Mueter and Norcross: Spatial and temporal patterns in the demersal fish community off Alaska 



561 



from the measured distance trawled and estimates or 

 measurements of the width of the net opening. Taxa to 

 include in the analysis were selected according to the 

 following criteria: 1) all species consistently identified to 

 species level and occurring at least at 1% of the stations, 2) 

 taxa that were not consistently identified to species level 

 were combined by genus or, if necessary, by family, 3) if, 

 after grouping taxa, a species, genus, or family was pres- 

 ent at less than 1% of the stations within each year, it was 

 not included in the analysis. 



As an initial estimate of abundance trends, we com- 

 puted gulf-wide averages of CPUE for each species by 

 year. CPUEs from the Japanese trawls used in 1984 and 

 1987 were adjusted to the US standard trawl gear by us- 

 ing fishing power coefficients provided in Tables 28 and 

 31 in Munro and Hoff (1995). Following standard NMFS 

 methods (Martin, 1997), we averaged estimated CPUEs 

 by stratum, weighted the averages by stratum area, and 

 combined them to obtain area-wide averages. 



From a haul-by-taxon matrix of CPUE data, we com- 

 puted univariate and multivariate indices to examine 

 spatial and temporal patterns in community structure. 

 For each haul we computed species richness (number of 

 species), species diversity (Shannon- Wiener index), and 

 total CPUE of all groundfish taxa combined as univariate 

 measures of community structure. We considered three in- 

 dices of species diversity (Shannon-Wiener, Simpson's D, 

 Fisher's a) and chose the Shannon-Wiener index because 

 it showed little dependence on sample size in simulations, 

 was approximately normally distributed (which facilitates 

 statistical comparisons), and is widely used in the ecologi- 

 cal literature. In addition, we computed multivariate indi- 

 ces of species composition based on Bray-Curtis dissimi- 

 larities of root-root transformed CPUEs and nonmetric 

 multidimensional scaling (NMDS) as described in Mueter 

 and Norcross (1999). Indices of species composition were 

 computed separately for each year and were related to 



explanatory variables to describe the major gradients in 

 species composition by year To examine trends in species 

 composition over time, catches were averaged by stratum 

 instead of using individual hauls, reducing the 3911 hauls 

 to 240 within-year strata (5 years x 48 strata). An NMDS 

 ordination of the stratum-by-species matrix was then used 

 to test specifically for significant trends in species compo- 

 sition over time. We computed the linear combination of 

 ordination axes that maximized the correlation with year 

 and tested whether the correlation was significantly high- 

 er than would be expected by chance by using randomiza- 

 tion tests. This linear combination was used as an index 

 representing changes in species composition over time 

 (time index). Species that were most strongly associated 

 with the time index were identified based on scatterplots 

 and Spearman rank correlations between the index and 

 individual species abundances, and changes in these spe- 

 cies were examined in more detail. 



Depth, geographic location, bottom temperature (mea- 

 sured by sensors attached to the gear), Julian day, year, 

 gear type, and area swept were included in the analysis 

 as explanatory variables. Geographic location was repre- 

 sented in the analysis either as a categorical variable by 

 using the five statistical areas depicted in Figure 1 or by 

 using alongshore distances (AD). Alongshore distance was 

 computed by projecting each station onto a line that ap- 

 proximately followed the shelf break and measuring the 

 distance along this line from its origin in the southeast 

 part of the study area (km 0) to the westernmost point (km 

 2600). Temperature was not measured for all hauls (Table 

 1); thus all regressions were done both with and without 

 temperature included. Gear type was included in all mod- 

 els to account for potential differences among gear types. 

 Area swept was only included for examining trends in spe- 

 cies richness and diversity because the number of species 

 per haul was expected to increase with the area sampled. 

 All other indices were based on standardized CPUEs. 



