580 



Fishery Bulletin 103(4) 



BTS 



■MM, 



EITS 



1 



04 06 



P-value 



Figure 4 



Histograms of P-values of linear regression between fish density and 

 temperature calculated for all clusters. Circled bars represent the clus- 

 ters with strongest relationship. 



each survey (Fig. 4) and the groups of clusters with 

 the strongest relationships between fish biomass and 

 temperature were chosen for further investigations. 

 These groups consisted of 21 clusters from BT surveys 

 with P-values between 0.000 and 0.066 and 15 clusters 

 from EIT surveys with P-values between 0.004 and 

 0.138. Using linear regression models (biomass against 

 temperature), we calculated the predicted percentage 

 of the total pollock biomass for each of these clusters 

 (Table 1) within the temperature range observed during 

 surveys (Fig. 5). 



To evaluate a spatial scale on which biomass redistri- 

 bution occurred for the EIT surveys, we calculated mean 

 distance between clusters of negative and positive slope 

 (Table 2). To obtain these values, we generated 100 ran- 

 dom points within each of the clusters and calculated 

 the mean distance between all possible pairs of points 

 from both clusters. We did not attempt to calculate this 

 distance for the BT surveys because of the much more 

 complicated nature of the BT cluster maps. 



Results 



Northward and inshore shifts in pollock distribution in 

 warmer years were found in the EBS for all length cat- 

 egories. The location and magnitude of these shifts and 

 distance between clusters differed with the survey type 

 and length categories. In the present study we address 

 changes in pollock distribution by length category within 

 each survey. 



Echo-integration trawl survey 



The biomass of pollock <20 cm in cluster F2 near Zem- 

 chung Canyon at latitude 59°N decreased (with increas- 

 ing temperature) from about 59% of the total biomass 

 of pollock in the coldest year to 0% in the warmest 

 year (Fig. 2A). This decrease was partially offset by 

 the increase in pollock biomass in area F3, northwest 

 of the Pribilof Islands. The relatively weak relationship 

 (P-value=0.324) between pollock biomass and temper- 

 ature in cluster Fl (north of F2) was caused by the 

 extremely high abundance of <20 cm pollock within 

 cluster F4 during 1997. Therefore the percentage of total 



