Reilly et al : Diets of pelagic juvenile Sebastes off central California 



507 



also collected at 100 m. Nets were fished for 15 min at 

 depth during the night, ~30 min after simset, or before 

 sunrise. 



Five specimens of each species were randomly sub- 

 sampled from each haul for dietary analysis. General- 

 ly, no samples were taken if fewer than five individuals 

 were taken in a haul. Specimens were tentatively iden- 

 tified to species and preserved whole in 10% buffered 

 formalin, usually within 1 hour of collection. Identifica- 

 tions were later verified ashore with meristics keys 

 (Matarese et al. 1989, Moreland and Reilly 1991); 

 samples were transferred to 70% isopropyl alcohol 

 within 1 month of collection. Standard length (SL) was 

 later measured to the nearest 0.1 mm. Stomachs were 

 removed and stored in 70% isopropyl alcohol until 

 examined. 



Stomach contents were examined with a dissecting 

 microscope. Empty stomachs were noted and the 

 digestive state of each prey item was coded on a scale 

 of 1-3, with 3 representing digestion too advanced for 

 identification. All prey types were identified to the 

 lowest possible taxonomic level and counted. When 

 possible, a subsample of all prey types was measured 

 along the longest axis with an ocular micrometer. 

 Heads or eyes were used to obtain total counts when 

 food items were fragmented. For each rockfish species, 

 the proportion of prey types in the diet was calculated 

 as the percentage of total prey numbers consumed in 

 a year, summed over all the individuals examined for 

 stomach contents. 



A ranking index, modified from 

 Hobson (1974), was calculated for the 

 major food items. The index (Ir) is the 

 product of proportional frequency of 

 occurrence and percent by number, 

 calculated for all specimens of a spe- 

 cies in a year. To quantify dietary 

 overlap among species, the index of 

 Colwell and Futuyma (1971) was used, 

 that is, 



■'ih 



1.0 - 0.5 



Pij - Phj 



j = i 



where pij and Phi are the numerical 

 proportions of prey j = 1 . . . N found in 

 the diets of species i and h, respective- 

 ly. The index has a minimum value of 

 zero, when no overlap occurs, and a 

 maximum value of one, when all prey 

 are shared in equal proportions by the 

 two species. 

 Multivariate analysis of variance 

 (MANOVA) was used to examine relationships among 

 latitude, bottom depth, and the diets of chilipepper, 

 shortbelly, and widow rockfish (Green 1978, SAS 1985). 

 Although only in 1987 were there sufficient data to 

 analyze the diets of all three species, adequate samples 

 of shortbelly rockfish were obtained during all years 

 (1984-87). Thus, examination of overall variation in 

 diet through time, vis-a-vis latitude and depth, was 

 limited to shortbelly rockfish. Analyses were confined 

 to the three prey types of highest frequency of occur- 

 rence during the year examined, which varied among 

 the different species and years. The numerical propor- 

 tions of the three prey types (the dependent variables) 

 were arcsine-transformed (Sokal and Rohlf 1981) prior 

 to MANOVA testing. Latitude, depth, and a latitude- 

 depth interaction term were the independent variables. 

 Station latitude was classified as either north or south 

 of lat. 37°20'N. Similarly, station depth was divided 

 into deep (>100m) or shallow (<100m) categories. 

 Data for chilipepper, shortbelly, and widow rockfish 

 sampled in 1987 were also divided into large (>1987 

 median SL) and small (<1987 median SL) size-classes 

 to examine diet variation as a function of fish size. 

 Shortbelly rockfish were sufficiently numerous during 

 all years to analyze diet variation as a function of 

 predator size. Prey types for this analysis were again 

 limited to the three prey categories with the highest 

 overall frequencies of occurrence in a year, and the 

 dependent variables were the arcsine-transformed 

 numerical proportions in the diet. 



