Lorenzo and Pa|uelo: Biology of Promethichthys prometheus 



93 



Figure 1 



Location of sampling areas (•) in the Canary Islands. 



The sex ratios (malesrfemales) for the whole 

 sample, for different size classes, for quarters of a 

 year, and for depth strata were calculated. The re- 

 productive season was determined on the basis of 

 monthly variations of the gonadosomatic index ( GSI ) 

 according to Anderson and Gutreuter (1983). The 

 length at sexual maturity (length at which fifty per- 

 cent of the specimens became mature I was estimated 

 by means of a logistic function that was fitted to the 

 proportion of the mature individuals (stages III, IV, and 

 V) by using a nonlinear regression (Saila et al., 1988). 



The ratio of total length to total weight was calcu- 

 lated over the whole period for males and females 

 separately, as well as for the population as a whole, 

 by applying a linear regression (Ricker, 1973). Age 

 was determined by interpreting growth rings on the 

 otoliths; whole otoliths were placed in a watch glass 

 with a blackened bottom and containing glycerin and 

 examined under a compound microscope with re- 

 flected light. Counts for each specimen were per- 

 formed at least twice and only coincident readings 

 were accepted. An index of average percent error 

 (APE) developed by Beamish and Fournier (1981) 

 was used to compare the precision of age determina- 

 tions. Ageing was validated indirectly by examina- 

 tion of monthly changes in appearance of the margins 

 of the otoliths ( Morales-Nin, 1987 ). The date 1 July was 

 considered as birthdate to assign the individual ages 

 to age groups. The von Bertalanffy growth curve was 

 fitted to data of the resulting age-length key by means 

 of the Marquardt's algorithm for nonlinear least 

 squares parameter estimation (Saila et al., 1988). 



Length-frequency data were converted to age fre- 

 quencies by using the estimated von Bertalanffy 

 growth parameters (Pauly, 1983, 1984). The rate of 



total mortality (Z) was calculated from the length 

 converted catch curve by using ELEFAN program 

 (Gayanilo et al., 1988). The rate of natural mortality 

 (M) was determined, from the equation of Pauly 

 ( 1980). Following estimation of Z and M, the rate of 

 fishing mortality \F) was calculated by substraction. 

 The length at first capture was estimated from the 

 selection ogive generated from the length converted 

 catch curve (Pauly, 1984). 



Results 



The size-frequency distribution showed a length 

 range of 36 to 80 cm TL in the catches, with a main 

 distribution between 56 and 66 cm (Fig. 2). 



Of the 776 fish examined, 282 (36.3%) were male, 

 491 (63.3%) female. The sex of the remaining 3 (0.4%) 

 individuals could not be identified macroscopically 

 because they were immature and had very thin, 

 translucent gonads. The overall ratio of males to fe- 

 males was 1:1.74 and x^ analysis revealed this to be 

 significantly different from a 1:1 ratio (Table 1). Fe- 

 males predominated in all size intervals. Sex ratios 

 for males and females grouped into 5-cm length 

 classes had significant departures from the 1:1 ratio 

 for all size intervals (Table 1). The ratio of males to 

 females varied throughout the period of study, but 

 there were no significant differences from the 1:1 ratio 

 during the spring and summer months (Table 2). 

 There was a relationship between the sex of roudi es- 

 colar and depth; males predominated at 600 to 800 m 

 depths, females at 300 to 500 m (Table 3). 



The GSI showed higher values for females than 

 for males (Fig. 3). The same temporal variation pat- 



