FISHERY BULLETIN: VOL. 72. NO. 2 



menhaden population during the experiment. 

 Neither photographic nor commercial fishing data 

 indicated a major change in P on main days, which 

 lends credibility to this assumption. As defined, 

 D can have two possible outcomes: yes, menha- 

 den are present and no, menhaden are not present. 

 In the analysis, areas where menhaden were 

 detected were assigned a value of 1 and areas 

 where fish were not detected were assigned a 

 value of 0. Although/) is clearly a discontinuous 

 dependent variable, the statistical techniques 

 used in the analyses converted it into a continuous 

 variable ranging from about to 1. The general 

 interpretation applied to predicted values is that 

 as the values approached 1, the chance of finding 

 fish increased proportionately. 



Regression techniques were used exclusively to 

 define relationships between the abundance and/ 

 or distribution of menhaden and available 

 measurements of oceanographic parameters. 

 Because remotely sensed oceanographic data 

 were not available, environmental conditions 

 where fish were detected had to be interpolated 

 and, in some cases, extrapolated from nearby 

 sea-truth sampling stations. This procedure 

 probably introduced experimental error into the 

 analyses and, as such, may have obscured subtle 

 relationships. 



Results 



Photographically sensed menhaden distribution 

 and abundance {A^JP) and distribution (D) in- 

 formation were regressed against available 



oceanographic parameter measurements (Table 

 1). These analyses reflect only those data col- 

 lected on 7 August, 25 August, and 28 September 

 1972 (i.e., main days) from the Mississippi Sound 

 portion of the study area. Forel-Ule color data 

 were not collected on 7 August 1972; consequently, 

 color analysis was limited to 25 August and 28 

 September. Clouds and cloud shadow obscured 

 portions of the Sound on 25 August and 28 

 September; these areas were ignored in the 

 analysis. 



In general, the two approaches, i.e., relative 

 abundance and distribution dependent vari- 

 ables, gave similar results. The type of relation- 

 ship, either positive or negative, was the same 

 in every case. Their precision varied, how- 

 ever, which affected level of significance. Of the 

 two approaches, relationships derived using dis- 

 tribution as the dependent variable probably are 

 the most reliable. Recent work has shown that 

 there may have been a variable bias associated 

 with the photographic sensor system used to 

 obtain the fisheries data (Benigno and Kem- 

 merer, 1973). The bias appeared to relate to 

 school size and atmospheric conditions and ap- 

 parently affected the number of schools detected 

 more than where they were detected. 



Assignment of biological significance to these 

 correlations is difficult in that the parameters may 

 be serving as indices of unmeasured parameters. 

 In other words, there is a question of concomi- 

 tance. Nevertheless, there does appear to be 

 support for the distribution significant (^ 90% 

 confidence level) correlations presented in Table 

 1. Menhaden fishermen frequently are frustrated 



Table 1. — Correlations between menhaden relative abundance (Ax.yIP) and 

 distribution (D) estimates and selected oceanographic parameters (E). 



' 90% significance level 

 '■ 95% significance level 

 *• 99% significance level 



380 



