762 



Fishery Bulletin 90(4). 1992 



70 71 72 73 74 75 76 77 11 



YEAR 



z 

 o 



00 



< 



o 



QD 



78 80 82 84 86 88 90 

 YEAR 



Figure 2 



Pelagic armorhead Pseudopentaceros wheeleri 

 biomass on Southeast Hancock Seamount dur- 

 ing periods of (A) high abundance (1970-78) and 

 (B) low abundance (1979-90). Prior to 1985, 

 when routine stock-assessment surveys were ini- 

 tiated, the biomass estimates are annual means 

 and are shown at the beginning of the year. 

 Starting in 1985, the biomass estimates are for 

 each stock-assessment survey and are shown for 

 the appropriate month. 



tion on SE Hancock seamount was suffi- 

 ciently small and the fishing effort was suf- 

 ficiently large to allow use of the Leslie 

 method to estimate both the mean catch- 

 ability of the trawlers (cy) and the biomass 

 at the initiation of each fishing season 

 (Bq, i) during the period just prior to re- 

 placement of commercial trawling by 

 research longlining. The time-series was 

 merged by using the estimate of qj and Pf 

 to compute biomass from Japanese trawl 

 CPUE and by using the estimates of Bq t to 

 estimate cy and thereby compute biomass 

 from longline CPUE. Thus, the Leslie 

 method provided the means to merge the 

 two time-series and to express the resulting 

 time-series in terms of biomass. 



Because success of this procedure rests on 

 the successful application of the Leslie 



method, any biases in the biomass estimates are likely the result 

 of violations of the underlying assumptions. One of the most im- 

 portant assumptions of the Leslie method is that the change in 

 size of the population is solely due to removals by the fishery. In 

 practice, this requires that the population is closed to immigra- 

 tion and the fishery is sufficiently short and intense so that the 

 effect of natural mortality is negligible. In nearly all cases exam- 

 ined, these requirements were met; that is, fishing usually oc- 

 curred well after the spring peak in recruitment of pelagic juve- 

 niles (Table 1; Boehlert and Sasaki 1988) and the catch was usually 

 obtained in 1 or 2 months and represented a large fraction of the 

 estimated biomass (Table 1). One further indicator of the suc- 

 cessful application of the Leslie method is the significance of the 

 slope of the regression, or q. Although the estimates of q obtained 

 using the Leslie method were not always significantly greater than 

 zero, they were remarkably similar among years for each vessel 

 that had fished repeatedly (Table 1). Such similarity was used 

 as justification for using the non-significant estimates in later 

 calculations. 



Bias could also result from violation of another assumption of 

 the Leslie method, that the entire population is equally \ailnerable 

 to the sampling gear. Such bias was considered likely when ini- 

 tial biomass estimates from Period 2 seemed too small to be con- 

 sistent with the observed catches. This apparent inconsistency was 

 examined statistically by testing the equality of two estimators 

 of recruitment, one that included catch (Rj ; Eq. 4), and one that 

 did not (R2; Eq. 5). The bias was confirmed since in 4 of the 5 

 years examined, R] was significantly (P<0.05) greater than R2, 

 a condition that could occur only if the biomass estimates were 

 too small. 



The most likely explanation for the underestimation of armor- 

 head biomass is that the Bot estimates do not include the entire 

 population and instead include only the fishable population or the 

 part actually exposed to trawls. This result was surprising because 

 we believed that the population would be sufficiently mixed by 

 the nocturnal vertical migration so that all armorhead would be 

 equally vulnerable even though the trawls were topographically 

 restricted to only a part of the armorhead depth range. Our 



