Kitadd and Tezuka Survey designs for estimating recreational fishery catch 



237 



Although the difference in the catch rate between day- and 

 season-permit anglers was unknown, the influence of this 

 bias on the total catch estimates would be minor In order 

 to check the bias, however, one could conduct a logbook 

 survey of day-permit anglers. Sixteen anglers of the total 

 sample ( 13'/r I in our study did not return logbooks and 

 therefore may have caused a bias in our estimates; however 

 no attempt was made to evaluate the difference between 

 nonrespondents and respondents. The angler sample was 

 drawn arbitrarily by the cooperatives but was not a random 

 sample in the strictest sense. If cooperative anglers tended 

 to be selected, this could have been a source of bias. 



The source of variation in total catch is the variation 

 in the catch of the sampling unit, including differences in 

 fishing days, skill of the anglers, and the number of an- 

 glers that a party boat could accommodate. A stratified 

 sampling scheme based on categories of anglers or boats is 

 effective for this situation. The weakest point in the use of 

 logbook surveys, perhaps, is that the catch data are report- 



ed by those who catch the fish and by boat owners with 

 monetary interests. To what extent the anglers might have 

 exaggerated or under- reported their catch is not known. 

 Party boat owners may record lower than actual catches to 

 reduce taxes. To examine this possible source of bias, on- 

 site sui-veys should be conducted. For the ayu fishery in 

 the Nakagawa River, an access point survey may be prac- 

 tical (Pollock et al., 1994). When comparatively complete 

 lists of boat owners and anglers are available, logbook sur- 

 veys based on these lists, combined with on-site surveys, 

 are appropriate. 



Longitudinal and stratifled survey designs 



Longitudinal surveys taking repeated measures on the 

 same random sample over time are better than successive 

 independent surveys if the designer's goal is to show sta- 

 tistically significant differences in the estimates between 

 time intervals. Monthly estimates showing seasonal trends 



