478 



Fishery Bulletin 103(3) 



a 

 — 



fa w 

 8a 



C o 



M o 



Q 5 



x  



— -a 



i -a 



ai 



J3 Bo 



■So: 



« 



- 

 fa 



Q 

 fa 



H S 



£ '3 



«? a; 



H ? 



n 

 fa 



E-i 3 



to 



CO 



d 



CM 



d 



i-l Tf 



CO ^H 



d d 



o 



CM 



in 



d 



to 

 in 



O CO 



O CO 



to c- 



d d 



- 

 d 



o 

 o 

 o 

 d 



CN 



d 



CN 



d 



to 



d 



o 

 o 

 o 

 d 



o 

 cm 

 m 



d 



co c- -* 

 CO CO o 

 O »-H, CN 



odd 



o CO r-l 



m co cn 

 cn co ^r 

 odd 



- 

 in 



CM i-l <H 



co m co 



in 



d 



co 

 d 



o 

 o 



CM 



d 



o 

 d 



CN .-I 



en m 



CM CO 



d d 



of 1992 and 1993. Tukey post hoc tests on the yellowfin 

 sole all-site and habitat indices showed that 1991 yel- 

 lowfin abundance was greater than that of 1994. All 

 three indices showed that Pacific halibut abundance 

 was greater in 1995 than in 1991 and 1993. Individual 

 indices also indicated that Pacific halibut abundance 

 was greater in 1995 than in 1992, 1994, and 1996. 



Discussion 



The Chiniak Bay multispecies survey was designed 

 to estimate the abundance of four species with equal 

 emphasis. Because the distribution of species varied 

 greatly throughout the bay, what might have been an 

 optimal stratification for individual species was com- 

 promised to develop a stratification scheme that was as 

 effective as possible for all target species. Because we do 

 not believe sampling was optimal for any one of the spe- 

 cies, a poststratification method of analysis was investi- 

 gated to increase the precision of abundance estimates 

 for each species individually and to account for possible 

 bias due to the uneven and nonrandom distribution of 

 sampling sites over space and time. 



The need for stratification and the concern about 

 the distribution of sampling sites arise because of the 

 varying distributions of species in the study region. 

 Knowledge of the spatial distributions of species is im- 

 portant when estimating abundance from trawl surveys. 

 A random distribution of individuals is often taken as 

 a starting point for defining spatial distributions in 

 ecology (Taylor et al.. 1978). It is also a primary as- 

 sumption for many survey sampling designs and analy- 

 sis measures. The assumption of randomly distributed 

 individuals often is not appropriate, however, because 

 the concentration of fish varies over time and space in 

 relation to environmental factors (Murawski and Finn, 

 1988; Gadomski and Caddell, 1991; Reichert and van 

 der Veer, 1991; Norcross et al., 1999). If habitat (Fiedler 

 and Reilly, 1994; Reilly and Fiedler, 1994) and related 

 spatial population density distributions (Buckland and 

 Anganuzzi, 1988) are not accounted for when calculat- 

 ing abundance estimates, precision can decrease and 

 results can be seriously biased. Inaccurate results can 

 have strong management repercussions. 



In situations such as that of the present study, where 

 the sample does not properly represent the population, 

 poststratification is appropriate (Scheaffer et al., 1996). 

 By comparing poststratified and unstratified estimates 

 of abundance, we found that in every species-year com- 

 bination for which the three estimates of abundance dif- 

 fered (Fig. 3), the poststratified estimates reduced the 

 effect of the disproportion of samples allocated between 

 habitat and nonhabitat areas and between high and low 

 fish-density areas. For instance, in 1992, a dispropor- 

 tionately large number of samples were taken in Pacific 

 halibut habitat (Table 2). We suspect, therefore, that 

 the unstratified estimate of abundance was an overes- 

 timate of true population abundance. The disproportion- 

 ately large number of samples taken in Pacific halibut 



