POLACHECK: YELLOWFIN TUNA CATCH RATES 



mates for different stratifications of the data may 

 indicate possible sources of bias and can provide 

 some indication of the robustness of any temporal 

 trends suggested by any single set of estimates. 



Another approach for dealing with possible biases 

 due to unequal distribution of fishing effort is to 

 calculate standardized catch rates using a general 

 linear model (Gulland 1956b; Robson 1966; Allen and 

 Punsly 1984). The advantage of this approach is that 

 well-developed, standard statistical procedures can 

 be employed to test for significant differences in 

 catch rates over time where the effect of other fac- 

 tors on catch rates have been taken into account. 

 Disadvantages of this approach include: 1) the data 

 may be nonnormal even when transformed, 2) ef- 

 fects may not be simply additive (or multiplicative 

 if a logarithmic transformation is used), and 3) the 

 design matrix is almost always unbalanced and 

 incomplete. 



Extensive attempts were made to fit a general 

 linear model to the catch rate data presented here. 

 While the model was successful in greatly reducing 

 the total sums of squares (e.g., an i?'^ as high as 

 0.80), in all cases, the models included significant 

 and large interaction effects between year and area, 

 and between year and season. Such interactions are 

 an indication of changes in availability and distri- 

 bution between years and are not surprising given 

 the large El Nino of 1983. When large interaction 

 terms exist in a model, particularly when it is un- 

 balanced and incomplete, direct interpretation of the 

 main effects (in this case year) is problematical. An 

 alternative to estimating the main effects in this 

 situation is to develop fjy -models (Searle 1971) to 

 compare directly the average effect between those 

 combination of cells which are of interest. Concep- 

 tually this approach is similar to the stratified means 

 approach developed above, but the calculation of the 

 variance for the stratified means makes no assump- 

 tion about the equality of the variance between cells. 

 Because of the similarity of these two approaches 

 and the problems with traditional general linear 

 model estimates for unbalanced and incomplete 

 data, the results of the general linear model have 

 not been included in the present paper. 



that yellowfin tuna are a homogeneous stock with 

 respect to the two fisheries. For this analysis, it is 

 important that relatively fine scale temporal and 

 area strata be used in order that differences in abun- 

 dance between areas and time do not mask any rela- 

 tionship. Comparison of quarterly longline and purse 

 seine catch rates are made for each individual 2.5° 

 X 10° rectangular area in which there were at least 

 five quarters with a reasonable amount of effort by 

 both gears (i.e., 5 days of purse seine effort and 

 20,000 longline hooks). 



The second approach involves the comparison of 

 changes in longline catch rates in different areas to 

 the purse seine catches that have occurred within 

 these areas. This approach is a direct test of whether 

 any reduction in longline catch rates can be detected 

 as a result of the large catches by purse seiners. A 

 fundamental assumption of this approach is that the 

 stocks of yellowfin tuna within the areas being com- 

 pared are largely spatially distinct or mixing only 

 slowly. If the stock being fished is a homogeneous 

 mixture, then no purse seine-induced differences 

 between areas would occur. 



For this second approach the percentage change 

 in the average 1984-85 longline catch rate, relative 

 to the average 1979-81 catch rate, are calculated 

 for each of the 2.5° x 10° rectangular areas. The 

 average catch rates within an area for the periods 

 1979-81 and 1984-85 were calculated as the sim- 

 ple average of the quarterly rates for an area. The 

 percentage changes between these two periods are 

 then examined in relation to past purse seine catches 

 that have occurred in these areas. These two time 

 periods were chosen for this comparison in order to 

 see whether there has been differential and consis- 

 tent long-term changes in abundances, and if so, 

 whether these changes can be related to the distri- 

 bution of purse seine catches. 



It should be noted that these two approaches are 

 meant to test for specific, possible localized inter- 

 actions (either temporal or spatial). They are not 

 meant as an exhaustive examination of the inter- 

 actions between these two gears, but as feasible 

 analyses given the short time series and limits of 

 the current data. 



Interactions 



The relationship between the longline and purse 

 seine fisheries is considered in detail from two dif- 

 ferent approaches. In the first, catch rates of purse 

 seiners and longliners operating in the same area 

 during the same time period are compared. In this 

 case a strong positive relationship would suggest 



RESULTS 



Purse Seine Catch and Effort 



Effort by Japanese purse seiners increased steadi- 

 ly through the first half of 1982 to around 450 days 

 per month (Fig. lA). Since 1982, levels of effort have 

 remained relatively steady and have fluctuated 



125 



