494 



Fishery Bulletin 98(3) 



About 739c of 463 mature females were reproductively 

 active (stages 4—6; including 28'? stages 5-6 imminent 

 spawners) and 27% were spent or resting (stage 7, Table 

 1). Of a total 130 spawning females, 33 had ovaries con- 

 taining oocytes that were hydrating or hydrated, indicat- 

 ing imminent spawning. The ovaries of 122 fish contained 

 POFs, suggesting ovulation within several days prior to 

 capture. The exact age of POFs was inestimable because 

 the likely temperature-dependent (Fitzhugh and Hettler, 

 1995) degradation rate of swordfish POFs is unknown and 

 because gonad samples were collected from fish that had 

 been hooked for differing lengths of time at varying water 

 temperatures. Twenty-five of the 130 fish had ovaries con- 

 taining both unovulated, hydrated oocytes and POFs from 

 a previous spawning (Fig. 3, A and B). Of a total 395 

 mature (stage >4) males, 65% were reproductively active 

 (stage 4) and 35% were spent or resting (stage 5, Table 1). 



Oocyte size and developmental stage 



Ovarian developmental stage was closely related to dia- 

 meter of the largest size class mode of oocytes present 

 in ovaries. The diameters of formalin-fixed whole oocytes 

 (OD, in pm) and oocyte cross sections (XS, in eyepiece units 

 [epul, where 1 epu=0.1544 pm) on histological slides were 

 best related by the linear regression 



0D= 11.9-I-6.320XS 



(r2=0.948, n = 182, P<0.001). 



With logistic regression, ovarian development as an indi- 

 cator of reproductive activity (where active=stages 4—6; 

 inactive=stages 1-3, 7) was predicted with 97.5% accuracy 

 for 828 fish collected during both spawning and nonspawn- 

 ing periods. The predictive relationship was 



ln(p/i 1-pl) = -6.318 + 0.0180OI>, 



where OD = whole oocyte diameter (in pm); and 

 p = the probability cf active stages 4-6. 



The threshold (p=0.5) for stages 4-6 is predicted at a whole 

 oocyte diameter of 35 1 pm. Oocyte size distributions are related 

 graphically to ovarian stages in Figure 4. Median oocyte dia- 

 meters for the respective stages are listed in Table 1. 



Body lengths at sexual maturity 



Estimates of L^q with all females were indistinguishable (<1 

 cm different: G-test; P=0.3) from those where females were 

 used that had been caught during the 5-mo. (March-July) 

 spawning periods when reproductively active females ( stages 

 4-6) were present. Lr^^ estimates for males were improved 

 if immature (most likely less-than-year-old) fish, many of 

 which were available only during nonspawning periods, 

 were included. We therefore felt it preferable to describe 

 Z/r,|,'s using all gonad sample fish collected year-round. 



Z,,o'swere 102.0± 2.5 (95% CDcm EFLfor males(r-'=0.98; 

 n=506) and 143.6 ± 2.8 cm EFL for females (r-^=0.97; «=816; 

 Fig. 5). Fits to the logistic model were, as follows: 



Males: % Mature = 100/(1 -i- ea4.40ii-o.i4i2£FLi) 

 Females: % Mature = 100/( 1 -i- e'i4 8569-o.i034£FL.) 



Females attained >95% sexual maturity at 173 cm EFL. 

 The corresponding value for males was 123 cm EFL. 

 The smallest reproductively active female whose ovaries 

 contained hydrated oocytes or postovultory follicles was 

 134 cm EFL. Sample sizes ranged from 2 to 45 fish of each 

 sex per 5-cm EFL class; only 6 out of 36 and 12 out of 31 

 classes contained <10 fish for females and males, respec- 

 tively. Median sample sizes were 22 (for females) and 17 

 (for males) for each 5-cm class (Fig. 5). 



Indices of spawning readiness 



Reproductive readiness (verified by histological staging of 

 ovaries ) of female swordfish in the Hawaii-based fishery could 

 be predicted by logistic regressions with body size (length, 

 weight) and ovary weight (or with length- or weight-based 

 gonad indices). Although the spawning activity of female 

 swordfish was predictable from body size and ovary weight 

 alone, the accuracy of prediction was improved if a measure 

 of ovary maturation (such as size of the largest oocytes pres- 

 ent) was included. Overall, about 96% of 95 females (range 

 8-295 kg, 76-247 cm EFL; Table 2A) were correctly classified 

 (with reference to histological stages 4—6) as either reproduc- 

 tively active ( 1 ) or not ( ) by the logistic regression 



ln(p/(l-p)) = \r\{RW) + In(GW) -t- \n{EV), 



where RW = round weight in kg; 



GW = gonad weight in g; and 



EV = oocyte volume in mm-^ = 4/3;rr'^; and 



r = diameter/2 (Table 2A). 



The fit of this model was similar to those where oocyte 

 volume plus any of several length- or weight-based gonad 

 indices were used. Among the latter, the index, G7(2') = 

 \\n{GW)f\n(RW)]. was best (90% accurate), and overall 

 classification accuracy increased to 97% if oocyte volume 

 was added to the model (Table 2B). The index, G/(2) = 

 [\n(GW)f\r\(EFL)] of Hinton et al. (1997), plus oocyte 

 volume, showed comparable accuracy (94%; Table 2C), and 

 retained 87% accuracy even if oocyte volume was excluded 

 (Table 2D). The latter length-based gonad index itself 

 (Table 2D) should have the widest applicability, despite 

 some loss in predictive power because an oocyte volume 

 term is lacking, because data on swordfish body lengths 

 and gonad weights are often available but body weight and 

 egg size data are not. We also provide (Table 3) several key 

 regressions of swordfish length-on-length and weight-on- 

 length, summarized from Uchiyama et al. ( 1999), to enable 

 conversions between the different index-based predictors 

 of reproductive readiness. 



Temporal and spatial spawning patterns 



Swordfish caught by the Hawaii-based longline fishery 

 during March 1994-June 1997 included both reproduc- 

 tively inactive and active adult fish, in addition to imma- 



