FISHERY BULLETIN: VOL. 76, NO. 2 



Hudson River from 6 December 1974 through 20 

 March 1975 were processed. 



Three estimates of stock contribution, i.e., "as- 

 classified," "adjusted," and "iterative" estimates, 

 were calculated for collections of legal-sized, 

 sublegal-sized, and overwintering striped bass by 

 geographical and temporal strata. Sublegal-sized 

 <406.5 mm or 16 in FL) and overwintering striped 

 bass collected in New York waters were not con- 

 sidered to be a part of the coastal fishery and were 

 analyzed separately. In each stratum, the 

 percentage of striped bass allocated to a stock pro- 

 vided an estimate of that stock's relative contribu- 

 tion. Mean 1975 estimates of stock contribution of 

 legal-sized striped bass were calculated by averag- 

 ing strata estimates within periods then averag- 

 ing across the six periods. Relative contribution 

 estimates by age were also obtained. 



The influence of the Hudson stock in coastal 

 strata adjacent to the Hudson River was investi- 

 gated by comparing the relative contribution of 

 Hudson, Chesapeake, and Roanoke stocks within 

 "inner" and "outer" zones designed by the U.S. 

 Nuclear Regulatory Commission.'*^ The inner 

 zone encompassed western Long Island Sound 

 (stratum 5), the New York Bight (stratum 7), and 

 northern New Jersey (stratum 8-1), whereas the 

 outer zone encompassed the remaining waters 

 from Cape May, N.J., to Maine (strata 1 to 4, 6, 8-2, 

 8-3). Estimates of relative contribution for inner 

 and outer zones were calculated for each period by 

 summing the number of Hudson-, Chesapeake-, 

 and Roanoke-classified fish within appropriate 

 strata. Mean estimates of contribution within 

 each zone were calculated for the year by averag- 

 ing across temporal strata. 



RESULTS AND DISCUSSION 



Establishment of Discriminant Functions 



Five characters were established as the charac- 

 ter set best able to discriminate among Hudson, 

 Chesapeake, and Roanoke stocks. They are, in 

 order of importance (as established by stepwise 

 linear discriminant analysis): 1) the ratio of snout 

 length/internostril width, 2) the scale ratio of first 

 to second annulus/focus to first annulus measure, 



'"U.S. Nuclear Regulatory Commission. 1975. Final environ- 

 mental statement related to operation of Indian Point Nuclear 

 Generating Plant Unit no. 3 Consolidated Edison Company of 

 New York, Inc. Office of Nuclear Reactor Regulation. Docket no. 

 50-286, Vol. 1:V-166-V-178. 



3) a character index (Raney and deSylva 1953), 4) 

 the upper-arm gill raker count (which includes 

 rudimentary rakers), and 5) the lateral line scale 

 count. The character index, i.e., the sum of left and 

 right pectoral, second dorsal, and anal fin rays, 

 was used since Grove et al. (1976) demonstrated 

 that individual fin ray characters did not add sig- 

 nificant discriminatory ability. 



The five characters satisfied the criterion for 

 independence with fish length in each stock with 

 only one exception. The snout length/internostril 

 width ratio for the Roanoke stock has a coefficient 

 of determination of nearly 0.20 but was retained 

 because its distribution had the least overlap 

 among spawning stocks of all characters, thus 

 making it a potentially good discriminator. 



Results of the test of homogeneity indicated that 

 only the Hudson stock was homogeneous among 

 ages and between males and females. Significant 

 differences (a = 0.05) were found among ages and 

 between sexes in the Chesapeake spawning stock 

 and among ages in the Roanoke spawning stock. 

 Differences found in the Chesapeake spawning 

 stock may have resulted from pooling collections 

 from its four major tributaries. 



Quadratic functions (Table 2) were used to dis- 

 criminate among stocks as a result of the investi- 

 gation of underlying assumptions of discriminant 

 analysis. Significant differences (a = 0.05) were 

 found among covariance matrices of Hudson, 

 Chesapeake, and Roanoke spawning stocks which 

 suggested that quadratic functions would better 

 discriminate among these stocks than linear func- 

 tions. Histograms suggested that no radical de- 

 parture of multivariate normality was evident, 

 although normality of individual characters does 

 not assure multivariate normality of the character 

 set. Therefore multivariate normality of the 

 character sets was assumed. 



Percentage of spawning-stock specimens cor- 

 rectly classified by the quadratic functions and 

 estimated stock percentages resulting from the 

 use of these functions closely agreed with results 

 obtained by the cross-validation procedure (Table 

 3). For the total set of collections, 76. 8^^ of Hudson 

 specimens, 67.T7( of Chesapeake specimens, and 

 85. 9^^ of Roanoke specimens were correctly clas- 

 sified, resulting in an overall correct classification 

 of 74.4%. This was similar to overall percentages 

 of 73.2 and 77.1 obtained for the cross-validation 

 subsets. Estimated relative percentages for each 

 stock varied <3 percentage points among the total 

 set and cross-validated subsets, whereas varia- 



340 



