FISHERY BULLETIN: VOL. 87, NO. 1 



available data, the analyses in this present paper 

 are based on catch rates for Japanese vessels. 

 These data form the most extensive and complete 

 set of data currently available to the SPC. Records 

 of daily fishing activity go back as far as the second 

 half of 1978 for the longline fishery and 1979 for 

 the purse seine fishery. However, these earliest data 

 are not complete and need to be interpreted with 

 caution. 



The stock or subpopulation structure and their 

 geographic limits for yellowfin tuna in the Pacific 

 are unknown despite considerable tagging and gene- 

 tic research (Cole 1980). However, a single stock 

 spanning the entire Pacific is considered unlikely. 

 For the present paper, the geographic boundaries 

 used for the western Pacific are from long. 130°E 

 to 180°E and from lat. 10°S to 15°N. For the Japa- 

 nese purse seine fishery, this area encompasses vir- 

 tually all of the reported catch and effort data. For 

 the Japanese longline fishery, this represents an 

 area in which the fishery has been relatively con- 

 sistent and its reporting fairly complete. 



Catch Rates 



Catch rates or catch per unit effort are calculated 

 below as a measure of relative abundance. An ex- 

 tensive literature exists on the use of catch rates 

 as abundance indices (Gulland 1956a; Beverton and 

 Holt 1957; Paloheimo and Dickie 1964; Allen and 

 Punsley 1984). However, the question of the rela- 

 tion between catch rates and abundance for these 

 yellowrfin tuna fisheries needs further research (see 

 Discussion). 



For longlining, the effort measure used here is the 

 number of hooks set (in thousands). Catch is 

 reported as the number of fish caught. For purse 

 seining, the effort measure used is the number of 

 days in which vessels made a set or were actively 

 searching for schools of tuna. The catch is recorded 

 in metric tons. In the earliest purse seine data, there 

 may be an underestimation of effort, as it is not 

 clear whether days in which vessels were searching 

 for fish, but did not catch any, were accurately 

 reported. 



The average catch rates and their variances with- 

 in any statistical stratum were calculated as the 

 weighted mean of the observed catch rates for all 

 cruises within the stratum. Thus, an individual 

 cruise's catch rate within a stratum constitutes the 

 primary sampling unit or replicate in the analyses 

 below. The weights used were equal to a vessel's 

 fishing effort. For the estimates of the mean catch 

 rate, this is equivalent to the sum of the total catch 



divided by the sum of the total effort within a 

 stratum. 



Various temporal and areal stratifications of the 

 data have been considered. Monthly, quarterly, and 

 annual stratifications are examined. When the data 

 were stratified by area, geographic strata were 

 defined as rectangular areas of 2.5° of latitude and 

 10° of longitude. These strata were chosen because 

 preliminary analyses indicated that there was much 

 greater variation both in effort and catch rates lati- 

 tudinally than longitudinally. If smaller areas are 

 selected, there tends to be too little data in many 

 of the strata for meaningful analysis. 



There are two statistical reasons for stratifying 

 data: 1) to eliminate biases due to unequal distribu- 

 tion of sampling effort in strata with different 

 means, and 2) to reduce the variance associated with 

 the estimate of the mean. The first reason is a 

 primary concern in calculating catch rates from 

 fisheries data since the distribution of fishing effort 

 both spatially and temporally is likely to be related 

 to catch rates (i.e., fishermen probably concentrate 

 on when and where the fishing is best). 



In order to estimate an average catch rate for time 

 periods and areas of interest, the estimates of the 

 catch rates in the various strata need to be com- 

 bined. For stratified data, an estimate of the aver- 

 age catch rate across strata is the weighted mean 

 of the average catch rate within each stratum, 

 where the weights are proportional to the magni- 

 tude of a stratum (Snedecor and Cochran 1967). The 

 geographical and temporal stratifications presented 

 below were considered to be equal in area and time. 

 (This is not strictly true both because of land masses 

 and differences in the length of a degree of longi- 

 tude at different latitudes. For two of the geograph- 

 ical strata, the amount of land area of Papua New 

 Guinea is large and these two strata should prob- 

 ably be given smaller weight in any extensions or 

 refinements to the estimates presented below.) 

 When all strata are of equal magnitude, the aver- 

 age catch rate across strata is the simple average 

 of the within-strata estimates. Similarly, in this 

 situation, an estimate of the variance is the average 

 of the variance estimates for each stratum (Snede- 

 cor and Cochran 1967). 



Because catch and effort statistics are not derived 

 from a well-designed and controlled sampling ex- 

 periment, there is not an a priori single best esti- 

 mate for the average catch rate covering large areas 

 and time periods. Thus, when considering estimates 

 of the annual average catch rates, a set of different 

 estimates based on various areal and temporal strat- 

 ifications are presented. Comparison of the esti- 



124 



