SOMERTON AND KOBAYASHI: FISH LARVAE CATCHES IN PLANKTON NETS 



intervals >4.25 and <6.75 mm. This result is 

 surprising because the two nets were identical 

 except for mesh size which, by itself, is unlikely 

 to influence avoidance. However, the catch of 

 nehu eggs in the test net was significantly less 

 than in the standard net (paired ^test, P < 

 0.05), indicating, to the extent egg catches can 

 be used as a measure of filtration volume, that 

 the test net was likely clogged by abundant fila- 

 mentous algae in Pearl Harbor during sampling. 

 Such clogging not only decreased the apparent 

 amount of water entering the test net but also 

 may have slowed the sinking rate, allowing 

 larvae to avoid the net more easily. 



The estimated retention probabilities for the 

 standard net (P,.s) during the day increased from 

 about 0.35 for 2.5 mm larvae to nearly 1.00 for 

 6.0 mm larvae (Fig. 1). At night, however, the 

 standard net appeared to have higher retention 

 probabilities in the smallest length intervals 

 (Pre', Fig. 1). When this difference was examined 

 statistically (two-sample f-test), P,.e and P,.s 

 were not significantly different at P < 0.05 in 

 any length interval, but they were significantly 

 different at P < 0.15 within the two smallest 

 length intervals. Although the differences be- 

 tween P,.,, and P,,, are relatively small, consider- 

 ing the same net and method of deployment were 

 used during both day and night sampling, it is 

 surprising that any differences were detected. 

 One possible explanation is that the density of 

 small larvae, rather than the retention probabil- 

 ity, was higher at night. This could have oc- 

 curred either because, by chance alone, the den- 

 sity of small larvae was higher in the patches 

 sampled at night or because the mean density 

 was higher as a result of eggs hatching between 

 the day and night sampling. However, the addi- 

 tion of new larvae is unlikely because, at the 

 time of year when our sampling occurred 

 (March), nehu eggs hatch during the morning 

 and new larvae would therefore have been 

 equally available to both our day and night sam- 

 pling (Clarke 1989). A second explanation is that 

 the greater retention of small larvae at night is 

 real and at least partially due to morphological 

 changes increasing the catchabihty of larvae be- 

 tween the day and night sampling periods. Evi- 

 dence for this is weak; however, Clarke (1989) 

 reported that during March nehu larvae display 

 considerable development of their eyes, mouth, 

 and pectoral fins between midday and early 

 evening of their thu'd day of life. Development of 

 such features might increase catchability rela- 

 tive to equal-sized, but undeveloped, larvae. 



Regardless of the reasons, over some length 

 ranges, P,.,, ^P,,. in the entry experiment and 

 Pes "^Pir in the i-etention experiment; both cases 

 are violations of the assumptions implicitly made 

 when entry and retention probabilities are esti- 

 mated as simple catch ratios (i.e., catch of stan- 

 dard net/catch of test net). The effect of ignoring 

 this can be judged from plots of entry and reten- 

 tion probabilities estimated from simple catch 

 ratios and estimates of Pes and P,,, using our 

 method (Fig. 2). Entry probabilities estimated 

 from simple catch ratios are similar to Pes for 

 larvae >4.0 mm but are increasingly less than 

 Pes at smaller lengths. This region of underesti- 

 mation corresponds approximately to the length 

 interval in which the retention of larvae differed 

 between day and night (Fig. 1). Retention prob- 

 abilities estimated from simple catch ratios are 

 similar to P,s at larval lengths <4 and >8 mm, 

 but are considerably larger than P,., at inter- 

 mediate lengths. Again, this region of overesti- 

 mation corresponds approximately to the length 

 region in which avoidance differed between the 

 standard and test nets (Fig. 1). Violation of the 

 assumptions therefore leads to bias in estimates 

 of entry and retention probabilities based on 

 simple catch ratios. 



The success of a net comparison, however, 

 also depends upon the validity of several other 

 assumptions. Foremost are the assumptions that 

 no avoidance of the test net occurred in the entry 

 experiment (Pee = 1.0) and no extrusion through 

 the test net occurred in the retention experiment 

 (P„. = 1.0). Violations of these assumptions lead 

 to positive bias in the estimates of P,.., and P,.,. 

 For the entry experiment, P^g was definitely 

 higher than Pes over a broad range of sizes be- 

 cause more larvae were caught at night and a 

 sizable fraction of the catch was larger than the 

 largest larva caught during the day (Fig. 3). But 

 no evidence indicates Pee remained equal to 1.0 

 for size intervals <14.0 mm (the size of the 

 largest larvae caught during the day), as is re- 

 quired to obtain unbiased estimates. For the re- 

 tention experiment, P„. was definitely higher 

 than P„. because more small larvae were cap- 

 tui-ed with the test net (Fig. 3). But, again, no 

 evidence indicates P,.,. remained equal to 1.0 for 

 size intervals >2.00 mm, the smallest size cate- 

 gory. 



Still another assumption is that, in each of the 

 two experiments, the standard and test nets 

 both sampled the same population of larvae. For 

 the retention experiment, the assumption is cer- 

 tainly valid because the two nets were deployed 



451 



