SEN: SAMPLING COMMERCIAL FISH LANDINGS 



be prevented from taking a sample from another 

 sort by the skipper who may not like some of his 

 fish being cut and otoliths removed for biological 

 studies. This may happen at ports where either pro- 

 cessing facilities are inadequate or fish are bought 

 by local merchants immediately after landing. The 

 question arises if failure to sample from all cat- 

 egories of a sample landing as originally planned 

 would cause appreciable bias and loss in efficiency 

 in the estimates of species catch and its distribution 

 and whether a more efficient method could be 

 developed that is operationally feasible. This point 

 has been examined in the present paper. 



The present technique of selecting a cluster (box) 

 of fish as second stage sampling unit is preferred 

 to random selection of a specified number of in- 

 dividual fish because in practice the potential of per- 

 sonal bias of the sampler could be considerable. 

 Often fish chosen by the latter technique are ones 

 closest to the sampler or those that fell in a certain 

 position. Tomlinson (1971) felt that in this approach 

 the sampler may tend to choose a fish with certain 

 qualities and thus may introduce procedural bias. 



The selection of a representative cluster would de- 

 pend whether samples after sorting on the vessel 

 come from bins, strap boxes, or off conveyor belts. 

 Buyers from small markets occasionally select fish 

 from the top of bins. Hence, to avoid this bias, it 

 is preferable to select the cluster from the conveyor 

 belt which exposes unsorted fish from the lower por- 

 tion of the bin. However, where small market buyers 

 do not buy fish, a cluster may be selected from a 

 bin. Where many bins are present a systematic sam- 

 ple of two clusters, preferably from the beginning 

 and end of the trip may be selected. Where fish are 

 graded on a conveyor belt before they enter the 

 plant (e.g., Fieldslanding at Eureka) the sampler 

 should try to intercept the landings prior to sec- 

 ondary sorting or obtain separate weights for each 

 subsort category. In general, selection of a cluster 

 for a market category should be done before any 

 presorting is done at the port. 



It has been pointed out earlier that bias may result 

 from personal selection of fish within a cluster. If 

 the sampler were to select a number of clusters with 

 few fish per cluster, a cluster will on the average 

 contain more big fish. This would lead to high non- 

 sampling bias. Sometimes, the top few fish in a bin 

 are selected and put there to impress small buyers. 

 The resulting bias in selection can be avoided by 

 taking all the fish in a cluster (e.g., 50 lb) from one 

 side of the box. 



For obtaining reliable and comprehensive infor- 

 mation on population characteristics, it is essential 



for the sampler to maintain good relationships with 

 both the skipper and the buyer; this will depend to 

 a large extent on the expertise of the sampler gained 

 in the course of the field work. 



SUMMARY 



1. The sampling scheme at a port during a month 

 with poststratification of sampled trips into 

 categories and subsampling of clusters from 

 each category (see sections on Estimation with 

 poststratification and Estimation ignoring 

 category variation) is not workable for esti- 

 mating rockfish catch since some of the 

 categories may be missed in sampling due to in- 

 adequate field staff and/or management 

 problems. 



2. For other commercial fish where the above 

 problem does not exist and landing weights by 

 categories are not available at the end of the 

 season, the methods (see sections on Estima- 

 tion with poststratification and Estimation ig- 

 noring category variation) are recommended, 

 e.g., for single species where the categories are 

 based on size. 



3. For estimating the catch of rockfish, a two- 

 stage sampling plan is recommended with boat 

 trips as first stage units poststratified into 

 categories and clusters subsampled from a 

 category; estimates are based on categories as 

 domains of study with landing weights available 

 for each category. A minimum of four landings 

 or boat trips should be used for each category, 

 to provide efficient estimates. With few categ- 

 ories, this number is likely to be large. 



Where only one category is subsampled for 



each boat in the sample, v(Y 3R ) = ^ V (YX In 



3 



all other cases Equation (13) should be used. 



4. The design described in the above paragraph is 

 recommended for use in other fisheries where 

 landing weights are available for each category. 

 Equations (9) and (21) are recommended for the 

 estimation of catch according as the clusters are 

 of equal or unequal size. Equations have been 

 provided for the more practical case when 

 cluster weight can be treated as constant with- 

 in a category but different among catego- 

 ries. 



5. Estimates of species catch by sex and age based 

 on method 1 are less efficient than those based 

 on method 2 which is based on categories as do- 

 mains of study (Tables 2, 3). 



6. Method 2 is preferred to method 1 when there 



419 



