74 



Abstract— In trawl surveys a duster 

 of fish are caught at each station, and 

 fish caught together tend to have more 

 similar characteristics, such as length, 

 age, stomach contents etc., than those 

 in the entire population. When this 

 is the case, the effective sample size 

 for estimates of the frequency distri- 

 bution of a population characteristic 

 can, therefore, be much smaller than 

 the number of fish sampled during a 

 survey. As examples, it is shown that 

 the effective sample size for estimates 

 of length-frequency distributions gen- 

 erated by trawl surveys conducted in 

 the Barents Sea, off Namibia, and off 

 South Africa is on average approxi- 

 mately one fish per tow. Thus many 

 more fish than necessary are measured 

 at each station (location). One way to 

 increase the effective sample size for 

 these sui"veys and, hence, increase the 

 precision of the length-frequency esti- 

 mates, is to reduce tow duration and 

 use the time saved to collect samples at 

 more stations. 



Assessing the precision of frequency distributions 

 estimated from trawl-survey samples 



Michael Pennington 



Institute of Marine Research 

 Department ol Marine Resources 

 Nordnesgaten 33 

 N-5005 Bergen, Norway 

 E mail address michaeliaimrno 



Liza-Mare Burmeister 



Ministry of Fishenes and Marine Resources of Namibia, NatMIRC 

 PO Box 912 

 Swakopmund, Namibia 



Vidar Hjellvik 



Institute ol Manne Research 

 Department of Marine Resources 

 Nordnesgaten 33 

 N-5005 Bergen, Norway 



Manuscript accepted 21 August 2001. 

 Fish. Bull. 100:74-80 (2002). 



Survey-based assessments often appear 

 to provide a more accurate prognosis of 

 the status of a fish stock than catch- 

 based assessments (Nakken, 1998; Pen- 

 nington and Stromme, 1998: Kors- 

 brekke et al., 2001). Aji advantage that 

 sui-vey-based assessments have over 

 those based on commercial catch sta- 

 tistics is that the uncertainties asso- 

 ciated with survey estimates can be 

 studied and quantified, and based on 

 such research, survey methods, and 

 ultimately stock assessments, can be 

 improved (Godo, 1994). In contrast, it is 

 generally difficult to determine either 

 the accuracy or the precision of esti- 

 mates based on commercial catch data, 

 and it is not clear how to improve, at a 

 reasonable cost, the collection of catch 

 data so that these data would more 

 accurately reflect the mortality caused 

 by fishing (Christensen, 1996). 



Trawl surveys provide estimates of 

 the abundance or relative abundance of 

 a fish stock and estimates of the relative 

 frequency of various population charac- 

 teristics, such as length, age, and stom- 

 ach contents. In our study we examined 

 the precision of survey-based estimates 

 of the length-frequency distributions of 

 cod and haddock in the Barents Sea, 

 hake off South Africa, and hake off 



Namibia. The focus was on length, 

 but the results are relevant for es- 

 timating the frequency distribution 

 of other population characteristics. 



Materials and methods 



Survey length data 



Bottom trawl survey length data for 

 Northeast Artie cod ^Gadun mohua) 

 and Northeast Arctic haddock (Mela- 

 nogrammus aeglefinusY were collected 

 during the Institute of Marine Research 

 (Norway) winter and summer surveys 

 in the Barents Sea. The surveys were 

 stratified systematic surveys and at 

 each station the trawl was towed for 30 

 minutes. - 



Also known as "Atlantic cod" and "had- 

 dock," respectively, according to Common 

 and scientific names of fishes from the 

 United States and Canada. 1991. Am. 

 Fish. Soc. Spec. publ. 20. Besthesda, MD, 

 183 p. 

  Aglen, A. 1999. Report on the demersal 

 fish sui-veys in the Barents Sea and Sval- 

 bard area during summer/autumn 1996 

 and 1997. Unpubl. manuscr. Fisket og 

 Havet NR. 7-1999. Institute of Marine 

 Research. PO Box 1870 Nordnes. N-5817 

 Bergen. Norway. 



