692 



Fishei7 Bulletin 100(4) 



determined semirandomly according to the ratio of blocks 

 per stratum. Bobbins were not used, so that trawling was 

 limited to nonreef substrata. A 180-foot German trawl was 

 used with 25-mm mesh (bar) liner attached to the bag. 

 Trawl duration was limited to 30 minutes; results of shorter 

 trawls (caused by technical reasons or to hitting reef) were 

 standardized to that time. Winter and summer cruises 

 were undertaken on the west coast between July 1985 and 

 January 1991; and autumn and winter-spring cruises on 

 the south coast between September 1987 and April 1996. 



Pelagic biomass surveys (PBS) are hydroacoustic sur- 

 veys designed to estimate annual spawner biomass and re- 

 cruitment strength of small, pelagic clupeoids — E. japon- 

 iciis,S. sagax. and round herring, £/n/;»e;/s whiteheadi (see 

 Hampton, 1992, for details). Acoustic targets are identified 

 by directed trawling with an Engels 308 midwater trawl 

 fitted with a codend liner of 8-mm mesh. Small T. atun 

 (<30cm) shoal with the pelagic species while larger in- 

 dividuals feed on them. As a result snoek are regularly 

 caught by trawl on PBSs, albeit in small numbers (<30 

 per trawl). These surveys provide qualitative data on the 

 distribution of juvenile snoek. A total of 80 PBSs, including 

 more than 1500 midwater trawls, were completed be- 

 tween 1985 and 1997. Although the sui-vey area varied con- 

 siderably, with the extremes being Liideritz (26°39'S; 

 15°09'E) in southern Namibia and Cape St Lucia (28°31'S; 

 32°25'E) in northern KwaZulu-Natal, regions 0-5 of the 

 present study were consistently covered. 



Migration of T. atun was studied by using spatial and 

 temporal variation in magnitude and size composition of 

 catches. Owners of commercial line boats and trawlers 

 are required to submit daily catch returns to Marine and 

 Coastal Management (MCM). The monthly catches of snoek 

 reported by the trawl and line fisheries in each of the seven 

 regions (1986-97) were expressed as percentages of the 

 respective annual totals. Mean monthly handline catch per 

 unit of effort (kg^oatyday) and trawl CPUE (kg/hour) by 50- 

 m depth zones were calculated for each region. Catch rates 

 recorded during fishery-independent DBSs were mapped 

 separately for summer and winter cruises and presented as 

 mean numbers of snoek per 30-min trawl, per grid block. 



Fish sampled for biological purposes were measured to 

 the nearest 1 mm (fork length), weighed, to the nearest 

 gram (fish <500 g), or the nearest 10 g (fish >500 g), cut 

 open, and sex was determined. Gonads were removed, 

 assigned a macroscopic index of maturity (see Table 1), and 

 weighed to the nearest 0.1 g. Random samples (/?=.5-30) 

 of each macroscopic ovarian stage were fixed in Bouins 

 solution for 48 hours and then stored in 70% ethanol for 

 microscopic verification. They were routinely embedded 

 in paraffin wax, sectioned to 3-7 pm, and stained with 

 haemotoxylin and eosin. Stomach contents were analyzed 

 fresh, and prey items were identified to the lowest possible 

 taxon and weighed (wet) to the nearest 0.1 g. Bait was rec- 

 ognized easily and discarded. 



The size at 50% maturity (Lr,,,* for males and females 

 was estimated by PROBIT analysis (SPSS, 1992) of the 

 fractions of mature fish (gonad stage 3-i-) per 5-cm length 

 class (midpoint), sampled during the breeding season. 

 Upper and lower 95% confidence limits were calculated 



by the software package by using maximum-likelihood 

 estimation. 



Seasonal patterns of reproduction were established by 

 calculating gonadosomatic indices (GSIs) and the monthly 

 percent frequency of each maturity stage, for fish >Lc^q: 



GSI = gonad weight I 



fish weight - 



gonad weight + 



stomach content weight 



xlOO. 



The extent of the spawning area was determined by com- 

 puting the percent frequency of each maturity stage for 

 females i>Lc,Q) that were sampled in each region during 

 peak spawning (June-October). Snoek spawn on the trawl 

 grounds (see below); thus spawning grounds were further 

 delineated by mapping commercial trawl CPUE (stan- 

 dardized by using general linear modelling to account for 

 vessel size ) during June-October ( 1986-97 ) with a 20 x 20 

 mile grid system. Commercial trawl CPUE was also used 

 to establish the depth distribution of snoek on the trawl 

 grounds, and hence the depth of the spawning area. Sex 

 ratios were tested statistically for significant deviations 

 from equality with chi-square tests. 



Nursery areas were delineated by comparing the 

 length-frequency distributions of snoek caught in each 

 region 1 ) during pelagic and demersal biomass surveys, 

 2) by the line fishery, and iii) by Cape gannets, Morus 

 capensis. Because trawling activities during PBSs were 

 aimed at small pelagic clupeoids, it was not possible to 

 use CPUE data from these cruises to analyze spatial pat- 

 terns in juvenile snoek abundance. However, gannets from 

 Lambert's Bay and Malgas Island (Fig. 1) have feeding 

 distributions that, although large, are separated at Cape 

 Columbine and do not overlap to any large extent (Berruti, 

 1987). Previous studies (Crawford et al., 1992; Berruti et 

 al., 1993; Crawford, 1998) have demonstrated that gan- 

 net diet adequately reflects temporal and spatial patterns 

 in prey abundance. Relative abundance of early-juvenile 

 snoek north and south of Cape Columbine was therefore 

 estimated by comparing rates of snoek consumption by 

 the two colonies. Stomach contents of gannets at the two 

 colonies have been sampled on a monthly basis since 1978 

 (see Berruti et al., 1993, for methods). Lengths of snoek 

 prey were obtained from undigested specimens, and snoek 

 consumption by the two colonies was compared annually 

 (1978-97) using the mean mass per stomach containing 

 food and percentage frequency of occurrence (i.e. percent- 

 age of stomachs with food that contained snoek). 



Dietary importance of snoek prey was assessed by per- 

 centage frequency of occurrence (%F), which provides an 

 indication of how often a particular item is selected within 

 a population (Hynes, 1950), and by percentage by mass 

 (%M) (Windell and Bowen, 1978), a measure of the energy 

 contribution of that item (Macdonald and Green, 1983). An 

 index of relative importance ilRI) was calculated for each 

 prey category / as the product of %M, and %F,. To facilitate 

 comparisons of prey importance between analyses (e.g. sep- 

 arated spatially, temporally or according to predator size), 

 this was expressed as a percentage (Cortes, 1997): 



