Abstract.— Density and biomass 

 estimates from a shrimp trawl and a 

 visual survey are compared. Results 

 indicate that the visual survey gives 

 estimates eight to nine times larger 

 than the trawl survey. However, 

 there are important variations by 

 species. Estimates of fish size from 

 the visual censuses are larger than 

 fish sizes observed in the trawl catch. 

 These results suggest that the catch- 

 ability of fish by shrimp trawls may 

 be lower than usually thought for 

 multispecies tropical stocks. In shrimp 

 trawl fisheries, this may have impor- 

 tant consequences for stock assess- 

 ment and for evaluation of param- 

 eters such as fishing mortality. 

 Trawling should be adequate for 

 qualitative and semiquantitative 

 stock assessment, but will need to be 

 compared with other methods for 

 quantitative studies on multispecific 

 stocks. 



Comparison Between Fish Bycatch 

 from Slirimp Trawlnet and Visual 

 Censuses in St. Vincent Bay, 

 Mew Caledonia 



Michel Kulbicki 

 Laurent Wantiez 



ORSTOM, Centre de Noumea 



B P A5 Noumea Cedex, New Caledonia 



Manuscript accepted 29 June 1990. 

 Fishery Bulletin, U.S. 88:667-67.5. 



Shrimp trawl fisheries are the source 

 of a large fish bycatch in the tropical 

 Indo-Pacific (Aoyama 1973, Gran- 

 tham 1980, Villoso and Hermosa 

 1982, Chong 1984). One question con- 

 cerning these fisheries is how much 

 damage they cause to fish stocks. To 

 answer this question, it is useful to 

 assess the catchability of fish by 

 shrimp trawls. Many methods exist 

 to assess this parameter. Most use 

 the same gear under different condi- 

 tions, such as double codends, split 

 codends, and repHcate hauls with dif- 

 ferent mesh sizes (Macket 1973, Gul- 

 land 1975). The bias of such methods 

 may be acceptable for single-species 

 populations but has seldom been 

 studied in relation to tropical multi- 

 species stocks. Another approach is 

 the cross-evaluation of trawling and 

 echosurveys (Doubleday 1976, Dines 

 1982, Dickie et al. 1983), but this 

 method is not well adapted to tropical 

 multispecies stocks. Another way 

 is to correlate the results of two 

 methods, such as with a TV camera 

 and otter trawl (Uzmann et al. 1977), 

 submersible and longlines (Ralston et 

 al. 1986, Richards and Schnute 1986, 

 Grimes et al. 1982), visual census and 

 longlines (Kulbicki 1988) or poisoning 

 and beamtrawl (Gray and Bell 1986). 

 The use of several methods has the 

 advantage of allowing some evalua- 

 tion of the biases of each method. In 

 the present study it was possible to 

 compare the results of visual cen- 

 suses and shrimp trawl bycatches. 



This work was originally designed to 

 estimate the catchability of the net (q 

 = number of fish caught/number of 

 fish present), but it also provided in- 

 formation on the biology of the fish. 



Material and methods 



Location 



The study was performed during 

 August 1986 on the trawling grounds 

 of the South Bay in St. Vincent Bay, 

 in the southwest lagoon of New Cale- 

 donia. The trawling gi-ounds cover 15 

 km-, or approximatively one-tenth 

 of the Bay area. A total of eight sta- 

 tions were sampled (Fig. 1), one of 

 which was sampled at night (Station 

 n°6). 



Visual census 



A 200-m transect line was laid on the 

 bottom with a buoy set at each end 

 of the line. The transect was divided 

 into 10-m sections. For each section, 

 two divers— one on each side of the 

 line— counted all fish, estimated their 

 length, and estimated the perpen- 

 dicular distance of the fish to the 

 transect. Length was given in 2-cm 

 size classes for fish less than 20 cm, 

 5-cm size classes for fish between 20 

 and 50 cm, and 10-cm size classes for 

 fish larger than 50 cm. Only fish less 

 than 1.5 m above the bottom were 

 counted. The distance from the fish 

 to the transect was recorded in 1-m 



667 



