480 



Fishery Bulletin 90(3), 1992 



Previous observations of H. ensifer from a submer- 

 sible found higher densities on flat, silty, sandy areas 

 than over low-relief, rocky outcroppings (Gooding et 

 al. 1988). Although statistical analysis of substrate 

 associations were not conducted for H. ensifer in this 

 study, we did observe a similar substrate association. 

 Heterocarpus ensifer were abundant at the Oahu dive 

 sites at depths (500-800 m), although this is deeper than 

 their reported optimum range of 300-600 m (Gooding 

 1984). The substrate on these two dives was flat, cor- 

 alline sand with few isolated, low-profile features (e.g., 

 sea anemones, small rocks) around which the shrimp 

 appeared to concentrate. No rocky outcroppings were 

 observed on these dives. Very few H. ensifer were 

 observed at the three Kona dive sites, where the bot- 

 tom was steep and composed largely of rocky rubble 

 with few sandy patches, though the dive depths again 

 were deeper than the optimum range for this species. 



The substrate associations of H. laevigatiis appeared 

 to differ from those of H. ensifer. Although the differ- 

 ences in substrate particle size were not significant, the 

 ANOVA test revealed significantly higher densities on 

 volcanic compared with coralline substrates (Table 3) 

 with data from all dive sites pooled. The significant 

 results for substrate type, however, must be viewed 

 with caution because of the significance of dive site to 

 H. laevigatus density and the unbalanced sample de- 

 sign. Not only were all substrate types not present on 

 a single dive, but those types present were not found 

 in equal proportions on any dive. Therefore, differences 

 in density attributed to substrate type may actually be 

 a reflection of differences related by some other factor 

 to dive site. In particular, the Kona dive sites were 

 largely volcanic, and the majority of the H. laevigatus 

 observed were from Kona sites 2 and 3. Although Kona 

 site 1 also was largely volcanic, the volcanic rocks 

 differed from those observed at sites 2 and 3, in that 

 the appearance was of a more recent rock slide (sharper 

 edges vs. weathered). This apparent instability may 

 be responsible for the low shrimp density observed at 

 site 1. Other aspects of the bottom, such as slope, 

 substrate complexity, stability, and current patterns 

 may be of considerable importance and should be in- 

 vestigated in future work on the substrate associations 

 of H. laevigatus. 



With visual censusing techniques, there is always a 

 concern regarding the reliability of abundance esti- 

 mates. Various factors, including sampling techniques, 

 species behavior, and physical conditions, can bias 

 results (Colton and Alevizon 1981, Sale and Douglas 

 1981, Brock 1982, Ralston et al. 1986, Matlock et al. 

 1991). Some authors believe that density estimates 

 based on direct visual surveys, though often much 

 higher, are more reliable than those estimated from 

 fishing gear catches (Uzmann et al. 1977, Powles and 



Barans 1980, Kulbicki and Wantiez 1990). Individuals 

 of the target species, H. laevigatus, were easily counted 

 because they were in the open and reacted almost with 

 indifference to the presence of the submersible, and 

 because the low, uncomplicated relief at the study sites 

 offered little opportunity for their concealment. Avoid- 

 ance of the submersible by the shrimp seems unlikely. 

 Observed densities were much greater than expected, 

 yet these would be underestimates if avoidance oc- 

 curred. We cannot discount the possibility of bias in 

 our density estimates caused by attraction of shrimp 

 to the baited container placed at the beginning of our 

 dives. However, we observed no increased density gra- 

 dient in the vicinity of the container, and density obser- 

 vations were taken well away from the container site 

 ( > 100 m), presumably outside the drawing range of the 

 bait, leading us to believe that bias due to this source 

 was small. 



Recalculation of exploitable biomass for the main 

 Hawaiian Islands using Ralston and Tagami (1992) data 

 and methods, but substituting the q value obtained in 

 this study, would lead to a 33-fold increase in the esti- 

 mate of exploitable biomass (~9000t instead of 271 1). 

 Just as Ralston and Tagami (1992) suggest that their 

 estimate may be too low, we suggest that 9000 1 may 

 be unreasonably high, considering the preliminary 

 nature of this estimate and the failure of the Hawaiian 

 fishery that was at least partly due to drops in catch 

 rates at annual yields of <200t (Tagami and Barrows 

 1988). The acceptance of either of these estimates 

 would drastically affect management decisions, and 

 careful evaluation of these two values must be made. 

 Contributions to the difference between the two esti- 

 mates may be from three sources: actual differences 

 in catchability for the two studies related to differences 

 in time and study locations, error in our estimate of 

 q, and error in the Ralston and Tagami (1992) estimate. 



The estimation of q can be influenced by a variety 

 of factors including currents, water turbidity and tem- 

 perature, type of bait, soak time of fishing gear, and 

 density of the target species (Morgan 1974, Richards 

 and Schnute 1986, Miller 1990). For the two studies 

 involved in this discussion, many of the potential 

 sources of error were standardized. Both studies used 

 the same traps, same bait, and same soak times. They 

 did not, however, conduct studies at the same location 

 or time, and the range of catch rates encountered dif- 

 fered for the two studies. In both studies it is assumed 

 that catchability is constant for all catch rates and den- 

 sities involved, but this may not be true, particularly 

 between studies. Unfortunately, we are unable to 

 evaluate the extent of the error involved from these 

 sources. 



The estimate of q presented in this study could also 

 be biased. Sources of potential bias include lack of 



