Somerton and Kikkawa: Population dynamics of Pseudopentaceros wheelen 



765 



the most representative value. For both cohorts, log-relative 

 abundance was clearly a linear function of time, and M is 

 therefore invariant with age (Fig. 6A). 



When M was estimated separately for each sex, considering 

 the 1986 cohort alone, the value for males (0.037/mo) was 

 significantly different (ANCOVA, P<0.001) from the value for 

 females (0.045/mo). Furthermore, the intercept of the regres- 

 sion, i.e., the log-relative abundance at the time of recruitment) 

 appeared to be smaller for males (4.27) than for females (4.94), 

 but the difference could not be tested because of the strong 

 difference in slopes (Fig. 6B). Taken together, these results in- 

 dicate that, at least for the 1986 cohort, females recruited to 

 the seamounts in greater abundance than males but subsequent- 

 ly died at a greater rate. It is also possible that the higher mor- 

 tality rate of females was primarily restricted to the first year 

 of residence on the seamount (Fig. 6B). 



Since a sexual difference in mortality seemed inexplicable to 

 us, we examined the possibility that longlines preferentially 

 selected fat females. This was done by examining whether the 

 ratio of females to males, expressed as proportion female, 

 changed with FI similarly for trawls as for longlines. Since the 

 FI values of females and males decrease identically with time, 

 any change in female proportion with FI would indicate either 

 selective sampling or differential mortality, depending on 

 whether one or both gear types showed the change. To test 

 for such changes, female proportion was regressed on FI for 

 various samples. When these regressions were performed on 

 the longline samples, all 10 had a significant (P<0.05) positive 

 slope. When the regressions were performed on the research 

 trawl samples, four of five had a significant positive slope. Thus 

 a sampling bias is unlikely, unless both gears produced a 

 similar bias. 



The estimate of natural mortality rate (0.54/yr) is more than 

 twice that reported in Borets (0.25/yr; 1975). His estimate, 

 however, was based on age data that were likely biased for two 

 reasons. First, the age range [i.e., 7 yr, ages 5-12], reported 

 in Borets (1975) appears to be excessive when compared with 

 the range estimated from modal progression through the 

 research FI histograms (4-5 yr). Second, the mean age of the 

 catch between 1968 and 1974 reported in Borets (1975) did not 

 decrease as would be expected in a developing fishery. On the 

 other hand, our estimate of natural mortality rate was less than 

 the rate implied in the studies of Uchida and Tagami (1984), 

 Humphreys et. al. (1989), and Uchiyama and Sampaga (1990), 

 which all suggested that armorhead were semelparous and, like 

 Pacific salmon Oncorhynchus spp., died soon after spawning. 



Fishing mortality 



Fishing mortality rate (F) can be estimated only between 1978 

 and 1983 when the total catch and effort on SE Hancock Sea- 

 mount are known with reasonable certainty, based on the U.S. 

 observer program. Over this period, F, which was calculated 

 as the estimated value of Qj x total annual effort, averaged 

 1.03/yr or roughly twice the natural mortality rate. The mean 



exploitation rate (which was approximated as 

 (F/Z) (1-e-Z) where Z = F-i-M), was ~0.50. 

 Therefore, provided that no recruitment oc- 

 curred and that fishing effort was continuous 

 throughout the year, then an average of 

 roughly 50% of the population at SE Hancock 

 Seamount was removed annually by the fishery 

 over this period. 



Recruitment 



Annual recruitment to SE Hancock Seamount 

 was extremely intermittent between 1978 and 

 1990. When expressed in metric tons, the 1980 

 recruitment clearly dominated the entire 

 record (Fig. 7A). However, the total biomass 

 changed considerably over this period, and, 

 when expressed as a percentage of the total 

 biomass, the 1986 recruitment and— to a lesser 

 extent— the 1988 and 1990 recruitments were 

 also relatively important (Fig. 7B). For some 



