BARKLEY: SELECTIVITY OF TOWED-NET SAMPLERS 



% OF CATCH 



Figure 9. — One possible analysis of skipjack catch data, Figure 8, assuming that the 

 length-frequency curve is determined by population density. Left panel : Lower curve, 

 speed-frequency data, Figure 8. The straight line represents a hypothetical population 

 density curve. Upper line shows P^ obtained by correcting catches for population changes. 

 Right panel: Values of P^ from left panel, plotted on linear coordinates. 



be somewhat less than one-seventh of the volume 

 sampled by the trawl, because of the increased 

 resistance to flow of the finer mesh. Catches of 

 the six largest class intervals of skipjack all ex- 

 ceed the "expected" catch (theoretical curve for 

 Xq/R = 4) by factors of seven or more, as would 

 be expected if in fact these larger fish were cap- 

 tured by the main body of the trawl, rather than 

 by the cod end. 



The evidence suggests that curve B on Fig- 

 ure 10 represents a reasonable analysis of skip- 

 jack catches with the Cobb trawl. This implies 

 that avoidance and mesh losses are dominant 

 factors determining the size-frequency curves 

 (Figure 8). The largest skipjack, 31.5 to 43.5 

 mm in length, were captured by the main body 

 of the Cobb trawl, with Pc values between 0.05 



and 0.009. Their reaction distance, Xo, was ZR, 

 or some 12 m. Some losses through the coarse 

 mesh are apparent for animals smaller than 

 39.5 mm in length. Skipjack smaller than 31.5 

 mm in length were captured by the cod end of 

 the trawl, at Pc values ranging from about 0.3 

 to 0.03, corresponding to reaction distances of 

 AR, or 6 m. Some losses through the finer mesh 

 are apparent for animals smaller than 17.5 mm; 

 estimated mesh losses reach 96% for animals 

 7.5 mm in length. As was pointed out earlier, 

 the above estimates of reaction distances are de- 

 pendent on the assumed escape speeds, 10 body 

 lengths per second. If the animals move faster 

 than this, Xo values will diminish proportion- 

 ately, and vice versa. 



Estimates of Pc are not affected by escape 



815 



