Kulbicki and Wantiez- Estimates of fish stocks by shrimp trawl and visual survey off New Caledonia 



673 



Because of the low number of stations, this difference 

 in sampled area may also be important to explain the 

 low correlation between the two methods for density 

 or biomass estimates of fish with a patchy distribution. 

 Conversely, comparisons of the density or biomass for 

 fishes with a regular distribution should not be affected. 



An important result from the present survey is the 

 magnitude of the difference between trawl and transect 

 estimates of density. This difference, which is a factor 

 of 9.7 for all species and a factor of 7.9 for the main 

 species, certainly indicates that shrimp trawls poorly 

 sample the bottom fish fauna in St. Vincent Bay. Visual 

 transects are usually performed as strip transects, 

 where distance of the fish to the transect line is not 

 taken into account (Harmelin-Vivien et al. 1985). This 

 underestimates density, especially if the width chosen 

 is large (Burnham et al. 1980). The method chosen in 

 the present survey is more accurate; however, it is not 

 possible to know by how much one underestimates or 

 overestimates the "real" density. Fish numbers (Har- 

 melin-Vivien et al. 1985) and distances are both under- 

 estimated but play opposite roles in the density estima- 

 tion. Even in the unlikely case of a large overestimate 

 of fish density (let's suppose by 100%) by transects, the 

 catchability of the trawl would still be 0.25 for the main 

 species and 0.21 for all species. 



Several surveys indicate that trawls may not be as 

 efficient as usually thought. Thus Uzmann et al. (1977) 

 found that estimates of fish density from submersibles 

 were 8.0 times greater than density estimates from 

 otter trawls, this ratio varying between 0.8 and 18 

 depending on the species. Similarly, Gray and Bell 

 (1986) found that poisoning indicated densities 2.5 to 

 6.5 times greater than a beam trawl over seagrass 

 beds. However, in some conditions trawling may be 

 more efficient, as shown by Harden Jones et al. (1977) 

 who used accoustical tags to estimate that 44% of the 

 tagged flatfishes Pleuronectes platessa were caught by 

 a Granton otter trawl. Uzmann et al. (1977) had found 

 that only 7% of the flatfishes seen from a submersible 

 were caught by the otter trawl. Serebrov (1986) used 

 submersibles to show that the catchability of an otter 



trawl varied with the size of the fish, mesh size, and 

 overall catch composition. 



In the case of one or a few species, catchability of 

 trawls may vary between 0.5 and 1.0. Such values of 

 catchability are widely used, even in multispecies 

 tropical fisheries, for which Pauly (1982) states that a 

 value of 0.5 is "realistic" and Gulland (1979) even sug- 

 gests that a catchability of 1.0 should apply to the 

 eastern Indian Ocean trawl fisheries. In many of these 

 fisheries, shrimp trawls of design similar to the one 

 used in the present experiment are used to get a mixed 

 catch of shrimp and fish (Grantham 1980, Poiner and 

 Harris 1986, Lamboeuf 1987, Sainsbury 1987). Our 

 results indicate a catchability near 0.1 (0.103 for all 

 species and 0.127 for the main species). Such a value 

 may need further testing; but if it proved correct, it 

 would have important implications in the management 

 of these tropical multispecies trawl fisheries. In par- 

 ticular, depending on whether one uses a catchability 

 of 1.0 or 0.1. stock size will increase by a factor of 10. 

 This would also affect the estimated fishing mortality 

 (Pauly 1982) and, as a result, most of the equations used 

 in stock management. What is more important, in our 

 opinion, is that if one applied a catchability of 0.1 in 

 most of these tropical trawl fisheries of the Indo- 

 Pacific, the current models would be unable to explain 

 the long-term decline of the catch seen in these fish- 

 eries. This could mean that trawling induces detrimen- 

 tal changes in fish populations which are beyond the 

 simple removal of fish. In particular, habitat changes 

 may be important, as noted l.iy Poiner and Harris (1986) 

 and Sainsbury (1987). 



Another question arising from the present work is 

 the value of the catch per unit effort (CPUE) of trawls 

 as an indicator of population abundance. Our results 

 show a poor correlation between abundance estimates 

 from trawls and visual transects at the population level, 

 but good correlations for a few selected species. Nu- 

 merous studies have examined the correlation of CPUE 

 by trawling and acoustic surveys. Most of these indicate 

 good correlations (see Tesler (1977), Olsen et al. (1977), 

 or Thorne (1977a) among others) but important varia- 



