Indices of brood year abundance. — Troll catch, troll CPUE, 

 and ocean sport catch off Washington were highly correlated 

 with 1) number of late-run (October-November) coho salmon 

 per mile of spawning stream on the lower Columbia River and 2) 

 Washington gill net CPUE in the Columbia River below Bon- 

 neville Dam during October through November in year i— 3. 



Hatchery production. —Troll catch, troll CPUE, and ocean 

 sport catch in nearly all areas were not significantly associated 

 with hatchery releases (either in total or by production area) and 

 a number of the correlations were negative. 



Environmental conditions. — We found no consistently signifi- 

 cant associations between ocean fishing success and any en- 

 vironmental factor. The highest correlation was between troll 

 catch in the Columbia River district and the index of upwelling 

 off Cascade Head in April of year /' - 1 , and was nearly signifi- 

 cant. 



Average size and growth increments. — We found no significant 

 relationships between indices of fishing success and these fac- 

 tors. 



Jack counts. — Troll catches off the central and southern 

 Washington coast were significantly associated with Columbia 

 River jack returns in the previous year. 



Prediction of ocean catches. — Most of the Washington ocean 

 catch is made south of La Push. The best predictive equation 

 for annual troll catch in the Columbia River district was 



In R = In a - 0.4334S + O.OlllA', + 0.0057AT 2 + In S (2) 



where R was troll catch in year i, S was the count of late-run 

 Columbia River coho salmon per mile of spawning stream in 

 year i— 3, X, was the Columbia River jack count in year /— 1, 

 and X 2 was the index of upwelling off Cascade Head during 

 April of year /— 1. This multiple regression explained 89% of 

 the variation in annual troll catch during 1966-75, with an 

 average percentage error of the predicted catches of ±15% 

 (Table 6). The best model for annual troll catch in the Coastal 

 Washington district was 



Table 7.— Comparison of annual coho salmon troll catches in the 

 Coastal Washington district with those predicted from Equation (3). 





Actual catch 



Predicted catch 



Percentage 



Year 



(fish) 



(fish) 



error 



1966 



431.592 



448,500 



+ 3.9 



1967 



360,046 



278,000 



-22.8 



1968 



377,690 



461,500 



+ 22.2 



1969 



200,665 



283,500 



+ 41.3 



1970 



475,657 



624,000 



+ 31.2 



1971 



640,230 



672,000 



+ 5.0 



1972 



248,329 



292,500 



+ 17.8 



1973 



484,833 



333,500 



-31.2 



1974 



642,614 



476,000 



-25.9 



1975 



453,620 



395,500 



-12.8 



HOO-i 



1000- 



900- 



X 800- 

 10 



fe TOO- 



in 

 o 



Z 600- 



<t 



CO 



r> 



g 500- 



400 



300 



| ACTUAL 

 □ PREDICTED 



Average error is 15% 



In R = In a - 0.0642S + 0.0O78A', + In S 



(3) 



Figure 6. — Annual commercial catch of coho salmon in the Columbia River and 

 Coastal Washington areas, 1966-75 and that predicted from Equations (2) and (3). 

 Numbers at the tops of the columns indicate percentage error. 



where R was annual catch and S and X, were as defined above. 

 This regression also accounted for about 89% of the variation in 

 troll catch during 1966-75, with an average error of about 



±21% (Table 7). Figure 6 illustrates the comparison of the 

 predicted catches combined for both districts with the catches 

 actually reported. 



Table 6. — Comparison of annual coho salmon troll catches in 

 the Columbia River district with those predicted from Equation (2). 





Actual catch 



Predicted catch 



Percentage 



Year 



(fish) 



(fish) 



error 



1966 



183.73! 



162,500 



-11.6 



1967 



236,429 



235,500 



- 0.4 



1968 



184,825 



178,000 



- 3.7 



1969 



157,791 



167,500 



+ 6.2 



1970 



205.541 



243,500 



+ 18.5 



1971 



469,629 



427.500 



- 9.0 



1972 



216,233 



163,000 



-24.6 



1973 



103,715 



155,000 



^49.4 



1974 



186,230 



207,000 



+ 11.2 



1975 



159,810 



131.000 



-18.0 



DISCUSSION 



Although ocean fishing success for pink salmon is roughly 

 associated with brood year abundance of major stocks, the rela- 

 tionship has little predictive value for specific ocean fishing 

 areas. The relative abundance and contribution of each stock in 

 such areas probably vary widely in different cycle years. Until 

 stock composition in specific ocean areas is better defined, the 

 predictive value of spawner-recruit models based on brood year 

 abundance of specific stocks will remain limited. The rate of 

 ocean exploitation is increasing and, combined with extreme 

 fluctuations in cycle year abundance, makes accurate 

 forecasting of ocean catches of pink salmon very difficult. 



10 



