758 



Fishery Bulletin 90|4). 1992 



Types of data and preliminary analysis 



Since the types of data available for describing the 

 population dynamics of armorhead have changed with 

 time, it is convenient to separate the entire 1970-90 

 interval into three periods: (1) 1985-90, when NMFS 

 stock surveys were conducted but no commercial fish- 

 ing occurred; (2) 1978-84, when regulated Japanese 

 fishing occurred vdth U.S. observers aboard the 

 vessels; and (3) 1970-77, when unregulated Japanese 

 and Soviet fishing occurred. 



Period 1 In the period 1985-90, NMFS conducted 10 

 armorhead stock-survey cruises to SE Hancock Sea- 

 mount. Although bottom trawls were occasionally used, 

 the primary sampling gear was a bottom longline. 

 Unlike trawls, longlines could be used on the steep 

 flanks of the seamount and allow sampling of the en- 

 tire population. Longlines consisted of 30 rigid poles 

 (droppers), each with 5 equally-spaced hooks on short 

 leaders, attached at 18 m intervals along a 600 m 

 groundline (Shiota 1987). On all cruises, longlines were 

 set perpendicular to the depth contours to maximize 

 the depth range sampled and were fished with the same 

 bait (squid), hook size (no. 20 circle), soak time (1 hr), 

 and fishing period (0800-1830 hr) to maintain constant 

 catchability. Starting in 1986, however, catchability 

 changed slightly when hook timers (small timing 

 devices that are activated when a fish strikes the hook; 

 Somerton et al. 1989) were installed on the leaders. To 

 estimate the effect of timers on armorhead catchabil- 

 ity, a comparison experiment was conducted in 1990 

 in which droppers were alternated with and without 

 timers along the longline. A correction coefficient ac- 

 counting for the effect of timers on catchability was 

 then estimated as the ratio of the armorhead catches 

 for droppers with timers to those without timers (this 

 ratio was 0.77). 



Since preliminary information indicated that armor- 

 head density varied with depth on the seamount, stock 

 surveys were based on a depth-stratified sampling 

 design. Fishing depths were estimated by recording a 

 depth profile of the bottom as the longline was set, then 

 partitioning the measured distance between the term- 

 inal anchors into 30 equal intervals (the number of 

 droppers). To help correct for possible differences 

 between fathometer depths and actual fishing depths 

 due to horizontal drift while the longline sank, max- 

 imum depth recorders were placed on both anchors and 

 at the midpoint of all longline sets. Recorded maximum 

 depths were used instead of fathometer depths to 

 determine where the anchors and midpoint lay along 

 each depth profile. 



When longlines were retrieved, the species identity 

 of each captured fish was recorded along with the 

 number of the hook on which it was caught. All fish 



from each 5-dropper segment of the longline were then 

 placed together into a basket for later collection of the 

 following biological attributes: sex, fork length (FL, 

 mm), and body depth (BD, mm) which is the shortest 

 distance between the bases of the first anal spine and 

 the dorsal fin. In 1985, body weight (W, g) was also 

 measured on some specimens in addition to body depth. 

 Equations predicting BD from W and FL, and predict- 

 ing W from BD and FL, were calculated from these 

 data by using multiple regression. These equations are: 



Females (n 436) 

 BD 86.69 - 0.19FL + O.IOW (R- 0.91) 



W -936.25 + 9.06BD + 2.49FL (i?^ 0.90) 



Males (n 476) 

 BD 75.55 - 0.18FL + O.IOW (R'~ 0.85) 



W -934.02 + 7.52BD + 2.82FL (R- 0.87) 



Although the depth distribution of armorhead on the 

 longline could be determined unambiguously with the 

 sampling procedure used, this was not true for the 

 depth distribution of any of the measured or derived 

 biological attributes, because the catch from each 

 5-dropper segment was aggregated before the attri- 

 butes were measured. As a means of approximating 

 such depth distributions, the biological attributes of in- 

 dividual fish within each segment group were randomly 

 assigned to the capture depths within the segment. 



The relative abundance of armorhead during each 

 stock-assessment cruise was expressed as the mean 

 catch in numbers per hook (U) estimated as a weighted 

 average over four depth strata (<265, 265-300, 301- 

 400, 401-500 m). Algebraically, (U) is 



U = 



ZUiAi 



i = l 

 4 



i=l 



(1) 



where Ui is the catch per hook, and Aj is the bottom 

 area in depth stratum i. Values of Uj were corrected 

 for the influence of hook timers, and values of Aj were 

 estimated as planar areas between the strata depth 

 boundaries measured on a bathymetric map of the SE 

 Hancock Seamount (Fig. 1). 



Since armorhead begin to lose weight after arriving 

 at the seamounts, we examined an index of relative 

 fatness (FI), defined as body depth divided by fork 

 length, as an index of post-recruitment age. Frequency 

 distributions of FI were calculated as weighted aver- 

 ages, where the weighting factors were proportional 

 to the estimated abundance of armorhead in each depth 

 stratum. Algebraically, this is expressed as 



