Sarda et al.: Spatio-temporal structure of a population of Ansteus antennatus 



601 



individuals were considered to represent the early 

 juvenile stages (L) and were analyzed separately to 

 obtain information on the earliest stages of recruitment. 

 Spatio-temporal changes in population structure 

 were analyzed by using factorial correspondence 

 analysis (FCA) (Benzecri, 1980; Greenacre, 1984). 

 Data were logarithmically transformed and a data 

 matrix was constructed by using the three habitats 

 sampled and the following variables: number of males 

 (M), number of females (F), number of juveniles (J), 

 number of smaller juveniles (L), and total number. 

 Because the LS below 1,000 m is a stable habitat 

 that, unlike the US and MS, is less affected by sea- 



sonal factors (Hopkins, 1984; Tyler, 1988; Gage and 

 Tyler, 1991) and is not exploited by the fishery 

 (Demestre, 1986; Sarda and Martin, 1986), the analy- 

 sis was repeated excluding samples taken on the LS 

 (Table 2). 



Multifactorial nonparametric analysis of variance 

 (Zar, 1984) was used to calculate differences between 

 variable and habitat combinations in each season 

 (spring, summer, fall, and winter). The Mood non- 

 parametric median test (Conover, 1980) was also used 

 to determine differences in total number of individu- 

 als by habitat. 



Results 



The first two inertial axes of the correspondence 

 analysis explained 63.14% and 26.84% of the vari- 

 ance in abundance of shrimp, respectively (Fig. 2). 

 Along the first axis, LS samples were associated with 

 a higher proportion of males and early juvenile stages 

 (L), which did not exhibit any marked seasonal pat- 

 tern in the deepest region. However, the samples 

 collected in summer on both the US and MS were 

 mainly females and were located in the region oppo- 

 site the LS samples (Fig. 2). In summary, the first 

 axis was mainly related to depth and juveniles pre- 

 dominated in the US and MS samples collected in 

 spring, autumn, and winter. 



The results of the analysis, excluding samples 

 taken on the LS (Fig. 3), yielded a cluster of data 

 points for spring samples associated with the pres- 

 ence of juveniles (recruits); in contrast, data points 

 for summer samples were again associated with adult 

 females. Autumn and winter samples were less 

 clearly discriminated but shifted to the right from 

 the origin and were associated with males. This may 

 be interpreted as a seasonal effect that alters the 

 population structure and is discriminated by the first 

 axis, which explained 75.22% of the total variance 

 (Fig. 3A). The distribution of juveniles appeared to 

 be linked to the topography of the submarine can- 

 yons (black points in Fig. 3B), mainly during winter 

 and spring, while the proportion of males rose on the 

 middle slope out to the canyon in autumn and win- 

 ter (open symbols. Fig. 3B). Juveniles and early ju- 

 venile stages (L) were caught in the deepest region 

 sampled (Table 2 and Fig. 2). In summer females were 

 distributed extensively on both the MS and the US. 

 In this second analysis the first axis may be related 

 to seasonality. The second axis (20.27%) probably 

 represents a depth-related component, because the 

 samples from the MS were taken at slightly greater 

 depths ( 600-650 m ) than those taken from the US 

 (400-500 m). 



