502 



Fishery Bulletin 90(3). 1992 



There is evidence to show that estimates of crusta- 

 cean population size obtained through survey removal 

 methods like the Leslie and DeLury methods (Ricker 

 1975) may severely misrepresent the actual size of the 

 population. In his study, Morrissy (1975) found that 

 DeLury estimates of population number of Cherax 

 tenuimanus were anywhere from 39 to 53% of those 

 based on a complete count of the population. Similar- 

 ly, DeLury estimates of the population density of 

 Panulirus cygnus were 25% of those estimated from 

 diver counts (Morgan 1974a). 



The most likely origin of bias in these situations is 

 that not all individuals in the exploitable portion of the 

 population are equally vulnerable (sensu Morrissy 1973) 

 to the gear. For example, when sampling with baited 

 drop nets, the catch of C. tenuimanus in intermolt, ex- 

 pressed as a known fraction of the actual population, 

 was much higher than the catch in a premolt condition; 

 individuals in molt stages immediately preceding and 

 following ecdysis were not caught at all (Morrissy 

 1975). 



It is possible that a similar bias was operating dur- 

 ing the depletion study at the Kaulakahi Channel study 

 site. The presence of shrimp in the exploitable portion 

 of the population, which were less susceptible to trap- 

 ping, would result in overestimation of catchability and 

 underestimation of biomass. Factors such as molt stage 

 (Morgan 1974b, Morrissy 1975), sex (Morrissy 1973), 

 and feeding history (Sainte-Marie 1987) are known to 

 cause variation in vulnerability to trapping. Although 

 social dominance, mediated through differences in size, 

 can affect catchability (Morrissy 1973, Chittleborough 

 1974), it is unlikely to have substantially biased our 

 results because the catch size-structure remained un- 

 changed during the experiment (Fig. 3). Also, Gooding 

 et al. (1988) did not see "any overt aggressive behav- 

 ior" among H. ensifer that were active around baits. 

 Lastly, seasonal and interannual alterations in catch- 

 ability due to temperature (Chittleborough 1970, 

 Morgan 1974b), salinity (Morgan 1974b), and food 

 availability (Chittleborough 1970) operate over longer 

 time-scales than the depletion experiment. 



Two other lines of evidence support the premise that 

 shrimp biomass was underestimated. Although results 

 presented in Figure 4 indicate that 12 days of trapping 

 dropped the catch rate to 48% of its starting value 

 (10.87 kg/trap-night), it had risen to 19.73 kg/trap-night 

 when resampled 47 days later (data from 9-11 July 

 1986 cruise; see Table 1). If catchability is estimated 

 from the decline in catch rate that occurred between 

 the beginning of the depletion study (22.84 kg/trap- 

 night) to the time the site was resampled 2 months 

 later, and we assume the decline was due only to trap 

 removals (1499.00kg), we obtain q = 2.4624 ha/trap- 

 night. This represents a 74% reduction in the estimate, 



which in turn would inflate biomass estimates by a fac- 

 tor of 3.85 (i.e., Bitot = 1050 MT). This relative bias is 

 similar to that reported by Morgan (1974a) for Panu- 

 lirus cygnus (see above). 



From submersible observations of H. laevigatus den- 

 sity, Moffitt and Parrish (1992) obtained q = 0.2895 

 ha/trap-night for the same traps we used, amounting 

 to a 33-fold difference relative to our Leslie analysis. 

 They, too, expressed concerns about bias in catchability 

 estimates derived from depletion experiments due to 

 variable susceptibility to the gear. Conversely, their 

 estimate of catchability was based on comparing site- 

 specific March 1988 trap catches with submersible 

 observations made during August, even though H. lae- 

 vigatus undergoes seasonal vertical migrations (King 

 1983, Dailey and Ralston 1986). In addition, at the start 

 of each dive, they deployed a baited trap in the area 

 of the submersible. Both factors could lead to under- 

 estimation of catchability. 



It is clear that biased estimates of q will result if the 

 probability of capture is not uniform among shrimp. 

 In an attempt to solve this problem, Quinn (1987) 

 developed a depletion model that explicitly incorpor- 

 ated a term for non-constant catchability. Application 

 of his model to Pacific halibut effectively accounted for 

 short-term trends in q, but auxiliary estimates of fish- 

 ing and natural mortality were required, data that are 

 unavailable here. 



The primary objective of this study was to determine 

 the exploitable biomass of H. laevigatus in the main 

 Hawaiian Is. (MHI). Even if shrimp biomass were as 

 high 1050 MT, rather than 271 MT (see above), our 

 results indicate that the MHI stock is much smaller 

 than previously believed, and that prior estimates of 

 maximum sustainable yield (MSY) are much too high. 

 For example, Struhsaker and Aasted (1974), using 

 figures from a fishery for H. reedi off the coast of ChUe, 

 speculated that H. ensifer in Hawaii could sustain a 

 level of production equal to 10-20 kg/ha ■ yr-^. If 

 H. laevigatus were assumed to be equally productive 

 (e.g.. Anon. 1979), then, given there are ~350,000ha 

 of prime habitat at 458-640 m (250-350 fm) in the MHI 

 alone (Table 2), the resulting estimate of MSY exceeds 

 stock biomass many times over. Moreover, catch rates 

 of//, laevigatus in the distant Northwestern Hawaiian 

 Is. (Nihoa to Kure), which represent a similar amount 

 of shrimp habitat as the MHI, are no more than half 

 those observed in the MHI (Gooding 1984, Tagami and 

 Barrows 1988, Tagami and Ralston 1988). 



Our estimates of 0, = 1.01 and ©(^ = 0.74 are much 

 lower than those given in Dailey and Ralston (1986), 

 who reported 0g =2.9 and Q^, = 4.3. By requiring CL^ 

 to be greater than the modal size, they effectively con- 

 strained to values much greater than unity. A similar 

 requirement was imposed by Moffitt and Polovina 



