296 



Fishery Bulletin 97(2), 1999 



stations at roughly monthly intervals using either 

 paired 0.6 1-m bongo nets fitted with 333-|im mesh 

 nets, an 8 + 2 m rectangular mid-water trawl (RMT) 

 fitted with 1600-|i m and 333-|j m mesh nets, or a paired 

 1.4 m'^ rectangular frame net, fitted with 333-|am mesh 

 nets. Depth information for both the bongo and frame 

 net was estimated from cable angles and lengths 

 deployed. The RMT was fully equipped and provided 

 continuous, real time depth, temperature, salinity, 

 and volume filtered data. The station exhibiting the 

 highest concentration of cod larvae was resampled 

 with a BIONESS sampler equipped with ten 1-m^, 

 333-|im mesh nets, that was deployed to sample dis- 

 crete 5-m depth strata in the upper 25 m of the wa- 

 ter column and 10-m depth strata at deeper depths. 

 Three of the 29 cruises were designed to track a patch 

 of eggs and larvae over smaller spatial scales for up 

 to 20 days in order to track how traits changed over 

 time. On these cruises we deployed principally a 

 BIONESS sampler and bongo nets. However, because 

 we were attempting to sample continuously from the 

 same patch of water, stations were distributed irregu- 

 larly in space. 



All net samples were sorted and cod eggs were re- 

 moved on board ship. Late-stage eggs that appeared 

 healthy and undamaged by the collection process 

 were videotaped under a dissecting microscope at 

 6— 50x magnification. Individual eggs were incubated 

 separately on a 12-h light: 12-h dark cycle and at near- 

 ambient temperature. Light from blue incandescent 

 bulbs mimicked the light environment at depth. 

 Nursery temperatures were recorded daily. All eggs 

 from the same cruise were incubated at the same 

 temperature. However, because sea temperatures 

 varied across the sampling grid, and with depth, 

 there were unavoidable differences between incuba- 

 tion and ambient temperatures for individual eggs. 

 Vials were checked every 12 hours for hatching. When 

 a larva hatched, it was immediately videotaped and 

 stored in liquid nitrogen prior to otolith extraction 

 and analysis. 



In the laboratory, lapillar and sagittal otoliths were 

 removed from larvae and mounted in cyanoacrylic 

 cement. In most larvae we could remove and classify 

 successfully all four otoliths. Otoliths were examined 

 under bright field illumination under a compound 

 microscope at 60-1 OOOx magnification. Oil immer- 

 sion was required for the higher magnifications. 



Videotape recordings and otolith images were ana- 

 lyzed in the laboratory by using an image analysis sys- 

 tem (Optimas v3.11, Bioscan Corporation, Seattle WA). 

 We staged each egg according to Thompson and Riley's 

 (1981) system. Egg diameters were calculated from 

 three digitized points on the circumferences of eggs. 

 From these points the diameter was calculated as 



Diameter  



a be 



4^s{s - a)(s - b){s - c) 



where a. b, and c = the lengths of the chords con- 

 necting the three points; and 

 s - 2{a + b + c). 



Larvae that hatched from these eggs were measured 

 to standard length (SL). Only undamaged otoliths were 

 analyzed. Several measures were taken to describe 

 otolith size. At this early stage, otoliths were essen- 

 tially spherical in cross section. We measured the cross- 

 sectioned area of both lapillar and sagittal otoliths. We 

 also measured the radius of the otolith at hatching be- 

 cause this is the most common measurement used when 

 estimating larval hatching sizes from otolith data. 



Analysis of the data collected depended upon the 

 purpose to which the analysis was being put. For 

 univariate analyses to determine seasonal trends in 

 a trait, data were aggregated to provide a mean value 

 for each deployment before analysis. This approach 

 reflects the sampling design employed in the field, 

 and thus the deployment is the appropriate sampling 

 unit. However, for bivariate analyses to determine 

 the correlation among measures of otolith size and 

 between otolith size and fish size, the individual fish 

 is the appropriate sampling unit, and thus these 

 analyses were conducted at the individual level. 



Results 



We identified and incubated 650 cod eggs from April 

 1991 to May 1993. Of this total, 259 (39.9%) success- 

 fully hatched. Otoliths from a random sample of 73 

 of these larvae were used in our analyses. We ob- 

 tained reliable measurements from both lapilli on 

 56 larvae (76.7%) and from both sagittae for 59 

 (80.1%) larvae. The distribution of data, by month 

 and year, is given in Table 1. 



We examined the correlation structure in the data 

 for area of otolith and radius of otolith at hatching. 

 Estimates of otolith area were correlated among 

 otolith types, but there were no significant correla- 

 tions among otoliths from the same side of the body 

 (Fig. 2). Moreover, the correlation among lapilli was 

 greater than that for sagittae (Fig. 2). An identical 

 pattern was found with respect to radius of otolith 

 at hatching (Fig. 3). 



We regressed each measure of otolith size on lar- 

 val size at hatching, using only otoliths from the left 

 side of the body. The area of the lapillus (LA) and SL 

 at hatching were significantly and positively related 

 (Fig 4A). Overall, longer larvae have bigger lapilli. 

 The 95% CIs around the predicted mean were nar- 



