Somerton and Kikkawa Population dynamics of Pseudopentaceros wheelen 



763 



finding, however, indicates either that mixing is minimal 

 or that the rate of mixing is relatively low compared with 

 the 2- to 4-week duration of a typical Japanese fishing trip. 



Although the bias in Bot was corrected by estimating 

 the proportion of the stock vulnerable to trawling (Pf = 

 0.27), adequacy of this correction rests on the assump- 

 tion that Pf does not vary with time. However, Pf may 

 vary with time because the depth distribution of armor- 

 head may vary. For example, the proportion of the popula- 

 tion occurring in the shallowest depth stratum (<250m) 

 averaged 15% over the 10 research cruises, but ranged 

 from to 40%. It is unclear if such variation in the 

 daytime distribution is reflected in the nighttime distribu- 

 tion, because armorhead do not feed at night and there- 

 fore cannot be sampled effectively with longlines (M.P. 

 Seki and D.A. Somerton, NMFS Honolulu Lab., unpubl. 

 data). Bias in the biomass estimates could additionally 

 occur if Pf depends on the degree of mixing of deep and 

 shallow fish, because Pf would likely be larger when 

 fishing periods were longer and less intense. Since fishing 

 periods tended to be longer during Period 3 than in Period 

 2, this would lead to an overestimate of biomass during 

 Period 3. 



Precision in the estimates of biomass, which is ex- 

 pressed as the coefficient of variation (CV) to compensate 

 for the large range in biomass, was smallest in Period 1 

 (Fig. 3), because longline CPUE estimates were more 

 precise than trawl CPUE estimates. Expressed different- 

 ly, based on the mean CV over the period 1970-84 (ex- 

 cluding 1971 when data were not sufficient to estimate 

 the variance of Ut ), the 95% confidence intervals for B*x 

 was ± 1.70 B*x, which in all years includes zero. Over the 

 period 1985-90, however, the 95% confidence interval 

 was ±0.29 B*x, and never included zero. 



Post-recruitment ageing 



The estimates of natural mortality rate and annual 

 recruitment required estimates of the age distribution. 

 Although the ages of armorhead can be determined using 

 either daily or annual growth increments on their otoliths 

 (Uchiyama and Sampaga 1990), they are easily obtainable 

 only for individuals in the pelagic phase of their life 

 history, because somatic growth ceases once armorhead 

 recruit to the seamount (Humphreys et al. 1989) and 

 growth increments become so closely spaced that they are 

 exceedingly difficult to count (R. Humphreys, NMFS 

 Honolulu Lab., pers. commun.). Thus, we expressed age 

 on a scale relative to the presumed time of recruitment. 

 Such post-recruitment ages were based on the decrease 

 in FI over time. 



Frequency histograms of FI display modes which can 

 be tracked sequentially from one histogram to the next 

 over time as they move from the right (high FI or fat) 

 to the left (low FI or lean). Two examples of this are the 



large mode that appeared in 1980 and could be 

 followed until 1984 (Fig. 4A) and the large mode that 

 appeared in 1986 and could be followed until 1990 

 (Fig. 4B). Since the first appearance of these modes 

 was always associated with an increase in CPUE 

 (Fig. 2B), they were interpreted to represent cohorts 

 of fish that had recruited to the seamount. 



To further substantiate our interpretation of the 

 modes, the rate of decrease in FI was examined for 

 consistency both over time and between the two 

 presumed year-classes. Plots of FI versus time ap- 

 peared to have slight curvature (Fig. 5), but for both 

 the 1980 and 1986 year-classes the curvature was 

 not significant (P>0.05). Changes in FI, therefore, 

 are proportional to changes in post-recruitment age. 

 The rates of decrease in FI of the 1986 (0.00169/mo) 

 and the 1980 (0.00157/mo) year-classes did not dif- 

 fer significantly (ANCOVA, P>0.05). In addition, 

 sexual equality in the rate of decrease in FI was 

 tested for the 1986 year-class alone, and the male 

 rate was not significantly different (ANCOVA, P> 

 0.05) from the female rate. Taken together, these 

 findings indicate that all armorhead decrease in FI 

 at approximately the same rate and that once estab- 

 lished by the recruitment of a strong year-class, the 

 coherency of an FI mode should be preserved over 

 time. 



Natural mortality 



Natural mortality was estimated from the change 

 in the relative abundance of two cohorts over time 

 during a period when no commercial fishing oc- 

 curred. The first of these cohorts, which consisted 



