formed data. Differences between population means 

 were tested by the Students' t-test on log,,, transformed 

 data (Sokal & Rohlf 1969). 



Results 



Vertical distribution of larvae 

 Well-mixed water In April 1981, recently-hatched 

 haddock larvae (2-5 mm) and fewer larger larvae 

 (6-8 mm) were distributed throughout the water 

 column (Fig. 2A). Results of the ANOVA for the 2- 

 5 mm haddock (Table 2) showed a significant effect of 

 depth (p<0.05) but not of time of day. 



Cod larvae (up to 8mm) also were distributed 

 through the water column (Fig. 2B). In several verti- 

 cal profiles the distributions are weakly bimodal. Dur- 

 ing the day, larger larvae (9-29 mm) were most abun- 

 dant at mid-depth, 30-40 m, whereas at night they 

 were more abundant above 30 m. Ratios in night-day 

 abundance increased from -2/10 m 2 or less for the 

 smaller larvae (<14 mm ) to 4-6/10 m 2 for the 14-19 mm 

 and 20-29 mm size-classes. ANOVA for cod larvae 

 showed a significant (p<0.05) interaction effect of time 

 X size, which was expected since abundance increases 

 at night as cod get larger (Table 2A). A significant 

 (p<0.05) depth x time interaction occurred only for 

 size-classes >9-13mm. 



