Moffitt and Parrish- Assessment of exploitable biomass of Heterocarpus laevigatas in Hawaiian Is . Part 2 



481 



representative catch rates for study sites, error in den- 

 sity estimation, incompatibility of CPUE and density 

 estimates collected at different points in time, and the 

 few data pairs involved. Not only were shrimp catches 

 for each dive site based on a single trap-night of ef- 

 fort, but also traps generally were set at depths greater 

 than the observed range of maximum abundance deter- 

 mined from our submersible observations (~750m vs. 

 550-675 m). This results in no estimate of error for 

 CPUE estimates at each site and no way to determine 

 whether the traps were set within the range of max- 

 imum shrimp abundance at the time of trapping. If the 

 traps were not set within this zone, yields at our sites 

 may underrepresent relative shrimp abundance, 

 leading to a lower-than-actual estimate of q. Another 

 potential source of error in our q estimate is the ac- 

 curacy of our density estimate. Confidence limits on 

 our density estimates are quite broad, allowing for a 

 fair degree of error. At many dive sites this problem 

 is related to the few observations of density taken 

 within the zone of maximum abundance. Additional 

 problems in density estimation associated with the 

 presence of bait in the water during the dives have 

 already been addressed. The question of compatibility 

 between data pairs of density estimates and CPUE 

 values obtained at different points in time stems from 

 possible changes in density or in q over time. Although 

 trap catch rates in the Mariana Archipelago did not 

 vary significantly on a seasonal basis (Polovina et al. 

 1985), suggesting that q does not vary seasonally, 

 H. laevigatus may undergo temporal changes in depth 

 range on either a diurnal or seasonal basis (King 1984, 

 Dailey and Ralston 1986). If such movements do occur 

 (the evidence is not strong) and depth range expands 

 or contracts during these changes, densities observed 

 during midday periods in February and August may 

 differ from those occurring during trapping in March. 

 Finally, although the fit is quite good, our estimate of 

 q is based on only five data pairs covering limited values 

 of CPUE and density. 



The potential error in the Ralston and Tagami (1992) 

 q estimate depends on the appropriateness of using 

 their habitat area estimate in normalizing q and the 

 validity of the assumption of constant catchability for 

 all members of the population. These error sources are 

 not necessarily greater than those discussed above, but 

 are much easier to quantify. Even when an accurate 

 estimate of biomass is obtained for a study site, calcula- 

 tion of a normalized q is dependent on the estimated 

 habitat area of the study site. Estimated habitat area 

 is apt to be larger when depth range is estimated from 

 trap catches, as opposed to visual surveys, because of 

 the ability of the traps to draw shrimp outside of their 

 normal depth range. Recalculating the habitat area of 

 the Ralston and Tagami study site using the observed 



depth range (550-675 m) instead of the reported range 

 (420-640 m) results in a reduction in area to 63% of the 

 original value (748 ha instead of 1187 ha). Normalizing 

 q with this reduced study-site area estimate gives an 

 adjusted q value of 5.999 ha/trap-night (CI 2.6709- 

 9.3271 ha/trap-night. The ratio of this adjusted value 

 to the q obtained in the present study is 20.7 instead 

 of the original 32.7. Ralston and Tagami (1992) 

 discuss the effect on their q value of a large portion 

 of the population not being susceptible to trap capture 

 for the duration of the study period. They supply 

 evidence that their original estimate of catchability may 

 have been four times too high, resulting in a four-fold 

 underestimation of exploitable biomass. Other authors 

 have reported similar overestimates of catchability 

 resulting from depletion studies (Morgan 1974, Mor- 

 rissy 1975, Miller 1990). A further four-fold reduction 

 of the Ralston and Tagami q value results in a ratio 

 of 5.9 relative to our q value and only 1.8 for the ex- 

 tremes of the 95% CL (the minimum value for the 

 Ralston and Tagami confidence interval compared with 

 our maximum value). Coupling these quantifiable fac- 

 tors with the non-quantifiable factors discussed above 

 could bring the two estimates of catchability into 

 agreement. 



Because of the importance of accurate estimates of 

 exploitable biomass to the management process, it 

 would be desirable to conduct a combined technique 

 survey of the H. laevigatus resource. This should in- 

 clude direct visual density estimation with a trapping 

 study conducted at the same time and place to obtain 

 a reliable estimate of catchability and thereby exploit- 

 able biomass. Until that time, the expanded exploitable 

 biomass estimate (1050 1) for the main Hawaiian Islands 

 as presented by Ralston and Tagami (1992) should be 

 accepted for management purposes as a reasonable, 

 conservative approximation. 



Acknowledgments 



We would like to thank the staff of the Hawaii Under- 

 sea Research Laboratory for their support during the 

 field portion of this study. We have also appreciated 

 the comments and suggestions made by various 

 reviewers, including M.G. King, S. Ralston, W.B. 

 Saunders, M.P. Seki, and D.T. Tagami, which have 

 helped to form this paper. 



Citations 



Brock, R.E. 



1982 A critique of the visual census method for assessing cor- 

 al reef fish populations. Bull. Mar. Sci. 32:269-276. 



