684 



Fishery Bulletin 100(4) 



sharks, other fish, etc.) at about 762 t (Ariz and Gaertner, 

 1999). 



Based on the ecological provinces in the oceans estab- 

 hshed by Longhurst (1998), the eastern tropical Atlantic 

 Ocean (from 25°N to 15°S and from 35°W to the African 

 coasts) was stratified by quarters into two areas: 



the Senegalese area (from latitude r2°N to 25°N), 

 the remaining areas, termed "equatorial" areas. 



Owing to time constraints during the set (and bearing in 

 mind that this program was directed at bigeye tuna), it 

 was very difficult for the observer to accomplish some addi- 

 tional bycatch tasks. Consequently, in some circumstances 

 the billfish species may not have been correctly identified. 

 For this reason we established two groups: the blue marlin 

 [Makaira nigricans) and the white marlin iTetrapterus 

 albidus) in one group and the sailfishes (Istiophorus 

 albicans) and the longbill spearfish (T. pfluegeri) in the 

 second group. The weight ranges of billfishes captured as 

 incidental catch by the purse seiners were approximately 

 130-150 kg for blue marlin, 10-20 kg for sailfish and 

 45-70 kg for nonidentified marlins (there were no weight 

 estimates for white marlin and longbill spearfish because 

 of their low numbers in the bycatch). 



On the basis of a study made on tuna size classes by 

 set type in this fishery (Pallares and Petit, 1998), the sets 

 made on whales and on whale sharks iRhiniodon typiis) 

 were classified as school sets (i.e. sets made without FADs) 

 and as FAD sets, respectively. In contrast, it must be 

 stressed that seamounts constitute a specific environ- 

 ment for small size classes of tuna species. In the Atlantic 

 Ocean 2000-3000 t of tunas can be taken yearly on some 

 seamounts (Fonteneau, 1991). Because large pelagic spe- 

 cies can be concentrated on seamounts (e.g. sphyrnids, 

 Klimley et al., 1988), we distinguished the sets made on 

 seamounts from the usual tuna fishing modes (i.e. school 

 sets and FAD sets). 



Methods 



For the two groups of billfishes (i.e. sailfishes and marlins) 

 the total bycatch taken by the European tuna purse-seine 



fishery in the eastern Atlantic Ocean was estimated for 

 a period of 12 months. The period between October 1997 

 and September 1998 was considered the best represen- 

 tative standard year for the observer program because 

 it included the best coverage in terms of tuna catches 

 (approximately 17% of the total tuna catch is taken by the 

 European purse seiners from October 1997 to September 

 1998). Assuming that billfish bycatch was proportional 

 to the tuna catch, the observed bycatch for each billfish 

 group was raised to the total European purse-seine catch 

 with the use of a raising factor RF^, ,, as in the following 

 equation: 



^'S^^ = ZSX^^-*'S'^' 



sijk"* 



where TBC 



total bycatch for the group s during a stan- 

 dard year; 



RF^ 1^ = total purse seine tuna catch,^|(./observed 

 aboard purse seine tuna catch,^^; 



BC\^ii, = bycatch for the group .s, in area (', quarter j, 

 and fishing mode k\ 



and / = 1, 2;j = 1, 2, 3, 4; ^ = school set, FAD set, set 

 on seamount. 



Sources of uncertainty in the calculation of the billfish 

 bycatch are caused by 1 ) changes in fishing strategies 

 adopted by the fishermen over the year, e.g. the probability 

 of choosing the fishing mode k and 2) some features of the 

 bycatch species, e.g. the conditional probability for a given 

 group of billfish s to be present in a set of type k, as well 

 as the probability of obtaining .v tons of group s in the 

 set (Table 1). To account for some of these uncertainties 

 our approach differed from that of Perkins and Edwards 

 (1996), who used a modified negative binomial distribu- 

 tion to model bycatch per set. We used computer-intensive 

 methods, such as a Monte Carlo simulation, to estimate the 

 total bycatch generated by the European tuna purse-seine 

 fishery. Monte Carlo methods of simulation introduce a 

 large number of random inputs into a model while record- 

 ing the range of outputs (Gaertner et al., 1996; Shelton et 

 al., 1997). 



In the present analysis, the model mimics the fishing 

 operations made by the purse seiners over a standard 



