480 



Fishery Bulletin 9 1 (3), 1993 



SQUA (10.1%) 



other (3.7' 

 RAJA (3 2%) 

 BOSS (3.6%) 



MICR (2-6%) 

 HANS (2.0%) 

 LARS (3.9%) 



ACER (11.0%) 



GUNN(23 9%) 



GIBB (21 9%) 



PSEU (14 0%) 



6 



Figure 2 



Pie charts illustrating the average composition of the demer- 

 sal fish community around South Georgia Island during the 

 AMLR surveys. (A) Composition based on the biomasses of 

 the fish collected; (B) composition based on the numerical 

 abundance of the fish collected. 



Table 2 



Diversity of South Georgia Island Demersal Fish Commu- 

 nity. R represents species richness. The # symbol indicates 

 that the diversity values were calculated on a numerical abun- 

 dance basis. The wt. symbol indicates that the diversity val- 

 ues were calculated on a biomass basis. H' is the 

 Shannon-Wiener information index. V is an index of species 

 evenness. 



The distribution of species abundances at most sta- 

 tions was described well by the logarithmic series model 

 (Fig. 4). However, 27% of the stations ( 11) in the 1988- 

 89 biomass data set fit the logarithmic series model 

 poorly, while those of earlier surveys had less than 

 half that proportion of poor fitting stations. The bio- 

 mass data sets generally had more poor fitting sta- 

 tions than did the numerical data sets. 



The normal cluster analyses clearly demonstrated 

 that there was little spatial structure in the demersal 

 fish community. Extensive chaining (Boesch, 1977; Jain 

 and Dubes, 1988), frequent reversals (crossovers) 

 (Kobayashi, 1987), and low number of significant clus- 

 ters were indicative of the absence of spatial structure 

 (Fig. 5, page 484). The 1986-87 and 1987-88 commu- 

 nities displayed slightly more significant structure than 

 that in 1988-89, but chaining and reversals were preva- 

 lent there as well. The nodal analyses demonstrated 

 that the species groups occurred with similar constancy 

 throughout the region and fidelity was not strong for 

 any particular group (Fig. 6, page 484). 



Species composition of significant clusters was basi- 

 cally the same as that for the region-wide species as- 

 semblage (Fig. 2), except for those clusters formed by 

 a small number of samples with rare species (Fig. 7, 

 page 485). Clusters containing samples from Shag 

 Rocks had Patagonothen brevicauda as an additional 

 component. Close examination of the clusterings re- 

 vealed that the presence or absence of the rarest spe- 

 cies was the basis for what little structure was present. 

 When these rare species (those which occurred at 5% 

 or fewer stations) were removed, all or nearly all sig- 

 nificant structure disappeared. The lack of structure 

 was consistent from year to year. 



Clustering of the length distribution of each species 

 revealed little structure there as well. Notothenia 

 gibberifrons, C. gunnari, andiV. larseni showed a weak 

 separation of large and small fish; large fish associ- 

 ated with deeper water and small fish with shallower 

 stations during 1987-88 (Figures 8, A and B; 9A). 

 Only N. larseni continued to show that trend in 1988- 

 89 (Fig. 9B). Notothenia larseni also displayed a tem- 

 poral separation in size, with larger fish caught later 

 in the cruise in both surveys (Fig. 9C). Muraenolepis 

 microps showed a similar weak trend of larger 

 fish later in the cruise during the 1988-89 survey 

 (Fig. 8C). 



The gradient analyses revealed several significant, 

 but weak trends. Skewer analysis detected significant 

 trends in longitude in all three years. Significant trends 

 in latitude and depth were identified in 1987-88 and 

 in 1988-89. In 1986-87 and 1988-89, significant tern- 



