402 



Fishery Bulletin 98(2) 



wings, 8.9-cm middle, and a 32.0-mm codend liner), 

 with one of two anchovy seines in surface water 

 (250.0 X 34.0 m or 20.0 m, 25.0-mm stretch mesh), 

 or with a small salmon fry seine in shallow water 

 (50.0 X 8.0 m, 3.0-mm stretch mesh deployed from 

 a 6-m skiff equipped with a 70-horsepower engine). 

 Fish collections were directed by the acoustic survey 

 and we attempted to sample every observed school. 

 The catchability of the fish species sampled may vary 

 with time of day, season, and between nets. However, 

 we assumed that these collections reflected the actual 

 species and age cohort ratios present within the acous- 

 tically observed fish school. A total of 220, 122, and 60 

 fish collections were made in 15, 29, and 28 locations 

 in October 1995, March 1996, and July 1996, respec- 

 tively. Each net collection was characterized by spe- 

 cies, and 1000 individuals of the dominant fish species, 

 usually Pacific hemng, were randomly sampled. Fork 

 lengths (mm) for 550 fish were measured immediately 

 after capture, and 450 fish were frozen and later their 

 fork length and wet weight (g) were measured in the 

 laboratory. After subsampling the remaining fish were 

 released unharmed from the seine. 



A length-dependent scaling constant was used to 

 convert estimated target strength (TS -10 log^^w) 

 from units of reflected acoustic energy (dB) to units 

 of biomass density (kg/m*): 



TS - 10 logioU' = -6.0 logiijL - 24.2 dB m-/kg, 



where L = the mean fork length (cm) of the fish col- 

 lected in the area (Thome, 1977; 1983a 

 1983b; Thorne and Thomas, 1990). 



This equation differs from the more standard regres- 

 sion equation (Foote, 1987) because it derives the 

 target strength as a proportion of weight. Thome's 

 equation was developed for echo integration primar- 

 ily with Pacific herring surveys from Alaska and 

 Puget Sound (Thorne, 1983a). 



For walleye pollock, that have a physoclistic swim 

 bladder, the standard equation: 



TS = 20 log|„r - 66.0 dB 



was used (Foote and Traynor, 1988). 



The acoustic estimates of Pacific herring and wall- 

 eye pollock school densities were derived from these 

 target strength equations. Many physical and biolog- 

 ical variables (including morphologic features [such 

 as physostomic, physoclistic, fat content], the orienta- 

 tion of the fish, water temperature, and depth) affect 

 target strength (Thorne, 1983b; Foote, 1987; Rose and 

 Leggett, 1988; Thome and Thomas, 1990; MacLennan 



and Simmonds, 1992; Misund et al., 1995; Huse and 

 Ona, 1996; McClatchie et al, 1996; Ona and Mitson, 

 1996; Misund, 1997; Misund et al, 1998). 



Echo integration measurements were converted 

 into data cells with lengths of 120 m, 40 m, or 20 m 

 and width and depth of 1 m for the October 1995, 

 March 1996, and July 1996 surveys, respectively. 

 Cell length was determined by using the simultane- 

 ously recorded latitude and longitude from the GPS 

 navigational system. 



Species proportion and size modes per species were 

 determined from the fish collections. The species pro- 

 portions, based on the number of individuals per fish 

 species in the random subsample, were multiplied 

 by the echo integration densities (kg/m'^) and then 

 converted by using length-weight regressions into 

 the number of Pacific herring per size mode, or the 

 number of walleye pollock. Walleye pollock were not 

 divided into size modes because the standard devia- 

 tions of the mean fork lengths of individual collec- 

 tions indicated that aggregations were unimodal. 



A group of data cells was considered to be a fish 

 school if the sum of the absolute differences between 

 latitudes and longitudes of adjacent cells was >0.009°. 

 We concluded that cells containing the equivalent of 

 <0.5 fish/m-^ were probably zooplankton on the basis 

 of frequency distributions of the data, and these cells 

 were removed from the data set (MacLennan and 

 Simmonds, 1992; Gunderson, 1993). If fish located 

 near the bottom were difficult to distinguish acousti- 

 cally, data cells for the bottom 5 m were removed. 



Cohorts of Pacific herring and walleye pollock 

 large-scale spatial distributions were examined by 

 using circular statistics (Batschelet, 1981). The angle 

 (0"=true north) for each data cell was determined 

 from an origin in the center of Prince William Sound 

 ( 60.6000 = N, 146.9000°W). These angles represent 

 distributions along the survey transect line and are 

 influenced by inequalities in shoreline distance and 

 sampling bias. Angle frequency distributions were 

 compared with random distributions and with the 

 distributions of other herring size modes along the 

 same transect by using a chi-squared test at the 

 5'7( level of significance (Batschelet, 1981). Expected 

 values were grouped according to Cochran's rule, 

 which states that <20'/f of the expected frequencies 

 should have a value <5 (Sokal and Rohlf, 1981). 



Nursei"y areas were determined by examining the 

 relation between juvenile fish spatial distributions 

 and coastline stnacture. Bays were defined statisti- 

 cally from passages or open coast by calculating the 

 sum of the three nearest shore distances (X3NSD). To 

 determine this value, first the distance between the 

 center of each fish school and the nearest shore was 

 measured. The second distance was determined by 



